Parts Inspection System Based On Deep Learning And Machine Vision Research And Implementation

At present,with the continuous progress of computer vision technology,research and application of related technologies such as unmanned driving,behavior recognition,smart city,and intelligent production have been widely carried out,and rich results have been achieved.Target detection is the key technology of these applications.Taking full advantage of this technology can effectively reduce the consumption of manpower and has important practical and economic significance.At present,most machinery factories still rely on manual sorting in the sorting process of parts,and the efficiency is relatively low.Workers in the production workshop are prone to errors in the identification of some parts with similar shapes and sizes.How to pass effective distinguishable data,For example,contour,key point,size,etc.,to complete the automatic identification of parts on the work table is a difficult point to improve the efficiency of the identification operation.Relies entirely on manual recognition,strong subjectivity and heavy workload,and has high requirements for workers' skill proficiency.New employees are prone to misdetection at the beginning of their work,so design a design that can help workers identify workpieces and make some difficult to distinguish The system that gives the similarity evaluation and size information of the workpiece is of practical significance.After on-site investigation of local parts production and processing enterprises,after detailed analysis of actual needs,this paper studies and implements a system based on deep learning and machine vision algorithm,and proposes to use deep learning-based target detection method to target industrial parts data Characteristics,by adjusting the existing Mask-RCNN network structure,the accuracy of target recognition is improved,and the output information of the mask branch is used for information extraction and processing,combined with a variety of edge detection algorithms to finally measure the size information of the detected target,And the size information as a basis for identification,you can improve the sorting efficiency of workers.Based on the above research content,a part inspection system based on deep learning and machine vision is designed and implemented under the Mac Os system.The feasibility of combining deep learning and machine vision measurement with two types of algorithms is comprehensively evaluated through qualitative and quantitative methods.Experimental results show that the method proposed in this paper achieves multi-target detection and segmentation,as well as labeling of size information.The classification accuracy is high and the size error is low.It can effectively help workers reduce labor intensity and improve sorting accuracy.