Carlos R. del Blanco

Advanced background modeling with RGB-D sensors through classifiers combination and inter-frame foreground prediction

An innovative background modeling technique that is able to accurately segment foreground regions in RGB-D imagery (RGB plus depth) has been presented in this paper. The technique is based on a Bayesian framework that efficiently fuses different sources of information to segment the foreground. In particular, the final segmentation is obtained by considering a prediction of the foreground regions, carried out by a novel Bayesian Network with a depth-based dynamic model, and, by considering two independent depth and color-based mixture of Gaussians background models. The efficient Bayesian combination of all these data reduces the noise and uncertainties introduced by the color and depth features and the corresponding models. As a result, more compact segmentations, and refined foreground object silhouettes are obtained. Experimental results with different databases suggest that the proposed technique outperforms existing state-of-the-art algorithms.

Multi-sensor background subtraction by fusing multiple region-based probabilistic classifiers

We use RGB-D cameras data for foreground/background segmentation.Pixel level and region level background models based on color and depth data.Foreground region prediction, based on depth based histograms.Fusion of region based classifiers as mixture of experts. In the recent years, the computer vision community has shown great interest on depth-based applications thanks to the performance and flexibility of the new generation of RGB-D imagery. In this paper, we present an efficient background subtraction algorithm based on the fusion of multiple region-based classifiers that processes depth and color data provided by RGB-D cameras. Foreground objects are detected by combining a region-based foreground prediction (based on depth data) with different background models (based on a Mixture of Gaussian algorithm) providing color and depth descriptions of the scene at pixel and region level. The information given by these modules is fused in a mixture of experts fashion to improve the foreground detection accuracy. The main contributions of the paper are the region-based models of both background and foreground, built from the depth and color data. The obtained results using different database sequences demonstrate that the proposed approach leads to a higher detection accuracy with respect to existing state-of-the-art techniques.

An advanced Bayesian model for the visual tracking of multiple interacting objects

Visual tracking of multiple objects is a key component of many visual-based systems. While there are reliable algorithms for tracking a single object in constrained scenarios, the object tracking is still a challenge in uncontrolled situations involving multiple interacting objects that have a complex dynamics. In this article, a novel Bayesian model for tracking multiple interacting objects in unrestricted situations is proposed. This is accomplished by means of an advanced object dynamic model that predicts possible interactive behaviors, which in turn depend on the inference of potential events of object occlusion. The proposed tracking model can also handle false and missing detections that are typical from visual object detectors operating in uncontrolled scenarios. On the other hand, a Rao-Blackwellization technique has been used to improve the accuracy of the estimated object trajectories, which is a fundamental aspect in the tracking of multiple objects due to its high dimensionality. Excellent results have been obtained using a publicly available database, proving the efficiency of the proposed approach.

Visual tracking of multiple interacting objects through Rao-Blackwellized Data Association Particle Filtering

A multiple object visual tracking framework is presented, which is able to manage complex object interactions, missing detections and clutter. The main contribution is the ability to deal with complex situations in which the interacting objects can change their dynamics while they are occluded. This is achieved by explicitly estimating putative locations of the occluded objects. The tracking is modeled by a Rao-Blackwellized Data Association Particle Filter (RBDAPF), which has a tractable substructure that allows to analytically compute the object positions, while the object-measurement associations are approximated by Particle Filtering. Besides improving the accuracy, this filter decomposition reduces the computational cost, since the complexity with the number of objects becomes linear instead of exponential. The Particle Filter efficiently manages the measurements from visible and occluded objects, the clutter, and missing measurements to estimate the correct data associations that lead to a robust tracking. Experimental results on surveillance videos show that the proposed RBDAPF framework is able to track multiple interacting objects in complex situations.

Robust tracking in aerial imagery based on an ego-motion Bayesian model

A novel strategy for object tracking in aerial imagery is presented, which is able to deal with complex situations where the camera ego-motion cannot be reliably estimated due to the aperture problem (related to low structured scenes), the strong ego-motion, and/or the presence of independent moving objects. The proposed algorithm is based on a complex modeling of the dynamic information, which simulates both the object and the camera dynamics to predict the putative object locations. In this model, the camera dynamics is probabilistically formulated as a weighted set of affine transformations that represent possible camera ego-motions. This dynamic model is used in a Particle Filter framework to distinguish the actual object location among the multiple candidates, that result from complex cluttered backgrounds, and the presence of several moving objects. The proposed strategy has been tested with the aerial FLIR AMCOM dataset, and its performance has been also compared with other tracking techniques to demonstrate its efficiency.

New generation of human machine interfaces for controlling UAV through depth-based gesture recognition

New forms of natural interactions between human operators and UAVs (Unmanned Aerial Vehicle) are demanded by the military industry to achieve a better balance of the UAV control and the burden of the human operator. In this work, a human machine interface (HMI) based on a novel gesture recognition system using depth imagery is proposed for the control of UAVs. Hand gesture recognition based on depth imagery is a promising approach for HMIs because it is more intuitive, natural, and non-intrusive than other alternatives using complex controllers. The proposed system is based on a Support Vector Machine (SVM) classifier that uses spatio-temporal depth descriptors as input features. The designed descriptor is based on a variation of the Local Binary Pattern (LBP) technique to efficiently work with depth video sequences. Other major consideration is the especial hand sign language used for the UAV control. A tradeoff between the use of natural hand signs and the minimization of the inter-sign interference has been established. Promising results have been achieved in a depth based database of hand gestures especially developed for the validation of the proposed system.

Segmentation-tracking feedback approach for high-performance video surveillance applications

Here, a novel and efficient feedback system for moving object segmentation and tracking is proposed. Through the use of non-parametric background-foreground modeling, moving objects are correctly detected in unfavorable situations such as dynamic backgrounds or illumination changes. After detection, objects are tracked by an original multi-object Bayesian tracking algorithm, which achieves satisfactory results under partial and total occlusions. Updating the previously detected foreground data from the information provided by the tracker, the foreground modeling is improved, reducing the color similarity problem.

Inference of Complex Trajectories by Means of a Multibehavior and Multiobject Tracking Algorithm

Visual tracking of multiple objects is a fundamental aspect of many video-based systems. Today, there are reliable algorithms that can track a small number of objects in restricted situations. However, the tracking of a large number of objects in uncontrolled situations involving interacting objects with complex dynamics is still a challenge. In this situation, the typical assumptions of linearity and independence of object motions are not fulfilled, causing a low tracking performance. This paper proposes a novel Bayesian tracking algorithm for interacting objects that are able to reliably simulate several object behaviors with an uncalibrated camera, which can be positioned in an arbitrary perspective. Three different models of object behavior are used to simulate and predict the object dynamics, where the proportion of hypotheses of each possible behavior of an object depends on the dynamics (position, velocity, etc.) of the other objects in the scene. Experimental results on public databases prove the reliability and robustness of the proposed tracking algorithm in the presence of object interactions.

Structured learning via convolutional neural networks for vehicle detection

One of the main tasks in a vision-based traffic monitoring system is the detection of vehicles. Recently, deep neural networks have been successfully applied to this end, outperforming previous approaches. However, most of these works generally rely on complex and high-computational region proposal networks. Others employ deep neural networks as a segmentation strategy to achieve a semantic representation of the object of interest, which has to be up-sampled later. In this paper, a new design for a convolutional neural network is applied to vehicle detection in highways for traffic monitoring. This network generates a spatially structured output that encodes the vehicle locations. Promising results have been obtained in the GRAM-RTM dataset.

High-quality region-based foreground segmentation using a spatial grid of SVM classifiers

This paper presents a novel background modeling system that uses a spatial grid of Support Vector Machines classifiers for segmenting moving objects, which is a key step in many video-based consumer applications. The system is able to adapt to a large range of dynamic background situations since no parametric model or statistical distribution are assumed. This is achieved by using a different classifier per image region that learns the specific appearance of that scene region and its variations (illumination changes, dynamic backgrounds, etc.). The proposed system has been tested with a recent public database, outperforming other state-of-the-art algorithms.