Abstract: In order to solve the problems made by manual operation in video tracking algorithm, such as, approval difficulties, low accuracy and tracking failure, a novel method to calibrate the initial center of position and extract the size of the salient moving target based on human vision is proposed in this paper. This method utilizes the characteristics of the images brightness and the motion vector in forward and backward frames, which can adaptively obtain the optimal center point and the region of target. With the simulation results, the parameters of the target are acquired accurately by the proposed method. The initial localization method which has good performance even when the background and the target are both moving or the difference between them are small.

Key words: information processing, initial localization, visual saliency, moving target in video, tracking algorithm

摘要： 视频图像跟踪算法中，手动核准目标初始参数困难、精度不高且易导致跟踪失效，针对以上问题，提出一种利用人眼视觉获取显著运动目标中心位置并提取尺寸的方法。该方法结合图像亮度及视频前后帧运动特征，可自适应获取人眼敏感运动目标最佳中心点及区域范围。仿真结果表明，该方法能准确定位显著运动目标，在背景运动或与目标差距较小时仍具有良好的性能。

关键词: 信息处理技术, 初始定位, 视觉显著, 视频运动目标, 跟踪算法