Research On Tree Trunk Defect Detection Technology Using Ground Penetrating Radar

Affected by natural disasters,pest invasion,and other factors,the interior of the tree trunk is prone to decay and cavities.In recent years,ground-penetrating radar(GPR)detection technology has flourished in tree trunk nondestructive testing(NDT),attracting wide attention by its antiinterference solid ability and rapid detection speed.The traditional GPR imaging method hypothesizes the tree trunk as a circular structure and intercepts the radius span time window.It reconstructs the trunk cross-section with polar coordinates,which will lead to the deviation of defect position and limited detection accuracy.This paper mainly studied two imaging methods for locating internal defects in tree trunks by combining signal processing,migration algorithm,deep learning,and a multivariate collaborative detection scheme for living standing trees.The main contributions are as follows:(1)There are two main problems in traditional GPR tree trunk tomography: one assumes that the trunk is a cylindrical structure,the other detects the internal defects by mapping from inside to outside in polar coordinates,by default,it considers the defect area is close to the trunk cross-section center.All of them are inconsistent with the actual situation and do not apply to detecting the complexshaped tree trunks.To this end,a back-projection algorithm based on contour modification is proposed to obtain a tomographic image of the trunk section.This scheme is divided into three parts.First,it used the contour gauge to correct the outline shape of the trunk and used the arc length parameterization method to obtain the coordinates of the measurement points.Then,the collected GPR data was preprocessed to reduce the noise and correct the zero-point time of the radar wave.Finally,a tomographic image of the tree trunk was obtained by back-projection from the surface to the interior according to the location of scan points.Validation of the proposed method on 6 simulated and 3 laboratory samples shows that the suggested detection algorithm improves the detection accuracy more than 20% over conventional methods on laboratory samples,which is more consistent with the actual defect situation.(2)In response to the back-projection algorithm making it difficult to detect multiple defects inside the trunk and the accumulation of contour gauge errors,this paper proposed using lidar to extract the measurement section profile and then using the imaging algorithm to obtain the tomographic image of the trunk section.The method is mainly divided into four parts.First,the surface 3-dimensional structure of the tree trunk was constructed by laser radar,extracting the contour coordinates of the measurement section,which used the arc length parameterization method to locate the measure points.Then,the GPR data would be preprocessed,the singular value decomposition(SVD)algorithm was used to filter ringing noises and clutter.The time window of the B-scan data is also intercepted to reduce the noise interference further.After,the preprocessed radar signal would be reversed on the timeline and imaged using the reverse time migration(RTM)algorithm at all measurement points.Finally,establish a sample library about trunk cross-section RTM results to distinguish the defective areas and healthy xylem,and use Unet deep learning network to detect the defective regions within the trunk.Select 3 simulation samples and 3 laboratory logs for the experiment.The experimental data shows that the proposed method reached more than 80% detection accuracies on simulated samples,and the reconstruction accuracy exceeded 88% on the Lithocarpus glaber laboratory sample,better than the back-projection algorithm,indicating that the method could effectively locate early defects and correctly detect defects in smaller diameter cavities inside the tree trunk.(3)To detect the internal defects of live trees accurately,a multi-technology cooperation detection scheme combining visual inspection,GPR,stress wave,and micro-drilling resistance instrument is proposed.Verification of the presented plan using 19 landscape trees from Jiangnan University shows that the GPR detection is generally advantageous for internal defect localization in irregular complex-shaped tree trunks but inadequate for small-diameter samples.The stress wave detection in complex-shaped tree trunks possibly has position deviations and error detection but has apparent benefits for detecting small-diameter and circular samples.The micro drill resistive instrument can accurately judge the span and defect type on the drilling path.The test results showed that the multi-technology collaboration scheme greatly improved the detection accuracy of living wood.