Research and realization of video target detection system based on deep learning

Abstract

The era of big data increases the number and scale of videos day by day, which brings challenges to video target detection. Improving the efficiency and speed of video target detection is of practical significance to image object detection and recognition. Deep learning has been a popular neural network with multilayer structure in recent years. By learning a deep nonlinear network structure, the generalization ability of complex classification problems is improved. In order to achieve higher processing efficiency and accuracy, this topic will study the latest achievements of machine learning method — deep learning — to realize target category detection. In the background extraction step, an improved meaning-based background extraction algorithm and an interest region extraction algorithm to reduce image pixels are proposed. The test results show that the proposed algorithm has good performance of video target detection