一个基于图像的物流数据采集系统的研发
发布时间：2018-12-27 20:53
【摘要】：物流数据的采集是实现供应链物流信息化的基础。本文基于图像处理的方法进行物流数据的采集,将行人作为图像研究的运动目标进行统计。在运动目标检测、分类和跟踪的技术基础上,结合目标统计在车站、道路、超市出入口等公共场所的广泛应用需求,设计并实现了一个基于图像的物流数据采集系统。在运动目标检测方面,本文利用背景建模的方法对监控区域的背景进行实时更新,采用背景减除与帧间差分相结合的方法进行运动检测,提取出运动区域。在运动目标分类方面,本文采用一种简化了的目标分类方法,选取运动区域的高宽比和垂直投影直方图梯度作为特征对目标进行分类,提取出行人目标,作为后续跟踪统计的基础。在行人跟踪研究方面,针对Camshift算法在目标互相遮挡时易发生跟踪丢失的缺陷,采用引入Kalman预测和Camshift匹配搜索算法完成跟踪任务。最后,在前文研究的基础上,本文针对物流数据采集系统的应用需求,实现了在实际环境下对人体目标的检测提取和跟踪计数。实验证明本文方法实用可行,统计数据具有较高的正确率。
[Abstract]:The collection of logistics data is the basis of realizing supply chain logistics informatization. In this paper, the method of image processing is used to collect the material flow data, and the pedestrian is regarded as the moving object of image research. Based on the technology of moving target detection, classification and tracking, and combined with the extensive application of target statistics in public places such as stations, roads, supermarket entrances and exits, an image based logistics data acquisition system is designed and implemented. In the aspect of moving target detection, the background modeling method is used to update the background of the monitored region in real time, and the method of background subtraction and inter-frame difference is used to detect the motion, and the moving region is extracted. In the aspect of moving object classification, this paper adopts a simplified target classification method, selects the aspect ratio of moving area and vertical projection histogram gradient as the feature to classify the target, and extracts the travel target. As the basis of follow-up tracking statistics. In the research of line human tracking, the Camshift algorithm is used to complete the tracking task by introducing Kalman prediction and Camshift matching search algorithm, aiming at the defects of the Camshift algorithm which is prone to the tracking loss when the targets are occluded from each other. Finally, on the basis of the previous research, this paper aims at the application demand of the logistics data acquisition system, and realizes the detection, extraction and tracking counting of the human body target in the actual environment. The experimental results show that the method is practical and the statistical data have a high accuracy.
【学位授予单位】：武汉理工大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TP391.41
本文编号：2393588
[Abstract]:The collection of logistics data is the basis of realizing supply chain logistics informatization. In this paper, the method of image processing is used to collect the material flow data, and the pedestrian is regarded as the moving object of image research. Based on the technology of moving target detection, classification and tracking, and combined with the extensive application of target statistics in public places such as stations, roads, supermarket entrances and exits, an image based logistics data acquisition system is designed and implemented. In the aspect of moving target detection, the background modeling method is used to update the background of the monitored region in real time, and the method of background subtraction and inter-frame difference is used to detect the motion, and the moving region is extracted. In the aspect of moving object classification, this paper adopts a simplified target classification method, selects the aspect ratio of moving area and vertical projection histogram gradient as the feature to classify the target, and extracts the travel target. As the basis of follow-up tracking statistics. In the research of line human tracking, the Camshift algorithm is used to complete the tracking task by introducing Kalman prediction and Camshift matching search algorithm, aiming at the defects of the Camshift algorithm which is prone to the tracking loss when the targets are occluded from each other. Finally, on the basis of the previous research, this paper aims at the application demand of the logistics data acquisition system, and realizes the detection, extraction and tracking counting of the human body target in the actual environment. The experimental results show that the method is practical and the statistical data have a high accuracy.
【学位授予单位】：武汉理工大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TP391.41
【参考文献】
相关期刊论文 前7条
1 张国华;卜纪伟;薛雷;肖刚;潘汉;;基于像素级图像融合的Mean Shift目标跟踪[J];微型电脑应用;2011年04期
2 彭源;;图像局部特征的提取检测技术[J];装备制造技术;2010年02期
3 杨伟;柴奇;;基于mean-shift的多目标粒子滤波跟踪算法设计[J];航天电子对抗;2009年05期
4 许言午;曹先彬;乔红;;行人检测系统研究新进展及关键技术展望[J];电子学报;2008年05期
5 张会军;;基于图像的动目标检测技术[J];微计算机信息;2007年22期
6 李子青;;智能视频监控技术——自主创新引领未来[J];中国安防;2007年03期
7 侯志强;韩崇昭;;视觉跟踪技术综述[J];自动化学报;2006年04期
相关博士学位论文 前2条
1 孔晓东;智能视频监控技术研究[D];上海交通大学;2008年
2 朱胜利;Mean Shift及相关算法在视频跟踪中的研究[D];浙江大学;2006年
相关硕士学位论文 前1条
1 李菲;智能监控系统中行人流量统计功能的研究与实现[D];西安电子科技大学;2010年
，本文编号：2393588
本文链接：https://www.wllwen.com/guanlilunwen/wuliuguanlilunwen/2393588.html
