The Tianshan Expressway plays a crucial role in China’s “Belt and Road” strategy, yet the extreme climate of the Tianshan Mountains poses significant traffic safety risks, hindering local economic development. Efficient detection of hazardous road surface conditions (RSCs) is vital to address these challenges. The complexity and variability of RSCs in the region, exacerbated by harsh weather, make traditional surveillance methods inadequate for real-time monitoring. To overcome these limitations, a vision-based artificial intelligence approach is urgently needed to ensure effective, real-time detection of dangerous RSCs in the Tianshan road network. This paper analyzes the primary structures and architectures of mainstream neural networks and explores their performance for RSC recognition through a comprehensive set of experiments, filling a research gap. Additionally, T-Net, specifically designed for the Tianshan Expressway engineering project, is built upon the optimal architecture identified in this study. Leveraging the split-transform-merge structure paradigm and asymmetric convolution, the model excels in capturing detailed information by learning features across multiple dimensions and perspectives. Furthermore, the integration of channel, spatial, and multi-head attention modules enhances the weighting of key features, making the T-Net particularly effective in recognizing the characteristics of snow-covered and icy road surfaces. All models presented in this paper were trained on a custom RSC dataset, compiled from various sources. Experimental results indicate that the T-Net outperforms fourteen once state-of-the-art (SOTA) models and three models specifically designed for RSC recognition, with 97.44% accuracy and 9.79% loss on the validation set.

In summary, near-surface icing risk in complex alpine terrain is jointly controlled by freezing conditions, moisture supply, freeze–thaw transitions, and topographic energy processes. Traditional approaches relying on sparse station data or single temperature thresholds fail to capture spatial heterogeneity, and frequent cloud cover together with topographic errors severely limit the application of thermal infrared remote sensing. Taking the area along the Duku Highway in the Tianshan Mountains as the study region, a daily icing risk assessment framework at 250 m resolution was constructed using multi-source remote sensing, ERA5-Land reanalysis data, topographic correction, and an energy–moisture dual-constrained model. A diurnal temperature cycle model, the CAP index, and physics-constrained machine learning were integrated to reconstruct the daily minimum land surface temperature ($T_{s,\mathrm{m}\mathrm{i}\mathrm{n}}$) at 250 m resolution under all weather conditions. A probabilistic two-tier risk assessment model was then established by incorporating moisture, topography, and freeze–thaw transitions. The results show that high-risk zones occur primarily in valleys and topographically constrained corridors rather than the coldest elevations. Validation against Landsat LST (r = 0.886) and the Bayanbulak station (bias −0.76 °C, RMSE 5.62 °C, r = 0.91) confirms spatial and seasonal accuracy. Sensitivity and Monte Carlo analyses indicate the RiskScore is mainly controlled by the low-temperature weight, while upstream parameters are less influential. The framework is best applied as a screening and early-warning product to identify sub-kilometer potential icing corridors, complementing point measurements and short-range forecasts.