Research On Target Tracking Based On The Embedded System

In recent years, artificial intelligence has more than ever the high degree of attention. In the future, I believe that artificial intelligence will be more widely used in our production and life, leading technology and even the trend of social development. And one of the most important technology of artificial intelligence is to make intelligent machines will like our human eyes system, with the surrounding environment perception, comprehension. This is the category of machine vision. In the machine vision, the most important and application requirements of the extensive is the target tracking system. At the same time, the concept of industry 2.0 and intelligent hardware etc.., accelerate the development of intelligent hardware market and the traditional industrial. Today, the market for mobile hardware platform to meet the system of machine vision applications have a considerable demand, while the general embedded platform is difficult to meet its requirements for processing capacity.Based on computer and intelligent car system on pedestrian detection and tracking of the demand while satisfying the premise, the system power consumption, stability, small volume, low cost etc.., this paper mainly studied the target detection and target tracking algorithm on embedded system platform, to make it have effective and stable operation.This paper focuses on the research of HOG+SVM and DPM + Latent SVM target detection algorithms, which is very famous in recent years. After experimental verification, DPM + Latent SVM theory can meet the needs of the target detection demands. In the research of pedestrian tracking, because ASMS algorithm can’t effectively solve the occlusion problem appeared in the process of pedestrian tracking. So this paper mainly studies the based on temporal and spatial context for target occlusion theory and Kalman filter theory based on the predictor corrector. Through theoretical analysis and practical verification, it is found that the Kalman filter can effectively solve the ASMS algorithm pedestrian tracking occlusion problem appeared in the process. The experimental test shows that, in the pedestrian appeared serious occlusion algorithm still can effectively identify the location.Finally, DPM detection theory and the Kalman filtering theory to optimize the ASMS target tracking process has been effectively combined and fully optimized. And a large number of practice verification showed that the system runs well, be able to effectively carry out pedestrian detection and tracking.