Research Article
Detection of Multiple Small Moving Targets Against Complex Ground Background
@INPROCEEDINGS{10.1007/978-3-031-06368-8_20, author={Junhua Yan and Jingchun Qi and Xuyang Cai and Yin Zhang and Kun Zhang and Yue Ma}, title={Detection of Multiple Small Moving Targets Against Complex Ground Background}, proceedings={Wireless Mobile Communication and Healthcare. 10th EAI International Conference, MobiHealth 2021, Virtual Event, November 13--14, 2021, Proceedings}, proceedings_a={MOBIHEALTH}, year={2022}, month={6}, keywords={Small target detection Complex background Target motion information Trajectory association Trajectory feature}, doi={10.1007/978-3-031-06368-8_20} }- Junhua Yan
Jingchun Qi
Xuyang Cai
Yin Zhang
Kun Zhang
Yue Ma
Year: 2022
Detection of Multiple Small Moving Targets Against Complex Ground Background
MOBIHEALTH
Springer
DOI: 10.1007/978-3-031-06368-8_20
Abstract
To tackle the problem that it is difficult to detect small moving targets accurately against complex ground background, a target detection algorithm that combines target motion information and trajectory association is proposed. To tackle the problem of small target size, firstly, background motion compensation is performed to obtain the background motion parameters. Then, forward and backward motion history maps are calculated to fuse continuous difference images for enhanced motion information of small targets. Finally, morphology processing is used to obtain the area of small moving targets. To tackle the problem of complex background, the Kalman predictor is used to predict the target position, and the Hungarian matching algorithm is used to correlate targets to obtain the target trajectory. Then, based on the target trajectory, targets missed by detection are supplemented to improve the target recall rate and false alarm targets are filtered out to improve the target precision rate. Experimental results show that the proposed algorithm has good detection performance, with the recall rate higher than 93%, the precision rate higher than 92%, and the F-measure higher than 93%.
