<![CDATA[Developing a reliable person-following robot in low-light and high-temperature environments remains challenging because conventional vision-based sensors often experience unstable detection. This study proposes a thermal sensor fusion system using a GY-AMG8833 thermal sensor (8x8 pixels) for directional tracking and an MLX90614 infrared sensor for human body temperature validation. The robot is controlled by an ESP32 microcontroller implementing a Weighted Centroid algorithm to determine the target position based on heat distribution. To overcome the sensitivity of low-cost thermal sensors to environmental noise, especially when ambient temperature exceeds 31 degrees C, the proposed method applies a dynamic threshold offset of 1.5 degrees C above ambient temperature. In addition, a heat-area-based noise filter is implemented, where the robot automatically stops when the detected heat area exceeds 60 pixels, indicating non-target objects such as walls or reflective surfaces. Experimental results in air-conditioned and non-air-conditioned environments show that the proposed system improves tracking accuracy by up to 70%, reduces the false positive rate from 80% to 0%, and prevents collisions with static heat-emitting objects. These results demonstrate that the proposed approach provides safer, more stable, and more reliable autonomous robot navigation.]]>
Copyrights © 2026
