摘要
随着自动化、智能化技术的发展,铁路站场需要对集装箱进行自动装卸作业。考虑吊具在作业过程中会发生晃动,为了直接准确地定位集装箱相对吊具的位置,采用视频流对锁孔进行追踪识别及中心定位。在对第一帧图像进行背景分割并提取出集装箱图像后,首先运用基于支持向量机(SVM)的滑动窗口再定位检测算法获取集装箱锁孔区域图像,然后追踪视频流中的锁孔区域并提取对应的Canny边缘轮廓,最后通过基于轮廓端点的轮廓修补算法和锁孔中心定位算法实现追踪锁孔的中心定位。使用了1∶15的起重机和集装箱模型进行模拟实验,实验结果表明:基于SVM的滑动窗口再定位检测算法可准确地提取初始帧的锁孔区域,准确率达到99.65%;基于轮廓端点的Canny轮廓修补算法和锁孔中心定位算法可有效地进行锁孔轮廓修补和锁孔中心定位;MOSSE追踪算法更适用于集装箱锁孔追踪,并可以满足站场自动化的实时性要求。
As the development of automation and intelligent technology,automatic container yards now have been established to improve the efficiency of container handling.Accurately detecting and tracking the lock holes will boost the performance of a crane.Considering that the spreader will shake during the operation process,in order to directly and accurately locate the position of the container relative to the spreader,it was proposed to use the video stream to track and identify the keyhole and to locate the center.After the background image of the first frame was segmented and the container image was extracted,the Support Vector Machine(SVM)-based sliding window relocation detection algorithm was first used to obtain the image of the container keyhole area,and then the keyhole area in the video stream was tracked and the Canny edge contour was extracted.Finally,the center positioning of the tracking keyhole was realized by the contour patching algorithm based on contour endpoints and the keyhole center positioning algorithm.The simulation experiment was carried out using the 1∶15 crane and container model.The experimental results show that,the SVM-based sliding window relocation detection algorithm can accurately extract the keyhole area of the initial frame,and the accuracy reaches 99.65%.The Canny contour patching algorithm based on contour endpoints and the keyhole center positioning algorithm can effectively perform keyhole contour repair and keyhole center positioning.The MOSSE(Minimum Output of Squared Error filter)tracking algorithm is more suitable for container keyhole tracking and can meet the real-time requirements of station automation.
作者
张军
刁云峰
程文明
杜润
姜伟东
ZHANG Jun;DIAO Yunfeng;CHENG Wenming;DU Run;JIANG Weidong(College of Mechanical Engineering,Southwest Jiaotong University,Chengdu Sichuan 610031,China;Sichuan Key Laboratory of Rail Transit Operation and Maintenance Technology and Equipment,Chengdu Sichuan 610031,China)
出处
《计算机应用》
CSCD
北大核心
2019年第S02期216-220,共5页
journal of Computer Applications
基金
四川省重点研发项目(2019YFG0300,2019YFG0285)
中国铁路成都、武汉局集团有限公司科技研究开发计划重点课题(CX1804,18Y06)
关键词
集装箱自动吊装
锁孔追踪
机器视觉
轮廓修补
锁孔中心定位
container automatic handling
keyhole tracking
machine vision
contour patching
keyhole center positioning
