This paper presents a thermal image tracking system for search and rescue (SAR) missions using a drone. The system combines object detection and multiple-target tracking to identify and follow people in thermal videos. The You-Only-Look-Once (YOLO) detection model is used to detect people, while multiple-target tracking is performed through track initialization, maintenance, and termination. A bounding box gating rule is proposed to associate measurements with tracks, improving tracking accuracy. The system also incorporates a Kalman filter for track maintenance and track-to-track association to enhance tracking performance. Experiments with three videos of hikers in mountainous environments under challenging conditions show that the system achieves robust tracking results in terms of average total track life and average track purity, although the average mean track life is reduced in harsh environments. The proposed method is effective in tracking people in thermal images, even when objects are close, occluded, or the drone is moving arbitrarily. The system is suitable for SAR missions, surveillance, and other applications requiring thermal imaging and tracking.This paper presents a thermal image tracking system for search and rescue (SAR) missions using a drone. The system combines object detection and multiple-target tracking to identify and follow people in thermal videos. The You-Only-Look-Once (YOLO) detection model is used to detect people, while multiple-target tracking is performed through track initialization, maintenance, and termination. A bounding box gating rule is proposed to associate measurements with tracks, improving tracking accuracy. The system also incorporates a Kalman filter for track maintenance and track-to-track association to enhance tracking performance. Experiments with three videos of hikers in mountainous environments under challenging conditions show that the system achieves robust tracking results in terms of average total track life and average track purity, although the average mean track life is reduced in harsh environments. The proposed method is effective in tracking people in thermal images, even when objects are close, occluded, or the drone is moving arbitrarily. The system is suitable for SAR missions, surveillance, and other applications requiring thermal imaging and tracking.