Research Of Surface Defect Recognition Of Aluminum Profile Based On CNN

The surface quality of products is an important part of the overall quality of products,and plays an important role in improving the core competitiveness of products.Therefore,many manufacturers will strictly control the surface quality of products.At present,aluminum profile manufacturers generally use manual inspection methods in the quality inspection.In the detection of surface defects of aluminum profiles,J company still adopts the traditional manual detection.With the increasing degree of production automation,the shortcomings of manual detection methods become increasingly prominent.Aiming at the 10 types of surface defects commonly seen by J company in the production of aluminum profiles,this paper introduces deep learning methods into the recognition of aluminum surface defects,based on convolution neural network to identify the surface defects of aluminum profiles,and design the surface of aluminum profiles Defect detection system prototype.First,through a large amount of collection and reading of relevant literature,analyze the current status and problems faced by J company in the detection of product surface defects,review the current status of domestic and foreign product surface defect detection,and determine the use of deep learning convolutional neural network Recognize the surface defects of aluminum profiles.Compared with traditional machine learning algorithms,deep learning algorithms can automatically extract features without the need for additional manual intervention,and are more robust and generalized.Convolutional neural networks are representative networks in deep learning.Secondly,pre-process the aluminum profile data,and then construct different aluminum surface defect recognition models.For the problem of the small size of the aluminum profile data set,use the transfer learning method to fine-tune the different convolutional neural network models in the aluminum profile data.Training and testing on the set,compare the recognition effect of different aluminum profile surface defect identification models in the aluminum profile data set,and finally select the ResNet50-based aluminum profile surface defect identification model as the optimal model for the aluminum profile surface defect detection system prototype.Finally,an aluminum profile surface defect recognition model based on Res Net50 is called,and a prototype of aluminum profile surface defect detection system is designed through Py Qt5.The system can detect and identify the aluminum profile products produced by the specific production line,and save the statistics of the test results.Through the statistical results,the production quality of the product can be analyzed,and the defect rate and the proportion of defect types of any batch of products on the specific production line can be analyzed.This has a guiding role in the production and processing of subsequent products and is conducive to improving the production process of the product.Process.The interface of the system is simple and friendly,and the operation is simple,which provides a certain reference and reference value for J company to realize the online automatic detection of surface defects on aluminum profiles.