In recent aircraft noise survey in Japan, noise data is associated with each aircraft by flight log or by radio information including transponder signals. Especially, above Tokyo metropolitan area, flight tracks are tangled extremely each other, therefore assessments from various perspectives such as departure / arrival airport, used runway, aircraft model, and operator have been demanded for determining noise policies. However, for military aircrafts, it is not easy to identify their information with the same way as commercial aircrafts, because their flight logs are not disclosed and many of them do not emit transponder signals like commercial aircrafts. Therefore, manned 24 hours survey around air bases have been necessary to obtain flight information of military aircrafts. In this paper, we propose an AI-based analysis using captured aircraft images for obtaining actual flight data of military aircrafts. In the past trials, we could determine the takeoff/landing time and the aircraft model by the above method. Associating these information and noise data measured at monitoring stations, details of noise characteristics around the air base can be clearly grasped. Advanced analysis of the causes of noise impact will lead effective and concrete countermeasures.