The presence of foreign objects in conveyor systems significantly contributes to conveyor belt damage and hinders efficient production in enterprises such as mines and ports. To address the reliability challenges posed by low illumination conditions on foreign object detection algorithms, this paper introduces an image adaptive enhancement algorithm based on white balance correction. By developing an end-to-end neural network that simulates the camera’s white balance and gamma correction processes, we achieve an inverse solution to the camera imaging process. Additionally, network training facilitates the updating of weights for the simulated imaging parameters, thereby establishing a correlation between low-illumination images and their enhanced counterparts. Objective image quality evaluation metrics, including PSNR (peak signal-to-noise ratio) and SSIM (structure similarity index measure) demonstrate that the proposed image enhancement algorithm significantly improves image quality and enriches target detail features, with greater improvements observed at lower image brightness levels. When integrated into the YOLOv5 foreign object detection algorithm, experimental results reveal that the enhanced detection algorithm performs better under low illumination conditions, achieving higher detection accuracy on the CUMT-Belt public dataset and substantially reducing the chances of missed detection. This method offers valuable technical support for the intelligent development of belt conveyors and target detection in other low illumination scenarios.
Zuo et al. (Mon,) studied this question.
Synapse has enriched 5 closely related papers on similar clinical questions. Consider them for comparative context: