A Method for Generating Real-Time Indoor Illuminance Maps Based on Images in a Sensorless Lighting Environment

In the field of system lighting control, which aims to improve light quality and reduce energy consumption, analyzing indoor illuminance is an essential technology. Researchers have collected and analyzed illuminance data using multiple light sensors in a room. However, operating various sensing devices and related equipment incurs additional maintenance costs and causes user inconvenience, thereby hindering the widespread adoption of such system lighting technologies. Achieving a higher level of natural light-integrated lighting technology requires methods for analyzing indoor illuminance with minimal intervention from physical sensors, yet research in this area has been severely lacking. Therefore, this study proposes a method for generating image-based indoor illuminance maps in a sensorless lighting environment. Thus, an indoor light environment dataset was constructed by collecting illuminance and floor images from various zones in an experimental environment with natural light. Through dataset-based analysis, color elements in images that were highly correlated with illuminance were identified, and a CNN-based deep learning model was trained to estimate illuminance at each point using indoor images as input. A general method for calculating real-time illuminance based on an indoor image has been proposed. In experiments, this method demonstrated a Mean Absolute Percentage Error (MAPE) of within 6% for individual points during daytime hours, and 11% daily, even when indoor illuminance fluctuated significantly.