Wood Detection By CT And Its Three-Dimensional Structure Reconstruction Based On Multifractal Spectrum

With the development of non-destructive testing technology and digital image processing technology, wood non-destructive testing technology has also achieved a rapid development. In this paper, a variety of commonly used wood nondestructive testing methods were given comparative and comprehensive evaluation, computed tomography (CT) was selected, which is a branch of radiation detection, as a means of wood nondestructive testing. Compared with old nondestructive testing methods, the CT imaging has the characteristics of visual image accuracy, high resolution and no overlap. CT is honored as the best means of wood non-destructive testing.The research described basic principles of CT testing in the wood, the achievements of CT in the wood nondestructive testing, and research progress of image processing technology in wood nondestructive testing. The paper studied the CT image reconstruction method and the basic principles, concluded the advantages and the limitations of CT imaging, found the best CT parameters in the fast wood testing. Based on the definition of CT number, the predicted models of green density and moisture content were built. Based on digital image processing and fractal, multifractal spectrum was applied to wood inner defects recognition, the range of multifractal spectrum was set for wood image detection. At the same time, linear interpolation was used in wood three-dimensional reconstruction. Combined with pseudo-color processing technology, wood surface and internal structure were reconstructed, and the method was applied to the wood virtual sawing for the further improvements.Research results were summarized as follows:(1) Current wood non-destructive testing technology at home and abroad were analyzed, summarized and compared, computed tomography was selected as the means of wood physical testing. Through the study of wood nondestructive testing status and development trend at home and abroad, CT has a good application prospect in wood defects detection and physical properties testing.(2) Compared with different CT reconstruction algorithms, analytic method was selected as the basic method of CT image reconstruction. Analytic method has the advantages of the fast computing speed and high precision reconstruction.(3) In order to achieve fast, accurate wood CT detection, improve the wood CT scanning efficiency, save time and cost, and build CT-based wood nondestructive testing system, the wood CT scanning parameters were fixed by the scanning images and calculation of 27 log species:the scanning thickness is 14mm, scanning voltage is 80KV-120KV, scanning current is 100mA, window level is-300Hu, window width is 1000Hu. Suitable window level and slice thickness can reduce the influence of image artifacts and partial volume effect in the wood defects image. At the same time they can improve the efficiency of wood CT scanning, save time and cost.(4) By the definition of CT number, the closed relationship model was derived between wood CT number and its mass attenuation coefficient. And based on the relationship, the predicted mathematic model of CT number and green wood density was built, which expand the scope of wood nondestructive testing from defects detection to wood physical properties detection. By statistic analysis the predicted mathematic model D=0.0009H+0.9564 (R2= 0.968) of CT number-wood density was built by the CT number derived from 28 green wood species in Heilongjiang Fangzhang forestry bureau. The average estimated difference was 0.017 g/cm3 and the average error rate was 2.64% by testing model.(5) On the basis of CT number-green wood density model and the relationship of wood density and moisture content, the predicted mathematic model of CT number and green wood moisture content was built. The predicted mathematic model M=-0.0864H+78.001 (R2=0.9203) of CT number-wood density was built by the CT number derived from 28 wood species in Heilongjiang Fangzhang forestry bureau. The average estimated difference was 3.02% and the average error rate was 2.90% by testing model.(6) Multifractal spectrum theory was selected as the edge detection method in the wood scanning image. Through the experiment of large number of Fraxinus mandshurica CT images, the complete and effective multifractal spectrum scope for wood defects detection was found,1.0≤f(a)≤1.3. And the larger the scope of the spectrum is, the more accurate the detected edge is. Compared with classic edge gradient operators, the multifractal spectrum detection has the advantage of locality and high accuracy.(7) Taking into account the characteristics of the wood image and detection accuracy, linear interpolation method was used to wood CT three-dimensional reconstruction combined with multifractal spectrum in wood feature extraction. On the basis of wood three-dimensional image, coordinate transformation gray-level interpolation technique was used in wood virtual cutting, and pseudo-color processing technique was used to wood three-dimensional image. Through the experiments of birch and poplar wood three-dimensional reconstruction samples, the algorithms in the paper has the advantages of better three-dimensional reconstruction effect, fast computing speed, clear and direct image, which provides an effective method for virtue cutting and reasonable utility of wood.