Adaptive Kalman Filtering for Histogram-Based Appearance Learning in Infrared Imagery

Abstract

Targets of interest in video acquired from imaging infrared sensors often exhibit profound appearance variations due to a variety of factors, including complex target maneuvers, ego-motion of the sensor platform, background clutter, etc., making it difficult to maintain a reliable detection process and track lock over extended time periods. Two key issues in overcoming this problem are how to represent the target and how to learn its appearance online. In this paper, we adopt a recent appearance model that estimates the pixel intensity histograms as well as the distribution of local standard deviations in both the foreground and background regions for robust target representation. Appearance learning is then cast as an adaptive Kalman filtering problem where the process and measurement noise variances are both unknown. We formulate this problem using both covariance matching and, for the first time in a visual tracking application, the recent autocovariance least-squares (ALS) method. Although convergence of the ALS algorithm is guaranteed only for the case of globally wide sense stationary process and measurement noises, we demonstrate for the first time that the technique can often be applied with great effectiveness under the much weaker assumption of piecewise stationarity. The performance advantages of the ALS method relative to the classical covariance matching are illustrated by means of simulated stationary and nonstationary systems. Against real data, our results show that the ALS-based algorithm outperforms the covariance matching as well as the traditional histogram similarity-based methods, achieving sub-pixel tracking accuracy against the well-known AMCOM closure sequences and the recent SENSIAC automatic target recognition dataset.

Robust infrared target tracking based on particle filter with embedded saliency detection

“…To evaluate the performance of the proposed method, we conducted experiments on the AMCOM FLIR dataset 1 in which the same 10 video sequences used in [29] are used. Besides, we choose the same target initialization settings (position and size) and video lengths detailed in Table II as in [29].…”

“…Besides, we choose the same target initialization settings (position and size) and video lengths detailed in Table II as in [29]. In our method, we use different number of samples for saliency model and eigen space model, i.e., only N e ¼ 100 samples are used for the eigen space model and N s ¼ 300 for the saliency model which is faster to calculate.…”

“…(13) is ð1; 1; 0:001; 0:001Þ. In all compared methods, the target is initialized by using the ground truth information in the first frame as is done in [29]. During the evaluation, all the algorithms are run 50 times on each sequence with each setup since particle filter method is a Monte Carlo method.…”

“…the intensity variance. Venkataraman et al in [29] proposed an appearance learning based histogram model to capture the size variance and the appearance change. For object tracking, it can be also benefited from object retrieval [15,10,17] and classification methods [16,9], which can learn or find related objects in the video tracking [32,8].…”

The Study of Visual Measurement Based on GPS Navigatio

“…However, it has loopholes, such as positioning error caused by signal attenuation due to building block and signal instability indoors and in some areas. Kalman filter algorithm, centrali zed Kalman filter algorithm or dispersed Kalman filter algorithm are some of the solutions to enhance the fault tolerance [1][2][3][4]. Deri L. and his team carried out an indepth study of GPS navigation technology [5][6][7].…”