New AI drone finds victims in forest, darkness

After the disaster, finding survivors in the wilderness has been a battle against time, and rescue teams are fighting challenging terrain and weather conditions. Now, researchers have developed an artificial intelligence system that greatly improves how drones detect people in search and rescue missions, even in dense forests or darkness.

A team from Northwestern Polytechnic University and Yan’An University in China created a dual-view method that combines conventional photographic lenses with infrared imaging, allowing drones to spot victims that are almost invisible from above.

“Our research helps develop more efficient airman detection for search and rescue missions,” said Dr. Xiangqing Zhang, principal investigator of the study. “By integrating AI with multimodal data, we designed a system that can improve detection capabilities in complex environments, making SAR operations more efficient and reliable.”

The technology addresses some long-standing challenges that hinder aerial detection efforts. Traditional drone cameras struggle with people who are partially hidden by trees, motionless or integrated into their surroundings. Small objects (such as a person viewing from a high altitude) are particularly difficult to reliably identify AI systems.

To address these problems, the researchers built a custom unmanned helicopter equipped with synchronous visible and infrared cameras. The system can detect a hot signature of a person who is invisible to conventional cameras due to camouflage, darkness, or partial covering.

The team has compiled an extensive dataset called VTSAR that contains thousands of images captured in a variety of environments, from urban areas to wilderness and sea environments. This dataset is the basis for training its detection algorithm, which has impressive accuracy in testing.

What distinguishes this study is the fusion of multiple AI methods. The system combines object-aware approaches to deal with proportional changes, as well as information fusion techniques that work under different lighting conditions. To perform rescue operations in remote areas with limited computing resources, lightweight algorithms ensure that the system can operate efficiently on the drone itself.

Apart from search and rescue, the technology could transform other applications, including disaster response, security monitoring and wildlife protection, providing a powerful example of how AI can enhance critical human security operations.

Reading Research

Related