Pedestrian Detection Based on Laplace Operator Image Enhancement Algorithm and Faster R-CNN

Abstract

Pedestrian detection is an important branch of computer vision. Many car manufactures have used this technology in real situation. Recently, deep learning has become the best method of pedestrian detection, and the advantage of deep neural network is that it can use statistical method to extract high-level features from raw sensory data and to obtain effective feature. Currently, Faster R-CNN is a typical framework, which commonly be used in the field of image processing. However, in order to achieve better performance in pedestrian detection, Faster R-CNN requires a large number of high-quality training samples. Due to the change of light and pedestrian density, the quality of the collected image is poor. Based on this problem, our research introduces Laplacian operator, it can enhance local image comparison. By introducing the Laplace operator, the proposed method can effectively preprocess the samples of the Faster R-CNN. The real data experiments verify the effectively of this algorithm as well as good robustness to the interference.