Accessible Thermal Night Vision Solutions to Protect Vulnerable Communities in Low-Visibility Conditions

This paper introduces a real-time human detection system with YOLOv8-seg coupled with thermal visualization for improving detection accuracy in nighttime environments. The approach utilizes an average image brightness-based night-time detection algorithm that initiates contrast-limited adaptive histogram equalization (CLAHE) to amplify dark frames dynamically. YOLOv8-seg identifies people using a confidence level, and regions found are postprocessed with thermal impact in the form of the INFERNO colormap, bounding boxes slightly minimized for display accuracy. Live webcam feed is processed by the system and superimposed with labeled regions of detected humans that are rendered as thermal color. It blends computer vision approaches, deep learning, and image enhancement methods for enhancing human identification in surveillance and safety contexts, especially in the context of night-time or low-visibility scenes. learning, and image enhancement methods for enhancing human identification in surveillance and safety contexts, especially in the context of night-time or low-visibility scenes.