Abstract:
The decay and cavities inside the fruit trees are the principal factors, leading to shortening their lifespan and even death. Long-term exposure to natural erosion, pest invasion, or artificial damage can easily cause cavities in the trunk, which seriously threatens the health of fruit trees and results in negative impacts on the fruit quality. Therefore, fruit trees should be inspected periodically to detect internal defects, and then take measures to protect the fruit trees. Ground-penetrating radar (GPR) presents entirely nondestructive testing with the characteristics of portability and high efficiency, often suitable for the field of fruit tree evaluation. A circumferential movement can also be made to identify the cavity along the tree trunk surface. The transmission position of the radar signal can be adjusted through the moving distance to realize multiple fixed-point measurements. The receiving signals of all detecting points are utilized to analyze the cavities of the measurement area. However, the irregular trunk profile has already been a great challenge to position the GPR measurement points. In this study, a novel detection was proposed for the tree cavities to construct the trunk contour coordinate system, and then locate testing points. Image affine transform and marching squares were applied to extract the profile of the trunk. The integral was also used to estimate the relative position of measuring points on the surface of the trunk, according to the coordinates of the cross-sectional profile. Signal preprocessing was utilized to extract the effective signals from the cluttered raw radar data, further reducing the influence of noise signals on the imaging for nondestructive detection of GPR. In sequential preprocessing, the specific steps included the time-varying gain on the radar signal to enhance the clarity, the removal of the direct wave to correct the size of the time window, filtering the ringing noise using background removal and singular value decomposition, and labeling the signal position of the cavity using a threshold. The migration imaging was combined with the coordinates of measuring points and the effective radar signal after preprocessing to construct the radar wave diffraction surface during the image reconstruction of the internal cavity in the trunk. All the diffraction surfaces were then superimposed to highlight the cavity part, where the image morphology was characterized to eliminate the interference data. A field experiment was conducted on the circular simulation and various actual irregular tree trunk samples. The cosine similarity was used to evaluate the reconstruction data. The accuracy was estimated via the overlap between the reconstruction and the actual tree trunks. Experimental results show that the contour acquisition presented an error of less than 0.6% for the image of a circular tree trunk, while less than 3% for the irregular trunk. In the current measurement software TreeWin, the reconstruction error of circular tree trunk contour was about 0.6%, while the error of irregular tree trunk was about 9.5%. It indicated that the contour acquisition scheme was more suitable for the actual ancient tree section. In the detection of location and size of the cavity, a construction accuracy of 95.41% was achieved for the simulation, and 87.54% for the actual tree trunks, much higher than 73.64% and 65.02% obtained by the TreeWin software, respectively. Correspondingly, the irregular tree trunk contour posed a great influence on the measurement of GPR. The radar combined with the migration was highly consistent with the actual situation suitable for ancient tree protection.