Abstract:

Slow speed and low accuracy of detection are the common drawbacks of traditional video moving target detection methods. In order to overcome these shortcomings, a new detection algorithm of moving targets is presented which integrates frame difference with variational level set method. The motion region is initialized through this improved frame difference method, and then it is used as initial level to evolve by the level set equation without re-initialization. In the evolution process, image intensity and optical flow information are used to control level set curve stop at the object boundary. Experimental results show that the proposed algorithm is an effective video rigid moving object detection method with high detection speed and high accuracy.

Key words: moving object detection, level set, frame difference

摘要：

检测速度慢、准确度低是传统视频运动目标检测方法普遍存在的问题，为克服以上缺点，结合帧间差分和变分水平集方法提出一种新的运动目标检测算法。通过改进的帧差法快速初始化运动区域，并将其作为初始水平代入无需重新初始化的水平集演化方程进行演化，利用强度和光流信息控制水平集演化最终停止在目标边界处。实验结果表明，该算法具有检测速度快、准确性高的特点，是一种有效的视频刚体运动目标检测方法。

关键词: 运动目标检测, 水平集, 帧间差分

CLC Number: 

• TP311.52