In low-light or dark environments, surveillance technology is crucial for advancing the safety and security of humans in computer vision. Low illumination images or videos pose challenges for human image recognition and personnel detection accuracy for detecting human abnormal activity. To solve this challenge, we introduce a low-light image enhancement technique for human safety and security. We utilize the local image enhancement module maps for low light to normal light of the images at the pixel level while conserving the spatial specifics. As well a transformer-based global adjustment module is used to refine the improved images, preventing over-brightening, under-illumination, and color distortions. Additionally, a feature similarity loss constrains target features to minimize adverse effects on detection. The proposed model achieves superior performance in other state-of-the-art methods in low-light environments while achieving competitive performance in normal conditions. This approach significantly enhances visualization and detection performance, contributing to improved safety monitoring.