Research And Implementation Of Intelligent Detection System Based On Deep Learning

With the increasing demand and requirement for product detection system in industrial production,there are many intelligent industrial detection systems on the market,which are applied in different fields.Among them,defect detection based on product surface image classification is one of the most popular application scenarios nowadays.How to design an intelligent detection system with good detection effect,high stability and releasing manpower has become a hot spot in the field of industrial production.The traditional method of product surface defect detection relies on manual visual inspection.Some enterprises and scientific research institutes also use the principle of pattern recognition to propose or design a defect detection system relying on machines.Compared with the traditional method relying on manual visual inspection or tactile inspection,using machines instead of manual inspection releases human resources and solves the problem of simple repetitive work done by people on the production line.However,there are still some shortcomings in these systems:(1)The reliability of defect detection by pattern recognition method can not meet people's expectations,and the patterns of unqualified products are often diverse,and these detection machines can only detect some unqualified products.(2)For different products to be inspected,the traditional detection system needs to grasp the characteristics of different products manually and design corresponding detection methods.The transplantation cost is high,and the later maintenance cost is huge.(3)The operation experience is not good,the traditional detection system lacks a unified visual management platform,which can not easily view the real-time situation of detection.In order to solve the above problems,an intelligent detection system with good detection effect,good system portability and convenient management operation is designed.In this paper,based on the convolutional neural network algorithm in deep learning,the following work is done.(1)The characteristics of image data that may be processed in detection are analyzed.IThe convolution neural network model is used to optimize the activation function and automatically extract the features of product image,which further improves the performance of the model in the application of image classification and meets the detection requirements in industrial production.(2)The image classification model based on convolution neural network does not need to extract artificial features for different products.It only needs to take pictures of products to train the model.When the program calls the model,the image features will be automatically extracted.The whole training process does not need manual participation,which not only improves the flexibility of the algorithm,but also reduces the system The maintenance cost of the period.(3)Intelligent testing system platform,which integrates model management,file management,equipment configuration and data analysis,can effectively carry out product testing,improve the monitoring of product quality by producers,and verify the validity of the research results.