摘要
针对模式识别在智能交通领域的实际工程应用,提出了一种提取运动车辆轮廓线的精细跟踪算法.首先,通过冗余离散小波变换法提取运动区域,检测出相邻两帧图像内的运动变化从而确定运动对象的存在及其初始位置;其次,以当前帧运动区域为参考,通过改进的mean-shift算法在后续帧中跟踪运动对象的中心位置;最后,以mean-shift跟踪窗口作为目标初始轮廓线,采用自适应水平集法得到目标轮廓,从而精确定位运动对象位置.实验结果表明本文算法能够以轮廓线的方式以较高精确度跟踪运动车辆目标,目前已被市交通局科研单位采纳,具有一定的工程应用前景.
For the application of pattern recognition in the field of intelligent traffic engineering, a moving vehicles extraction and fine contour tracking algorithm was put forward. Through the extraction of movement region between two frames by redundant discrete wavelet transforms, the existence of moving object and its initial position was determined, and according to the movement region of current frame, the central location of the moving object in the follow-up frames was tracked by an improved mean-shift method. The mean shift tracking window was used as an initial target contour line to obtain the precise contour of moving object by adaptive level set method, and then the precise location of the moving object was determined. Experimental results showed that this algorithm could track the moving vehicles with high accuracy, and it had a certain actual application prospect.
出处
《应用基础与工程科学学报》
EI
CSCD
2010年第2期343-351,共9页
Journal of Basic Science and Engineering
基金
天津市公安交通局科研基金(公科[2005]16号)
天津市应用基础及前沿技术研究计划基金(09JCYBJC07700)
