A Robot Sorting System Based On Vision

In recent years,with the proposed of the Industry 4.0 concept and the Made in China 2025 plan,the intelligent functions and automatization of industrial has been greatly speed up.During the process of industrial automation production,robots usually cope with work that is dangerous,burdensome,boring.But,with the rapid developments of computer technology,image processing technology,pattern recognition technology,robots can not only cope with simple and repetitive working,but also can liberate people from the complex,high-precision position.In the industrial automation production,sorting is one of the important links,but also relatively difficult to be replaced by robot.In order to solve this problem,this paper designs an industrial robot sorting system based on vision,which is used to replace the traditional mechanical and manual sorting.The utility model has the advantages of wide adaptation,high flexibility and portability,and the main work of this paper can be summarized as follows:(1)According to the technical requirements and the main parts of the sorting process,a robot sorting system based on vision has been designed.In the model selection,using IRB360 sorting robot of ABB as the actuator,the imaging source DFK33G445 industrial camera as the image acquisition equipment,with the computer and transfer mechanism set up the entire sorting operation platform.(2)In order to reduce the possible dust,noise pollution,and ensure the quality of the image,the image process of filtering and denoising has been taken.At the same time,the effects of different features on the image are studied.By comparing the characteristics of the corner information,edge information and contour information of the workpiece,we found that workpiece can be described best by the contour information.And it is found that the template matching detection algorithm based on contour information has better effect and anti-interference ability than gray template matching algorithm.(3)For overcome the shortcoming of the traditional target detection and recognition algorithm,such as poor anti-interference,weak adaptability and strong dependence on light source,this paper introduces the machine learning method to achieve the target detection.Firstly,the HOG features of the target samples are extracted,and then the samples are trained by SVM to get the classifier.Because the SVM is a two-point classifier,so we use the method of 1 V all to achieve multi-target detection.(4)Combined with the previous work,the image detection algorithm combined with the CamShift target tracking algorithm is used to prepare a set of sorting algorithm for moving workpiece.When the target is detected,the position of the target is tracked in real time and the coordinate information is recorded.And the coordinate information is sent to the robot controller when the robot is ready.Finally,this paper build a sorting system framework based on the LabView software and write a recognition and tracking algorithm of moving workpiece based on Opencv,and both apply to the sorting operation platform.