Human tracking in thermal images using adaptive particle filters with online random forest learning

Abstract

This paper presents a fast and robust human tracking method to use in a moving long-wave infrared thermal camera under poor illumination with the existence of shadows and cluttered backgrounds. To improve the human tracking performance while minimizing the computation time, this study proposes an online learning of classifiers based on particle filters and combination of a local intensity distribution (LID) with oriented centersymmetric local binary patterns (OCS-LBP). Specifically, we design a realtime random forest (RF), which is the ensemble of decision trees for confidence estimation, and confidences of the RF are converted into a likelihood function of the target state. First, the target model is selected by the user and particles are sampled. Then, RFs are generated using the positive and negative examples with LID and OCS-LBP features by online learning. The learned RF classifiers are used to detect the most likely target position in the subsequent frame in the next stage. Then, the RFs are learned again by means of fast retraining with the tracked object and background appearance in the new frame. The proposed algorithm is successfully applied to various thermal videos as tests and its tracking performance is better than those of other methods. © The Authors. Published by SPIE under a Creative Commons Attribution 3.0 Unported License. Distribution or reproduction of this work in whole or in part requires full attribution of the original publication, including its DOI.

PTB-TIR: A Thermal Infrared Pedestrian Tracking Benchmark

“…In [6], the authors also fuse the intensity and edge information using a relative discriminative coefficient. In [7], the authors coalesce local intensity distribution (LID) and oriented center symmetric local binary patterns (OCS-LBP) to represent the TIR pedestrian. Furthermore, some other features are often used in TIR pedestrian tracking, such as regions of interest (ROI) histograms [5], speeded up robust features (SURF) [28], and the histogram of oriented gradients (HOG) [33].…”

“…For example, Xu et al [4] use the mean shift to get the final tracked target from a series of target candidates generated from the Kalman filter. Ko et al [7] exploit a random forest to obtain the tracked target from candidates. Some other classification methods such as SVM [30], boosting [35], and multiple instance learning (MIL) [36] are also suitable for TIR pedestrian tracking.…”

Recognizing pedestrian’s unsafe behaviors in far-infrared imagery at night

“…In general, FIR cameras can detect relative differences in the amounts of thermal energy emitted or reflected from different body parts of a pedestrian in a scene, regardless of the illumination [7]. Therefore, it allows robust detection of pedestrian bodies in both indoor and outdoor environments.…”

Far-infrared pedestrian detection for advanced driver assistance systems using scene context

“…[2][3][4] On one hand, the FIR camera is well-suited for pedestrian detection even in a dark environment, since it detects the amount of thermal radiation emitted from objects without depending on the illumination of the scene. [2][3][4] On one hand, the FIR camera is well-suited for pedestrian detection even in a dark environment, since it detects the amount of thermal radiation emitted from objects without depending on the illumination of the scene.…”