一种图像差分与特征匹配的货物检测系统研究

A Cargo Detection System Based on Image Difference and Feature Matching

针对自动化立体仓库货物检测难度大、实时性差等缺陷,提出一种基于图像差分与特征匹配的货物检测系统。通过四旋翼无人机结合无线图像系统获取货物图像,并传输至上位机软件系统进行检测,通过将实时图像与模板图像差分结合自适应阈值滤波方法检测出货物缺失部分;通过改进的ORB算法结合网格运动统计方法完成图像的精准匹配,采用网格划分统计特征点的方法检测货物缺陷。实验和现场系统运行测试表明,所研究的货物检测系统,在货物缺失、缺陷检测时正确率分别达到95.0%和93.3%,具有较高的准确率。

In order to solve the difficulty of cargo detection and poor real-time performance in automated warehouse,a cargo detection system based on image difference and feature matching is proposed.The cargo image is obtained by Quad rotor UAV combined with wireless image system,and transmitted to the software system for detection.The cargo missing is detected by image difference combined with adaptive threshold filtering method;the improved orb algorithm combined with the grid motion statistics method is used to accomplish the accurate image matching and cargo defects is detected with the method of grid division statistical feature points.Experiments and field tests show that the accuracy of the proposed system is 95.0%and 93.3%respectively.