The Research On Wood Surface Defect Detection Based On Computer Vision

Wood is an indispensable resources which holds a significant position in national economy of our country.But wood surface defects affect the quality and use of it.So the detection of wood surface defects is very important to improve the utilization rate. At present, wood surface defects is detected by manual method in our country.But it is difficult to execute the method quickly and accurately.On the other hand,there always been some problems in systems which imported from abroad like:high price,improper configuration and operation of the device for our actual manufacturing situation.So designing a wood surface defect detection system which fit the situation of China has practical value and social significance.Based on the study of development of automatic wood surface defects detection technology,the paper focus on researching theories and methods of automatic wood surface defects according to the categories and characteristic of wood surface defects. These methods include segmentation of wood surface defects image,feature extraction and classification and recognition.The main work are listed as follow:1.After analyzing the traditional segmentation method of wood surface defects images,a method based on wavelet transform and mathematical morphology is presented.Wavelet transform is used to suppress the interference information.Then bottom-hat transform of morphology and threshold are applied to get seeds in the wavelet reconstructed images.Eventually the wood defects can be segmented through region-growth algorithm.This method can effectively and accurately detects the wood surface defects and meet the requirement of wood processing.2.The segmentation images of wood-surface defects are extracted from the wood image.Then features can be sorted out according to the particularity of the wood surface defects.The features are classified into geometric features,moment invariant features,grey texture features,region features.After that,we use the feature selection algorithm to find out the features which contribute most to the classification.These features formed foundation for classification and recognition.3.The principle and method of LS-SVM are studied and applied to automatic detection.Three kinds of wood defect are experimented with 240 samples in this paper. The distinguishability ratio reach 94.67 percent.The experimental results show that the method is feasible.The correctness and feasibility of the methods are all proved by experimental data. The proposed methods are successfully applied in the wood defect recognition. Therefore, the study of this project has strong practicability.