The Design And Implementation Of Video Target Tracking System

Countries all over the world have put much higher request to public safety in the21st century. How to position the interested video targets rapidly and accurately in the case of the rotation or occlusion becomes the research direction of scientific research workers. Domestic and foreign scholars have done a lot of research in video tracking and provided many effective tracking algorithms. Due to the ability to solve the non-linear and non-Gaussian problems, the particle filter algorithm has been more extensively studied and rapid developed in the field of video object tracking than all the other existing algorithms.TI’s Davinci technology is mainly designed for digital video system solutions.It makes the implementation of digital video productions much easier.As the classical Davinci technology-based SoC processor, DM6446has a dual-core structure of ARM+DSP and opened software architecture.It is the right platform for video object tracking system.In order to overcome the limitations of using single feature information to track the video target, we choose the particle filter tracking algorithm which use the fusion information of color and texture instead. Through the simulation experiments, it shows that the algorithm can track single video target effectively in the event of the rotation or occlusion with static background.According to the Davinci software architecture and development process, we package the target tracking algorithm library and conform the Codec Engine which is in keeping with xDM algorithm interface.On the ARM side, we designed multi-thread applications based on Linux to realize the function of system control and scheduling. Finally, we build a demonstration system which is able to capture video, track target and display the tracking result in the same time. The experimental results show that the system can achieve the desired objectives well both in the terms of real-time and accurate performances.