针对TLD(Tracking Learning Detection)在跟踪过程中由于自身光敏感性造成的对快速运动目标以及产生形变的目标的跟踪识别率较低问题,提出了一种改进的TLD方法,在跟踪模块中引入金字塔光流法,通过构建金字塔图像,减小图像尺寸,从而使运动目标在图像中的速度降低,亮度变化减小,并使得Lucas-Kanade方法的前提条件成立。在开源的科学工程计算软件SCILAB6.0.1上进行实验,结果表明,改进的TLD在光线变化和快速运动目标的跟踪准确性得到了提高,跟踪算法的可靠性得到了加强;同时将算法封装为算法模型,可通过简单的参数输入实现目标跟踪目的。

In order to solve the problem of low recognition rate for tracking fast moving targets and deformed targets due to their photosensitivity in the tracking process of TLD (Tracking Learning Detection), an improved TLD method is proposed, and a pyramid is introduced into the tracking module. The optical flow method reduces the image size by constructing a pyramid image, which reduces the speed of the moving object in the image and reduces the brightness change, and makes the premise of the Lucas-Kanade method true. Experiments are conducted on the open-source scientific engineering calculation software SCILAB 6.0.1, and the results show that the improved TLD improves the tracking accuracy of light changes and fast moving targets, and the reliability of the tracking algorithm is enhanced. At the same time, this algorithm is packaged as an algorithm model, which can achieve target tracking purpose through simple parameter input.