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

Abstract

Tracking infrared (IR) small targets is a vital component of many computer vision applications, including IR precise guidance, early warning, and IR remote sensing. Various complicated scenes, however, present significant challenges to the tracking task. To solve this problem, we present a novel 3D spatiotemporal difference-of-Gaussians (DoG) filter-based algorithm for tracking small targets in IR videos of various complex scenes. First, biologically inspired 3D DoG filters are proposed for IR small target tracking, which are capable of accounting for spatial and temporal information. Then, based on such filters, an effective and robust tracker is constructed to track the small targets, which are spatiotemporally distinguishable from background clutter. Extensive experiments show that our approach tracks the small targets accurately and robustly in realistic IR videos of complex backgrounds that present unique difficulties to other approaches.

Saliency detection using mutual consistency-guided spatial cues combination

“…In previous literature [14][15][16][17], luminance is always regarded as an important and distinctive visual characteristic of infrared images. Most methods for infrared computer vision applications, including infrared image segmentation [14], target detection [15], and target tracking [16], assume that objects of interest, such as human beings, ground vehicles, aircraft, are hotter than backgrounds.…”

Infrared Small Target Detection Using a Temporal Variance and Spatial Patch Contrast Filter

“…Since speckle noise and broken clouds with similar intensity and shape to the target are the major false alarm sources after processed by single-frame detection methods, various multi-frame detection methods are proposed to remove these regions using temporal information, such as 3-D Matched filtering [18], [19], Dynamic Programming method [20], [21], Maximum Likelihood method [22], and Markov Random Field (MRF) [23], and Convolution Neural Network (CNN) based methods [24]- [26]. However, these methods cannot detect targets with sub-pixel motion in adjacent frames.…”

Multi-feature local sparse representation for infrared pedestrian tracking

“…Decades of study on this issue have generated a series of approaches [2][3][4][5][6][7][8][9][10][11][12][13]. Thereinto, particle filter (PF) has gotten particular attention for the capability of solving non-linear and non-Gaussian questions [2][3][4].…”

“…Mean shift-based tracking technique has been put forward as an expeditous technique [6][7][8][9][10]. In [11], spatial-temporal filters have been designed to track infrared target. The dense structural learning has been proposed to train a classifier with dense samples through Fourier techniques for infrared object tracking [12].…”