Infrared target tracking with kernel-based performance metric and eigenvalue-based similarity measure

Abstract

An infrared target tracking framework is presented that consists of three main parts: mean shift tracking, its tracking performance evaluation, and position correction. The mean shift tracking algorithm, which is a widely used kernel-based method, has been developed for the initial tracking for its efficiency and effectiveness. A performance evaluation module is applied for the online evaluation of its tracking performance with a kernel- based metric to unify the tracking and performance metric within a kernel-based tracking framework. Then the tracking performance evaluation result is input into a controller in which a decision is made whether to trigger a position correction process. The position correction module employs a matching method with a new eigenvalue-based similarity measure computed from a local complexity degree weighted covariance matrix. Experimental results on real-life infrared image sequences are presented to demonstrate the efficacy of the proposed method.

Two Algorithms for the Detection and Tracking of Moving Vehicle Targets in Aerial Infrared Image Sequences

“…In this study, we aimed to produce a tracking method that is highly robust and accurate by drawing on the idea of Tracking Learning Detection (TLD) [30][31][32] and combining target tracking with real-time detection in order to realize real-time updating of the target model. Based on these ideas and the characteristics of infrared images, we investigated the use of kernel-based tracking theory [33], which has previously performed well in infrared target tracking [23,26]. As a result, a novel tracking algorithm referred to as the Kernel-Based Mean Shift Target Tracking Based on Detection Updates (MSDU) is proposed to realize stable target tracking in infrared aerial sequences.…”

“…For example, Ling et al [23] defined the evaluation criterion for the tracking effect and searched for the relatively accurate region similar to the reference region by maximizing the eigenvalues of the covariance matrix of the local complexity when the tracking error was large. Based on active contours, Salah et al [24] combined a kernel photometric tracking term and a model-free shape tracking term to track several objects independently and accurately in infrared image sequences.…”

Spatiotemporal difference-of-Gaussians filters for robust infrared small target tracking in various complex scenes

“…In PF approaches, IR target tracking is thought as a recursive Bayesian estimate problem, and the estimated states represent some parameters of the tracked targets, such as positions and velocity [13]. Equally popular are mean shift (MS) methods [14][15][16], which represent IR target appearances by kernel weighted gray histogram and use the mean shift procedure to identify the most likely position of the target in the next frame [17]. Both PF and mean shift methods rely on models of target appearance, which are used to locate the target as time evolves.…”

Salient point region covariance descriptor for target tracking

“…In Ref. 37, the characteristics of the eigenvalues of the weighted covariance matrix are used for the position correction task. The weighted covariance matrix proposed in that work is based on the pixel-wise intensity statistics of the reference image and the scene image.…”