Adaptive intelligent three-dimensional modeling method for hazard-avoidance routing in linear infrastructure

Linear infrastructures, due to their extensive spatial reach, making hazard avoidance a critical and decisive aspect of route planning. However, current hazard avoidance design methods remain largely dependent on existing regional maps and engineering experience, resulting in high subjectivity, limited risk quantification, and insufficient visualization capabilities. These limitations significantly impede the efficiency and reliability of hazard avoidance design for linear infrastructure. To address these issues, this paper proposes an adaptive intelligent three-dimensional (3D) modeling method for hazard-avoidance routing in linear infrastructure. The final output is a 3D route information model that integrates route alignment, environmental context, and hazard distribution. A case study on the Qinghai-Tibet Plateau validates this method, demonstrating a significant improvement in hazard susceptibility prediction accuracy compared to conventional machine learning and transformer-based methods. The optimized route avoids two high-risk zones and closely matches the actual highway alignment, aided by adaptive cost weighting. The deviation between the modeled terrain and the digital elevation model is limited to just 0.007 meters, enabling intuitive and precise comparison among alternative route options. This method effectively supports hazard-avoidance design in linear infrastructure projects and provides a methodological reference for similar applications in other complex environment conditions.