5D target trajectory detection via intelligent monocular visual tracking system in real-time with air-target orientation recognition

A night-and-day operational monocular visual tracking system (VTS) for a gun turret system is proposed. Its functions include detection and tracking an air-target in realistic noisy environment as well as automatic aiming of the turret gun at the target. For detection purpose, a real-time integrated detector is proposed to obtain the 2D target location. In addition, we retrieve the other 3D orientational information of the target via the model based pattern recognition. Thus, the task of 5D target trajectory detection can be achieved in our monocular VTS. However, in this paper only the first 2D target location tracking is implemented. Next, a two-layer structure fuzzy controller, named recurrent fuzzy visual gain controller, based on the usual perspective visual model but with additional consideration of the target velocity is proposed. Inside this controller, a mixed Kalman filter with IMM algorithm is also developed to estimate the 2D target position when the target is temporarily missing.