Abstract: Wood, a material widely used in construction, furniture, and transportation, faces significant challenges due to defects that threaten the health of trees, reduce the utility value of timber, and can cause irreparable damage to ancient and venerable trees. Wood defects significantly affect the structural integrity of trees, making them vulnerable to environmental stressors and potentially leading to their decline. Therefore, non-destructive testing (NDT) technologies are critical in ensuring the efficient utilization of timber and providing essential protection for heritage trees. Among various NDT methods, stress wave tomography has gained considerable attention for its ability to detect wood defects by analyzing the propagation speed of stress waves through the wood and presenting the detection results in the form of images. This technique offers several key advantages, such as safety, portability, interference resistance, and remote transmission capabilities, making it highly suitable for complex environments where traditional testing methods may be impractical. However, traditional stress wave tomography has inherent limitations. It often fails to account for the effects of wood anisotropy on the propagation of stress waves, which can lead to inaccurate representations of the internal structure of tree trunks. Specifically, the varying densities and elastic properties of wood grains in different directions result in uneven stress wave propagation, causing discrepancies in detecting internal defects and mapping actual cavities. To address these challenges, this study proposes an intersection fitting-based defect detection (IFDD) method for tree health assessment. The quality of stress wave tomography images is largely dependent on the accuracy of wave speed calculations, which are influenced by the anisotropic nature of wood. To improve this accuracy, the proposed method involves the placement of multiple sensors uniformly distributed across the tree’s cross-section to collect stress wave data. The stress wave velocities collected from these sensors are corrected and normalized based on a deviation rate model, which compensates for the uneven propagation speeds caused by wood anisotropy. This correction ensures that the stress wave data more accurately reflects the internal structure of the tree, thereby enhancing the precision of the imaging process.In this method, the imaging area is subdivided into multiple grid cells, and the speed of each stress wave ray is refitted based on the intersection speeds at which the rays cross the grid cells. A corresponding stress wave propagation ray diagram is then drawn, representing the corrected wave speeds. When constructing the ray diagram, appropriate threshold values are selected to categorize the wood’s health condition, and different velocity intervals are color-coded to visually differentiate between healthy, decayed, and hollow regions within the tree. The estimated speed for each grid cell is calculated using a weighted average method, with the weights determined by the proportion of rays passing through each cell. In cases where certain grid cells lack sufficient reference speed values due to limited ray coverage, a nearest-neighbor interpolation algorithm is employed to fill in these gaps, ensuring the completeness and accuracy of the data. Abnormal grid cells are subjected to image pooling to mitigate the influence of outlier data. Finally, the defect status of the tree is determined by analyzing the speed data for each grid cell, and the internal structure, including healthy, decayed, and hollow areas, is visualized using image processing techniques.This study tested five log samples to evaluate the effectiveness of the proposed algorithm. The evaluation criteria included the proportion of the defect area after image segmentation and the degree of overlap with the actual defect shape. The results demonstrated that the proposed imaging algorithm achieved an overall average relative error of 8.25%, an accuracy rate of 93.19%, a precision rate of 80.37%, and a recall rate of 82.30%. The locations and sizes of the detected defect areas closely matched the actual conditions within the logs, demonstrating the improved accuracy of the proposed method. By refining the wave speed model and improving data processing techniques, the proposed algorithm enhances the robustness of defect detection against crack interference and improves the identification of defect regions, particularly in complex structures.Despite these advances, the intersection fitting-based tomography algorithm still exhibits certain limitations, particularly in detecting micro-cracks and decay. These limitations highlight the need for further research to improve and optimize the algorithm. Future research efforts will focus on enhancing the robustness and imaging accuracy of the intersection fitting-based tomography algorithm, thereby laying a solid foundation for the development of three-dimensional imaging technologies that can further advance the field of tree health assessment and conservation.