Wood Defects Image Processing Based On Fractal And Mathematical Morphology

On the condition of nondestructive appearance and structure of log, testing internal defects of log correctly is significant for selecting log, and it is an effective method to utilize forest resource. X-ray imaging technique was selected as the detecting method, log x-ray image collection system was designed. Study on the log x-ray imaging rule, log images were obtained according to the intensity differences of the ray through logs. In order to reduce the influenced by signal transmission, digital image processing technology was selected to pre-processing the log x-ray images in this paper. The methods included histogram equalization, smooth filtering and so on. So that the images after pre-processing were more suitable for human's eyes, at the same time enhanced and beneficial to later processing (fractal method and mathematical morphology method).Multi-scale fractal feature (DMl) was selected to detect the log image with rotten knot after pre-processing in this paper. Segmented log x-ray images into 15×15 sub areas, to make sure that normal regions and defect regions of the log image were possessed of different sub areas. Selectedε1= 3 andε2= 5 as measuring scale, calculate the (DMF) values of each sub areas one by one. The calculate result show that the multi-scale fractal feature values of the normal regions and the defected regions were different. Usually the (DMF) values of the normal regions are smaller than the defected regions. In this experiment, the(DMF) values of the normal sub areas are between 0.020 and 0.070, the (DMF) values of the rotten knot sub areas are between 0.300 and 0.600. So when we analyzed x-ray log image, the multi-scale fractal feature could be as reference to distinguish the defections from the normal regions.Multi-scale mathematical morphology dilation method was selected to detect the log image with crack and knots in this paper. A group of diamond (SE) with scale B1=1, B2=2,B3=3,B4=4, B5=5 were selected to dilate the original image. Choose the weighted valueα1=α2=α3=α4=α5=0.2, and the resulting image was obtained with the weight sum. The experimental result shows that, this method is able to overcome the noise successfully and preserve the detail information in the log image. The global thresholding method was selected to compute a suitable threshold of the image, and a binary image was produced from the smoothed image, and then the crack and knots in the log image are apparent, furthermore, the center coordinates and percent areas of the defects can be calculated.This experiment collected a large number of log sample images with rotten knot, crack and knots. It can be known from the experimental results that these two methods can improve the effective rate of log detection. This study provides an effective method for log x-ray nondestructive detection and at the same time provides a new way for the development of digital image processing.