The video people detection based on neural network

In order to solve the problem of pedestrian detection algorithm, which presents the weak real-time video monitoring and the recognition rate for improvement, a kind of HOG-neural network pedestrian detection method based on gaussian mixture background deduction, HOG feature and BP neural network, is presented in this paper. Compared with the classical HOG-SVM method, many experiment results show that this method has a great increase in real-time, and certain promotion in terms of recognition rate, by training and testing the INRIA pedestrian database under the environment of Opencv.