Research Of Moving Object Tracking Based On Convolutional Neural Network

With the advent of the information age, moving object tracking has become a hot spot in the field of computer vision, and has wide application value in many fields. Although many moving object tracking algorithms have been proposed, there are many difficulties in the process of actual tracking, such as illumination variation,occlusion, motion blur, scale variation, and their own variation and so on. Therefore,the development of object tracking technology is still facing a huge challenge. The theories and methods of deep learning provide a new opportunity for the research of the object tracking, and also provide the main theoretical framework of the research on the moving object tracking algorithm. The main contents of this paper are as follows:(1) The basic knowledge of moving object tracking technique is studied, the paper start from the object tracking method, then introduce the basic knowledge of object tracking classification and the traditional feature extraction methods.(2) The basic theory of convolution neural network is studied. Firstly, on the basis of analyzing the structure of artificial neural network, the paper introduce the structural characteristics and the training process of the convolution neural network.Then the method of feature extraction of convolution neural network is introduced and compare with the traditional feature extraction and BP feature extraction methods, and the result is obviously superior to the two feature extraction methods.(3) A improved moving object tracking algorithm based on convolution neural network is proposed. The moving object tracking algorithm based on convolution neural network is a tracking algorithm combining deep feature extraction, particle filter with classifier. Firstly, the PCA eigenvector is extracted from the local image dataset using principal component analysis (PCA), and then the convolution neural network is used to extract the deep feature by using the PCA eigenvector. Finally, the particle filter motion estimation and classifier are used to identify and track the target.The experimental results show that the improved algorithm of moving object tracking can overcome the interference between the external disturbance and the target itself in the tracking process, and this algorithm is superior to the current mainstream tracking algorithm in terms of accuracy and success rates.