Research On Automatic Disassembly Information Model Of Used Product Based On Machine Vision

Intelligent remanufacturing is the direction of innovation and development of remanufacturing technology.Integrating the new generation of artificial intelligence technology into the disassembly process will help to improve the automation and intelligence of remanufacturing disassembly and realize flexible large-scale disassembly of waste products.In this paper,aiming at the small batch,diversification and flexibility of waste products in the disassembly line,machine vision technology is added to the disassembly process to obtain the assembly model of waste products based on multi-angle three-dimensional reconstruction method.A disassembly discrimination method for rusted bolts of used products based on depth learning +support vector machine is proposed,which can automatically detect the position of bolt connections for key parts and distinguish the disassembly of rusted bolts.Through computer simulation analysis technology,the disassembly information model is established to automatically analyze the spatial geometric disassembly of the parts of the waste product after reconstruction,and the disassembly sequence planning of waste product based on improved ant colony algorithm is proposed by combining the bolt disassembly status information.It provides an effective way for deep automatic disassembly.The research content of this paper is as follows:(1)Obtain three-dimensional model of waste products by three-dimensional reconstruction method based on machine vision.Sparse point cloud of waste products is reconstructed by SFM algorithm,and multi-angle two-dimensional image is restored to three-dimensional space points.PMVS algorithm is used to reconstruct sparse point cloud into dense point cloud to obtain more surface information of used products.The surface shell model of used products is obtained by surface reconstruction with Poisson surface reconstruction algorithm.After entity transformation,the part segmentation is completed step by step and the reconstructed assembly model of waste products is obtained.(2)A disassembly discrimination method for rusted bolts of used products based on machine vision is proposed.Bolt detection model of waste products based on YOLOv5 is used to automatically detect and cut bolt targets of different waste products.After image preprocessing,the target image is segmented by Hough gradient circle detection algorithm to remove background interference.The disassembly discrimination model of rust Bolts Based on SVM is used to complete the disassembly discrimination of rust bolts and obtain the disassembly status information of key parts for the disassembly of waste products.(3)The information model of scrap product disassembly is established and the scrap product disassembly sequence planning considering bolt disassembly status is proposed.Based on the four interference matrix models,comprehensive interference inspection results and detailed disassembly path angles of parts are obtained on the basis of ensuring disassembly stability.The disassembly prototype system is developed to automatically extract the disassembly path information and disassembly status information of the components taking the waste product assembly after three-dimensional reconstruction as an example.Based on improved ant colony algorithm and considering the disassembly status of parts,the disassembly sequence of used products is automatically generated.