Research On Goods Identification And Picking Method By Vision Robot Of Modern Logistics System

With the functions of storage, transmission, distribution and management, automated warehouse has the advantages of more storage capacity, less floor area, easy manipulation and can be integrated with ERP system information. Based on Modern logistics control experimental center of Shandong University, an automated goods identification and picking system is proposed in this paper by introducing the computer vision technology and robot control technology into automated warehouse. Based on this system, the automatic goods positioning, automatic goods identification and automatic goods storage and retrieval management are realized, to eventually realizing a fully automatic and intelligent unmanned warehouse.In the goods' image processing area, the researches as follows have been done. (1) Goods' image acquisition. An image acquisition system which is integrated by CCD camera, image acquisition card and personal computer is designed to get the goods' images. (2) Goods' image pre-processing. In order to make the images more suitable for further processing, a goods' image pre-processing algorithm is proposed. Because of this algorithm's characteristics, the noise is filtered effectively and the edge information is preserved mostly. (3) Edge detection and contour extraction of goods' images. Considering the characteristics of the goods images, an edge detection and contour extraction method is proposed. Firstly the goods' images are detected roughly to locate the goods' general positions. Secondly these positions are detected detailedly to find the goods' edge. At last, the goods' single-pixel wide contour is extracted by the contour extraction method. And the positioning of goods is realized.In the goods' recognition area, the researches as follows have been done. (1) A goods' feature extraction method is proposed to make the features meeting the rotation invariance and translation invariance. The kernel-weighted color histogram, entropy value and contour's Fourier descriptor, which belong to the color feature, texture feature and frequency domain feature respectively, are selected for goods' recognition. And the advantages of each feature are analyzed based on the experiment results. (2) Multiple classifier combination. For improving the correct recognition rate and the reliability level, the serial integration of multiple classifiers is used in this paper. The kernel-weighted color histogram, contour's Fourier descriptor and entropy value are used as the characteristics of the first, second and third classifier. Then the discrimination methods and the parameters of each classifier are decided according to the goods' features. And the goods' recognition is realized.In the SK-6 robot control area, with the comprehensive use of local programming and remote command, a SK-6 robot control method is proposed. By the combination of real-time control and fixed works, the robot is controlled to finish the image shooting and goods picking work successfully.The experiments show that the goods' recognition algorithm is reliable and the SK-6 robot control scheme is feasible. The visual positioning accuracy meets the technical requirements of goods picking, goods storage and retrieval operations. With the correct control of goods' automatic drawing and automatic handling, the expected design request is achieved and this method can be applied to the automatic goods storage and retrieval operation of automated warehouse.