In this paper,the application of infrared thermal imaging technology in the ADAS system is studied,aiming at the contradiction between driving safety required by the advanced driver assistance system (ADAS) and that existing vehicle sensor kits cannot adequately detect vehicles or pedestrians.First,the necessity of adding an infrared camera to a typical car sensor kit is discussed,and then preprocessing the images while performing model training combined with the YOLOv5 algorithm for target detection under deep learning.Finally the experimental data are used to prove that the algorithm can better detect and classify traffic targets in a complex driving environment,thus helping the ADAS system to achieve target detection with both accuracy and real time.