International Journal of Computational Vision and Robotics (44 papers in press)
Adaptive Robust Control of a Four-Cable-Driven Parallel Robot
by Arash Kiani, Seyed Kamal-e-ddin Mousavi Mashhadi
Abstract: This present study introduces an adaptive control strategy for Four-cable robots. An adaptive sliding mode control to overcome the uncertainties of the system as well as avoidance of estimating an upper bound of the system uncertainties is presented. The proposed controller is designed based on the Lyapunov stability theory. Therefore, it ensures the stability of the closed-loop system and makes the tracking error converge to zero. In this robot, the cables can only pull the end-effector but not push it; therefore we present a simple mathematical solution to design a positive tension controller for the cable suspended robot with redundant cables. The properties of the proposed method such as high performance tracking, disturbance rejection and insensitivity to parameter variations are demonstrated by simulation.
Keywords: Cable robots; Adaptive sliding mode control; Lyapunov stability Adaptive Inverse Dynamic Control; Positive cable tensions;.
Computing disparity map using Minimum Sum Belief Propagation for stereo pair images
by Chitra Suresh, Kushal R. Tuckley
Abstract: Stereo matching between two images is done by computing disparity of all points on the object. The process involves identifying corresponding points in stereo image and finding the horizontal shift. Presently there is no method that finds the shift in the corresponding points between left and right images, this is due to non-availability of procedure to identify the group of pixel in the right and left image of the same object. The available local methods either uses window or feature to find shift in a stereo image. In these methods, finalizing size of the window or deciding the correct feature remains an unresolved issue. On the other hand, global methods uses graph theory and probability theory to find the shift efficiently.
The belief propagation algorithm is one of the global method devised to offer computationally efficient approach with good results. This paper has applied "Minimum Sum Belief Propagation" method for message updates with linear "Quadratic function "for computation of horizontal shift in stereo image. The results with the computational estimations are presented hereby and based on these results, suggestive comments on effectiveness of update which indicate strategy versus type of the image are also mentioned.
Keywords: Stereo Image; Parallax Effect; Stereo Matching; Markov Random Field; Belief Propagation; Disparity Map.
Extended Opinion Lexicon and ML based Sentiment Analysis of tweets : A novel Approach towards Accurate Classifier
by Gaurav Dubey, Santosh Kumar, Sunil Kumar, Pavas Navaney
Abstract: Micro-blogging, today has become a very trendy communication tool among internet users. Millions of users share their opinions on diverse aspects of life which are rich sources for opinion mining. This paper addresses the sentiment analysis of twitter data on Demonetization. An new approach to sentiment analysis based on extended opinion lexicon-based-scores has been presented in this paper. Na
Keywords: Sentiment analysis; NLP; opinion mining; lexicon.
Adaptive Neuro-Fuzzy Inference System based On-the-Move Terrain Classification for Autonomous Wheeled Mobile Robots
by Rakesh Kumar Sidharthan, Ramkumar Kannan, Seshadhri Srinivasan
Abstract: Building intelligence in autonomous robots to classify heterogeneous terrains on-the-move is a challenging task, but a pivotal feature required for accomplishing safety critical missions. This paper proposes an adaptive neuro-fuzzy inference system for online terrain classification in wheeled mobile robot using the steady state behavior of robot wheel on the terrain. The key idea is to model the wheel-terrain interactions as a parametric varying system, whose steady state behaviors are characterized by the terrain type. The proposed method use the steady state gains and the corresponding input command to robot wheel for identifying the terrain type. Our results show that the proposed approach has classification accuracy of 95.2 % for the trained terrains, whereas 94.2 % and 93.8 % are observed in robust and adaptive testing, respectively. Additionally a customized graphical user interface is developed in order to provide easy access to the researchers for terrain identification.
Keywords: Autonomous robots;Adaptive neuro-fuzzy inference system;Terrain classification;Wheel-terrain interactions;User interface.
Characterizing Local Feature Descriptors for Face Sketch to Photo Matching
by Samsul Setumin, Shahrel Azmin Suandi
Abstract: Sketch and photo are from a different modality. Inter-modality matching approach requires right feature representation to represent both images so that the modality gap can be neglected. Improper feature selection may result in low recognition rate. There are many local descriptors have been proposed in the literature, but it is unclear which descriptors are more appropriate for inter-modality matching. In this paper, we attempt to characterize local feature descriptors for face sketch to photo matching. Our evaluation for the characterization uses Cumulative Match Curve (CMC), and we compare seven different descriptors that are LBP, MLBP, HOG, PHOG, SIFT, SURF and DAISY. The evaluation focuses only on a viewed sketch. Based on the experiments, we observed that gradient-based descriptors gave higher accuracy as compared to the others. Out of five popular distance metrics evaluated, L1 gives a better result as compared to the other similarity distance measures.
Keywords: Local feature descriptors; sketch to photo; matching; forensic sketch;face recognition.
Extended COCOMO: Robust and Interpretable Neuro-fuzzy modeling
by Shailesh Tiwari
Abstract: The software project management activities deal with a set of tasks which spans in every phases of the software development life cycle. Prediction of Software development efforts is one of the crucial activities in software project management. Various software cost estimation are developed by the researchers in last few decades. But still, search for the perfect model for software cost estimation has become most difficult task of the organisations dealing in software development. Constructive Cost Model (COCOMO) is one most acceptable model in recent years. This paper presents the extended version of COCOMO. This extension is done with the help of two very popular methods i.e. artificial neural networks (ANN) and fuzzy logic, which ultimately provide the foundation for effort assessment models. Firstly, the expert judgement about model is used for validation, and overpowers the common software engineering black box problem that arises widely in ANN-based solutions. Moreover, we choose the best combination of one of the three membership functions for continuous-rating values which reduce the variance while estimating the cost of similar projects. The validation, using 93 NASA projects Dataset, shows that the model significantly improves the estimation accuracy in terms of mean magnitude of relative error (MMRE) by 10.104 %relative to other known estimation models.rn
Keywords: Fuzzy Logic; Neural Network; COCOMO; Neuro-Fuzzy Software Effort Estimation; NASA projects Dataset; and Mean Magnitude of Relative Error.
Structured Learning and Prediction in Face Sketch Gender Classification and Recognition.
by Khalid Ounachad, Mohamed Oualla, Abdlghani Souhar, Abdelalim Sadiq
Abstract: Structured prediction methods have become an attractive tool for many machine learning applications. For this raison, the objective of this paper is to identify the gender of the human being by using their face sketch applying a structured learning approach. We used a deep geometric descriptor as features and the gender as labels, and structured learning and prediction approach as matching. The basic idea is to extract perfect face ratios for the face sketch as a feature and the labels are the gender. To extract perfect face ratios, we use the landmarks point in the face then sixteen features will be extract. The training and the testing tasks are applied to CUHK Face Sketch dataset (CUFS). An experimental evaluation demonstrates the satisfactory performance of our approach on CUFS and the recognition rate reaches more than 98%.
Keywords: Structured Learning; Prediction; Face Sketch; Face Sketch Recognition; Facial Gender Recognition; Perfect Face Ratios.
Mathematical Variable Detection in Document Images
by Bui Phong, Hoang Manh Thang, Le Thi Lan
Abstract: Mathematical expression detection in documents is a prerequisite step for developing a mathematical retrieval system that has attracted many researches recently. In the detecting process, one challenging issue is the detection of variables. The similar properties of variables and narrative text cause many errors in the detection in existing approaches. In the paper, a novel detection methodology of variables in inline mathematical expressions is proposed. The merit of the method is that it can operate directly on the variable images without the employment of character recognition. The proposed method uses the features of Projection Profile of images and the fine-tuning of different machine learning algorithms in the detection process. The achieved accuracy varies from 86.14\% to 94\% for the detection of variables in inline expressions in document images in various public benchmark datasets. The performance comparison with existing methods demonstrates the effectiveness of the proposed method.
Keywords: Document analysis; Mathematical expression extraction; Italic detection; Machine learning.
Taylor Rate-Distortion trade-off and Adaptive block search for HEVC Encoding
by Anitha Kumari R.D, Narendranath Udupa A
Abstract: The advancement in High efficiency video coding (HEVC) is adapted for defining the subsequent generation compression model for offering efficient compression without affecting the image quality. The HEVC offers improved performance than the existing compression models. This work develops an approach for video compression by proposing weighted entropy coding and adaptive block search based Rate-Distortion (R-D) trade-off. A new R-D trade-off, named Taylor R-D trade-off, is designed using Taylor series. The adaptive block search algorithm is proposed for initiating the block search process of motion estimation in video coding by selecting the optimal block using the Hexagon Based Tree Search Algorithm (HBTSA), along with the Taylor R-D trade-off. Initially, the frames are extorted from the input video. Then, the video frames are divided into macroblocks to perform the adaptive block search. Further, the suitable blocks are selected and given to the encoding process by weighted Context-Adaptive Binary Arithmetic Coding (CABAC) that employs a weighted entropy function to persist the video quality after the compression. The results evaluate that the proposed HBTSA method shows improved PSNR and SSIM using Football, coast guard, garden, and, tennis with values 42.717dB, and 0.991, respectively.
Keywords: Video coding; HEVC; R-D trade-off; Taylor series; Adaptive block search.
Guidance Based Improved Depth Upsampling With Better Initial Estimate
by Chandra Shaker Balure, Ramesh Kini M
Abstract: Like optical images, depth images are also gaining popularity because of its use in many applications like robot navigation, augmented reality, 3DTV and more. The commercially available depth cameras generate depth images which suffer from low spatial resolution, corrupted with noise, and missing regions. Such images need to be super-resolved, denoised and inpainted before using them to have better accuracy. Super-resolution (SR) techniques can be used to produce a high-resolution output. Since SR is an ill-posed inverse problem, a good initial estimate is always a good regulariser to find the optimal solution. We propose an initial estimate as part of our SR pipeline, esp. x8, which will helps in quick convergence and accurate output. We propose a cascade approach by combining residual interpolation (RI) method with anisotropic total generalised variation (ATGV) method, both uses HR guidance image. The improvements are shown qualitative and quantitative with different levels of noise.
Keywords: super-resolution; depth image; initial estimate; interpolation; cascade.
Automated System for Road Extraction & Traffic Volume Estimation for Traffic Jam Detection
by Jyoti Parsola, Durgaprasad Gangodkar, Ankush Mittal
Abstract: Efficient vehicle detection and traffic density estimation for traffic congestion is one of the essential tasks of traffic surveillance and it has been solved up to some extent. However there is still a need for better solution which effectively and efficiently estimates traffic congestion. In this paper we propose an automated surveillance system for vehicle detection from a traffic scene. Moreover the proposed system performs the functions like identifications of path followed by vehicle, estimation of traffic volume, identification of moving direction of vehicle, traffic analysis and reports traffic jam. These functions are not collectively discussed by a single author. Our proposed system performs all the function discussed aforementioned. Rather than extracting features of the roads or creating a model our system directly extracts road region from the road scene by motion segmentation of vehicle. Further based on the movement of vehicle, path is plotted. Vehicular density is computed with respect to the corresponding road of moving vehicle. A traffic jam alert is generated based on the intensity of traffic density. Traffic density is categorized in heavy, medium or low based on the flow of the traffic. The performance of our proposed system is evaluated using various benchmark datasets captured in various road scenarios (urban, highway) which, shows the ability of proposed method to work in any road conditions and thus makes it suitable for deployment.
Keywords: Intelligent Transportation System; Traffic Density Estimation; Congestion Detection.
Comprehensive analysis of a diverse group of features and development of Vision-Based Two-Level Hand Detector under Practical Environment Conditions
by Songhita Misra, Rabul Laskar
Abstract: Developing a bare-hand detection system for practical environment conditions is a complex and challenging task. Factors such as change in appearance, uneven illumination, and complex background add up to the difficulty in detecting the target hand. Present study newly explored 13 color-texture and integrates them with texture models to develop robust two-level hand detector under practical conditions mentioned above. Color-texture and texture models are assessed using multiple classification tools and employed in two subsequent levels such that the second level only classifies the optimal sub-windows classified in the first level. The analysis showed that the proposed two-level detection system detects the hand with 53.4% higher accuracy than the baseline model which the integrated motion detection and skin filtering method, under the practical conditions. With five times lower time-complexity than the baseline model, the proposed system can be used to detect hand in both static as well as dynamic gesture systems.
Keywords: Two-level hand detection system; Complex background; Positional variation; AdaBoost classifier; Color-texture features.
Design of Filter for Image De-noising using Discrete Wavelet Transform for ASIP
by Mood Venkanna, Rameswar Rao, Chandra Sekhar
Abstract: Application Specific Instruction Set Processors (ASIP) is a customized processor for user specific application. Though a significant research has been done on this, still it is most promising technology, due to lack of efficient methodologies for designing the processor configuration according to the applications. Again ASIP solution explores the trade-off between the dedicated hardware design and flexibility among software. It endeavours to fulfil the functionality of an algorithmic with lowpower costs and less complexity. In this paper, an approach is considered to design a processor for image de-noising.Thedesign of suitable filter is an important task for the transmission and real-time processing. Designing ASIPs requires a suitable design of custom datapath, simultaneously modify the instruction-set, decoder including the compiler. We present an ASIP based on custom architecture design using the Discrete Wavelet Transform (DWT) as a filter. It startswith the general purpose datapathlike MIPS. Itcustomizes the datapathiteratively for better power utilization, usable area and performance. All the experiments have been synthesized using Xilinx FPGA andalso verified in Spartan board. The subjective evaluations of the filter isanalysed through various figures. Further it is implemented in HDL to support the customized processor.
Keywords: Image Filtering; Wavelet Transform; ASIP; Impulse noise; DSP;FPGA; VHDL.
A Novel Approach for Secured Multimodal Biometric Authentication based on Data Fusion Technique
by Gayatri Bokade, Rajendra Kanphade
Abstract: The upcoming biometric technology is focusing more on multiple biometric traits to authenticate the user for security, access control and Universal Identity. This is because utmost considered powerful biometric when used solely, grieves from spoof attacks, intra-class erraticism, noise, vulnerability etc. In the arena of biometrics, the integration of evidences offered by multiple biometric is considered as an effective mode of enhancing the authentication accuracy and security. This research work proposes an authentication technique for a multimodal biometric scheme using three traits i.e., Face, Ear and Palmprint at feature level fusion. This Novel method utilizes the raw data fusion technique to create unique pattern for each registered user. Even with the use of three different biometrics traits, the template is created with extreme low dimension and by using a single algorithm. The proposed system provides security with reduced computational complexity and improved robustness.
Keywords: authentication; multimodal biometric; raw data fusion; Face; Ear; Palmprint; computational complexity; robustness.
ARP Cache Poisoning: Detection, Mitigation & Prevention Schemes
by Jayati Bhardwaj, Virendra Yadav, Munesh Chandra Trivedi, Anurag Sen
Abstract: Providing security to the networks is of utmost importance for all kinds of users. The fundamental of any communication network are implemented protocols. Hence ensuring security at the protocol level is point of concern. Major communication protocols like IP and ARP lack mechanisms for protection against malicious activities.ARP is a network communication protocol employed for mapping a network address to a MAC address at the data link layer of the IP suite. However, the absence of authentication process in the ARP protocol allows vulnerabilities like ARP Cache Poisoning or Arp Spoofing to take place. This allows malicious nodes to associate its MAC address with the IP address of host and hence resulting in the exposure of network to several severe attacks like DoS, MITM, Session hijacking and many more. With the ongoing increasing number of attacks a lot number of detection, prevention and mitigation schemes have been proposed regarding the scope of the problem. However, there is no universally accepted benchmark scheme that reaches to the solution at fullest. This paper presents a comprehensive review of all those schemes along with their associated strengths and weaknesses. Also a comparative evaluation of schemes is included in the paper for further insight into the development of improvised solutions to the above stated problem. This evaluation leads to a summary of all the requirements to be needed for a novel approach leading to the solution of the mentioned problem
Keywords: ARP Cache Poisoning; MAC address; Proxy ARP; Public Key Cryptography; Spoofing.
Incremental Approach for Multi-Modal Face Expression Recognition System using Deep Neural Networks
by Anand Handa, Rashi Agarwal, Narendra Kohli
Abstract: Facial Expression Recognition (FER) plays a vital role in building human-machine interaction systems. The ability to recognize facial expressions and emotions automatically and efficiently helps in building novel applications such as Human Machine Interaction system, Human-Robot interaction, driving safety and health care. Face helps in depicting a wide range of information about a persons identity, sex, age, mental state and emotional state. Despite significant work and improvement in this field, the facial expression is still one of the most challenging tasks. Convolutional Neural Network (CNN) and Deep Convolutional Neural Network (DCNN) has evolved as an efficient tool for facial expression recognition models but they differ significantly in terms of their network configuration and architecture. There exists a variety of bottlenecks in existing facial expression recognition systems such as they lack in generalizing their algorithms over different databases. Hence, in this paper, we propose a model based on DCNN to overcome these challenges which exist in recognition of emotions and validated our results on a variety of well-known databases in three steps. Firstly, the proposed model focuses on the selection of an appropriate activation function depending on its accuracy and training loss over a database. Secondly, an incremental strategy is used in which deeper models are developed simultaneously from shallower networks to increase the accuracy with less training loss. Lastly, by an ensemble of CNN and DCNNs, the model achieves an accuracy of 74.15% for FER2013, 96.20% for CK+ and 98.25% for JAFFE databases, outperforming previous work.
Keywords: Activation functions; Deep Neural Network (DNN); ConvolutionalrnNeural Network (CNN).
HADEM-MACS: A Hybrid approach for detection and extraction of objects in movement by multimedia autonomous computer systems
by Elie FUTE T.
Abstract: Nowadays, multimedia information become an inescapable medium for the validation of applications such as identification, localization and objects tracking. These give rise to many processing methods that, after collecting multimedia data (images, videos), continue with a preprocessing in order to reduce noise, finally it finishes with processing in order to extract objects, more precisely the form of object that capture our domain of interest. The first stage consists to detect objects in movement in the scene. This detection passes through a background modeling. Model based on mixture of Gaussian is commonly used. However, this approach is subject to resources consumption mainly processing unit and memory. We present in this paper a hybrid approach of detection and extraction of objects in movement by a multimedia wireless sensor network. It is based on an improved frame difference, an adapted mixture of Gaussian and a simplified shadow removal.
Keywords: Background subtraction; movement detection; mixture of Gaussian;
neural network; multimedia.
Perceptual Image Quality Assessment Based on Gradient Similarity and Ruderman Operator
by Zianou AHMED SEGHIR
Abstract: In this work, a new metric for image quality assessment is suggested, which provides more suppleness than previous measures in using Ruderman operator, visual region of interest and gradient similarity. Firstly, the luminance distortion between the reference and test images is determined. Secondly, the gradient similarity is computed by using canny filter and proposed gradient mask. Thirdly, the test and reference images are transformed using Ruderman operator. Fourthly, the visual region of interest is calculated by employing entropy operator. Lastly, the dissimilarity between the reference and test images is obtained, by combining all previous metrics: luminance distortion measure, gradient similarity measures, Ruderman measure and visual region of interest measure. Experimental comparison demonstrates the effectiveness of the proposed method.
Keywords: Ruderman operator; image quality assessment (IQA); gradient similarity.
Human Skin Ringworm Detection Using Wavelet and Curvelet Transforms: A Comparative Study
by Manas Saha, Mrinal Kanti Naskar, B. N. Chatterji
Abstract: The common human skin disease called ringworm is investigated in the light of computer vision. Two distinct methodologies are developed for its detection. The first methodology implements three level multi-wavelet decomposition of the skin images and subsequent evaluation of the approximation and detail subband energies which act as the texture characterizing features. The second methodology incorporates the curvelet to segment the circular protrusion of the skin images especially with ringworms followed by statistical texture investigation by gray level co-occurrence matrix (GLCM). After feature extraction by both the methodologies, binary classifier called the support vector machine (SVM) recognizes the images as ringworm with detection accuracy of around 87% and 80% for the first and second methodologies respectively. In addition, the performance indexing parameters of SVM classification like sensitivity, specificity, Positive Predictive Value (PPV) and Negative Predictive Value (NPV) which are not previously addressed are evaluated. Both the methodologies are comprehensively demonstrated and compared to select the better one. The selected method is then compared with the available technique and commented upon.
Keywords: Multiresolution; Wavelet; Curvelet; Approximation subband; Detail subband; Energy signature.
A Wikipedia-based Semantic Tensor Space Model for Text Analytics
by Han-joon Kim, Jae-Young Chang
Abstract: This paper proposes a 3rd-order tensor space model that represents textual documents, which contains the concept space independently of the document and term spaces. In the vector space model (VSM), a document is represented as a vector in which each dimension corresponds to a term. In contrast, the model described here represents a document as a matrix. Most current text mining algorithms only take vectors as their input, but they suffer from term independence and loss of term senses issues. To overcome these problems, we incorporate the concept as a distinct space in the VSM. For this, it is necessary to produce the concept vector for each term that occurs in a given document, which is related to word sense disambiguation. As an external knowledge source for concept weighting, we employ the Wikipedia encyclopedia, which has been evaluated as world knowledge and used to improve many text-mining algorithms. Through experiments using two popular document corpora, we demonstrate the superiority of the model in terms of text clustering and text classification.
Keywords: tensor space model; vector space model; text mining; concepts; Wikipedia.
COLOUR THRESHOLDING BASED AUTOMATIC Ki67 COUNTING PROCEDURE FOR IMMUNOHISTOCHEMICAL STAINING IN MENINGIOMA
by FAHMI AKMAL DZULKIFLI, MOHD YUSOFF MASHOR, HASNAN JAAFAR
Abstract: Image processing are widely used by medical experts since it can help them by providing extra visualization for early detection and treatment. Nuclei or cell counting represents a critical part of the histopathological analysis. Nuclei segmentation is the initial step in cell counting and is very challenging, especially in determining between the normal and abnormal cell nuclei. This is due to the variation of the cell shape and size. The Ki67 is a nuclear protein that was widely used among the pathologists to measure the proliferation of tumour cells. Generally, the pathologists use the manual counting technique for counting the Ki67 cells. However, the counting results has poor reliability and lack of accuracy. The current study aimed to propose an automatic Ki67 cell counting for meningioma images by using the colour thresholding approach. The proposed method has been tested on 12 photomicrographs of meningiomas. The performance of the proposed method was compared to the manually segmented images, which have been validated in prior by the medical expert. The results showed that the proposed method was able to segment the immunostained positive and immunostained negative Ki67 cells with an average accuracy of more than 90%. For counting results, the proposed system produced good results in counting the Ki67 cells with an average relative accuracy of 0.91 for positive Ki67 cells and 0.89 for negative cells. Furthermore, the average time of executing the proposed algorithm was fast at 24 seconds per image.
Keywords: Automated Counting; Colour Thresholding; Image Segmentation; Immunohistochemical Staining; Ki67 Cell; Meningioma.
Blob Analysis of an Automatic Vision Guided System for Fruit Picking and Placing Robot
by Tresna Dewi, Zarqa Mulya, Pola Risma, Yurni Oktarina
Abstract: Agriculture has a strategic role in improving the economic development in a country. As the population grows, the demand to feed the nation. The agriculture strategy needs to be improved adopting automation for better handling of the harvest. The automation can be accommodated by robotics application started by implementing a pick and place robot to move the product. This paper presents the blob analysis method as the visual cue for a pick and place robot handling fruit. Blob analysis is used to detect fruit based on color and shape that is processed by filtering objects and extracting blob using morphological operators. The main controller of the robot is an Arduino Mega that moves the robot based on the input from processed image using Phyton and OpenCV in Raspberry Pi. The captured images are facilitated by a PI Camera functioning as an "eye" for the robot. The experiment was conducted to prove the effectiveness of the proposed method where the average time of picking and placing fruit is 6.69 s for fruit in Position 1 with a range of 332-334 of x and 255-266 of y coordinates respectively, and 7.63 s for Position 2 with a range of 475-576 in x and 205-206 of y coordinates. The image plane considered in this study is 600 x 480 pixel frame. The experiment shows that the proposed method is effective as an automatic vision guided system for fruit picking and placing robot.
Keywords: Agriculture robot; Blob analysis; Pick and place robot; Visual cue.
Automatic Defect Inspection System for Beer Bottles based on Deep Residual Learning
by Qiaokang Liang, Shao Xiang, Jianyong Long, Dan Zhang, Gianmarc Coppola, Wei Sun, Yaonan Wang
Abstract: Recyclable beer bottles are increasingly popular due to the cost effectiveness in recent years. Prior to refill, they need to be scrubbed and sanitized, which requires quality inspection. Automatic detection of defects in recyclable beer bottles would reduce both the cost of the production process and the time spent in the quality inspection. A novel approach is proposed for automatic detection of defects occurring on the beer bottles by deep residual learning. This method extracts the characteristic information of beer bottle defects through the deep learning network and realizes the classification of defect characters. In this work, the recognition of 3 kinds of common defects (defective body, defective mouth, and defective bottom) is realized, and the promising result demonstrated that the proposed method is capable of inspecting defects of beer bottles with outstanding accuracy. Particularly, a state-of-the-art Convolutional Neural Network (CNN) was applied to the detection of beer bottle defects, which improved the accuracy of beer bottle detection comparing with traditional methods. Experimental results show that the new approach satisfies the requirement of defect detection and is able to improve the production efficiency.
Keywords: Detection of defects; Deep learning; Convolutional Neural Network; Quality inspection.
An Optimal Automatic Brain Tumor Detection Using Fuzzy Co-Clustering Algorithm
by Heena Hooda, Om Prakash Verma
Abstract: This paper presents an automatic technique to detect brain tumor from real time Magnetic Resonance Imaging images. Traditionally, tumor is mapped manually by the radiologist in the lab and this process is very time consuming and prone to human error. The proposed strategy makes use of fuzzy co-clustering for images algorithm for initial segmentation of MRI images. The parameters in the FCCI algorithm are optimized using Particle Swarm Optimization technique. The intra-cranial mask is extracted from the MRI scan images by using the intensity difference as a measure to differentiate between the extra-cranial and intra-cranial region. The segmented image along with the intra-cranial mask is used to detect and extract brain tumor and calculate the area of tumor from MRI scan images of the brain. The performance of the algorithm is evaluated on the basis of match score, accuracy score, dice score and Jaccard's similarity coefficient. The result of the proposed approach corresponds to the ground truth in tracking the total area of the tumor and is seen to outperform most of the techniques. The real time database is taken from Rajiv Gandhi Cancer Institute & Research Centre, Delhi, India and results are validated by the radiologist.
Keywords: Brain MRI images; Tumor Detection; Clustering; Particle Swarm Optimization; Morphological operation,Fuzzy Co-Clustering Algorithm.
Maximum Entropy Based Semi Supervised Learning for Automatic Detection and Recognition of Objects Using Deep Convnets.
by Vipul Sharma, Roohie Naaz Mir
Abstract: Object detection and localisation is one of the major research areas in computer vision that is growing very rapidly. Currently, there is a plethora of pre-trained models for object detection including YOLO, mask RCNN, RCNN, fast RCNN, multi-box, etc. In this paper, we proposed a new framework for object detection called maximum entropy-based semisupervised learning for automatic detection and recognition of objects. The main objective of this paper is to recognise objects from a number of visual object classes in a realistic scene simultaneously. The major operations of our proposed approach are preprocessing, localisation, segmentation and object detection. In preprocessing, three processes namely noise reduction, intensity normalisation, and morphology are considered. Then localisation and object segmentation is performed using maximum entropy in which optimal threshold is detected and in the end, object detection is performed using deep ConvNet. The performance of the proposed framework is evaluated using MATLAB-R2018b and it is compared with some previous state of the art techniques in terms of localisation error, detection and segmentation accuracy along with computation time
Keywords: maximum entropy; object detection; weakly supervised learning; deep convolutional neural networks; segmentation and localisation.
Multi source Retinal Fundus Image Classification Using Convolution Neural Networks Fusion and Gabor-Based Texture Representation
by Radia Touahri, Nabiha AZIZI, Nacer Eddine Hammami, MONTHER ALDWAIRI, Nacer Eddine Benzebouchi, Ouided Moumene
Abstract: Glaucoma is one of the most known irreversible chronic eye disease that leads to permanent blindness but its earlier diagnosis can be treated. Convolutional Neural Networks (CNNs), a branch of deep learning, have an impressive record for applications in image analysis and interpretation, including medical imaging. This necessity is justified by their capacity and adaptability to extract pertinent features automatically from the original image. In other hand, the use of Ensemble learning algorithms has an important impact to improve the classification rate. In this paper, a two stage based image processing and ensemble learning approach is proposed for automated glaucoma diagnosis. In the first stage, the generation of different modalities from original images is adopted by the application of advanced image processing techniques especially gabor filter based texture image. Next, each dataset constructing from the corresponding modality will be learned by an individual Cnn classifier. Aggregation techniques will be then applied to generate the final decision taking into account the outputs of all Cnns classifiers. Experiments were carried out on Rime-one dataset for glaucoma diagnosis. The obtained results proved the superiority of the proposed ensemble learning system compared to the existing studies with classification accuracy of 89.63%.
Keywords: deep learning; ensemble Classifier Fusion; Convolution Neural Networks; Glaucoma diagnosis; Gabor filter.
Majority voting based hybrid feature selection in machine learning paradigm for epilepsy detection using EEG
by Sunandan Mandal, Bikesh Kumar Singh, Kavita Thakur
Abstract: This article presents a combination of statistical and Discrete Wavelet Transform (DWT) based features for the identification of epileptic seizures in Electroencephalogram (EEG) signals. A total of 150 quantitative features are extracted from EEG signals. A multi-criteria hybrid feature selection is proposed by combining 6 feature ranking methods using the majority voting technique to identify the most relevant EEG markers. Kernel-based support vector machine is used to evaluate the proposed approach along with a hybrid classifier namely Support Vector Neural Network (SVNN) which is combination of Support Vector Machine (SVM) and Artificial Neural Network (ANN). For performance evaluation of the proposed method a benchmarked database is used. A comparative study of various types of SVM and SVNN with 10-fold and hold-out cross-validation techniques is conducted. The highest classification accuracy (CA) of 98.18 % and 100% sensitivity is achieved with a fine Gaussian SVM classifier with hold-out data division protocol.
Keywords: EEG quantitative features; epilepsy; wavelet transform; multi-criteria feature selection; classification.
A Feature-Based Approach for Digital Camera Identification using Photo-Response Non-Uniformity Noise
by Megha Borole, Satish Kolhe
Abstract: Source camera identification of an image is an emerging field of digital forensics. To identify the source camera through which the image is captured, Photo-Response Non-Uniformity (PRNU) noise is used as a camera fingerprint, as it is a unique characteristic that distinguishes images taken from the similar cameras. This paper presents a feature-based approach to identify the source camera. The input image is denoised using the denoising filter and from this denoised image, PRNU noise pattern is extracted. These PRNU noise patterns are represented by Hus invariants, which are perpetual under image scaling, translation, and rotation. These features are fed to Fuzzy Min-Max Neural Network (FMNN) for training and classification for digital camera identification. The proposed approach has the ability to identify the cameras capturing the same scene.
Keywords: Camera Identification; Photo-Response Non-Uniformity (PRNU); Fuzzy Min-Max Neural Network (FMNN).
An Integrative Approach for Path Planning and Tracking of Shape Aware Mobile Robot in Structured Environment Using Vision Sensor
by Sangram Keshari Das, Sabyasachi Dash, B.K. Rout
Abstract: A shape aware path planning algorithm is necessary for real time execution of a task by a mobile robot whereas path planning algorithms available in literature consider mobile robot as a point. Current work proposes a shape-aware A* path planning approach with a heuristic function to accommodate the shape of mobile robot. In this paper, the detection and tracking of the robot has been carried out for the mobile robot while performing a task in a structured environment. For implementation and validation, an overhead camera is used to capture the task space and the obstacles in ROS platform and the captured images are processed using OpenCV software for detection, and tracking using KLT and Kalman filter algorithms for different test scenarios. The proposed approach accurately detect and track the shape aware mobile robot with error % ranging from 6 - 10% in different cases.
Keywords: Path planning; Obstacle detection; Shape aware Algorithm; Robot Operating System (ROS); OpenCV.
Design of an ICT Convergence Farm Machinery for an Automatic Agricultural Planter
by Byungchul Kim, Jaesu Jang, Sangjo Kim, Seonmin Hwang, MoonSun Shin
Abstract: Recently, ICT technology such as information technology and automatic control technology has been applied into agriculture, and the era of conversion to smart agriculture aimed at improving productivity and improving quality of agriculture has been reached. In particular, it is required to develop a technology that maximizes productivity through growth and quality control based on optimized parameters for each cultivated crop by applying a new automatic control system to the existing traditional agricultural field. In this paper, we propose and design a controller module of agriculture planter applying ICT convergence techniques in order to control rotating speed of various devices in real time. The planters with the controller are useful for saving time than the existing planters which has been dependent on mechanical type.
Keywords: Agricultural Machinery; Smart Farming; ICT convergence farm machinery; Agriculture Planters.
Malware Classifier for Dynamic Deep Learning Algorithm.
by Youngbok Cho
Abstract: This study proposes a malware classification model that can handle arbitrary length input data using the Microsoft Malware Classification Challenge dataset. We are based on imaging existing data from malware. The proposed model generates a lot of images when malware data is large, and generates a small image of small data. Dynamic RNN learns the generated image as time series data. The output value of the RNN is classified into malware by using only the highest weighted output by applying the Attention technique, and learning the RNN output value by Residual CNN again. Experiments on the proposed model showed a Micro-average F1 score of 92% in the validation data set. Experimental results show that the performance of a model capable of learning and classifying arbitrary length data can be verified without special feature extraction and dimension reduction
Keywords: : Convolution neural network; malware; deep learning; recurrent neural network; kaggle data.
A Review on Content Based Image Retrieval System: Present Trends and Future Challenges
by Narendra Kumar Rout, Mithilesh Atulkar, Mitul Kumar Ahirwal
Abstract: Recently, to get similar image with better accuracy is becoming matter into concern for Content based image retrieval (CBIR) system due to exponential rise in image databases. In CBIR system, a user takes a sample image as input query and retrieves its relevant similar images from a large database of images by the help of features based on color, shape and texture. This study is based on different databases used in the CBIR system, focusing on commonly used databases e.g., Wang, Corel and Brodatz database. Low level features are used in the CBIR system and importance of each feature was graded by their repute based on citations in various comparable studies. In the CBIR system, the features which are more frequently used for achieving better result have been further considered to search images from the image databases. Along with the feature, different weight assignment methods like individual weightage, equal assignment of weights and other assignment methods employed in the analysis of CBIR systems are also reported. However, the percentage weight assignment to the features of the image is calibrated based on its importance in doing accurate searches on the used image databases. This paper presents an intense review of CBIR systems and frequently used features with different weight assignment methods. Including this, the future challenge of entire study is user-free or automated weight assignment methods based on soft computing approaches are required in CBIR systems.
Keywords: CBIR; low level features; feature extraction methods; weight assignment method; Image database.
A Clustering Based Differential Evolution with Parapatric and Cross-Generation Selection
by Seong-Yoon Shin
Abstract: Differential evolution (DE) is one of the efficient Evolutionary algorithm (EA) for continuous optimization problems. It is commonly known that the mutation is one of the cores of the DE algorithm. However, the mutation strategies randomly selected from the current population cant be fully exploited to search the optimal solution, especially in the big data era. To provide some suitable parent individuals for the mutation strategies, it is essential to exploit the data-driven method for analyzing the population data. Tensor decomposition, proven to be an efficient data processing method, can be used to provide data-driven services. We propose a novel data-driven mutation strategy for parent individuals selection, namely tensor-based DE with parapatric and cross-generation (TPCDE). Firstly, we construct a third-order population tensor to represent the relationship among generation, individual, boundary partition. Then the population data is classified into multiple clusters by combing the tensor-based feature extraction approach and affinity propagation (AP) clustering algorithm. Finally, different parent individuals are selected from other clusters to guide evolution. To evaluate the effectiveness of the proposed TPCDE, a series of data-driven experiments are carried out on 13 benchmark functions. The experimental results indicate that TPCDE is an effective and efficient framework to enhance the performance of the DE algorithms.
Keywords: differential evolution; tensor; clustering; parapatric selection scheme; cross-generation selection scheme.
Novel video coding methods for versatile video coding
by Heeji Han, Daehyeok Gwon, Jaeryun Choe, Haechul Choi
Abstract: Versatile video coding (VVC), which is the next generation video coding standard, is being developed to provide greater coding efficiency than existing video coding standards. In VVC, various coding tools related to intra and inter prediction modes have been adopted. This paper introduces several methods that improve coding efficiency or reduce computational complexity on top of VVC adopted tools. The first method enhances the most probable mode list derivation with the statistics of the intra modes of neighbouring blocks. The second method reduces the number of contexts of the merge with motion vector difference mode. The third method excludes invalid block vector predictors early for the intra block copy mode to improve block vector coding. The experimental results show that the three proposed methods show coding efficiencies of -0.05% for all intra coding, -0.02% for random access, and -0.14% for random access coding scenarios, respectively.
Keywords: versatile video coding; intra prediction; inter prediction; most probable mode; intra block copy.
A high-quality frame rate up-conversion technique for Super SloMo
by Minseop Kim, Haechul Choi
Abstract: In this paper, we propose several methods to improve Super SloMo, a deep learning-based frame rate up-conversion technique for the temporal quality improvement of video. In the proposed methods, the training dataset and hyper-parameter are changed and trained to obtain optimal results while maintaining the existing network structure of Super SloMo. The first method improves the cognition of images when trained with the validation set of characteristics similar to the training set. The second method reduces video loss in all validation sets when trained by adjusting the hyper-parameters of the error function value. The experimental results show that the two proposed methods improved the Peak Signal-to-Noise Ratio and the Mean of the Structural Similarity index by 0.11 dB and 0.033% with the specialised training set and by 0.37dB and 0.077% via adjusting the reconstruction and warping loss parameters, respectively.
Keywords: frame rate up-conversion; deep learning; image processing.
Study on Hand Gesture Recognition with CNN-based Deep Learning
by Buemjun Kim, Kyounghee Lee
Abstract: Currently, natural user interface technology is actively studied to enable a computer to understand a users natural behaviours such as gestures and expressions. To recognize humans motions, while the existing approaches generally require additional facilities such as an infrared camera or motion detection sensors, this paper proposes a system based on deep learning to recognize a users hand gestures in normal images generated by common devices such as a webcam. A key feature of the proposed system is input image pre-processing to improve training efficiency and inference accuracy of a deep neural network. It performs black-white binarization process of each image to effectively distinguish a users hand area from the others. Then the proposed system trains a deep neural network by learning from those black-white scale images and makes an inference model to classify various hand signals. Our implementation shows the proposed system has a practicality to be applied for hand gesture recognition by correctly classifying a lot of hand signals such as rock-paper-scissor signs, decimal number 0~9 and Korean alphabet consonants. It is also shown that the confidence degree of those classifications can be considerably improved by the aforementioned image pre-processing. Our future work will be to extend our study to recognize a meaning of hand movements included in a series of continuous images to understand more complicated gestures such as a sign language.
Keywords: hand gesture recognition; convolutional neural network; deep learning; image binarization; natural user interface.
A New Framework for Contour Tracing using Euclidean Distance Mapping
by Raju G., Sobhana Mari S.
Abstract: In this paper a new fast, efficient and accurate contour extraction method, using eight sequential Euclidean distance map and connectivity criteria based on maximal disk, is proposed. The connectivity criterion is based on a set of point pairs along the image boundary pixels. The proposed algorithm generates a contour of an image with less number of iterations compared to many of the existing methods. The performance of the proposed algorithm is tested with a database of handwritten character images. In comparison to two standard contour tracing algorithms (the Moore method and the Canny edge detection method) the proposed algorithm found to give good quality contour images and require less computing time. Further, features extracted from contours of handwritten character images, generated using the proposed algorithm, resulted in better recognition accuracy.
Keywords: Contour tracing; Euclidean Distance Mapping; Medial Axis Transform; Handwritten Character Recognition.
Improving transmission method of Cluster Head Node in Two-level Wireless Sensor Network
by Jong-Yong Lee, Daesung Lee
Abstract: WSN is a wirelessly configured network of nodes equipped with sensors that can collect data in the real world. Sensor nodes have limited energy because they are configured wirelessly. Therefore, long-term use of the network should minimize energy consumption. There are many WSN protocols to increase energy efficiency, among which LEACH Protocol is typical. LEACH Protocol improves energy consumption by reducing the transmission distance of the sensor nodes. However, the transmission distance of nodes elected by the CH node has not been reduced. To improve this, TL- LEACH Protocol has been proposed. The second Clusters are composed of only the CH nodes, so that the transmission distance can be reduced. However, depending on the situation, the transmission distance may increase. In this paper, we applied the Dual-hop method to solve this problem and showed better network lifetime than the existing protocol.
Keywords: Cluster; Energy; Network; Protocol; Sensor; WSN.
A study on selection of optimized piping supports and E-BOM generation
by Jung-Cheol Kim, Il-Young Moon
Abstract: A piping support is an element installed to maintain the stability of a piping system in the design process of a construction plant project. The type and size of the components are determined according to pipe size and pipe load values. Pipe supports change depending on the type and characteristics of the fluid flowing through the pipe, the weight of the pipe and the installation environment. In particular, in the order production environment, the design is frequently changed at the request of the ordering party. Whenever a design change occurs, the type and specifications of the piping supports are changed. When the piping is changed, selection of the piping support must be repeated, which consumes a large amount of time and money. This study proposes a method for constructing a portable document format drawing and an optimised design bill of material without two-dimensional drawing design for piping supports that are frequently changed. This is achieved by a programme according to the characteristics of the piping.
Keywords: Computer aided design; Automatic drawing program; Selection program; Engineering bill of materials; Development of software; Hanger support.
A lossless blind image data hiding scheme for semi-fragile image watermark
by Amine Khaldi
Abstract: In this work, we propose a digital watermark approach that is invariable to rotation and guarantees the integrity of the inserted mark. The size of the resulting image is identical to the original image and the process guarantees acceptable transparency. For this, we have experimented eight substitution processes to conclude that the substitution of four bits gives a great capacity of insertion while guaranteeing an acceptable transparency for the insertion of text. However for the insertion of an image, it is recommended to substitute only one bit. This reduces the capacity but guarantees transparency of the watermarking process.
Keywords: digital watermarking; imperceptibility; robustness; digital image; least significant bit substitution.
Cucumber disease detection using adaptively regularised kernel-based fuzzy C-means and probabilistic neural network
by M.G. Jayanthi, Dandinashivara Revanna Shashikumar
Abstract: India is an agricultural country. The major position of India depends on agriculture. But due to diseases in leaves, there is a great loss to farmers. To avoid this problem, automatic disease detection of cucumber disease is proposed. The proposed methodology consists of three modules namely, segmentation, feature extraction cucumber disease detection. Initially, the cucumber diseases are segmented using adaptively regularised kernel-based fuzzy C-means (ARKFCM). Once the disease is segmented, the colour features are extracted using hue, saturation and value (HSV) technique and texture features are extracted using grey level co-occurrence matrix (GLCM) technique. After the feature extraction process, the extracted features are given to probabilistic neural network (PNN) to recognise the image as anthracnose, downy mildew and grey mould. Finally, the experimental results demonstrate that our method is efficient and powerful to recognise the cucumber diseased image and its performance is analysed in terms of accuracy, sensitivity and specificity.
Keywords: cucumber disease; segmentation; feature extraction; classification; ARKFCM; probabilistic neural network; PNN; sensitivity; specificity; accuracy.
Development of NN classifier for recognition of human moods
by Sandeep A. Awachar, Prashant V. Ingole
Abstract: Numerous works have been carried out in the field of recognition using artificial neural network, based on audio, video and image processing. Face recognition has been found to be one of the major topics of interest in this field. Recognition of moods opens its ways, when it comes to face recognition. Face recognition can pave its ways to new ideology of recognising human moods. This paper focuses on development of a NN classifier for recognition of human moods from facial expressions. Apart from a normal mood and the six basic moods (anger, disgust, fear, sad, surprise and happy), the three added moods viz. contempt, courage and desire forms the new and unique feature of this work. Gabor filter and principal component analysis have been utilised for feature extraction and neural network as classifier for classifying various moods on the basis of various face features related to the moods.
Keywords: facial expressions; moods; NN classifier; neural network; contempt; courage; Gabor; PCA.
Human computer interactive future framework: automation of human interaction and interfaces
by Anil Kumar Dubey, Mohan Lal Kolhe, Vikash Kumar Singh
Abstract: Contemporary era technological improvements and social living urges the advancement of technological implementations and to create an environment for automation of interaction platform. The sole aim being the reduction of labour work or human effort and optimising the throughput gained. The future generation is expecting such developments in current systems, but none can elucidate the particular area of improvements to be done for future models. In order to cater such requirements, a HCI future framework for automating the interaction and interfaces is proposed. The framework is divided into ten co-frames of interfaces with further classifications for facilitating the understanding of individual components. The activity and classes of co-frames is designed according to their properties and the need of people. The basic theme behind proposal is to pave path for initiating the technological developments and interface designing for the purpose of reducing the labour and increasing automation of work with motivating people for towards acceptance of newer systems in HCI to execute the work in an efficient manner.
Keywords: human-computer interaction/interfacing; HCI; framework; interaction; interface; automation; engineering.
Fusing pyramid histogram of gradients and optical flow for hand gesture recognition
by S.S. Suni, K. Gopakumar
Abstract: Human computer interaction systems based on hand gestures catch the eye of the research community for implementing natural communication between man and machines. However, different persons perform the same gestures differently in terms of velocity and motion scale. This poses a challenging issue in minimising the variations between different persons and maximises the coherence of the same gestures. In this paper, the original pyramid histogram of gradients in three orthogonal planes combining with the dense optical flow to create dynamic descriptor is explored in to discriminate features for recognition of hand gestures. The shape and motion features of images in a video sequence are captured to obtain the geometric and illumination invariant spatio-temporal feature descriptor for classification. A multiclass support vector machine classifier is used to recognise the hand gestures. The proposed method gives an excellent recognition rate and excels the existing approaches.
Keywords: human computer interaction; HCI; pyramid histogram of gradients; PHOG; optical flow; hand gesture recognition; multi-class support vector machine.