The Application of Neural Network in Multiple Object Tracking

Abstract

Multiple object tracking (MOT) is one of the basic issues in the field of video analysis and monitoring. It has great significance in areas of behavioral event understanding, traffic management and security preventation. Nowadays, there are increasingly deep researches on the application of neural networks in MOT, such as feature extraction, model formulation of both appearance and motion. Compared with the traditional MOT with the employment of hand-crafted features and the design of similarity function between detections, research in neural networks has shown competitive superiority and drawn wide attention from scholars. In this paper, we analyze the current trends and introduce the application of convolutional neural network and recurrent neural network in MOT. We can see that the neural network techniques in MOT have great potential and vast development prospects.