In view of the complex environment of railway signal room, poor image quality caused by dark light and low efficiency of manual inspection, this paper designs a set of key technical schemes for unmanned inspection of railway signal room in low illumination environment. This scheme combines the characteristics of 5G technology, and mainly does two parts of core work: one is a multi-resolution image enhancement model, and the other is a low-illumination target detection optimization algorithm. First, the image enhancement model is described. It uses the Laplacian pyramid decomposition method to layer the image, and then matches the adaptive enhancement strategy. Under the light conditions of 0.1 to 5Lux, whether it is a low-resolution image of 320 × 240 or a high-resolution image of 1920 × 1080, their SSIM can reach 0.89 and 0.92 respectively, and the image is much clearer. Then look at the target detection optimization algorithm, we add two things: centralized operation function and balanced focus loss. With this change, the effect is obvious - its mAP is 15.3 percentage points higher than the original basic FCOS algorithm, directly to 93.8 %. Later, experiments were carried out to verify the results. The results show that this technology can accurately identify the equipment, environmental abnormalities, and personnel in the computer room in the case of dark light. When it is actually used, the efficiency of inspection is particularly significant, which provides technical support for the intelligent operation and maintenance of railway computer rooms, and also helps to break through the bottlenecks of traditional manual inspection.
Wu Yuan (Thu,) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: