Dynamic Programming-Based Multiple Point Target Detection Using K-means Clustering Algorithm
The problem of detecting multiple point targets that provide a level of situation awareness for unmanned aerial vehicles is addressed. The proposed tracking system, based on the track-before-detect approach, is designed to track and detect multiple targets from a sequence of a vision sensor under low SNR conditions. The system achieves multiple point target detection in three steps. The first step is morphological filtering process based on grayscale morphology for extracting intensive point-like features within image frame. Such filters are derived from combinations of dilation and erosion operations. The second step is target detection and tracking based on a dynamic programming approach. The dynamic programming approach accumulates scores of the pixels from the image sequence of morphological filter outputs along possible target trajectories. The scores for the potential target trajectories can be accumulated by considering the temporally and spatially uncorrelated noise and smoothly moving targets with only gradually changes in direction and speed. The decision of the target presence and position is made in the third step with threshold parameters set to achieve appropriate probabilities of detection and false alarm. In this step, K-means algorithm is used for identifying position and number of targets in twodimensional space. The proposed track-before-detect approach using K-means clustering algorithm is applied to several image sequences containing different scenarios and noise conditions.
point target detection dynamic programming k-means clustering
Won Daeyeon Kim Keumseong Shim Sangwook Tahk Minjea
Department of Aerospace Engineering, KAIST, Daejeon, Korea Department of Aerospace Engineering,KAIST, Daejeon, Korea Department of Aerospace Engineering,KAIST,Daejeon,Korea
国际会议
2010 Asia-Pacific International Symposium on Aerospace Technology(2010 亚太航空航天技术研讨会 APISAT 2010)
西安
英文
732-735
2010-09-01（万方平台首次上网日期，不代表论文的发表时间）