Research On Defect Detection Method For The Surface Of Furniture Board

In recent years,the consumption of furniture plate has been increasing year by year.In order to meet the needs of domestic and foreign markets,the performance and scale of automation equipment have been significantly improved by manufacturers.The traditional detection method can not meet the current large-scale surface quality inspection of furniture panel.Machine vision,as a non-contact detection technology,has many advantages such as fast detection speed,stable performance,wide applicability and high precision,and is widely used in various fields.This paper takes furniture board as the research object,and combines machine vision technology,studies and designs a set of defect detection methods and recognition system suitable for the surface of furniture board,which can effectively identify the scratches,flaws,contusion and pressure wounds on the surface of the furniture board.The main contents are as follows:(1)Aiming at the problem that the random shallow texture may overlap or cross with the defect in the furniture board image,and the contrast between the defect area and the background area is not obvious,which is easy to interfere with the subsequent image segmentation,the gray piecewise linear transformation is used to improve the contrast of the image,and an improved local threshold segmentation method is proposed,which can quickly determine the appropriate size of the neighborhood block,At the same time,compared with other dynamic threshold segmentation methods,it has lower miss detection rate.(2)Aiming at the problem of feature extraction of furniture board defects,firstly,based on mathematical morphology,the structural elements suitable for binary image of furniture board and the defects of connection fracture are studied;Then,according to the contour extraction algorithm,the connected regions are separated and the polygon boundary information is solved,and the shape feature and gray feature of the defect are extracted,and the initial defect feature data set is established;On this basis,according to the variance filtering method and recursive feature elimination method,three kinds of defect feature data sets are constructed.(3)Aiming at the problem of defect classification of furniture board,firstly,the problem of unbalanced number of four kinds of defects is solved.Based on oversampling method,smote algorithm is used to amplify samples with small number of defects in feature space;Then,the support vector machine with different kernel functions and the random forest in ensemble learning algorithm are trained optimally,and the optimal classifier model and its feature data set are determined according to the results of the test set.(4)Based on Open CV and QT,a defect recognition software for furniture board image is developed.According to the image processing algorithm,the determined features and the optimal classifier model,a human-computer interaction interface for furniture board image is designed to verify the feasibility of the algorithm.