On Line Sorting Of Wood Based On Artificial Neural Network Classifier

Wood plate with its unique characteristics, has been closely related to human life. Especially in recent years, with the improvement of living standards, people are more advocating the pursuit of beauty and health, is also growing demand for wood materials. In today’s increasingly scarce of wood resources, and how to more effectively improve the utilization of wood become important research topic in the field of wood processing and automation. Before the timber using in the practical application, the wood surface monitoring is an important process in the production process, directly impacting on product quality and production efficiency. But due to outside interference and subjective effects, traditional manual method on the wood surface detection reliability is not high, and it also caused a large number of human, financial waste. With the increasing requirements of the increasing development and production of the level of modern technology, wood processing and the optimal degree of automation have become increasingly demanding. In this paper, computer image processing technology, timber grading standards and artificial neural network is a combination of wood plates online sorting, in order to effectively improve the sorting accuracy and sheet utilization.In this paper, we selected pine and oak wood plate which is widely used as a sample plate for image acquisition, used the Oscar F-810as the image acquisition device, and used36*15LED row of lights as the lighting, were collected520sample images, which including specialized collection with Setsuko defects of the wood plate image108, according to state sheet classification of the relevant provisions of the sample plate is divided into three levels of sorting.In this research, we select a color and defect characteristics of two features in the fusion to get the final separation characteristics. First of all, computing the three components of the image RGB space as five color characteristics, computing the R component of demand color moment as other three color characteristics, a total extraction of the eight parameters as color features; And the initial screening of color characteristics, to remove the redundant featuresI1=(G-B)/(G+B); Then, using the remaining seven color features and BP neural network classifier to classify, the way is get rid of one and two feature solid wood board image features fusion classification study, which remove a color characteristic combination of removing the combination of the two color features21kinds of contrast will be based on the classification accuracy of28kinds of combinations of features solid wood plate, and ultimately determine the color characteristics of the five kinds of sorting color characteristics of solid wood boards. Meanwhile, the use of the area of image segmentation based on genetic algorithms and morphological image segmentation to extract the knots of the wood sheet; finally we can get the six-dimensional integration characteristics of five color features and a defect characteristic as sorting defects in the number of features.On the basis of integration of feature extraction of plate image, we use artificial fish neural network, the FNN neural network as a timber sorting classifier to research the sheet classification and the compare the classification results.The results show that the artificial fish neural network classifier of wood plates in online sorting is better accuracy rate, which is95%, the classification time is0.0738s to meet the online system requirements. Application integration feature vector and artificial fish neural network classifier in wood line sorting system will eventually speed the fastest up to12m/s, the slowest of4.7m/s, line separation to good effect.