Abstract: Solid wood panels were crafted from natural timber into panels with specific dimensions, prized for their natural grain patterns and robust physical properties. These characteristics made them indispensable across various industries, including furniture manufacturing, interior decoration, and construction. A critical preliminary step in their production, longitudinal sawing, was employed to process these panels, where appropriate cutting strategies significantly enhanced both material utilization and the value of the final products. However, the process of longitudinal sawing in solid wood panel production encountered numerous challenges that hindered efficiency and quality.One major issue stemmed from the inherent imperfections in natural timber. During its growth, timber was frequently subjected to environmental factors and pest infestations, resulting in defects such as knots, cracks, and wormholes. If these flaws were not accurately detected and either avoided or removed during processing, they severely undermined the market value and durability of the solid wood panels. Furthermore, traditional longitudinal sawing methods typically relied on fixed-length cuts without the benefit of optimized layout planning. This lack of strategic planning led to substantial waste, as excess offcuts were produced, reducing the overall efficiency of the process and squandering valuable resources. Extensive research had been conducted on defect detection in timber and solid wood panels, offering efficient methods to identify these imperfections. Despite these advancements, achieving intelligent sawing demanded more than just pinpointing defects; it required the integration of these detection outcomes into an optimized layout planning process tailored specifically for longitudinal sawing. Existing systems, originally designed for optimized cross-sawing, had proven effective in reducing waste and boosting yield in that specific context. However, these systems were poorly suited for longitudinal sawing due to their high hardware costs and design incompatibilities, leaving a gap in addressing the unique demands of this cutting method.To overcome these challenges, a comprehensive system was developed with a focus on scots pine panels, aiming to improve both defect detection efficiency and material utilization. The initiative began with the collection of 1,000 images of the upper and lower surfaces of scots pine panels, captured through photography. After undergoing image processing and annotation, these images were combined with 653 defect images sourced from a public dataset, forming a robust collection of 11,571 images dedicated to defect detection analysis.Building upon this dataset, the RT-DETR (realtime-detection transformer) modelserved as the baseline for further development. Enhancements were introduced through self-developed modules—namely, CSP-MBDC (cross stage partial multibranch dilated convolution module), GADownsample (gated attention dowmsample), and GAUpsample (gated attention upsample)—resulting in the creation of the MBDGA-DETR(multibranch dilated convolution module-detection transformer) model, specifically engineered for detecting surface defects on scots pine panels. With the goal of maximizing panel utilization, the system then generated an optimized layout plan for longitudinal sawing, taking into account defect locations and sawing width constraints. To enhance usability, an interactive interface was crafted using PyQt6, enabling practical application and seamless user interaction.Experimental evaluations demonstrated the superiority of the MBDGA-DETR model over common YOLO series models and the baseline RT-DETR. The model achieved a precision of 96.1%, a recall of 93.0%, and an accuracy of 96.4%, while maintaining a detection speed of 104 frames per second. These metrics highlighted its exceptional precision and efficiency in identifying defects, setting a new standard for performance in this domain.The system was subsequently integrated into a longitudinal sawing device and rigorously tested on 50 scots pine panels under varying complexity conditions. In low-complexity scenarios, the average time to generate an optimized layout plan was 1.2 s, with a panel utilization rate of 90.1%. Under medium-complexity conditions, the generation time increased to 6.2 s, accompanied by a utilization rate of 84.8%. In high-complexity situations, the system required 10.1 s on average, achieving a utilization rate of 75.3%. These results underscored the system’s adaptability and effectiveness across diverse operational contexts.When compared to manual processing, the system showcased significant improvements, particularly under low-complexity conditions. It delivered an average solution generation time of 1.4 s and a utilization rate of 88.8%, substantially reducing processing time while enhancing material efficiency. This marked contrast with traditional methods emphasized the system’s capacity to streamline operations and conserve resources.In conclusion, this innovative system represented a transformative advancement in solid wood panel processing, particularly for scots pine panel production. By seamlessly integrating advanced image processing, the MBDGA-DETR model, and optimized layout planning, it offered a sustainable and efficient solution that addressed longstanding inefficiencies in the woodworking industry. The combination of cutting-edge technology and practical application not only elevated the quality and value of the panels but also contributed to a more resource-conscious approach to timber processing, paving the way for future developments in the field.