Application of Nighttime Low-Light Image Enhancement Technology for Safety Monitoring and Prevention at Wellsite Operations

Summary Wellsite operations have complex environments, often affected by insufficient lighting, varied light sources, and uneven light intensity. This results in monitoring images with low brightness, poor contrast, and blurred details, severely hindering the intelligent development of oilfield safety management. Traditional image enhancement methods, such as histogram equalization and gamma correction, typically rely on global brightness adjustment when handling complex scenes in nighttime low-light wellsite environments. These methods struggle to restore image details selectively, often leading to overexposure or underexposure in certain areas, and may also introduce noise or artifacts. Therefore, this paper proposes a nighttime low-light image enhancement method based on adaptive enhancement-zero-reference deep curve estimation (AE-ZERO-DCE), which improves the visualization and detail expression of images, enhancing the visual effects of monitoring images in complex lighting conditions and providing strong technical support for the intelligent safety management of oilfield operation sites. AE-ZERO-DCE primarily consists of two components: the improved deep curve estimation network (IDCE-Net) and the gated context aggregation layer (GCALayer). IDCE-Net automatically learns and generates a light-enhancement curve (LE-Curve) suitable for the current input image. The LE-Curve adjusts the brightness of each pixel in the picture, optimizing its brightness and contrast. GCALayer, through smooth dilated convolution and gated fusion subnetworks, extracts multiscale contextual information, removes haze and blur from the image, and restores details. Finally, combined with the Gamma transformation, it optimizes the image’s brightness and contrast, ensuring that the enhanced image maintains good visual effects under complex lighting conditions. Experimental results show that AE-ZERO-DCE significantly improves brightness, contrast, and detail preservation in nighttime low-light environments. Compared to existing methods such as Zero-DCE, Zero-DCE++, MIRNet, and EnlightGAN, AE-ZERO-DCE demonstrates notable performance gains—peak signal-to-noise ratio (PSNR) is improved by 44.8%, structural similarity index (SSIM) by 12.5%, and processing speed (frames per second; FPS) by 21.0%. This allows for the effective enhancement of image details and provides faster real-time processing capabilities. It can work in conjunction with object detection technology to more effectively detect safety violations, identify equipment anomalies, and prevent accidents, thereby improving operational safety and reducing risks. These advancements lay a solid foundation for accurate monitoring and safety management in wellsite operation environments, offering efficient and reliable technical support for the intelligent development of oil field safety management.