基于Camshift和SURF的目标跟踪系统  

Object tracking system based on Camshift and SURF

作　　者：陈杏源 郑烈心 裴海龙[1,2] 

机构地区：[1]华南理工大学自主系统与网络控制教育部重点实验室,广东广州510640 [2]华南理工大学自动化科学与工程学院,广东广州510640

出　　处：《计算机工程与设计》2016年第4期902-906,共5页Computer Engineering and Design

基　　金：国家自然科学基金项目(61174053);教育部科技创新工程重大项目培育基金项目(708069);国家973重点基础研究发展计划基金项目(2014CB845301/2/3);华南理工大学中央高校基本科研业务费基金项目(2014ZP0021)

摘　　要：针对动态的实时目标跟踪算法Camshift(continuously adaptive meanshift)在背景复杂或者存在较多与目标颜色相近的像素时,容易出现跟踪目标丢失的问题,研究并实现一套基于Camshift和SURF(speeded up robust features)算法的目标跟踪系统。使用双内核的DM3730为核心,连接网络摄像头实现图像的动态采集,在DM3730上实现SURF算法和Camshift算法的融合,并负责将采集到的视频压缩,通过网络传输到计算机,对得到的图像做进一步结果分析。实验结果表明,基于Camshift算法和SURF算法融合的目标跟踪系统在简单背景、有相似物体干扰和复杂背景等情况下都能够更准确快速地跟踪到目标,鲁棒性更强、效果更好。Camshift is a dynamic real-time target tracking algorithm based on color features.However,when similar color with the object exists in the background,it may loss the object.To solve the problem,an object tracking system based on Camshaft and SURF was proposed.The IP camera was used to capture the image,and the Camshift and SURF algorithm were accomplished on the dual-core DM3730 to solve shortages of Camshift.Meanwhile,the video was compressed and transmitted to the computer.Experimental results show that the object tracking system can work effectively in different kind of environments,such as simple or complex background,some similar objects disturbing in the environment.It is more robust and more effective.

关 键 词：CAMSHIFT 目标跟踪 快速鲁棒特征 网络摄像头 嵌入式 

分 类 号：TP391.4[自动化与计算机技术—计算机应用技术]