Research On The Recognition Technology Of Aluminum Profile Surface Defects Based On Deep Learning

Surface defects of metal materials will seriously affect the quality,safety,usability and aesthetics of products.Therefore,it is very important to identify the problematic materials in actual production.Traditional manual recognition and machine vision recognition based on image processing can not deal with complex and changeable surface defects in actual production.Based on the idea of artificial intelligence and in-depth learning,this paper classifies and detects the surface defects of aluminium profiles by convolution neural network of in-depth learning technology,taking the surface defects of aluminium profiles as a breakthrough point.Firstly,this paper introduces the research background of metal surface defect recognition and the research status at home and abroad,points out that taking the surface defect of aluminium profile as the cut-in point,and makes it clear that this paper mainly studies the classification and detection of surface defect of aluminium profile based on convolution neural network in deep learning.Then,the main knowledge points involved in introducing in-depth learning into the study of surface defects of aluminium profiles are introduced in detail,including convolution layer,pooling layer,Softmax regression layer,loss function,training process and main parameters of convolution neural network,and the experimental environment for training defect classification network and detection network is established,and the use of Google cloud as an experimental platform is pointed out.Next,an automatic end-to-end classification network for surface defects of aluminium profiles is proposed.The idea of migration learning solves the problem of less labeling data for defective images,and the idea of data enhancement strategy makes great use of the data derived from the Alibaba Cloud platform.In addition,the intrinsic feasibility of the classification network is proved by analyzing convolution kernels,feature images and sal iency maps.Subsequently,a self-adaptive and self-learning surface defect detection network for aluminum profiles based on deep convolution neural network is proposed.After analyzing the characteristics of defects,the feature pyramid is integrated into the defect detection network,and a multi-scale defect detection network is proposed to solve the problem of large size difference of surface defects.Finally,the work of this paper is summarized,and the further research content is given.