<![CDATA[This research develops an image processing-based floor stain detection system using grayscale conversion and binary thresholding. Two conversion approaches are compared: (i) a simple RGB grayscale formula and (ii) a built-in OpenCV function. The system uses a fixed intensity threshold of T=80 and classifies a floor as “dirty” if the detected area exceeds 20% of the image. Experiments are conducted on three floor types (plain, dark, patterned), five stain types (coffee, oil, ink, plastic, chalk), and two lighting conditions. Results show that the performance of both methods is very close with an average difference of ≈0.07% and a maximum of 0.6%; the simple formula is suitable for resource-limited devices, while OpenCV is more robust to color/lighting variations. The main contributions of this paper are (1) a practical comparison of two grayscale conversion pathways for cleanliness monitoring, (2) a simple decision rule based on the percentage of dirty area that aligns with cleanliness perception, and (3) an analysis of implementation implications for real-time systems in cleaning robots/IoT. Future directions include adaptive thresholding and morphology integration to improve reliability in dynamic environments. (Replace the current abstract paragraph containing T=80 and the 20% rule with the version above. The 20% policy reference is already explained in your manuscript).]]>
Copyrights © 2025
