Research On Image Recognition And Positioning Technology For Automatic Assembly System Of Train Equipment Box

This paper takes the realization of the automatic assembly of train equipment boxes as the research purpose,and carries out the overall design of the assembly system according to the current assembly status.In this paper,the system composition,handling path,control method,navigation method,alignment algorithm,etc.are designed,and the visual positioning part is studied in depth.The main research contents are as follows:(1)Based on the assembly environment and current status of the train equipment box,an automatic assembly plan was developed to minimize the changes to the existing environment and realize local automatic transformation.(2)The mechanical structure of the system is designed,and the software and hardware used are selected and explained.With multi-sensor fusion and visual positioning technology as the core,a three-level navigation method was designed to realize the equipment box handling and mounting hole alignment.(3)The calibration technology of the two cameras involved in visual alignment is studied.Based on the comparison of several classic camera calibration methods at present,Zhang's calibration method was used to calibrate the two positioning cameras respectively according to the characteristics of the research object.The calibration results and reprojection errors were obtained.(4)The third level of navigation is the precise positioning based on visual recognition positioning,which directly affects the success or failure of the equipment box assembly.This paper performs a series of processing on the collected images,including denoising filtering processing,threshold segmentation processing,edge detection and feature recognition.For each step of image processing,multiple algorithms are designed.For the collected images of this topic,programming experiments are performed under the environment of OpenCV 3.1.0 configured in Visual Studio2015.The results of the classical algorithm processing are compared and analyzed.Excellent plan.(5)A four-degree-of-freedom experimental platform is designed for the visual positioning involved in the third-level navigation.To link the image processing data with the physical coordinates of the experimental platform,hand-eye calibration and pixel equivalent calibration were performed.The deviation forms in the alignment are studied,and a four-degree-of-freedom platform alignment control algorithm is proposed.The feasibility of the image processing algorithm and the control algorithm is verified by experiments.