Drive recorders are widely used by police department for accident analysis. However, current methods rely on human observation. This results in high temporal and physical costs when estimating vehicle speed and inter-vehicle distance from images. In addition, the performance of drive recorders varies. Although many modern drive recorders are equipped with GNSS receivers, GNSS measurements are not available for all devices and are typically limited in temporal resolution. Therefore, it is difficult to obtain accurate, high-frequency vehicle speed directly. In this paper, we propose an image processing method that automatically estimates vehicle speed and inter-vehicle distance using only drive recorder images. Our method tracks the motion of the road surface and the preceding vehicle using image processing techniques. We conducted experiments using images captured during real-world driving with two vehicles. The results demonstrate that our method can automatically estimate vehicle speed and inter-vehicle distance. This suggests that the method has the potential to enable more efficient traffic accident analysis using only drive recorder images.

This study proposes a metric monocular depth estimation method for indoor RGB images and evaluates its application to 3D point cloud reconstruction. The proposed Swin Transformer-U-Net integrates local feature extraction with global contextual reasoning. Experiments using the NYU Depth v2 dataset show that the proposed method improves the RMSE from 0.702 m for Attention U-Net to 0.352 m. The reconstructed point clouds also show improved planar continuity and spatial consistency. These results indicate that the proposed method is effective for approximate indoor 3D reconstruction from a single RGB image, although it does not replace high-accuracy ranging sensors such as LiDAR or RGB-D cameras.