中文摘要：

为了提高连续自适应均值漂移跟踪方法在复杂背景中的跟踪性能，提出一种基于显著性色度特征的运动目标自动选取及跟踪方法。利用高斯混合模型确定出目标模板，根据目标模板与其背景区色度直方图的对比确定出目标的显著性色度等级，将目标模板中具有显著性色度等级的区域确定为跟踪目标。根据跟踪目标的色度直方图模型利用反向投影建立跟踪图像的概率分布图，采用自适应均值漂移方法实现目标跟踪。仿真结果表明：该方法能够有效提取目标的显著性色度等级，从而确定出易于区分背景的跟踪目标，可有效抑制背景对目标跟踪的影响，改善复杂背景情况下目标跟踪的性能，单帧平均跟踪时间小于15 ms，满足跟踪系统实时性的要求。

英文摘要：

In order to improve the tracking performance of the continuously adaptive Mean Shift tracking algorithm in the complex background, an automatic selection and tracking method of the motion target is proposed based on the saliency hue level. Gaussian mixture model is used to determine the target template, and the saliency hue level can be determined by comparing the hue histograms of the target template and the background. The area composed of pixels with the saliency hue level is selected as the tracking target. The probability distribution image of the tracking image can be got according to the hue histogram of the tracking target by the back projection. The continuously adaptive MeanShift tracking algorithm is used to perform the target tracking. Simulation results show that the method can extract the saliency hue level of target, and it is easy to distinguish the target from the background and restrain the information from the background to disturb the target tracking. Thus, the tracking performance can be improved under the complex background. The average tracking time of single frame is less than 15 ms, so the method can meet the real-time requirement of the tracking system.