摘要 :
It's difficult to detect LSS(Low-Small-Slow) target because of its overturning, distortion during movement. Aiming at this problem, this paper proposes a LSS target tracking algorithm based on optical flow detection and polynomial...
展开
It's difficult to detect LSS(Low-Small-Slow) target because of its overturning, distortion during movement. Aiming at this problem, this paper proposes a LSS target tracking algorithm based on optical flow detection and polynomial fitting relocation. Firstly, target is detected by optical flow method, and then the SVM classifier trained with hog feature and color histogram feature is used to eliminate the false target. Then use TBD strategy to process the target information in the first three frames to confirm the initial location of the target. In follow-up tracking process, the obtained target information is compared with polynomial fitting results to determine whether to trigger the relocation mechanism. In relocation mechanism, the hot spot region will be generated according to the polynomial fitting results, and the salient features of the hot spot region will be extracted to determine the target location. Through the performance on the public data set LaSOT and the image sequence collected by the author, the algorithm in this paper is insensitive to the maneuver and distortion of LSS targets, and the tracking effect is stable. Under certain constraints, the tracking accuracy of the algorithm can reach more than 96%, which has strong application value.
收起
摘要 :
It's difficult to detect LSS(Low-Small-Slow) target because of its overturning, distortion during movement. Aiming at this problem, this paper proposes a LSS target tracking algorithm based on optical flow detection and polynomial...
展开
It's difficult to detect LSS(Low-Small-Slow) target because of its overturning, distortion during movement. Aiming at this problem, this paper proposes a LSS target tracking algorithm based on optical flow detection and polynomial fitting relocation. Firstly, target is detected by optical flow method, and then the SVM classifier trained with hog feature and color histogram feature is used to eliminate the false target. Then use TBD strategy to process the target information in the first three frames to confirm the initial location of the target. In follow-up tracking process, the obtained target information is compared with polynomial fitting results to determine whether to trigger the relocation mechanism. In relocation mechanism, the hot spot region will be generated according to the polynomial fitting results, and the salient features of the hot spot region will be extracted to determine the target location. Through the performance on the public data set LaSOT and the image sequence collected by the author, the algorithm in this paper is insensitive to the maneuver and distortion of LSS targets, and the tracking effect is stable. Under certain constraints, the tracking accuracy of the algorithm can reach more than 96%, which has strong application value.
收起
摘要 :
Aiming at the difficulties of detecting and tracking ground targets, it puts forward a universal detecting and tracking algorithm of ground target based on the analysis and comparison of current detecting and tracking algorithms. ...
展开
Aiming at the difficulties of detecting and tracking ground targets, it puts forward a universal detecting and tracking algorithm of ground target based on the analysis and comparison of current detecting and tracking algorithms. Firstly it adopts inter-frame differencing algorithm and morphological filter algorithm to detect the potential targets. Then calculate the accurate position of targets based on integrated information measurement and layered voting. Finally the confidence was adopted in the observation probability of particle filter to realize the reasonable tracking of targets. The experimental result shows that the algorithm mentioned above is effective and universal for detection and tracking of ground targets.
收起
摘要 :
Aiming at the difficulties of detecting and tracking ground targets, it puts forward a universal detecting and tracking algorithm of ground target based on the analysis and comparison of current detecting and tracking algorithms. ...
展开
Aiming at the difficulties of detecting and tracking ground targets, it puts forward a universal detecting and tracking algorithm of ground target based on the analysis and comparison of current detecting and tracking algorithms. Firstly it adopts inter-frame differencing algorithm and morphological filter algorithm to detect the potential targets. Then calculate the accurate position of targets based on integrated information measurement and layered voting. Finally the confidence was adopted in the observation probability of particle filter to realize the reasonable tracking of targets. The experimental result shows that the algorithm mentioned above is effective and universal for detection and tracking of ground targets.
收起
摘要 :
In this paper the Staring Infrared Imaging System process is divided into two stages, the one is related with the target distance, another with the sensor only. The apparent radiance of the airborne target and the expression betwe...
展开
In this paper the Staring Infrared Imaging System process is divided into two stages, the one is related with the target distance, another with the sensor only. The apparent radiance of the airborne target and the expression between the pixel gray and the apparent radiance were deduced based on each stage respectively. Then the aerial surface target distance was deduced based on the single-station dual-band infrared imaging system. Lastly the algorithm was validated by the IR images got from the aerodrome, and the results and ranging errors were analyzed.
收起