Development Of Intelligent Identification System For Automatic Disassembly Of Waste Electrical Appliances

Modern intelligent disassembly puts forward higher requirements for disassembly of waste electrical appliances.Automatic classification intelligent disassembly and materials obtained from different disassembly scenarios have higher requirements for intelligent disassembly technology of waste electrical appliances.In the traditional dismantling process,the waste electrical appliances are mainly identified and classified by manual or mechanical methods.Due to the variety of waste electrical appliances,the size,color,shape and other factors are not fixed,the classification is time-consuming,the degree of automation is low,and the identification efficiency is low.The existing target recognition algorithms generally use neural network to identify the target,but do not extract the specific features of waste electrical appliances to identify and classify them.In consideration of the above problems,this text put forward a kind of intelligent identification system of waste electrical appliances based on machine vision.The main work and achievements are as below:(1)Target and background segmentation of waste electrical appliances.Firstly,this paper studies the techniques and methods of image semantic segmentation in the field of deep learning,compares the characteristics of different semantic segmentation and introduces the advantages and practicability of PSPNet semantic segmentation algorithm,and then introduces the principle and network structure of PSPNet.The PSPNet network was built based on TensorFlow deep learning framework under Windows 10 operating system,and the corresponding data set was made by collecting and labeling certain pictures of waste electrical appliances.Then,the PSPNet network was trained and tested by transfer learning.PSPNet semantic segmentation network was used to obtain the mask image of waste electrical appliances,and finally the mask operation was carried out to obtain the target image of waste electrical appliances segmented with background.(2)Feature extraction of waste electrical appliances.In this paper,Hu moment invariants of waste electrical appliances are extracted as the shape features of waste electrical appliances,and PCA is used to reduce the dimension of Hu moment invariants extracted.The improved AlexNet convolutional neural network is used as the network junction to extract deep features of waste electrical appliances,and the advantages of the improved AlexNet network compared with other convolutional neural networks for deep features of waste electrical appliances are introduced.Finally,the extracted shape features and deep features are fused,and the fused features are normalized,that is,the exclusive features of waste electrical appliances are extracted.(3)Recognition and classification based on extracted features.First introduces classification algorithms in machine learning and this article USES the SVM classifier advantage compared with other classification algorithms,and introduces the principle and parameters of the SVM structure,build more class SVM classifier in construction and introducing the improved loss function,after the normalization of characteristics of waste electrical appliance into the category structure to train the SVM classifier.Using the trained classifier to test,the recognition accuracy is as high as 91.21%,its precision meets the intelligent identification requirement of waste electrical appliances before automatic disassembly.(4)Development of intelligent identification system for waste electrical appliances.In terms of hardware,image acquisition and processing are carried out by external camera.In terms of software,Qt Desigenr graphical interface software is used in C++ language,the deep learning module is developed by Python,the image processing library is Opencv3.4.1 and PIL,and Mysql is used as the local database interface in Qt Creator.Navicate Premium software was used to visualize data,and finally pyinstaller was used to package the software.The corresponding software was developed and tested to realize and meet the requirements of intelligent identification of waste electrical appliances.