Search

According to the nonlinear maneuvering characteristics of flying targets, unscented Kalman filter is used to achieve a two-dimensional trajectory tracking and predicting in the coordinates of camera view field. By calculating the target centroid position in the imaging focal plane, the third-order constant acceleration motion model of two-dimensional is established. Because of the nonlinear characteristics of flying targets, unscented transform sampling is used to update the states matrix. It is not necessary to solve the complicated Jacobian matrix which cannot be avoided in the ordinary extended Kalman filtering. Unscented Kalman filter is used to update the target prediction matrix, and the next state of the target prediction matrix is used to be the trajectory prediction value. Model simulation and data processing results show that the algorithm brings a great convenience for real-time processing platform at the coordinate of camera view field.

Nowadays, infrared imaging systems play important roles in the field of civil and military. Especially small infrared target detecting and recognizing is one of the most widely use. The capability of target-detection algorithm is an important index of the system. This paper presents a novel algorithm for detecting a small moving target in infrared (IR) image sequences and finding its mass center, and recording the target moving track. In the target searching and recognizing algorithm of infrared image sequences, infrared image sequence is broken into frames, filtered by spatial filter algorithm, which helped to reduce granular noise. We use the Canny algorithm factor to find the edge of the target, and the result of detecting target edge is process by ecological open-loop filter method, including erosion and dilation algorithm with a same scale. Then, the candidate targets are recognized and saved temporarily. In order to get the mass centers of the candidate targets, the valid area of the candidate targets is defined by different weight valves, and then the mass centers are calculated by weighted average algorithm, and record per frame. After got several frames mass centers of the candidate targets, we get rid of the non-target mass centers by frame difference algorithm, and get the real mass center of the small moving infrared target. If the background is observed for enough time, the effect of frame difference algorithm is more efficiency. Finally, the moving track of the target is found out. The infrared (IR) image sequences used here are obtained through an IR camera in the laboratory, which uses a 288*384 silicon infrared image sensor produced by ULIS company. The methods referred above are realized and simulated on compute with Matlab. Theory analysis and experiments prove the method is reasonable and efficient.