The thermal images captured by night vision cameras are usually monochromatic, grainy, and somewhat blurry. According to the journal Nature published on the 26th, researchers from Purdue University and Los Alamos National Laboratory have developed a thermal assisted detection and ranging (HADAR) system, which trains artificial intelligence (AI) to determine the temperature, energy characteristics and physical texture of each pixel in the thermal image. The resulting image is almost as clear as the image taken by traditional cameras in daylight, It can help autonomous vehicles operate more safely in all outdoor conditions. In order to train AI, researchers use complex thermal imaging cameras and imaging sensors at night to capture data outdoors, which can display the energy emission of the entire electromagnetic spectrum. They also created computer simulations of outdoor environments for additional AI training. Researchers say that HADAR has learned to detect objects and estimate their distance from them, with an accuracy 10 times higher than relying solely on traditional night vision technology. It uses invisible infrared radiation to reconstruct nighttime scenes, with clarity similar to that of daytime. However, HADAR still faces challenges in practical applications, such as real-time data acquisition, dynamic blur, and cost issues, as bulky and expensive cameras and imaging devices need to be manufactured in smaller volumes and at lower costs. The process of collecting and processing data still takes about a minute, ideally within a few milliseconds, in order to ensure that autonomous vehicles use the system while driving. (Zhang Jiaxin) (Liaoxin News Agency)