Environmental monitoring and emergency response projects in urban-industrial areas increasingly rely on efficient collaboration between experts in control rooms and at incident locations, and citizens who live or work in the area. In the video accompanying this abstract we present a system that uses distributed sensor technology, Bayesian decision tools, and advanced map-based interfaces to facilitate collaboration between environmental experts and the public for environmental monitoring and early detection of chemical incidents.

FIT3D is a Toolbox built for Matlab that aims at unifying and distributing a set of tools that will allow the researcher to obtain a complete 3D model from a set of calibrated images. In this paper we motivate and present the structure of the toolbox in a tutorial and example based approach. Given its exibility and scope we believe that FIT3D represents an exciting opportunity for researchers that want to develop or test one particular method with real data without the need for extensive additional programming.

Changing image intensities causes problems for many computer vision applications operating in unconstrained environments. We propose generally applicable algorithms to correct for global differences in intensity between images recorded with a static or slowly moving camera, regardless of the cause of intensity variation. The proposed intensity correction is based on intensity-quotient estimation. Various intensity estimation methods are compared. Usability is evaluated with background classification as example application. For this application we introduced the PIPE error measure evaluating performance and robustness to parameter setting. Our approach retains local intensity information, is always operational and can cope with fast changes in intensity. We show that for intensity estimation, robustness to outliers is essential for dynamic scenes. For image sequences with changing intensity, the best performing algorithm (MofQ) improves foreground-background classification results up to a factor two to four on real data.

This contribution addresses the problem of bias in stereo based motion estimation. Using a biased estimator within a visual-odometry system will cause significant drift on large trajectories. This drift is often minimized by exploiting auxiliary sensors, (semi-)global optimization or loop-closing. In this paper it is shown that bias in the motion estimates can be caused by incorrect modeling of the uncertainties in landmark locations. Furthermore, there exists a relation between the bias, the true motion and the distribution of landmarks in space. Guided by these observations, a novel bias reduction technique has been developed. The core of the proposed method is computing the difference between motion estimates obtained using dissimilar heteroscedastic landmark uncertainty models. This approach is accurate, efficient and does not rely on auxiliary sensors, (semi-)global optimization or loop-closing. To show the real-world applicability of the proposed method, it has been tested on several data-sets including a challenging 5 km urban trajectory. The gain in performance is clearly noticeable.

During simultaneous localization and mapping, geometrical constraints are established between map features. These constraints, introduced through measurements and motion prediction, produce a bending effect in the event of closing a large loop. In this paper we present a discussion of the bending problem for trajectory based representations. Furthermore, we propose a generic approach to reduce the bending effect that exploits common geometrical constraints in human-made environments through the use of a heading based curvature measurement. We show by means of experimental results that our approach increases significantly both the global and local map accuracy.

In this paper, we present ongoing research concerning the interaction between users and autonomous mobile agents in the environmental monitoring domain. The overarching project, DIADEM, deals with developing a system that detects potentially hazardous situations in populated industrial areas using input from both a distributed sensor network and humans through mobile devices. We propose a model of interaction with the gas detection system where concerned citizens communicate with a mobile agent to inform the gas monitoring system about unusual smells via their mobile phones. Next, we present a preliminary user requirements analysis based on 40 phone calls from members of the public to an environmental monitoring agency. Finally, we introduce measures to study the delicate long-term social relationship between users and the gas monitoring system.

The capability to follow a person in a domestic environment is an important prerequisite for a robot companion. In this paper, a tracking algorithm is presented that makes it possible to follow a person using a small robot. This algorithm can track a person while moving around, regardless of the sometimes erratic movements of the legged robot. Robust performance is obtained by fusion of two algorithms, one based on salient features and one on color histograms. Re-initializing object histograms enables the system to track a person even when the illumination in the environment changes. By being able to re-initialize the system on run time using background subtraction, the system gains an extra level of robustness.

A novel robust visual-odometry technique, called EM-SE(3) is presented and compared against using the random sample consensus (RANSAC) for ego-motion estimation. In this contribution, stereo-vision is used to generate a number of minimal-set motion hypothesis. By using EM-SE(3), which involves expectation maximization on a local linearization of the rigid-body motion group SE(3), a distinction can be made between inlier and outlier motion hypothesis. At the same time a robust mean motion as well as its associated uncertainty can be computed on the selected inlier motion hypothesis. The data-sets used for evaluation consist of synthetic and large real-world urban scenes, including several independently moving objects. Using these data-sets, it will be shown that EM-SE(3) is both more accurate and more efficient than RANSAC.

We have developed a stereo vision based obstacle detection (OD) system that can be used to detect obstacles in off-road terrain during both day and night conditions. In order to acquire enough depth estimates for reliable OD during low visibility conditions, we propose a stereo disparity (depth) estimation approach that uses fine-to-coarse selection in a stereo image pyramid. This fine-to-coarse selection is based on a novel disparity validity metric that reflects the estimation reliability. Dense three-dimensional terrain data is reconstructed from the estimated stereo disparities. In our OD methods, several geometric properties, such as the terrain slope, are inspected to distinguish between obstacles and drivable terrain. This is achieved in a robust and efficient manner by considering the inherent uncertainty in stereo depth and using a hysteresis threshold. A large and varied collection of day- and nighttime images has been used to evaluate the performance of our system. The results show that our methods can reliably detect different types of obstacles in all tested conditions.

A novel algorithm to associate and trilaterate detections from multiple distributed radars is presented. The algorithm provides for flexible track state representations. The coordinate system of a track is switched from the measurement coordinates (range-Doppler) to cartesian coordinates when a detection from another sensor is associated to the track. In the case of multiple targets and false alarms we run into the complication of multiple association possibilities. These can be resolved by using a multi-hypothesis algorithm. In general, correctly formed tracks will have more likely associations. Therefore, hypotheses describing these tracks will be favored. Simulations with one or two targets and different false alarm rates show the need to preserve multiple hypotheses of the world state. Tracking performance for various false alarms rates is evaluated

The topic of this paper is the integration of expectation maximization (EM) background modeling and template matching using color histograms as templates to improve person tracking for surveillance applications. The tracked objects are humans, which are not rigid bodies. As such shape deformations of the objects must be allowed. For each frame, the decision has to be made which pixels belong to an object, and which do not. The integration of detection and tracking is done using a likelihood-based framework. This way the classification of pixels between background and object can be based on comparing likelihoods rather then separate thresholds. A demonstration of the proposed algorithm is given.

In this paper we consider a single camera attached to a moving vehicle. Based on the observations, we estimate the time-to-contact of vehicles around us. We propose three different approaches for the estimation of this parameter. The measurement input for these approaches respectively consists of: the horizontal position in the image plane of a point observed on a vehicle, the vertical position and the observed vehicle width. This measurement data can robustly be obtained in practice. It is possible to automatically select the approach that provides the most accurate result for a given situation. The influence of disturbances in the observations in practice is suppressed by means of temporal filtering. Due to the similar structure of the three approaches, the same temporal filtering technique can be applied. The practical value of the algorithms is illustrated by means of real video data.

We discuss the use of virtual reality (VR) techniques for interaction between humans and a multi-agent system in the context of RoboCup. The goal of RoboCup is to let teams of cooperating autonomous agents play a soccer match, using either robots or simulated players. We use RoboCup to study distributed collaborative applications, which allow multiple users at different geographic locations to cooperate, by interacting in real time through a shared simulation program. Our objective is to construct a VR environment in which humans at different locations can play along with a running RoboCup simulation in a natural way. The simulation system consists of the Soccer Server and a set of processes modeling the players. The server keeps track of the state of the game: provides the players with information on the game, and enforces the rules. The players request state information and autonomously calculate a behavior, sending the server commands that consist of accelerations, turns and kicks. The server discretizes time into slots, only one command is executed per time slot. We have developed a 3D visualization system that allows a user in a CAVE to interact with the soccer simulation software.

If a camera moves on a straight line, the optic flow field is a diverging vector field, of which the singularity is called "focus of expansion" (FOE). An object which is seen in this FOE is located on the future path of the camera. If the camera is also rotating, the future path is no longer a point in the image domain, but a line. All objects which are on the future path (and thus will cause collisions) are projected on this line. However, the reverse is not necessary true - not all points on the line are collision points. In this paper we derive how the optic flow can be used to compute which points in the image are projections of collision points.

Rendering objects transparently gives additional insight in complex and overlapping structures. However, traditional techniques for the rendering of transparent objects such as alpha blending are not very well suited for the rendering of multiple transparent objects in dynamic scenes. Screen door transparency is a technique to render transparent objects in a simple and efficient way: no sorting is required and intersecting polygons can be handled without further preprocessing. With this technique, polygons are rendered through a mask: only where the mask is present, pixels are set. However, artifacts such as incorrect opacities and distracting patterns can easily occur if the masks are not carefully designed. The requirements on the masks are considered. Next, three algorithms are presented for the generation of pixel masks. One algorithm is designed for the creation of small (e.g. 4/spl times/4) masks. The other two algorithms can be used for the creation of larger masks (e.g. 32/spl times/32). For each of these algorithms, results are presented and discussed.

The robot navigation task presented in this paper is to drive through the center of a corridor, based on a sequence of images from an on-board camera. Our measurements of the system state the distance to the wall and orientation of the wall, are derived from the optic flow. Whereas the structure of the environment is usually computed from the spatial derivatives of the optic flow, we used the structure contained in the temporal derivatives of the optic flow to compare the environment structure and hence the system state. The algorithm is used to control a 'remote brain' robot and results on the accuracy of the state estimates are presented.

A method is presented to correct an approximate polygonal sketch of an image object boundary by adjusting the location of each corner point in the sketch. The method is suitable for the segmentation of unknown images using interactive techniques, and for locating object boundaries in model-based segmentation. For each point, p, at which two lines meet in an open or closed polygonal sketch, a corner segmentation model is derived based on the angle, orientation and scale of the corner defined, and the image function f(x, y) in the immediate region about p. A corner template is then constructed and matched in a small neighborhood about p, thereby providing a corrected polygonal sketch. The segmentation model is correct in 95% of the cases, and the corner is accurately located.>

A method is presented to learn 3D grasping of objects with unknown dimensions using a monocular eye-in-hand manipulator. From a sequence of images a motion profile is generated to approach the object of unknown size. It is shown that monocular visual information suffices to control the deceleration of the robot manipulator. A strategy for generating learning samples is presented, and simulation results demonstrate the effectiveness of the method. I. Introduction Sensor based robot control systems can overcome many of the difficulties which are caused by unknown or uncertain models of the environment. Also, conventional sensor based control systems require explicit knowledge of the kinematics and dynamics of the robot arm and a careful calibration of the sensor system. We are interested in self-learning and adaptive systems, where an implicit model of the arm and sensor system is learned from the behaviour of the robot. Neurocomputational techniques have been successfully applied in t...

In this paper, a model-based scheme for recognizing hand-drawn symbols in schematic diagrams using attributed graph (AG) matching in the absence of any information concerning their pose (translation, rotation and scale) is described. The process of AG matching proceeds as follows. First, an observed AG (AG O ) is constructed from single-pixel-width line-representation of an observed symbol. Second, the pose of the AG O is estimated in terms of translation, rotation and scale with respect to the model AGs (AG M s ). The search space is effectively pruned by introducing the concept of control vertex and applying geometrical constraints in an early stage. In this step, a small number of candidate AG M s are selected. Third, correspondences between components of the observed AG after normalization (AG O N ) and those of the AG M s are found for the given poses. Fourth, distance measures between the AG O N and the AG M s are calculated, based upon the correspondences. Finally, the AG O N is classified as the AG M with the minimum distance. Experimental results for hand-drawn symbols with and without templates show that using AG matching is very efficient and successful for translation-, rotation- and scale-invariant recognition of hand-drawn symbols in schematic diagrams.

A basic minimum-square-error (MSE) transform for point pattern matching and a fast computational method for its implementation are described. Currently, the method is being used successfully in a number of research projects. A computational analysis shows that it is faster than that of F.C.A. Groen et al.>

An algorithm is proposed for skeletonization of 3-D images. The criterion to preserve connectivity is given in two versions: global and local. The latter allows local decisions in the erosion process. A table of the decisions for all possible configurations is given in this paper. The algorithm using this table can be directly implemented both on general purpose computers and on dedicated machinery.