基于帧差法与Mean-shift算法相结合的运动熔滴识别与跟踪方法  

Recognition and Tracking Algorithms of Moving Droplet Based on Inter-Frame Difference Method Combined with Mean-Shift

作　　者：张赛钰 朱小玲 汪衍广 叶佳 马国红[1] ZHANG Saiyu ZHUXiaoling WANG Yanguang YE Jia MA Guohong(School of Mechanical Engineering, Nanchang University, Nanchang 330031, China Shandong Huahai Group Co. Ltd. , Dezhou 253000, Shandong, China Applied Materials, Inc. , Santa Clara 95054, CA, USA)

机构地区：[1]南昌大学机电工程学院,南昌330031 [2]山东华海集团有限公司,山东德州253000 [3]应用材料公司,美国加州圣克拉拉95054

出　　处：《上海交通大学学报》2016年第10期1605-1608,共4页Journal of Shanghai Jiaotong University

基　　金：国家自然科学基金项目(61165008);教育部回国人员基金项目(13006199)资助

摘　　要：为了对焊接视频图像中的运动熔滴进行自动识别与跟踪,针对熔滴图像为灰度图像且背景单一的特点,提出了一种基于帧差法与Mean-shift算法相结合的方法.利用帧差法对视频图像的前2帧进行差分处理,获取目标窗口和中心位置并进行标定,以解决Mean-shift算法需要在起始帧手动框取目标的问题;结合基于灰度直方图的Mean-shift算法确定下1帧的目标模板位置,以实现对运动熔滴的自动识别与跟踪.结果表明,所提出的运动熔滴识别与跟踪方法能够对熔滴图像进行自动识别与跟踪,且具有良好的鲁棒性和实时性.In order to recognize and track moving droplets in welding video, a algorithm based on the interframe difference method and the mean-shift algorithm was proposed for gray-scale images whose background is single. The inter-frame difference method was used to differentiate the processing between the first two frames of the video image to obtain the target window and the center position and to calibrate the position. By doing so, the problem that the mean-shift algorithm needs to be manually framed to obtain the target at the initial frame was solved. Combined with the mean shift algorithm based on the gray-level his- togram, the position of the target template of the next frame was determined to realize the automatic identification and tracking of the moving droplets. The results show that the proposed algorithm can automati- cally perform real-time identification and tracking of the droplet images with good robustness.

关 键 词：熔滴图像 帧差法 MEAN-SHIFT算法 目标 识别 跟踪 

分 类 号：TG409[金属学及工艺—焊接]