Research And Development Of Online Statistical System For Enamel Production Based On Deep Learning

Image recognition technology is one of the key technologies for intelligent manufacturing and is also widely used in the automation of traditional production lines.In this paper,the production line of enamel products for multi-type mixed-line production is taken as an object,and an on-line product statistical system based on deep learning image recognition technology is researched and developed.The image preprocessing,target area positioning,image recognition and actual size calculation involved in the system are studied.The technology was studied and the Caffe framework was used to implement the prototype system development based on MATLAB.In this paper,the color image preprocessing technology is first studied.In the aspect of image enhancement,the algorithms such as histogram equalization,Laplacian operator,logarithmic transformation and gamma transformation are analyzed;in the noise removal,the neighborhood average method is used.Three kinds of algorithms were analyzed,median filtering method and Gaussian filtering method.Finally,according to the contrast effect of different preprocessing methods,the image enhancement algorithm based on Laplacian operator and the denoising method based on Gaussian filter were used to preprocess the image.After that,the statistical methods for classification and classification of enamel products were studied.Five common object detection networks in the image region location method were compared and analyzed,including R-CNN,Fast R-CNN,Faster R-CNN,YOLO,SSD,etc.,and Faster R-CNN algorithm was chosen to design the regional positioning model of enamel products.The analysis of the deep convolutional network and ROI pooling used for image classification,combined with the deep convolutional network adding ROI pooling layer and the actual size conversion algorithm of enamel products,designed the classification and identification model of enamel products;and classified statistics for enamel products.The model was evaluated for performance.Finally,the online statistics system for enamel products was developed.The system includes three parts: image acquisition,target area positioning,and target classification and statistics.The functional structure of the system was analyzed.A MATLAB-based online classification statistical prototype system for enamel products was developed to determine Real-time statistics on the types and quantities of hybrid enamel products.