Baerbel Mertsching

Evaluation of Visual Attention Models for Robots

This paper presents a new approach for providing visual attention on robot vision systems. Compared to other approaches our method is very fast as it processes regions rather than individual pixels. The proposed method first builds a list of regions by applying a shade and shadow tolerant segmentation step. The features of these regions are computed using their convex hulls in order to simplify and accelerate the processing. Feature values are stored within the records of respective regions instead of constructing a master map of attention. Then an algorithmic method is applied for finding the focus of attention in contrast to mathematical approaches used by existing models. Experiments conducted on simulated and real image data have not only demonstrated the validity of the proposed approach but have also led to the establishment of a comprehensive robotic vision system.

TV-L1 Optical Flow Estimation with Image Details Recovering Based on Modified Census Transform

This paper proposes an improved optical flow estimation approach based on the total variational L1 minimization technique with weighted median filter. Furthermore, recovering image details using modified census transform algorithm improves the overall accuracy of estimating large scale displacements optical flow. On the other hand, the use of the Taylor expansion approximation in most of the optical flow approaches limits the ability to estimate movement of fast objects. Hence, a coarse-to-fine scheme is used to overcome such a problem of the cost of losing small details in the interpolation process where initial values are propagated from the coarse level to the fine one. The proposed algorithm improves the accuracy of the estimation process by integrating the correspondence results of the modified census transform into the coarse-to-fine module in order to recover the lost details. The outcome of the proposed approach yields state-of-the-art results on the Middlebury optical flow evaluations.

Towards a Generic Control Architecture of Rescue Robot Systems

This paper proposes a generic control architecture for robot systems performing search and rescue operations. The multimodal attention is introduced as an important factor in the design and implementation of control architecture for robot systems. We also considered a reliable learning ability of robot system about an explored environment as another important aspect to increase the robustness of decision processes. These decision processes include safe navigation through cluttered environment, search for victims in complex scenarios and obtaining a map of targeted environment.

Fast Saliency-Based Motion Segmentation Algorithm for an Active Vision System

In this work, we propose a saliency-based approach for estimating and segmenting 3D motions of multiple moving objects represented by 2D motion vector fields (MVF). In order to overcome typical problems in autonomous mobile robotic vision such as noise, occlusions, and inhibition of the ego-motion defects of a moving camera head, a classification module has been implemented to define the global motion of the mounted camera. The proposed method achieves valuable reduction in computational time by applying a guided control module which limits the segmentation output to a flexible predefined threshold value. The computational enhancement is very important since the output of the motion segmentation approach is implemented in an active vision system.

Bearing only SLAM: a new particle filter based approach

In this paper a new method to address bearing-only SLAM using particle filters is proposed. We use a set of line pieces to model the uncertainties of landmarks and derive a proper formulation to modify the joint robot and landmark assumptions in the context of a particle filter approach.

Bearing only mobile robots' localization: Observability and formulation using SIS particle filters

In this paper the localization problem of mobile robots using bearing only measurement which is mostly used for mono vision robot, is investigated. Firstly the observability of the problem is discussed and then the position estimation of the robot using SIS particle filter is formulated. The performance of this method is discussed in comparison to the case that only range measurement is used and the case that both measurements are utilized.

Classification using scale and rotation tolerant shape signatures from convex hulls

A novel real-time approach for classification or identification of objects is presented here that is suitable for visual attention system of mobile robots. The proposed method constructs convex hulls for regions found in an image using a new external scanning technique. Then a cleaning step produces refined polygons that are in turn used for extracting shape signatures for the regions. In the training phase, shape signatures are collected from test data to find a mean signature for a particular object. A small database is created for all objects related to a specific context in which classification is to be performed. In classifying phase, signatures obtained from objects found in a given image are compared with those present in the database for identification. Nearest signature from the database to a given one is taken as identity of the later. Results have proved efficiency and accuracy of this method.

Real time stereo-based biologically inspired 3D motion classifier cells

In this paper, we present a real time biologically motivated 3D motion classifier cells integrating the depth information generated from a stereo input implemented in an active vision system. The proposed approach is accurately able to detect and estimate multiple interfered 3D complex motions under the absence of predefined spatial coherence. Moreover, the system has ability to examine the response of input 3D motion vector fields to a certain 3D motion patterns (3D motion classifier cells) such as motion in the Z direction representing movements towards the system, which is very important to overcome typical problem in autonomous mobile robotic vision such as collision detection and inhibition of the ego-motion defects of a moving camera head. The output of the algorithm is part in a multi-object segmentation approach implemented in an active vision system.