基于改进的tMHI的实时视频运动目标检测及跟踪

提出了一种对视频图像进行实时目标分割及跟踪的新方法。该方法利用基于时间片的运动历史图像（tMHI）的灰度阶梯轮廓，对存在的子运动区域进行包围划分并予以标记，实现视频图像中运动目标的实时分割，进而将每帧tMHI图像中各个运动区域同场景中运动目标连续关联起来，实现对多运动目标的轨迹跟踪。为了提高分割质量，对tMHI进行了改进处理，去除了大部分噪声干扰，取得了明显的改善效果。实验表明，该方法可以有效地分割并跟踪视频中的多个运动目标，鲁棒性好，检出率较高，并且处理速度较快，达到了实时性的要求，还解决了局部粘连的问题。

Improved tMHI based real-time video moving object detecting and tracing

A new method for segmenting and tracing real-time moving object from video image was purposed, this method using the step-down grey value silhouette of timed Motion History Image (tMHI). By enveloping and marking the motion silhouette regions, it can segment the moving objects from video image real-timely. By associating the motion silhouette regions in every tMHI with the moving objects in real scene, it achieved the trace of moving objects. To get high quality segmentation, the primary tMHI was improved by removing most of the noise. The method was demonstrated with a traffic video, and some moving objects were segmented effectively. It appeared that the method has good locating robust; high detecting rate and fast processing speed that satisfied real time usage. Part conglutination was basically avoided also.