Research On Dynamic Perception Of Objects In Metal Casting Process Based On Machine Vision

In the process of metal casting,oxide slag will form on the surface of the liquid metal solution at high temperature,which will affect the product quality if it is not removed in time.In the casting cooling stage,the metal is transformed from liquid to solid,so the robot operation needs to intelligently perceive the change of casting target under complex working conditions.Therefore,this thesis focuses on the visual perception of the intelligent slag removal system,taking zinc alloy metal ingots as the research object,and carries out research on three aspects: the perception of mold position,the perception of the thickness distribution of oxide slag in the mold,and the perception of industrial robot slag removal information.The design of the industrial robot intelligent slag removal operation system based on machine vision guidance is realized.The main work and innovations of this thesis are as follows:(1)Focusing on the problem of poor robustness of mold tracking due to casting occlusion in the metal casting process,a BS-Vibe algorithm is proposed to achieve accurate and efficient segmentation of the conveyor belt area.An edge extraction method based on SG smooth gray profile curve is proposed,so that the local directional feature on the edge generates an adaptive feature template,and the detection of metal molds is completed according to the similarity measure.The mold tracking method based on data association matrix is used to accurately label the detected molds.(2)Focusing on the problem of uneven distribution of oxide slag on the surface of the ingot,which leads to the uncertainty of the robot’s slag removal area and depth,a DTANet network is proposed to classify the oxide slag thickness.Cubic spline interpolation is used to extract the illumination component of a single mold image and correct it.According to the size of the mold model,the internal area of the single mold is divided into four sub-regions,and the trained DTANet network is used to classify the thickness of the oxide slag in each area of the mold.(3)Focusing on the problem that the robot slag removal area and depth cannot be accurately extracted,a method for dividing the slag removal area is proposed.According to the thickness distribution of oxide slag on the surface of the ingot,two types of slag removal points and slag removal areas are determined.According to the principle of equal volume,the extraction of slag removal depth information is completed in the slag removal area.Finally,according to the algorithm in this thesis,the hardware equipment selection is completed,the communication format between the vision module and the industrial robot is determined,and the software and hardware design of the industrial robot intelligent slag removal operation system based on machine vision guidance is completed.Figures 51,Tables 6,References 78.