Accuracy Analysis And Improvement Of Visual Positioning Method For Aircraft Skin Riveting Holes

In the detection process of combustible cartridges,traditional manual defect detection methods have problems such as low efficiency,poor detection quality,and high risk factors,which cannot meet the detection requirements of efficient and intelligent production.In response to the above problems,this paper uses machine vision methods to study the defect detection of combustible cartridges,proposes a convolutional neural network method based on texture features,and verifies the superiority of this method through comparative experiments.First,the relevant principles and selection principles of each component in the defect detection system are introduced respectively,and based on the physical characteristics of the surface of the combustible cartridge,appropriate hardware is selected to build a combustible cartridge visual defect detection system.Secondly,a feature analysis method using mutual information entropy to optimize the texture feature parameters of the gray level co-occurrence matrix is proposed.The influence of the construction factor of the gray level co-occurrence matrix on the feature parameters is analyzed through experiments;the maximum correlation and minimum redundancy between the texture feature parameters of the gray level co-occurrence matrix and the sample are calculated,and the feature parameters that can express the image texture are selected.Then,the related principles of artificial neural networks are explained and the principles of convolutional neural networks are explained.On the basis of the existing convolutional neural network,a convolutional neural network based on texture features is proposed.By introducing the feature parameters of the gray-level co-occurrence matrix,the generalization ability of the convolutional neural network is improved.The training set,the verification set and the test set are constructed using the actual collected combustible cartridge defect samples,and the network parameters are optimized.Finally,on the basis of the existing combustible cartridge defect sample set,the improved model was compared with the original model to compare the effects of the two methods in the detection of white spots and oil defects in combustible cartridges.The experimental results show that the improved model has better recognition effect and applicability in the edge area of white spot defects without reducing the oil stain recognition effect.