Papers by KAHLOUCHE Souhila
Centre de Développement des Technologies Avancées CDTA
Abstract: In this paper we try to develop an algorithm for visual obstacle avoidance of autonomou... more Abstract: In this paper we try to develop an algorithm for visual obstacle avoidance of autonomous mobile robot. The input of the algorithm is an image sequence grabbed by an embedded camera on the B21r robot in motion. Then, the optical flow information is extracted from the image sequence in order to be used in the navigation algorithm. The optical flow provides very important information about the robot environment, like: the obstacles disposition, the robot heading, the time to collision and the depth. The strategy consists in balancing the amount of left and right side flow to avoid obstacles, this technique allows robot navigation without any collision with obstacles. The robustness of the algorithm will be showed by some examples.
Human Activity Recognition Based on Ensemble Classifier Model
Lecture Notes in Electrical Engineering
This paper presents a new approach to image segmentation using genetic algorithm (GA) in conjunct... more This paper presents a new approach to image segmentation using genetic algorithm (GA) in conjunction with morphological operations. The proposed method consists to classify each pixel of the image in pixels belonging either to the object or to the background. It starts by a configuration of individuals randomly generated, representing possible segmentation of the image. The solutions are evaluated through an appropriate fitness function which measures the similarity of the individuals with the desired solution and the fittest ones are selected to reproduce in the next generation. Then, the population progresses to a stable representation where all the individual converges to an optimal solution which is the segmented image. We show the use of the morphological operations in the reproduction step of the GA applied on the selected individuals in order to exploit a priori image information. Tested on gray level images, the presented method has yield good results where objects are well ...
Human Activity Recognition Based on Ensemble Classifier Model
Real-Time Human Action Recognition Using Deep Learning Architecture
International Journal of Computational Intelligence and Applications
In this work, efficient human activity recognition (HAR) algorithm based on deep learning archite... more In this work, efficient human activity recognition (HAR) algorithm based on deep learning architecture is proposed to classify activities into seven different classes. In order to learn spatial and temporal features from only 3D skeleton data captured from a “Microsoft Kinect” camera, the proposed algorithm combines both convolution neural network (CNN) and long short-term memory (LSTM) architectures. This combination allows taking advantage of LSTM in modeling temporal data and of CNN in modeling spatial data. The captured skeleton sequences are used to create a specific dataset of interactive activities; these data are then transformed according to a view invariant and a symmetry criterion. To demonstrate the effectiveness of the developed algorithm, it has been tested on several public datasets and it has achieved and sometimes has overcome state-of-the-art performance. In order to verify the uncertainty of the proposed algorithm, some tools are provided and discussed to ensure i...
A social planning and navigation for tour-guide robot in human environment
2016 8th International Conference on Modelling, Identification and Control (ICMIC), 2016
The biggest challenges in modern robotics is service robots, which are able to execute many tasks... more The biggest challenges in modern robotics is service robots, which are able to execute many tasks for humans in their presence. This perspective naturally causes the problem of social navigation of mobile robots as well as human-robot interaction. In this article, we present the implementation of a navigation method on a mobile robot in indoor environment; we detail the algorithm and the implementation of human-robot interaction social rules. The principle is to define a goal for the robot, which plans a path that drives it to its goal, choosing the shortest, the smoothest, and the safety way, avoiding socially the dynamic obstacles. For this purpose, the robot uses a laser sensor for building environments maps, localization, and detection of new obstacles, and an RGB-D camera (Kinect sensor) for social avoidance.
Human pose recognition and tracking using RGB-D camera
2016 8th International Conference on Modelling, Identification and Control (ICMIC), 2016
In this work, we address the problem of human body pose recognition using RGB-D sensor, to perfor... more In this work, we address the problem of human body pose recognition using RGB-D sensor, to perform user tracking by a mobile robot. User's skeleton joints orientations are used in this approach to compute torso joint orientation, which is necessary to distinguish the four poses: Face, Back, Right of Left. In addition, a face recognition method is used to re-identify user when lost from camera field of view. Experiments were carried out to validate proposed methods, and obtained results show that targeted user following can be achieved successfully.
A genetic algorithm for geometric primitives extraction
Seventh International Symposium on Signal Processing and Its Applications, 2003. Proceedings., 2003
ABSTRACT This paper presents a geometric primitives extraction method from an image based on gene... more ABSTRACT This paper presents a geometric primitives extraction method from an image based on genetic algorithms. Our objective is to complete the classical Hough transform method which is a good extractor of segments from an image, since it isn't adapted for the detection of three or higher primitive parameters in regard of the run time and the memory space used. The main idea of the proposed method consists to consider the primitive extraction as an optimization problem in the sense that the detection of the primitive is done by the minimization of the size and the position of the considered primitive in the parameters space. In order to deal with small primitives localization errors, we have used the distance image obtained with a particular morphological operator: the chamfer distance. Some results of geometric primitives extraction for both real and synthetic images are presented.
On the Use of Optical Flow in Robot Navigation
2007 IEEE International Conference on Signal Processing and Communications, 2007
This work addresses the use of optical flow to supervise the navigation of mobile robot. The opti... more This work addresses the use of optical flow to supervise the navigation of mobile robot. The optical flow information is computed from the images sequence in order to be used in the navigation algorithm. The optical flow provides very important information about the robot environment, like: the obstacles disposition, the robot heading, the time to collision and the depth. The use of this information in autonomous robot navigation consists in balancing the amount of left and right side flow, this technique allows robot navigation without any collision with obstacles. The robustness of the developed algorithm will be showed by some examples.
A genetic algorithm for geometric primitives extraction
This paper presents a geometric primitives extraction method from an image based on genetic algor... more This paper presents a geometric primitives extraction method from an image based on genetic algorithms. Our objective is to complete the classical Hough transform method which is a good extractor of segments from an image, since it isn't adapted for the detection of three or higher primitive parameters in regard of the run time and the memory space used. The main idea of the proposed method consists to consider the primitive extraction as an optimization problem in the sense that the detection of the primitive is done by the minimization of the size and the position of the considered primitive in the parameters space. In order to deal with small primitives localization errors, we have used the distance image obtained with a particular morphological operator: the chamfer distance. Some results of geometric primitives extraction for both real and synthetic images are presented.
This paper presents a new approach to image segmentation using genetic algorithm (GA) in conjunct... more This paper presents a new approach to image segmentation using genetic algorithm (GA) in conjunction with morphological operations. The proposed method consists to classify each pixel of the image in pixels belonging either to the object or to the background. It starts by a configuration of individuals randomly generated, representing possible segmentation of the image. The solutions are evaluated through an appropriate fitness function which measures the similarity of the individuals with the desired solution and the fittest ones are selected to reproduce in the next generation. Then, the population progresses to a stable representation where all the individual converges to an optimal solution which is the segmented image. We show the use of the morphological operations in the reproduction step of the GA applied on the selected individuals in order to exploit a priori image information. Tested on gray level images, the presented method has yield good results where objects are well ...
On the Use of Optical Flow in Robot Navigation
This work addresses the use of optical flow to supervise the navigation of mobile robot. The opti... more This work addresses the use of optical flow to supervise the navigation of mobile robot. The optical flow information is computed from the images sequence in order to be used in the navigation algorithm. The optical flow provides very important information about the robot environment, like: the obstacles disposition, the robot heading, the time to collision and the depth. The use of this information in autonomous robot navigation consists in balancing the amount of left and right side flow, this technique allows robot navigation without any collision with obstacles. The robustness of the developed algorithm will be showed by some examples.
International Journal of Advanced Robotic Systems, 2007
In this paper we try to develop an algorithm for visual obstacle avoidance of autonomous mobile r... more In this paper we try to develop an algorithm for visual obstacle avoidance of autonomous mobile robot. The input of the algorithm is an image sequence grabbed by an embedded camera on the B21r robot in motion. Then, the optical flow information is extracted from the image sequence in order to be used in the navigation algorithm. The optical flow provides very important information about the robot environment, like: the obstacles disposition, the robot heading, the time to collision and the depth. The strategy consists in balancing the amount of left and right side flow to avoid obstacles, this technique allows robot navigation without any collision with obstacles. The robustness of the algorithm will be showed by some examples.
International Journal of Advanced Robotic …, 2007
In this paper we try to develop an algorithm for visual obstacle avoidance of autonomous mobile r... more In this paper we try to develop an algorithm for visual obstacle avoidance of autonomous mobile robot. The input of the algorithm is an image sequence grabbed by an embedded camera on the B21r robot in motion. Then, the optical flow information is extracted from the image sequence in order to be used in the navigation algorithm. The optical flow provides very important information about the robot environment, like: the obstacles disposition, the robot heading, the time to collision and the depth. The strategy consists in balancing the amount of left and right side flow to avoid obstacles, this technique allows robot navigation without any collision with obstacles. The robustness of the algorithm will be showed by some examples.
Uploads
Papers by KAHLOUCHE Souhila