Abstract: A method of imitation insect compound eye of vehicle breaking detection method on video of traffic is proposed, for automatic detection of traffic on general highway and express way. The foreground image is tracted with the technology of a new variational background and is divided into many small blocks with grid. The count of foreground points of every grid is got with checking every grid. And vehicle breaking automatic detection is done with mathematical modaling of vehicle breaking. Finally, several traffic videos of different environments are tested, and the results indicate the proposed method is efficient, high-detection-rate and robust.

摘要： 实现了对普通公路和高速上的停车事件自动监控，提出一种仿昆虫复眼的交通事件检测方法。利用新的背景差技术提取出包含视频中的目标的前景图，把整个前景图分割成许多小块，以块为单位检查每个网格内前景点的分布，并对每一个网格建一停车事件数学模型，实现对停车事件的自动检测。最后，对不同环境下的多组视频进行测试，结果证明该算法检测精度高且算法实时性好，具有较好的鲁棒性。