Dm6437-based Video Object Tracking Algorithm Implementation And Optimization

In recent decades, video processing technologies have been rapid development. As one of the key technologies, video target-tracking technology becomes important increasingly, and has recently attracted more and more institutions and researchers attention. However, it is difficult to meet large volume of data and real-time requirements for the most hardware platforms. While TI's DaVinci series of DSP processors have a powerful parallel processing capability and plenty of resource, and have a special advantage in real-time processing and products'rapid development, thus they are ideal choice for the implementation of video target-tracking algorithms.This paper mainly studies the Mean Shift algorithm which is popular in the video target-tracking algorithms, and focuses on the implementation and optimization of the algorithm based on the TI's DM6437. Specific work is as follows.1. Through referring to lots of domestic and international literatures, video target-tracking algorithm's development and research status are introduced, and TI-DSP special advantage in video processing is also expounded.2. Decribes a typical video target-tracking system signal flow and its various modules, expatiates on the pricinple of the Mean Shift tracking algorithm, simulates the Mean Shift algorithm by using the Matlab, analysis the advantage and disvantage of the algorithm with the simulating results.3. Using TI's CCS, the Mean Shift tracking algorithm is implemented based on the TI's DM6437 which is designed based on the DSP/BIOS structure. Meanwhile, the DSP program is optimized by use of the C64x+ DSP core features and a variety of optimization strategies. Experimental results show that the achievement of the good performance, and meet the algorithm's real-time requirement and codes'high effiecncy.