Research On Wood Defect Detection Method Using Ground Penetrating Radar

Forest is the precious resource for human beings and also a purifier for natural environment.In addition,wood products are widely used in the human society.To gurrante the quality of wood products,it is necessary to protect and manage the forest,and detect the defects of wood products.Efficent and nondestructive technology for wood defect detection has been the hot topic of academic research,and it attracts more and more attention from both academic and industrial fields.Traditionally,the detection of ancient trees and ancient wood structures depends on the professional experiences of engineers,and the accuracy and reliability of detection are limited.The high-resolutioned nondestructive evaluation methods are the future trends for the forest and wood detection.In this paper,we focus on the ground penetrating radar(GPR)application techonology for the protection of ancient wood structures and ancient trees.Combined with the image processing,signal processing and machine learning,we studied the GPR reflection principle in the wood samples with different water content.The positioning and identification method for internal targets of wooden components of ancient bridge,as well as a multi-technology cooperative nondestructive evaluation scheme for live trees were proposed.The main contributions are described as follows:(1)To precisely identify the region of interest of the ancient bridge timber with GPR,a method using traditional empirical mode decomposition(EMD)and dynamic time warping(DTW)was used.EMD was used to reduce the noise of the channel data,and the first intrinsic mode function(IMF)component of each channel data was chosen as the sample data.The average value of first IMF components of partial channel data in the area without obvious target echo information was used as the reference component.The range and type of the target were determined according to the difference between the GPR signal with and without the target,and the similarity between the first IMF component of each channel data and the reference component was calculated using DTW.This method was verified on two sets of samples,i.e.,the bridge timbers and the artificially-made wood core samples from the USDA Forest Service Forest Products Laboratory.The physical and defective status of the wood core samplse were different from those of the timbers,so the similarity of core samples can help determining the material properties of the target area on the timbers.The results demonstrated the effectiveness of the proposed method.(2)For the case of randomly deployed multi-objects in the wood timber,an adaptive migration imaging method for target localization using DTW was proposed.This method determined the size of sliding window according to the wave length of GPR echo and physical properties,and adptively processed the frequency wavenumber domain migration.For the objective points after migration,2-D maximum entropy segmentation method on image was used to correct the error of the target objective points.Combined with the histogram of oriented gradient(HOG)features,the target classification was conducted by the support vector machine(SVM)method.On the synthesized radar data set,and the actual defects of the wood timber samples,the experiments were finished to evaluate the proposed method.The results showed that this method achieved a good performance for identifing the complex targets inside the wood structure,and the classification accuracy was 95.73%.(3)To improve the accuracy of nondestructive evaluation of live trees,a new tomography imaging algorithm based on stress wave was developed.Furthermore,we present a multi-technology coorperative scheme for live trees based on ground penetrating radar,stress wave testing and other methods.The proposed imaging algorithm used the least square QR decomposition inversion method to solve the velocity distribution of stress wave,and applied a specially designed mechanism to correct the error of wave velocity.With the Hexcone model,we quantitively evaluate the generated tomography image based on the morphological image processing.We finished the nondestructive evaluation of 157 ancient trees in a historic city park in Yangzhou,Jiangsu Province,China.We compared the advantages and limitations of three techniques including stress wave,GPR and micro drilling resistance testing.The results showed that the detection accuracy of GPR was influenced by the testing environment and the diameter of trees,and the error may increase if the diameter of trees was small or the shape of cross section was irregular.With the visual checking,stress wave,micro drilling and GPR root system scanning,the proposed multi-technology cooperative evaluation scheme will improve the efficiency and accuracy of nondestructive evaluation of ancient trees.At present,the proposed evaluation scheme has been applied for protection of live trees in many cities.