Research On Wood Surface Defect Recognition Method Based On Machine Vision

Wood is widely used in furniture,decoration,musical instruments and other fields.Wood surface defects can better reflect the quality of wood.It is urgent to solve the problem of adaptability and effectiveness of hardware and software in machine vision recognition of wood apparent defects.In this paper,we explore and experiment.Aiming at the problems of different sizes and shapes of wood surface defects and complex and diverse textures.Based on Local Binary Patterns(LBP),Multi-scale Block Local Binary patterns(MBLBP)is introduced,and the window size of Histogram of Oriented Gradient(HOG)is improved to enrich the image feature information,The mean and variance features are further fused.In the Support Vector Machine(SVM)classifier,the recognition accuracy can reach 86.33%.There are a lot of redundant features in high-dimensional features.The weight coefficient based on pre classification is introduced in the Locality Preserving Projection(LPP)algorithm to reduce the dimension and pre classify the features;Then,Manhattan distance weighting is introduced into Linear Discriminant Analysis(LDA)to reduce the dimension of pre classification features and reduce noise interference.When the feature dimension is reduced to 200,the recognition accuracy can still be improved from 86.33% to 97.16%.In order to solve the problem of uneven distribution of wood defect samples,the linear kernel function and radial basis function are weighted and mixed in this paper.The mixed kernel function can make up for the shortcomings of single kernel function,and the grid search method is used to optimize the parameters of mixed kernel function.The experimental results show that the recognition accuracy of the optimized hybrid kernel function is improved from97.16% to 98.58% compared with the single kernel function.Finally,the whole algorithm is verified and tested by hardware and software platform.The experimental results show that the algorithm can effectively realize the recognition of wood surface defects,and provide a new solution for the recognition of wood surface defects.