基于背景复杂度分析的智能跟踪器研究

传统基于图像处理的成像目标跟踪系统中,为了提高目标跟踪的可靠性,更多关注的是目标本身的特性,通过分析目标在不同阶段的成像特点,采用相应的算法进行自动提取与跟踪。由于不同目标提取算法在不同复杂程度的背景条件下具有明显的优劣性,实际跟踪过程中需要同时依据目标以及背景两方面的特性才能做出最合理的选择,然而,目前对于影响跟踪的背景特性研究却显得不足。针对这一问题,提出了一种基于结构相似度的背景复杂度判决方法,结合背景复杂度,完成了算法间的自动切换,实现了一个根据目标特性与背景特性的自适应实时多算法跟踪器。仿真测试与外场试验表明,该方法显著提高了对目标跟踪的可靠性。

The study of intelligent tracking machine based on the complexity degree analysis of background

In the traditional target tracking system based on image processing, it is attached more attention to the target feature for improving the reliability of tracking. Through analyzing different features of target in different tracking phase, the most suitable algorithms are selected automatically for current processing. Due to target extraction or tracking algorithms have different tracking capability faced different background complexity degree, When making decision for algorithm selection, not only the target features but also the complexity degree of background should be considered. But the background characteristic which is important for tracking reliability is not studied sufficiently. For this problem, a complexity degree of background determination method based on the structural similarity is presented. Then a method based on the target and background features is introduced for tracking algorithm decision-making and a real-time multiple algorithms tracking machine based on the multi-core DSP processor is realized. The test results based on the emulated image and real image show the method proposed improved the tracking reliability remarkably.