Image Sequence Measures for Automatic Target Tracking

Abstract

Abstract-In the field of automatic target recognition and tracking, traditional image metrics focus on single images, ignoring the sequence information of multiple images. We show that measures extracted from image sequences are highly relevant concerning the performances of automatic target tracking algorithms. To compensate the current lack of image sequence characterization systems from the perspective of the target tracking difficulties, this paper proposes three new metrics for measuring image sequences: inter-frame change degree of texture, inter-frame change degree of target size and inter-frame change degree of target location. All are based on the fact that inter-frame change is the main cause interfering with target tracking in an image sequence. As image sequences are an important type of data in the field of automatic target recognition and tracking, it can be concluded that the work in this paper is a necessary supplement for the existing image measurement systems. Experimental results reported show that the proposed metrics are valid and useful.

Adaptive edge-based mean shift for drastic change gray target tracking

“…To evaluate and compare the performance of various tracking algorithms, tracking error which is calculated by the Euclidean distance as follows [44] is used to quantify the performance.…”

A novel method for quantifying target tracking difficulty of the infrared image sequence

“…While ordinary image metrics, such as peak signal-to-noise ratio (PSNR) [3], mean square errors (MSE) [3], V-Factor [4] and mean structural similarity (MSSM) [5] usually focus on measuring information losses in image communication and video compression [2]. Therefore, these metrics have not been applied to evaluate the performance of target tracking algorithms on the infrared image sequence.…”

“…It should be noted that the concept of ''image metric'' in the field of automatic target recognition and tracking is different from its normal concept for an ordinary image or video: they are conceived to describe the factors interfering with the performance of the ATR algorithms [2]. While ordinary image metrics, such as peak signal-to-noise ratio (PSNR) [3], mean square errors (MSE) [3], V-Factor [4] and mean structural similarity (MSSM) [5] usually focus on measuring information losses in image communication and video compression [2].…”

A Novel Sar Target Detection Algorithm Based on Contextual Knowledge

“…It has long been acknowledged that visual context plays an important role in visual perception of object [11,12]. Consequently, there has been increasing interest in recent years in developing computational models to improve object detection in images by exploiting contextual knowledge [13][14][15][16][17][18][19][20][21][22]. Blacknell et al [23,24] proposed a contextual knowledge-based target detection algorithm which takes the influence of context to the prior probabilities for the occurrence of targets or clutter into account and validates the algorithm using simulation image.…”