Research On Sorting System Of Multiple Robots Based On Machine Vision

With the continuous advancement of "Made in China 2025",the robotic intelligent sorting system based on machine vision is more and more widely used in the sorting industry with its high precision and high efficiency.This paper takes the sorting system consisting of machine vision and multiple Delta robots as the research object,and the effective method of recognition and location under two different working conditions is proposed.At the same time,the task assignment of multiple robots and the single robot sorting path are studied.The main research work and achievements of the full text are as follows.Firstly,a vision-based multi-robot sorting system is designed,including a vision system and a parallel robot system.According to the actual working environment,the selection of hardware including robots,cameras,lenses and light sources was determined and the development platform of vision system and robot control system was introduced.Those work provides a theoretical basis for the subsequent experimental platform construction.Secondly,the noise removal effects of four commonly used filtering algorithms are studied under certain working conditions.The watershed image segmentation algorithm based on distance transform is studied to solve the problem of image segmentation between mutual contact targets.Aiming at the identification and location of the target under different working conditions,an effective method is proposed.Under the simple working condition which is no overlap between the targets.The image matching recognition and localization algorithm based on the target contour features is studied.By extracting the target contour features,using the invariant moments for feature description and differential measurement,the target recognition is finally achieved.Get the target position information by calculating the contour average.For the more complicated working condition which is occlusion or overlap between the targets.The recognition algorithm based on a local feature of corner points is studied.The feature of the corner points is described by binary operators.Use Hamming distance to complete initial matching.The second matching is performed by the random sampling consistency algorithm,which improves the accuracy of target recognition.The target position information is obtained by calculating the average value of the position in the dense corner points distribution area..Then,aiming at the task assignment problem of multiple parallel robots,a method of interval balanced assignment based on target position sorting was proposed to reduce the missing rate of the sorting system and coordinate the sorting tasks of multiple parallel robots.Aiming at the optimization of dynamic target sorting path of a single parallel robot,an optimization method based on the shortest sorting path principle is proposed.Firstly,the mathematical model of sorting path based on dynamic target is established,the order of target fetching is determined by a certain sorting principle,and the sorting path in each case is calculated through the combination of target placement position,and the shortest sorting path is finally obtained.This method can not only reduce the missing rate of sorting,but also shorten the sorting path to improve the sorting efficiency of a single parallel robot.Finally,an intelligent sorting system of multiple parallel robots based on a single vision system was built.The communication between the vision system and the parallel robot system is completed,and the transmission of the target position information is realized.At the same time,the calibration of the sorting system and the target tracking on the conveyor belt are completed.Through object recognition and sorting experiments,the feasibility of the algorithm and sorting system is verified.