Title: HOG features and connected region analysis-based workpiece object detection algorithm
Authors: Ting Yu; Maoyi Tian
Addresses: School of Computer Science, Sichuan Technology and Business University, Chengdu city of Sichuan Province, 611745, China ' School of Applied Technology, XiHua University, Chengdu city of Sichuan Province, 610039, China
Abstract: In order to solve the problem of bearing workpiece object, namely, the insufficient detection ability of the algorithm caused by the complex edge features, a HOG features and connected region analysis-based workpiece object detection algorithm is proposed in this paper. Firstly, the target images of standard workpiece in the training set are meshed to calculate the pixel gradient in the grid, count the gradient histogram and complete the extraction and training of HOG features. Then interval division of the single peak threshold is refined, and a twothreshold segmentation mechanism is proposed to convert the two-valued image into a label image by combining the connected region analysis, and the evaluation of pixel attribute and the filtering of interference is conducted to achieve the purpose of accurately detecting the workpiece object. The experimental results show that the bearing workpiece object detection algorithm in this paper has higher accuracy and stability.
Keywords: workpiece object detection; image gradient; chromatography; edge feature; connected region; meshing; histogram.
DOI: 10.1504/IJICA.2020.107108
International Journal of Innovative Computing and Applications, 2020 Vol.11 No.2/3, pp.61 - 66
Received: 01 Mar 2019
Accepted: 01 Apr 2019
Published online: 04 May 2020 *