技術摘要(英)
This AI-based night-time image recognition technology optimizes model architecture and parameters using annotated night-time and long-distance datasets. By refining cross-layer feature integration, the model accurately captures vehicle contours and light sources. Furthermore, targeted fine-tuning addresses challenges like long-distance detection, glare, low light, and small objects, significantly improving recognition stability and accuracy in real-world nighttime environments.