Reviewed by Lexie CornerApr 22 2025
Researchers from Northwestern Polytechnical University and Yan’an University developed an AI-powered APD system to enhance the recognition of people in images captured by drones.
Challenges in Aerial Person Detection (APD). Image Credit: Journal of Remote Sensing
Unpredictable weather, rugged terrain, and limited resources contribute to the challenges of search and rescue (SaR) missions. Traditional methods rely heavily on experienced personnel, making the process time-consuming and labor-intensive.
Unmanned aerial vehicles (UAVs) offer a promising alternative, but their effectiveness is limited by the difficulty of detecting small or partially visible individuals in aerial imagery. This highlights the need for advanced APD technology to improve the precision and efficiency of rescue operations.
The new system addresses challenges such as occlusion, scale variation, and changing illumination conditions, thereby enhancing the accuracy and reliability of SaR operations, particularly in remote and inaccessible areas.
The research team developed the VTSaR dataset, which includes a range of locations, human activities, and capture angles. The dataset incorporates both visible and infrared images, as well as synthetic data, to provide a comprehensive benchmark for APD. Several detection algorithms were tested, resulting in noticeable improvements in detection accuracy and efficiency. The proposed system performed effectively under challenging conditions, surpassing previous technologies in managing occlusions, scale variations, and lighting changes.
Our research contributes to the development of more effective Aerial Person Detection for search and rescue missions. By integrating AI with multimodal data fusion, we have designed a system that improves detection capabilities in complex environments, making SaR operations more efficient and reliable.
Dr. Xiangqing Zhang, Study Lead Researcher, Northwestern Polytechnical University
The study utilized a custom-built unmanned aircraft equipped with a dual-camera gimbal system to capture aerial images from various habitats, including urban, suburban, maritime, and wilderness areas. The VTSaR dataset is divided into three versions: Unaligned VTSaR (UA-VTSaR), Aligned VTSaR (A-VTSaR), and Aligned Synthetic VTSaR (AS-VTSaR), comprising a total of 19,956 real-world and 54,749 synthetic occurrences.
Researchers evaluated models such as YOLOv8-s and EfficientViT, achieving a precision of 95.03 % and a mean average precision (mAP) of 94.91 %. The study highlighted the effectiveness of combining visible and infrared imagery to enhance detection performance across various environmental conditions.
Beyond SaR applications, the APD system has potential uses in disaster response, security monitoring, and law enforcement. Improved APD accuracy could assist in locating missing persons, monitoring high-risk areas, and responding more efficiently to emergencies. As AI and UAV technologies advance, the system may be adapted for other applications, such as wildlife monitoring and border surveillance, further enhancing safety and security.
This study outlines an AI-powered APD system that improves SaR efficiency by overcoming key technological challenges. By integrating AI-driven analysis with multimodal data fusion, the technology provides a more accurate and flexible method for locating people in complex environments. These advancements help optimize rescue operations and improve the likelihood of timely intervention in critical situations.
The study was partially funded by the National Natural Science Foundation of China under Grants 62171381.
Journal Reference:
Zhang, A., et al. (2025) Aerial Person Detection for Search and Rescue: Survey and Benchmarks. Journal of Remote Sensing. doi.org/10.34133/remotesensing.0474