In post-disaster mine rescue, extremely low illumination inside the mine severely degrades the quality of images captured by rescue robots, and conventional image enhancement techniques often struggle in such harsh environments. To address this problem, this paper proposes an image-enhancement method based on dynamic light source control. By constructing a robotic dynamic light source control system, the proposed method evaluates image quality scores under candidate light intensities using a trained prediction model and real-time feedback on the distance and incident angle between the light source and the target object, thereby selecting the optimal light intensity and improving image brightness, contrast, clarity, and detail visibility. Experimental results show that, compared with the strategy of capturing images under fixed light intensity and then applying post-processing enhancement, the proposed method improves image quality, better preserves natural appearance, and suppresses noise more effectively. It also improves the operational efficiency of robots in complex rescue environments without relying on computationally intensive post-processing.
Sun et al. (2026) studied this question.