Mokhtar Keche

Divided Difference Kalman Filter for indoor mobile localization

Knowledge of the positions of sensor nodes is crucial for numerous applications in wireless sensors network. In this paper, we propose to use the Divided Difference Kalman Filter (DDKF) as a solution for locating and tracking a mobile node. This approach is an alternative variant of the nonlinear Kalman filtering, already used in this type of applications. The advantage of this approach is that it does not require calculation of the Jacobian as for the Extended Kalman Filter (EKF) and it does not need to use several parameters, as for the Unscented Kalman Filter (UKF) whose accuracy is closely dependent on the good choice of such parameters. In this work, a comparative performance study of four localization methods is conducted, namely the DDKF, the EKF, the UKF and the Least Squares Kalman Filter (LS-KF), which is a method based on multilateration in the least squares sense, followed by a smoothing step, using Kalman filtering. This study reveals many advantages in favor of the DDKF which, when applied for indoor localization, provides up to 40% gain in terms of Root Mean Squares Errors (RMSE) in position estimation, as compared to the other considered methods and which has a location error that is less than 2 meters in 95% of the considered cases.

New Efficient Schemes for Adaptive Selection of the Update Time in the IMMJPDAF

The paper addresses the problem of adaptive manoeuvring targets tracking in clutter with a phased array radar. The tracking algorithm is based on the combination of the interacting multiple models (IMM) algorithm and the joint probabilistic data association filter (JPDAF), the resulting algorithm is named IMMJPDAF algorithm. Moreover, the phased array radar is a multifunction radar with the capability to select adaptively the sampling time interval; consequently, the tracking performance is improved. First, a complete comparative study between the IMMJPDAF algorithm and the multirate IMMJPDAF (MRIMMJPDAF) algorithm for tracking close manoeuvring targets with varying amounts of clutter density is presented. Then a description is made of the integration of a new fast method into the IMMJPDAF algorithm to adaptively select the next update time according to the targets motions. We call the resulting algorithm, the fast adaptive IMMJPDAF (FAIMMJPDAF) algorithm. Furthermore, an enhancement of the tracking accuracy in the FAIMMJPDAF algorithm is made by also taking into account the separation distance between targets in the selection of the next update time. The performance of the proposed algorithm, named improved fast adaptive IMMJPDAF (IFAIMMJPDAF) algorithm is assessed via Monte Carlo Simulations and compared with that of four algorithms that use an adaptive selection of the update time: FAIMMJPDAF algorithm, the adaptive IMMJPDAF that uses a modified version of the van Keuk criterion (MAIMMJPDAF), the adaptive IMMJPDAF that uses the original van Keuk method (AIMMJPDAF), and the IMMJPDAF (CIMMJPDAF) that uses a constant update time.

Nose tip localization on a three dimensional face across pose, expressions and occlusions variations in a Riemannian context

Nose tip localization is an important step for registration, preprocessing and recognition of 3D face data. In this paper, we propose a new approach for the nose tip detection that is robust to pose and expression variations and in presence of occlusions. From a rotated 3D face, we extract facial curves that are matched to a profile curve model. An optimal matching using the Riemannian geometry, based on the Elastic Shape Analysis is performed to obtain the accurate nose tip. The proposed method requires no training and can locate the nose tip in less than 6 seconds. Experiments are performed on the Bosphorus database. Quantitative analysis and comparison with the ground truth locations are provided. The results confirm that our approach achieves 97.68% with error no larger than 12 mm and 98.19% within 20 mm.

Real time Hough transform based track initiators in clutter

In both military and civilian surveillance systems such as Air Traffic Control (ATC) systems, tracking targets in clutter using radar involves dealing with a number of challenges, all related to real time decision and data fusion theories. A major challenge is to detect the real targets through the received measurements in real time and to activate the tracking process.This paper deals with real time automatic initiation of tracks in clutter. Among the proposed solutions in the literature to handle this problem, the conventional Hough transforms (HT) and the Modified Hough Transforms (MHT) have been shown to be effective as track initiator techniques in dense clutter. However, these techniques are computationally intensive. Inasmuch as the tracking of targets is a real time application, we propose in this work, to modify both the Hough transforms (HT) and the Modified Hough Transforms (MHT) so that they can work in real time in adverse environments. The resulting techniques are called Real Time HT (RTHT) and Real Time MHT (RTMHT). Monte Carlo simulations are used to assess the performance of the proposed techniques.

Moving Objects Localization by Local Regions Based Level Set: Application on Urban Traffic

In this paper, a novel method for locating multiple moving objects in a video sequences captured by a stationary camera is proposed. In order to determine the precise location of the objects in an image, a new local regions based level set model is carried out. The whole process consists of two main parts: the global detection and the fine localization. During the global detection, the presence or absence of an object in an image is determined by the Mixture of Gaussians method. For the fine localization, we propose to reformulate global energies by utilizing little squared windows centered on each point over a thin band surrounding the zero level set, hence the object contour can be reshaped into small local interior and exterior regions that lead to compute a family of adaptive local energies, which enables us to well localize the moving objects. Moreover, we propose to adapt the smoothness of the contours, and the accuracy of the objects’ perimeter of different shapes with an automatic stopping criterion. The proposed method has been tested on different real urban traffic videos, and the experiment results demonstrate that our algorithm can locate effectively and accurately the moving objects; optimize the results of the localized objects and also decrease the computations load.

Vision based system for driver drowsiness detection

Drowsiness of drivers is amongst the significant causes of road accidents. Every year, it increases the amounts of deaths and fatalities injuries globally. In this paper, a module for Advanced Driver Assistance System (ADAS) is presented to reduce the number of accidents due to drivers fatigue and hence increase the transportation safety; this system deals with automatic driver drowsiness detection based on visual information and Artificial Intelligence. We proposed an algorithm to locate, track, and analyze both the drivers face and eyes to measure PERCLOS, a scientifically supported measure of drowsiness associated with slow eye closure.

SSD and Histogram for Person Re-identification System

In this paper, we give the design and implementation of a system for person re-identification in a camera network, based on the appearance. This system seeks to construct an online database that contains the history of every person that enters the field of view of the cameras. This system is qualified to associate an identifier to each detected person, which keeps this identifier in the same camera and in other cameras even if he or she disappears and then appears again. Our system comprises a moving objects detection step that is implemented using the Mixture of Gaussians method and a proposed difference method, to improve the detection results. It also comprises a tracking step that is implemented using the sum of squared differences algorithm. The re-identification stage is realized using two steps: the intersection of tracking and detection for the temporal association, the histogram for comparison. The global system was tested on a real data set collected by three cameras. The experimental results show that our approach gives very satisfactory results.

Face recognition from unconstrained three-dimensional face images using multitask sparse representation

Abstract. We propose and evaluate a three-dimensional (3D) face recognition approach that applies the speeded up robust feature (SURF) algorithm to the depth representation of shape index map, under real-world conditions, using only a single gallery sample for each subject. First, the 3D scans are preprocessed, then SURF is applied on the shape index map to find interest points and their descriptors. Each 3D face scan is represented by keypoints descriptors, and a large dictionary is built from all the gallery descriptors. At the recognition step, descriptors of a probe face scan are sparsely represented by the dictionary. A multitask sparse representation classification is used to determine the identity of each probe face. The feasibility of the approach that uses the SURF algorithm on the shape index map for face identification/authentication is checked through an experimental investigation conducted on Bosphorus, University of Milano Bicocca, and CASIA 3D datasets. It achieves an overall rank one recognition rate of 97.75%, 80.85%, and 95.12%, respectively, on these datasets.

A new Decoupled Unbiased Converted Measurement Adaptive alpha-beta Filter developed for target tracking

Tracking a target in a Cartesian repair with polar radar measurements usually involves transformation of the measurements from a polar coordinate frame to a Cartesian coordinate one. In a large cross-range target tracking, the classical conversion method becomes inconsistent and is therefore inadequate. In this paper we propose a new decoupled tracking filter that uses Unbiased Converted Measurements within an adaptive αβ filter. The proposed filter, named the Decoupled Unbiased Converted Measurements Adaptive αβ Filter (UCMAαβF), while having a performance similar to that of the Unbiased Converted Measurements Kalman Filter (UCMKF), has the advantage to be simpler to implement for real time applications.

Adaptive FastIMM filter for tracking a maneuvering target using nonlinear measurements

A commonly encountered problem for tracking community is: target tracking in Cartesian plane while the measurements are delivered by radar in polar coordinates. In this paper we consider a maneuvering target tracking problem by using the Fast Interacting Multiple Model (FastIMM) algorithm. Based on the theoretical framework of the Debiased Converted Measurements Kalman Filter (DCMKF), we propose the use of the pseudo-static versions based on the well known αβ and αβγ filters. The performances of these nonlinear filters, have been assessed by means of Monte Carlo simulations and compared to that of the standard filter. The proposed filter is also suitable for real time implementation, which makes it a potential candidate for applications on embedded systems.