| With the "Made in China 2025" nearing its end,the manufacturing process in China’s manufacturing industry has become increasingly automated,modernized and intelligent.In the process of manufacturing factory assembly line,more and more robots replace the manual into the automated production.Machine vision-based robots have gradually become the main industrial robots in modern factories because of their intelligence,flexibility and efficiency.However,the current machine vision-related algorithms still have problems such as inaccurate detection or slow detection speed for objects with blurred boundaries and low contrast,and their application scope is still somewhat limited.In this paper,we design a robot loading and unloading system based on machine vision for the problem that it is difficult to realize automatic loading and unloading of tube-type materials in the current industrial production,and propose a recognition algorithm to meet the recognition accuracy and real-time of the vision system for the problem that it is difficult to recognize the stacked steel tubes with blurred boundary and low contrast,in order to improve the production efficiency of the factory.The main contents and innovation points of this paper are as follows:(1)Overall design and construction of robotic loading and unloading system.Firstly,we analyze the requirements of the loading and unloading process of pipe materials in the factory,design the overall scheme of the loading and unloading system,and analyze the key technologies of the system,point out the shortcomings of the template matching algorithm used in this project and propose improvement ideas;in addition,we establish a D-H model for the built robot,and analyze the forward and reverse kinematics of the robot through the D-H model,and analyze the working space of the robot In addition,the D-H model of the robot was established,and the positive and negative kinematics of the robot were analyzed by the D-H model,and the working space of the end-effector was analyzed.(2)Identification and positioning of material shelves.To address the problem of slow target matching using template matching algorithm for large field-of-view images,the YOLOv8 target detection algorithm is proposed to detect the material shelf in advance and crop the image according to the prediction frame as the input image of the template matching algorithm.The target detection dataset of the material shelf is established for the simulated experimental scenario,and the dataset is expanded by a data enhancement algorithm to improve the generalization of the algorithm;a model training environment is built based on the Py Torch framework and the model is trained based on the YOLO v8 algorithm.It is verified that the model trained in this paper achieves 100% detection accuracy for material shelves.(3)An improved rotation-invariant NCC algorithm is proposed.For the problem that the traditional rotation-invariant NCC algorithm has a large amount of operations in each layer of the pyramid in rotating target detection,the combination of using Hough transform and pyramid layering strategy is proposed to improve the real-time performance of the algorithm,and the robustness of the algorithm is enhanced by Gamma enhancement of the image.After experimental verification,the improved algorithm in this paper improves more than 57% in speed and has higher matching accuracy,and the recognition accuracy of this paper’s algorithm can reach more than 95% under different light environments,with high robustness.(4)Conduct experimental tests on the integration of robotic loading and unloading system.In the simulated experimental environment,the system designed in this paper was tested in terms of positioning accuracy,precision and real-time.The experimental results show that the robotic loading and unloading system designed in this paper has high recognition accuracy,recognition speed and positioning accuracy,and has strong engineering significance for improving the production efficiency of intelligent workshops. |
PDF Full Text Request