Previous [1] [2] [3] [4] [5]

Journal of Inforamtion Science and Engineering, Vol.6 No.1, pp.1-24 (March 1990)
Infrared Target Detection and Tracking by
Region Correspondence in Image Sequences

Kung-Puu Tang and Wen-Hsiang Tsai
Institute of Computer and Information Science
National Chiao Tung University
Hsinchu, Taiwan 30050, Republic of China

In this paper, a region correspondence approach is proposed for use in image sequence analysis to remove false alarms in target detection. Object segmentation techniques based on thresholding are performed first. A range of threshold values is determined according to different image backgrounds which are classified by gray level contrast and cooccurrence matrices. Hierarchical clustering is then used to construct effective feature-based tracking models. Local least-mean-square-error estimation is finally used to track the targets continuously. High target detection rates and no false alarm prove the feasibility of the proposed approach for real applications.

Keywords: target detection, target tracking, infrared images, image sequences, region correspondence, hierarchical clustering, target models, target motion

Received July 29, 1989; revised November 24, 1989.
Communicated by Lin-Shan Lee.