Research On Reinforcement Binding Point Location Method Based On Machine Vision

Taking the project of Fuzhou Science and Technology Bureau(2020-GX-16)as the background,this article has done some research,experiments,and analysis on the difficult problems encountered in the location recognition and positioning of steel bar binding points during the process of using machines instead of manual work.In this paper,two vision system schemes are designed for different steel bar binding platforms.After comparison,a converged binocular vision system built with two monocular cameras coupled with a gantry type 3-degree of freedom binding platform suitable for the needs of the factory project is selected for research;The shadow problem in the difficulty of binding point recognition is studied and a method of removing shadow using R-Y channel redundancy is proposed;Three identification schemes are used to identify the binding points of steel bars.After comparison,the most suitable skeleton method is selected for identification and optimization;Finally,a convergent binocular vision system is constructed to study the positioning method of the binding point,and calculate the spatial coordinates of the binding point in the left camera coordinate system.The main work of this article includes:(1)Two visual positioning systems were designed based on a single binocular camera(top single point shooting)and two monocular cameras(dual side converging shooting),respectively paired with six degrees of freedom arm binding systems and gantry type three degrees of freedom rectangular coordinate system arm binding systems.Designed and built a converged binocular vision system for the second gantry binding system,which meets the actual needs of the factory and expands the camera recognition range.(2)Taking the project as the background,the application of different image preprocessing methods in the processing of steel bar images was studied,and the shadow problem in steel bar images was studied.The concept and steps of converting color space and analyzing image features by channels were described.On this basis,a convenient method for removing steel bar shadows was proposed and applied to the visual system to solve the existing problems.Using closed operations and morphological methods to remove noise,analyzing several commonly used edge processing operators,and extracting more complete and clear steel edges.(3)Three methods,skeleton method,vertex intersection method,and neural network method,are used to quickly and accurately identify the binding points of steel bars.By comparing various types of visual recognition methods,the visual method skeleton method of the visual recognition positioning system in this paper is determined.Hough line optimization is performed on this recognition method to improve the recognition accuracy.(4)A binocular converging positioning system is designed for the gantry type three degrees of freedom binding system,and the binocular camera is calibrated to obtain the internal and external parameters of the camera.Then,the binocular camera is calibrated to obtain the projection matrices P1 and P2.Finally,the binding point is located using the parallax method,and the spatial three-dimensional coordinates of the binding point in the left camera coordinate system are calculated.After the binding experiment verification,the positioning results are within the binding error range,meeting the project requirements.