A Robust mRMR Based Pedestrian Detection Approach Using Shape Descriptor
Kaushal Kumar, Ritesh Kumar Mishra · Traitement du signal · 2019
Pedestrian detection is a fundamental problem in various computer vision applications and is addressed by complex solutions involving complex feature extractor and classification techniques.In this paper, feature selector is used along with Histogram of Significant Gradients (HSG) descriptor and linear SVM classifier to enhance the detection accuracy and reduce the processing time.A feature selector organizes the extracted features in decreasing order of their significance.A comparative study has been done employing different classifiers and feature selectors and a system is proposed with better detection capability.The proposed scheme is found to have a 50 % reduction in feature size, 67 % reduction in processing time and 3.93 % improvement in accuracy.