Pedestrian Detection Method Based on Deep Convolution Neural Network

Abstract

Abstract Compared with the traditional pedestrian detection technology, the pedestrian detection technology based on deep learning has achieved overwhelming advantages. However, due to the large scale of deep convolution network, the demand for dedicated processor limits the popularization of pedestrian detection system. To solve these problems, this paper proposes a deep convolution network model with moderate network scale, which improves the universality of the detection model on the premise of ensuring the detection accuracy. Based on the low dimensional shallow convolution neural network, the optimal network structure is found from three aspects of network layer number, sensor field size and characteristic graph, and the final network parameters are determined by guiding experimental evaluation. The pedestrian image to be detected is input into the above network model, and the pedestrian detection result is obtained. The pedestrian detection results on several popular pedestrian databases such as Daimler, MIT and INRIA show that the network structure designed in this paper not only has moderate network size, but also has good detection performance. Cross experiments also verify the robustness and generalization of the algorithm. The work of this article is funded by the undergraduate training program for innovation and entrepreneurship (202010373050, 202010373051, 202010373046).