Sparse Method Application Research In Wood Image Recognition

In the field of image processing,the first thing is to effective representation of images.With the development of technology,nowadays the image acquisition equipment performance is becoming better,the images information is more and more complex.Previous research about image representation method can not gradually met the requirement.So the sparse representation of image has been caused wide public concern over recent years.The sparse representation of image not only speed up data transformations but also improve computing speed.In recent years,the sparse coding theory has been intensively studied and fruitful,it has been widely used in pattern recognition,signal sparse reconstruction,image processing.For Sparse Coding theory perform well in the field of image processing such as face recognition,and in the wood defect image recognition is not widely used.So we put forward to apply Sparse Coding theory in wood defect image recognition experiment.L1 norm is the optimal solution of sparse representation and which compared to the traditional principal component analysis used L2 norm is more robust to noise and abnormal data.SURF(Speeded Up Robust Features)and LBP(Local Binary Pattern)are image texture feature descriptors,which are prominent to image texture feature.So after analyzing SURF,LBP and L1-Minimization algorithm,this thesis propose Automatic wood defects recognition based on Fast L1-Minimization Algorithm and SURF Algorithm and another Automatic wood defects recognition based on Fast L1-Minimization Algorithm and LBP Algorithm.Many experiments indicted that wood defects recognition algorithm this thesis proposed can obtained better results.In this thesis,the main work done is as follows:1.Propose automatic wood defects recognition based on fast L1-Minimization Algorithm and SURF Algorithm;We use wood image texture feature extracting SURF operator first,using Fast L1-Minimization Algorithm to get the best match texture feature vector,get the location of the test set descriptor in the training set to determine whether the defect description.The experiment result indicted that Automatic wood defects recognition based on Fast L1-Minimization Algorithm and SURF Algorithm can getting 0.912 correct to reach location of defects recognition abilities to wood surface defects.2.Propose automatic wood defects recognition based on fast L1-Minimization Algorithm and LBP Algorithm;we use the LBP algorithm to extract three layers cross-sectional features of different wood RGB images data and then use a fast L1-Minimization Algorithm to implement fast and accurate identification to judge whether the wood surface had defects or not,and where were defects located.The experiment result indicted that Automatic wood defects recognition based on fast L1-Minimization Algorithm and LBP Algorithm can getting 0.931 correct to reach location of defects recognition abilities to wood surface defects.