Detection of Industrial Heat Source in Remote Sensing Images Using Modified YOLO

Accurate detection of industrial heat sources holds paramount importance in efforts towards reducing carbon emissions and safeguarding the environment. Traditional approaches of identifying these sources primarily rely on field research and statistical registration of pertinent industrial establishments. However, the traditional approaches are time-consuming, inefficient, having a small monitoring range and difficult to update in real time. Leveraging the rapidity and precision inherent in YOLO's detection capabilities, coupled with the accessibility of optical remote sensing imagery, this study proposes a tailored YOLO network for swift industrial heat source detection. YOLO's unique approach to treating object detection as a regression problem significantly enhances target recognition speed. Experimental findings affirm the efficiency of the proposed methodology in effectively identifying multi-scale industrial heat sources in remote sensing imagery. Moreover, the modified model showcases a notable improvement of 5.1% in precision rate and 11.7% in recall rate compared to the baseline.