Research On Enhancement And Recognition Of Wood Veneer Defects Image Based On Digital Image Processing Technology

China’s demand for wood resources is huge,nearly 60% rely on imports,low comprehensive utilization rate makes wood resources waste serious,how to improve the utilization rate of wood,become an important issue that needs to be solved in China.Veneer as a kind of wood products,from the logs rotary cutting,as an important material of plywood,the demand is huge,the uneven quality of veneer affects its utilization rate and the quality of plywood,and knots,cracks as a common defect in the veneer,the veneer quality impact Therefore,it is especially important to identify the defects of veneer knots and cracks.At present,the production enterprises mainly rely on workers’ visual inspection of veneer quality,which is labor-intensive and easy to fatigue the eyes,resulting in low accuracy of manual inspection.Therefore,a method to replace manual identification of veneer defects by digital image processing technology is proposed.In this thesis,the image enhancement,feature extraction and construction of recognition model in digital image processing technology are investigated,and experiments are conducted to realize the image enhancement and recognition of veneer defects based on digital image processing technology by taking knots and cracks in veneer as the research objects.The main research contents of this thesis are as follows.(1)The principles of Gamma Correction,Adaptive Gamma Correction(AGC),Artificial Multiple-Exposure Image Fusion(AMEF)algorithms and image enhancement methods are introduced.AGC is used to However,due to the discontinuity of the enhancement function of the AGC algorithm,there is a pre-classification problem and the enhancement of images near the classification boundary is poor.A non-linear weight adjustment-based AGC algorithm is proposed to modify the contrast and luminance enhancement functions to continuous functions,and the improved AGC algorithm can effectively enhance the defective images located near the boundary with better performance than the AGC algorithm;combining the AMEF algorithm with the improved AGC algorithm,taking the advantages of each,the AMEF algorithm is first used to enhance the detail information of the defective images,and then the improved AGC algorithm is used to enhance the contrast and luminance of the defective images,and the proposed method is subjectively analyzed with the AMEF algorithm,the improved AGC algorithm,HE and GC algorithms.The results show that the visual quality of the defect image is improved,and the defect part and detail part are clearly visible.(2)The enhancement effect of veneer defect images is evaluated by scientific mathematical models,i.e.,Peak Signal to Noise Ratio(PSNR),Structure Similarity(SSIM),Entropy(Entropy,E),Root Mean Square Error(RMSE),Average Gradient(AG),and Enhancement Measure Evaluation(EME)are used to objectively evaluate the enhanced images and analyze them.The results show that the improved AGC algorithm scores higher than the AGC algorithm in these six indexes,and the improved method has better performance;the enhancement method proposed in this thesis achieves the highest scores in four indexes,PSNR,SSIM,AG and EME,and has better performance than other methods,and maximizes the quality of defective images.(3)The principles of two feature extraction algorithms,Histogram of Oriented Gradient(HOG)and Local Binary Pattern(LBP),and the feature extraction steps are introduced to extract the HOG and LBP features of veneer defect images,and the feature vectors are combined and used as the classifier’s input.(4)To identify and classify the wood veneer defect images,the principle of support vector machine and the role of several different kernel functions are firstly described.The SVM is used as the classifier,and the effects of three factors on the recognition accuracy are considered: the radial kernel function,the Sigmoid kernel function and the selection of its parameters,the pixel block(Cell Size)size of HOG and LBP,and the image quality.Experiments are conducted and analyzed for each factor,and it is concluded that when the RBF kernel is selected,the parameters are set to C=32,gamma=0.0625,and the Cell Size of both HOG and LBP is 16×16,and the optimal identification model of veneer nodes and cracks can be obtained by improving the image quality with the enhancement method proposed in this thesis.The average recognition rate of the three types of defects can reach93.46% with an average time of 5.38 s.