Multi-Object Detection and Tracking via Kernel Covariance Factorization in Thermal Video

摘要

This paper addresses the problem of joint object detection and tracking in thermal videos. Object detection is formulated as a sparse factorization task of a properly defined kernel covariance matrix. The support of these estimated factors is used to determine the indices of the pixels that form each object. A coordinate descent approach is utilized to determine the sparse factors, and extract the object pixels. For each object, the centroid pixel is subsequently tracked via Kalman filtering. A novel interplay between the sparse kernel covariance factorization scheme along with Kalman filtering is proposed to enable joint object detection and tracking, while a divide and conquer strategy is put forth to reduce computational complexity in tracking. Numerical tests demonstrate the improved tracking performance over existing alternatives.