International Journal of Computational Intelligence Studies (9 papers in press)

Special Issue on: Interdisciplinary Applications and Technologies of Computer Vision

• Performance prediction analysis of college aerobics course based on Back Propagation neural network  
  by Jianlin Su, Hao Zheng, Yanxi Chen 
  Abstract: To solve the problem of low accuracy of the performance prediction method of aerobics courses, the study proposes to combine the partial least squares regression (partial least squares
  Keywords: score prediction; BP neural network; relative error value; partial least squares regression.
  DOI: 10.1504/IJCISTUDIES.2022.10051976
   
  
• Research on the optimal charging method of parallel power batteries for smart electric vehicles  
  by Yanlin Li, Zhen Li 
  Abstract: To save charging cost on the premise of ensuring stable and efficient charging of electric vehicles, an optimal charging method for parallel power batteries of intelligent electric vehicles was proposed. Through mathematical model analysis, the RC network branch is added to construct a second-order RC equivalent circuit model. Optimise battery parameters based on CRUISE simulation platform. By optimising the control of battery current, current sharing and voltage stability are achieved. Finally, the local voltage equalising charging principle of the inverter is used to redesign the equalising charging of the electric vehicle battery. The experimental results show that the relative cost and charging stability of the proposed charging method are 0.58 and 0.62 respectively, which are better than the other two charging methods. The results show that the method can meet the requirements of intelligent electric vehicles for charging stability and efficiency, and has high application value.
  Keywords: electric vehicle; battery charging; equalisation charging; equivalent circuit model; charging power.
  DOI: 10.1504/IJCISTUDIES.2022.10051977
   
  

Special Issue on: Recent Trends in Pattern Recognition and Image Analysis

• A method of athlete's starting image posture correction based on deformation model and image restoration  
  by Xingbo Zhou, Yong Yang 
  Abstract: In order to improve the similarity between the starting pose and the actual structure and the signal to noise ratio (SNR) of the image, this paper proposed a new approach to correct the starting pose of the athlete based on deformation model and image repair. Firstly, dual Kinect sensors were used to collect the starting posture data of athletes and construct the three-dimensional image of the starting posture of athletes. Secondly, the method of attitude 3D image segmentation based on self-segmentation theory is used to obtain the attitude feature artefact region. Finally, after the artefact is repaired, the image attitude correction method based on B-spline deformation model completes the attitude correction. The test results show that the similarity between the image pose structure and the actual structure is as high as 0.98 and the minimum peak signal-to-noise ratio is 0.93 dB.
  Keywords: deformation model; image repair; athletes; start image; attitude correction; three-dimensional reconstruction.
  DOI: 10.1504/IJCISTUDIES.2023.10053255
   
  
• Research on basketball dunk motion recognition method based on characteristic point trajectory  
  by Yong Wang 
  Abstract: Because of the problems of low recognition accuracy and long recognition time in traditional basketball dunk motion recognition methods, a basketball dunk motion recognition method based on feature point trajectory is proposed. Firstly, the basketball dunk action image model is established, and the action image is obtained. Then, the multi-scale detail enhancement technology is used to preprocess the action image. The preprocessed basketball dunk action image is compared with the ideal image. According to the comparison results, the basketball dunk action feature points are extracted, and the basketball dunk action feature point trajectory is generated. Finally, the basketball dunk action recognition is realised based on the feature point trajectory. The experimental results show that the recognition accuracy of the proposed method is as high as 97.3%, the recognition time is only 10.7 s, and the recognition effect is good. It can effectively recognise the basketball dunk.
  Keywords: feature point trajectory; basketball dunk action; multi-scale detail enhancement; action recognition.
  DOI: 10.1504/IJCISTUDIES.2023.10053256
   
  

Special Issue on: Recent Trends in Pattern Recognition and Image Analysis (RTPRIA)

• Adaptive surveillance image enhancement algorithm based on wavelet transform  
  by Lan Li 
  Abstract: In order to improve the definition and signal-to-noise ratio of surveillance image, an adaptive surveillance image enhancement algorithm based on wavelet transform is proposed. First, FWT filter is used to decompose the monitoring image signal, and wavelet reconstruction is used to reconstruct the adaptive monitoring image. Secondly, Sobel operator is introduced to improve the NL means algorithm, and the improved NL means algorithm is used to remove the noise in the adaptive surveillance image. Finally, in the scale space, according to the grey calculation results, the adaptive surveillance image disparity map is decomposed and enhanced according to the decomposed disparity map. The experimental results show that the proposed enhancement algorithm can improve the definition and signal-to-noise ratio of the surveillance image, and the maximum signal-to-noise ratio is 61.5 dB.
  Keywords: wavelet transform; adaptive surveillance image; image enhancement; image denoising.
  DOI: 10.1504/IJCISTUDIES.2023.10053637
   
  
• Research on Target Recognition of UAV Remote Sensing Image Based on Improved Mask R-CNN Model  
  by Zufang Yang 
  Abstract: To solve the problems of low visible edge, low average gradient, low value signal-to-noise ratio, low overlap, high loss rate of image feature details and poor recognition quality existing in traditional methods, a target recognition method of UAV remote sensing image based on improved mask R-CNN model is proposed. Firstly, multi-scale Retinex and dark channel are used to process fog in UAV remote sensing images. Secondly, the mask R-CNN model is improved through pyramid balance strategy and bypass connection. Finally, the demobilised image is input into the improved model to realise remote sensing image target recognition. The experimental results show that the highest values of visible edge and average gradient of the method reach 0.988 and 992, the highest value of PSNR reaches 67.47, the highest value of overlap reaches 98.7%, and the loss rate of detail is below 1.2%, the recognition quality is high.
  Keywords: improved mask R-CNN model; UAV; remote sensing images; target recognition; restore the image; feature extraction.
  DOI: 10.1504/IJCISTUDIES.2023.10053638
   
  
• A Reinforcement Learning-Based Multimodal Scenario Hazardous Behavior Recognition Method  
  by Di Sun, Yanjing Li, Yuexia Han 
  Abstract: In order to solve the problems of low recognition accuracy, low accuracy of key frame extraction of dangerous behaviours, and long delay time in risk behaviour recognition methods in multimodal scenes, this paper studies a multimodal scene risk behaviour recognition method based on reinforcement learning. Firstly, static and dynamic dangerous behaviour features are extracted through long short-term memory artificial neural network; then the extracted dangerous behaviour key frame set is optimised by reinforcement learning; finally, a multi-sensory risk behaviour model library is constructed to complete the multi-modal scene dangerous behaviour recognition. Experiments show that the AUC area of the proposed method is closest to 1, which proves that the recognition accuracy is high, and the accuracy of the extracted key frames of dangerous behaviours is high. When the maximum number of video frames is 500 frames, the delay time of this method is 192 ms, and the delay time; the shortest, the higher the recognition efficiency.
  Keywords: reinforcement learning; multimodal scene; hazardous behaviour recognition; neural network.
  DOI: 10.1504/IJCISTUDIES.2023.10053639
   
  
• Facial Expression Recognition of High Jumpers Based on Wearable Multi physiological Parameter Collection  
  by Ruoqun Mou, Ding Lu 
  Abstract: In order to improve the recognition accuracy of athletes’ facial expression, a method of high jump athletes’ facial expression recognition based on wearable multi-physiological parameter collection is proposed. Firstly, through the wearable multi-physiological parameter acquisition device, the device is used to collect the high jump athletes’ heart rate, skin electrical activity parameters, body temperature and other multi-physiological parameters. Secondly, the baseline drift suppression method based on wavelet transform is used to suppress the baseline drift within multi-physiological parameter signals of high jumpers. Finally, the multi-physiological parameters of high jumpers without baseline drift are inputted into the convolutional neural network model, and the model is used to output the results of high jumpers’ facial expression recognition. The experimental results show that the research method can accurately recognise the facial expressions of high jumpers, with the highest recognition accuracy of 99.21%.
  Keywords: wearable; acquisition of multiple physiological parameters; high jumper; facial expression recognition; baseline drift; convolutional neural network.
  DOI: 10.1504/IJCISTUDIES.2023.10053640
   
  
• Color Image Cross modal Retrieval Method Based on Multi modal Visual Data Fusion  
  by Liu Xiangyuan 
  Abstract: Because the traditional colour image cross-modal retrieval methods have the problems of low retrieval accuracy and recall, and long retrieval time, a colour image cross-modal retrieval method based on multi-modal visual data fusion is proposed. First, collect multimodal visual colour images, and then use bilateral filtering method to filter the collected images to enhance colour images. Then, code the enhanced multimodal visual data, decompose the image modal features and colour modal features through cross-layer fusion encoder, and finally fuse the decoded two modal features. According to the multimodal visual data fusion results, cross-modal retrieval of colour images is performed by using two scale similarity measurement. The simulation results show that the proposed method has higher precision and recall rate for colour image cross-modal retrieval, and shorter retrieval time.
  Keywords: multimodal visual data fusion; colour image; cross-modal retrieval; YCbCr colour model; bilateral filtering method.
  DOI: 10.1504/IJCISTUDIES.2023.10054853