화재 현장에서의 RGB-T 카메라 기반 객체 인식 및 추적

This study investigates a method for accurately detecting and tracking the location and movement of objects in fire scenes using RGB and thermal images. The proposed coarse-to-fine fusion method is used to recognize objects, and the optical flow algorithm is then applied within the detected object areas to track their movement directions. The experiments were conducted using a mobile robot in a simulated fire environment, where object recognition and tracking were performed. The performance of the object detection model was evaluated using the standard COCO evaluation metrics, and the object tracking performance was verified by analyzing the changes in tracking vectors of accelerating objects in thermal images. This research provides a foundation for practical applications, such as security surveillance and rescue operations in disaster environments, by utilizing complex image data.