Research Of Wood Defects Identification By FBP Neural Network And Statistical Detection Of Log Rings

As a renewable resource with relatively long growth cycle,forest resources have been a scarce asset for human beings.We are grateful for the gift of nature,but also pay attention to ecological balance.In recent years,the global forest area has been reduced,and the greenhouse effect has been aggravated.It is particularly important to choose wood scientifically and improve the utilization rate of wood.This topic mainly of larch,Manchurian ash,two kinds of lumber of three kinds of typical defects(knots,decay,crack)for identification,log rings to spruce were detected,then age estimates.The research on the characteristics of wood defects effectively utilizes X ray nondestructive testing technology.Based on the defect area and other area of X-ray difference absorption attenuation,obtain defect images.The 13 feature values of the defect image were extracted by using the gray level co-occurrence matrix,and the input vectors of the neural network are given.Through the comparative analysis,this paper uses the quasi Newton algorithm as the network training algorithm,the concept of membership in fuzzy theory and BP neural network to identify the type of wood defects.The number of rings is the tree age,tree ring RGB image characteristics obtained by digital camera,which is based on the computer digital image processing technology of ring images using histogram equalization method for image enhancement.Canny edge detection,LOG edge detection and DOG edge detection are used to extract the tree rings.Finally,the Gauss difference operator(DOG)is selected to segment the image.Using the area threshold method and binarization reversal addition method respectively to remove the knots and cracks in the ring,get the ring line clear image,and then through two value matrix scanning method the statistics of the number of rings to get old.Research results show that the gray level co-occurrence matrix is utilized to extract the characteristic value can more fully reflect the features of trees of defects,at the same time combined with fuzzy BP neural network to identify the defect types can get the expected effect,recognition rate above 90%.Using the method of Gauss difference operator edge detection and area threshold method,we can get clear tree ring contour image,and calculate the tree age.