Automatic Identification Technology Of Waste Bucket Codes Based On Deep Learning

When testing barrels of solid nuclear waste,it is necessary to manually record the codes on the barrels of nuclear waste solid waste for classification.In this process,the radiation generated by nuclear waste will cause certain harm to the human body,and it is also not conducive to the improvement of economic benefits.Therefore,if the codes on the nuclear waste solid waste storage barrels can be accurately identified,and the unmanned and intelligent industrial scenarios can be realized,it can not only reduce the tedious and dangerous manual labor,but also effectively improve the efficiency and save social resources.This paper conducts an in-depth study on the specific application requirement of fully automatic identification of waste bucket codes.The specific research contents and results are as follows:(1)This paper studies and implements a Tesseract-based fully automatic identification method of waste storage bucket codes.First,use an industrial camera to align the waste storage bucket on the rotating platform,collect sequence image data and preprocess the image during one rotation,including image denoising,image binarization,and region screening based on connected domains,thereby segmenting the encoded region from the background image.Secondly,the Tesseract optical character recognition technology is used to detect the characters in the coding area.Finally,a scanning line-based coding sorting method is proposed,which removes redundant codes in sequence images and sorts them correctly,so as to realize automatic identification of waste bucket codes.This method has high recognition efficiency,and the codes in the images collected during the rotation can be recognized in real time,and the correct codes can be obtained within 1 second after the waste storage bucket stops rotating,so as to achieve the goal of real-time classification during assembly line operations.(2)This paper designs a CNN-based network model for the identification of waste bucket codes.The image size of the input layer of the network model is set to 16 × 16 pixels,the number of convolution layers is set to 3 layers,and the size of the convolution kernel is all set to 3 × 3.At the same time,32,64,and 64 convolution kernels are stacked in the first,second,and third convolutional layers,respectively.In addition,a batch normalization layer is added after each convolutional layer,and a leaky linear rectification function(Leaky Re LU)is selected as the activation function.When identifying the waste bucket codes,a CRAFT algorithm that introduces an attention module is used to locate and segment the codes.After that,the located code is sent to the waste bucket code identification network for identification.Finally,based on the scanning line-based coding sorting method,the redundant codes in the sequence images are removed and sorted correctly,so as to realize the automatic identification of the waste storage bucket codes.(3)In this paper,a set of automatic identification system of waste storage bucket codes for assembly line operation is developed.The hardware platform of the code recognition system for waste storage buckets was built,and the software system was developed using Python.The system encapsulates the two waste storage bucket code recognition algorithms proposed in this paper,which can realize the functions of code positioning,identification and sorting,so as to automatically identify the codes of waste storage buckets operating on the assembly line in real time.Finally,the system is tested and verified under the actual application environment.The experimental results show that the accuracy rate of the automatic identification algorithm of waste storage bucket coding based on Tesseract reaches92%,and the waste storage bucket coding based on the network model of waste storage bucket coding recognition network model is accurate.The accuracy rate of the fully automatic identification algorithm is 96%.