Design And Implementation Of Target Detection System For Logistics Picking

In recent years,the logistics industry has developed rapidly,and the volume of freight has increased rapidly year after year.How to keep the efficiency of the logistics industry in line with the increasing volume of freight is an important research topic.In the logistics work line,the "cargo-to-person" picking station is a work scenario to reduce the workload of workers,improve the efficiency of picking by workers,and realize the collaborative work of machinery and people.The "cargo-to-person" workstation requires a robotic arm to pick items on the conveyor belt.A precise,fast response target detection system is of great significance for the efficient and accurate work of the robotic arm.Therefore,for the application scenario of the "goods-to-person" picking station,we propose the process of automatic selection,which has great practical application value.Based on the research on the algorithm of target detection technology,a target detection system for logistics picking is designed.The system is designed to meet the application requirements of the "removal and picking" of the "goods-to-person" workstation.After the workstation,the logistics picking target detection system starts real-time detection,and the robot begins to perform the picking work,placing the goods in the designated outbound order box.The system uses Tensorflow as the framework for the implementation and training of the logistics picking target detection algorithm.It selects Faster RCNN algorithm model as the main network model of the target detection algorithm,and provides different feature extraction networks for users to choose according to different needs of users.The system uses a COMATRIX industrial camera as a collection tool for image data,uses Cobotsys to communicate with the camera and the lower computer,collects 2D images and depth maps,performs target detection on the 2D images,and combines the depth map information to obtain the three-dimensional coordinates of the capture points.In order to facilitate the use of the user,the system uses the Py Qt5 framework design to implement the graphical interface of the system.The mean average accuracy(m AP)and model derivation speed of faster RCNN have achieved ideal results in box data sets..The design and implementation of the logistics picking target detection system has been completed.The test results show that the logistics picking-oriented target detection system has a high accuracy and real-time detection effect in the "goods to people" workstation scenario.