基于Mean-Shift的复杂工业环境运动目标跟踪算法  

Improved Moving Object Tracking Approach Based on Mean-Shift in Complex Industrial Situations

作　　者：华聚良[1] 黄河燕[2] 王树梅[1] 

机构地区：[1]南京理工大学计算机科学与工程学院,南京210094 [2]北京理工大学计算机学院,北京100081

出　　处：《系统仿真学报》2014年第11期2600-2606,共7页Journal of System Simulation

摘　　要：随着近年来计算机及成像技术的高速发展,视频运动目标检测和跟踪已成为计算机视觉领域的一个研究热点。由于运动目标与摄像头位置的相对变化、室内光照的变化及噪声的存在,以及周围环境中相似物体的干扰,传统的Mean-Shift跟踪算法的跟踪效率、准确性和抗干扰能力均无法满足该应用的需要。为了解决这一问题,在传统Mean-Shift跟踪算法基础之上,提出了颜色直方图更新算法、运动信息融合等改进方案,在Mean-Shift迭代搜索过程中加入了速度矢量加权机制,使目标与相似背景能够有效地被区分开来,提高了系统的抗干扰能力。仿真实验结果表明算法有效提高了跟踪的效率及准确性,能够满足复杂工业环境中运动目标跟踪的需要。With the rapid development of the computing technology in recent years, detecting and tracking moving target in video has become a problem urgently to be solved in computer vision. Mean-Shift tracking algorithm has become one of the most commonly used algorithms for its excellent robustness and processing speed. Because of the change of the position against camera, the light in the warehouse, the noise, and the similar objects around, the efficiency, the accuracy and the anti-jamming capability of the traditional Mean-Shift tracking algorithm are not able to meet the requirement of the application of moving boxes in the logistical storages. In order to solve these problems, a series of improved solutions was put forward such as histogram optimization, histogram update and moving information amalgamation which could improve the efficiency, the accuracy and the anti-jamming capability. The moving vector weighing scheme could improve the discriminative capability of the algorithm and make the tracking more successful. Simulation experiments on some test videos validate that the proposed approach can meet the requirements of the targeted industrial application.

关 键 词：目标跟踪 MEAN-SHIFT算法 运动信息 直方图更新 

分 类 号：TP391.9[自动化与计算机技术—计算机应用技术]