Automatic Sorting System For Pipe Fittings Based On Deep Learning

In recent years,intelligent manufacturing and traditional manufacturing have been closely combined to form a new manufacturing industry form.Sorting is a relatively common way of working in the current manufacturing industry,and its operation scenarios are characterized by high risk and single operation.Therefore,many companies introduce robots to replace manual sorting.The traditional automatic sorting system is based on images,and realizes object classification by binarizing the target image and extracting edge contours.This method has problems such as low target object recognition rate and poor robustness.The sorting system has high recognition rate and good robustness.This paper designs and implements an automatic sorting system for pipe fittings based on deep learning.The main work is as follows:（1）Aiming at the problems of low recognition rate and unsatisfactory mask effect of the Mask R-CNN algorithm,this paper improves it as follows: increase the scale information,extract more low-level features of the pipe image,and improve the area generation according to the size information of the pipe The size of the candidate box in the network,thereby improving the training speed and network convergence effect.By adding an attention mechanism module,the recognition rate and masking effect of pipe fittings are improved.The experimental results show that the algorithm improves the recognition accuracy of pipe fittings and the effect of pipe contour masking,and has better real-time performance and robustness.（2）A method for judging the placement shape of the pipe fittings,the size of the same category,the grasping point and the attitude information is proposed.First,the Hough circle transformation method is used to judge the placement shape of the pipe fittings,and the robot is driven to convert the pipe fittings to be placed horizontally.Second,use the pipe profile mask profile to determine the pipe size.After judging the size of the pipe fitting,combine the centroid of the pipe fitting and the center point of the smallest outside rectangle as the pipe fitting grabbing point.Finally,the angle between the long axis of the minimum circumscribed rectangle of the pipe and the horizontal axis of the image is used as the rotation angle of the manipulator.The experimental results show that the scheme has good accuracy and real-time performance.（3）The automatic sorting system for pipe fittings is designed and realized.First,the pipe fittings are photographed to obtain their images,and the position information of the pipe fittings is calculated by the method proposed in this paper.The computer and the robot end establish TCP communication for data transmission,drive the robot to grasp and stack the pipe fittings,and drive the start-stop operation of the conveyor belt.The experimental results show that,The system has high picking and stacking accuracy.