Pedestrian detection based on improved HOG algorithm and SVM

The main research content of this paper is pedestrian detection in images. In order to enhance both computational efficiency and accuracy simultaneously, an improved HOG+SVM algorithm has been put forward. The HOG algorithm itself has a high computational accuracy, and the SVM approach guarantees both classification accuracy and efficiency. Through the utilization of median filtering, bilinear interpolation, and L2 norm normalization, the accuracy of the HOG algorithm is enhanced while guaranteeing computational efficiency. Finally, an experiment was designed to compare the improved HOG+SVM algorithm with the prevailing deep learning algorithm (CNN) at present. The experimental results indicate that the improved HOG+SVM approach enjoys superior computational speed and accuracy.