prof. Ing. Tomáš Svoboda, Ph.D.

We propose a data-driven method for simulating lidar sensors. The method reads computer-generated data, and (i) extracts geometrically simulated lidar point clouds and (ii) predicts the strength of the lidar response – lidar intensities. Qualitative valuation of the proposed pipeline demonstrates the ability to predict systematic ailures such as no/low responses on polished parts of car bodyworks and windows, for strong responses on reflective surfaces such as traffic signs and license/registration plates. We also experimentally show that enhancing the training set by such simulated data improves the segmentation accuracy on the real dataset with limited access to real data. Implementation of the resulting lidar simulator for the GTA V game, as well as the accompanying large dataset, is made publicly available.

We address the problem of self-supervised learning for predicting the shape of supporting terrain (i.e. the terrain which will provide rigid support for the robot during its traversal) from sparse input measurements. The learning method exploits two types of ground-truth labels: dense 2.5D maps and robot poses, both estimated by a usual SLAM procedure from offline recorded measurements. We show that robot poses are required because straightforward supervised learning from the 3D maps only suffers from: (i) exaggerated height of the supporting terrain caused by terrain flexibility (vegetation, shallow water, snow or sand) and (ii) missing or noisy measurements caused by high spectral absorbance or non-Lambertian reflectance of the measured surface. We address the learning from robot poses by introducing a novel KKT-loss, which emerges as the distance from necessary Karush-Kuhn-Tucker conditions for constrained local optima of a simplified first-principle model of the robot-terrain interaction. We experimentally verify that the proposed weakly supervised learning from ground-truth robot poses boosts the accuracy of predicted support heightmaps and increases the accuracy of estimated robot poses. All experiments are conducted on a dataset captured by a real platform. Both the dataset and codes which replicates experiments in the paper are made publicly available as a part of the submission.

The Subterranean Challenge (SubT) is a contest organised by the Defense Advanced Research Projects Agency (DARPA). The contest reflects the requirement of increasing safety and efficiency of underground search-and-rescue missions. In the SubT challenge, teams of mobile robots have to detect, localise and report positions of specific objects in an underground environment. This paper provides a description of the multi-robot heterogeneous exploration system of our CTU-CRAS team, which scored third place in the Tunnel Circuit round, surpassing the performance of all other non-DARPA-funded competitors. In addition to the description of the platforms, algorithms and strategies used, we also discuss the lessons-learned by participating at such contest.

We consider the problem of active victim segmentation during a search‐and‐rescue (SAR) exploration mission. The robot is equipped with a multimodal sensor suite consisting of a camera, lidar, and pan‐tilt thermal sensor. The robot enters an unknown scene, builds a 3D model incrementally, and the proposed method simultaneously (a) segments the victims from incomplete multimodal measurements and (b) controls the motion of the thermal camera. Both of these tasks are difficult due to the lack of natural training data and the limited number of real‐world trials. In particular, we overcome the absence of training data for the segmentation task by employing a manually designed generative model, which provides a semisynthetic training data set. The limited number of real‐world trials is tackled by self‐supervised initialization and optimization‐based guiding of the motion control learning. In addition to that, we provide a quantitative evaluation of the proposed method on several real testing scenarios using the real SAR robot. Finally, we also provide a data set which will allow for further development of algorithms on the real data.

Mobile-tracked robots are suitable for traversing rough terrain. However, standard exteroceptive localization methods (visual or laser SLAM) may be unreliable due to smoke, dust, fog, or insufficient lighting in harsh conditions of urban search and rescue missions. During extensive end-user evaluations in real-world conditions of such scenarios, we have observed that the accuracy of dead-reckoning localization suffers while traversing vertical obstacles. We propose to combine explicit modeling of robot kinematics and data-driven approach based on machine learning. The proposed method is experimentally verified indoors and outdoors traversing various obstacles. Indoors, a reference position has been recorded as well to assess the accuracy of our solution. The experimental dataset is released to the public to help the robotics community.

This paper presents a complete pipeline for learning continuous motion control policies for a mobile robot when only a non-differentiable physics simulator of robot-terrain interactions is available. The multi-modal state estimation of the robot is also complex and difficult to simulate, so we simultaneously learn a generative model which refines simulator outputs. We propose a coarse-to-fine learning paradigm, where the coarse motion planning is alternated with guided learning and policy transfer to the real robot. The policy is jointly optimized with the generative model. We evaluate the method on a real-world platform.

Mobile robots with complex morphology are essential for traversing rough terrains in Urban Search & Rescue missions. Since teleoperation of the complex morphology causes high cognitive load of the operator, the morphology is controlled autonomously. The autonomous control measures the robot state and surrounding terrain which is usually only partially observable, and thus the data are often incomplete. We marginalize the control over the missing measurements and evaluate an explicit safety condition. If the safety condition is violated, tactile terrain exploration by the body-mounted robotic arm gathers the missing data.

This paper presents a novel technique that allows for both computationally fast and sufficiently plausible simulation of vehicles with non-deformable tracks. The method is based on an effect we have called Contact Surface Motion. A comparison with several other methods for simulation of tracked vehicle dynamics is presented with the aim to evaluate methods that are available off-the-shelf or with minimum effort in general-purpose robotics simulators. The proposed method is implemented as a plugin for the open-source physics-based simulator Gazebo using the Open Dynamics Engine.

We propose an active 3D mapping method for depth sen- sors, which allow individual control of depth-measuring rays, such as the newly emerging solid-state lidars. The method simultaneously (i) learns to reconstruct a dense 3D occupancy map from sparse depth measurements, and (ii) optimizes the reactive control of depth-measuring rays. To make the first step towards the online control optimization, we propose a fast prioritized greedy algorithm, which needs to update its cost function in only a small fraction of possible rays. The approximation ratio of the greedy algorithm is derived. An experimental evaluation on the subset of the KITTI dataset demonstrates significant improvement in the 3D map accuracy when learning-to-reconstruct from sparse measurements is coupled with the optimization of depth measuring rays.

Registration of laser scans, or point clouds in general, is a crucial step of localization and mapping with mobile robots or in object modeling pipelines. A coarse alignment of the point clouds is generally needed before applying local methods such as the Iterative Closest Point (ICP) algorithm. We propose a feature-based approach to point cloud registration and evaluate the proposed method and its individual components on challenging real-world datasets. For a moderate overlap between the laser scans, the method provides a superior registration accuracy compared to state-of-the-art methods including Generalized ICP, 3D Normal-Distribution Transform, Fast Point-Feature Histograms, and 4-Points Congruent Sets. Compared to the surface normals, the points as the underlying features yield higher performance in both keypoint detection and establishing local reference frames. Moreover, sign disambiguation of the basis vectors proves to be an important aspect in creating repeatable local reference frames. A novel method for sign disambiguation is proposed which yields highly repeatable reference frames.

Policy Gradient methods require many real-world trials. Some of the trials may endanger the robot system and cause its rapid wear. Therefore, a safe or at least gentle-to-wear exploration is a desired property. We incorporate bounds on the probability of unwanted trials into the recent Contextual Relative Entropy Policy Search method. The proposed algorithm is evaluated on the task of autonomous flipper control for a real Search and Rescue rover platform.

Localization of mobile robots is still an important topic, especially in case of dynamically changing, complex environments such as in Urban Search & Rescue (USAR). In this paper we aim for improving the reliability and precision of localization of our multimodal data fusion algorithm. Multimodal data fusion requires resolving several issues such as significantly different sampling frequencies of the individual modalities. We compare our proposed solution with the well-proven and popular Rauch–Tung–Striebel smoother for the Extended Kalman filter. Furthermore, we improve the precision of our data fusion by incorporating scale estimation for the visual modality.

In this paper we demonstrate a combined hardware and software solution that enhances sensor suite and perception capabilities of a mobile robot intended for real Urban Search & Rescue missions. A common fail-case, when exploring unknown environment of a disaster site, is the outage o r deterioration of exteroceptive sensory measurements that the robot heavily relies on especially for localization and navigation purposes. Deprivation of visual and laser modalities caused by dense smoke motivated us to develop a novel solution comprised of force sensor arrays embedded into tracks of our platform. Furthermore, we also exploit a robotic arm for active perception in cases when the prediction based on force sensors is too uncertain. Beside the integration of hardware, we also propose a framework exploiting Gaussian proces ses followed by Gibb's sampling to process raw sensor measurements and provide probabilistic interpretation of the underlying terrain pro file. In the final, the profile is perceived by proprioceptive means only and successfully substitutes for the lack of exteroceptive measurements in the close vicinity of the robot, when traversing unknown and unseen obstacles. We evaluated our solution on real world terrains.

Urban search and rescue (USAR) missions for mobile robots require reliable state estimation systems resilient to conditions given by the dynamically changing environment. We design and evaluate a data fusion system for localization of a mobile skid-steer robot intended for USAR missions. We exploit a rich sensor suite including both proprioceptive (inertial measurement unit and tracks odometry) and exteroceptive sensors (omnidirectional camera and rotating laser rangefinder). To cope with the specificities of each sensing modality (such as significantly differing sampling frequencies), we introduce a novel fusion scheme based on an extended Kalman filter for six degree of freedom orientation and position estimation. We demonstrate the performance on field tests of more than 4.4 km driven under standard USAR conditions. Part of our datasets include ground truth positioning, indoor with a Vicon motion capture system and outdoor with a Leica theodolite tracker. The overall median accuracy of localization-achieved by combining all four modalities-was 1.2% and 1.4% of the total distance traveled for indoor and outdoor environments, respectively. To identify the true limits of the proposed data fusion, we propose and employ a novel experimental evaluation procedure based on failure case scenarios. In this way, we address the common issues such as slippage, reduced camera field of view, and limited laser rangefinder range, together with moving obstacles spoiling the metric map. We believe such a characterization of the failure cases is a first step toward identifying the behavior of state estimation under such conditions. We release all our datasets to the robotics community for possible benchmarking.

In USAR (Urban Search and Rescue) missions, robots are often required to operate in an unknown environment and with imprecise data coming from their sensors. However, it is highly desired that the robots only act in a safe manner and do not perform actions that could probably make damage to them. To train some tasks with the robot, we utilize reinforcement learning (RL). This machine learning method however requires the robot to perform actions leading to unknown states, which may be dangerous. We develop a framework for training a safety function which constrains possible actions to a subset of really safe actions. Our approach utilizes two basic concepts. First, a "core" of the safety function is given by a cautious simulator and possibly also by manually given examples. Second, a classifier training phase is performed (using Neyman-Pearson SVMs), which extends the safety function to the states where the simulator fails to recognize safe states.

Abstract This paper describes the project TRADR: Long- Term Human-Robot Teaming for Robot Assisted Disaster Response. Experience shows that any incident serious enough to require robot involvement will most likely involve a sequence of sorties over several hours, days and even months. TRADR focuses on the challenges that thus arise for the persistence of environment models, multi- robot action models, and human-robot teaming, in order to allow incremental capability improvement over the dura- tion of a mission. TRADR applies a user centric design approach to disaster response robotics, with use cases involving the response to a medium to large scale industrial accident by teams consisting of human rescuers and several robots (both ground and airborne). This paper describes the fundamentals of the project: the motivation, objectives and approach in contrast to related work.

This paper describes our experience in designing, developing and deploying systems for supporting human-robot teams during disaster response. It is based on R&D performed in the EU-funded project NIFTi. NIFTi aimed at building intelligent, collaborative robots that could work together with humans in exploring a disaster site, to make a situational assessment. To achieve this aim, NIFTi addressed key scientific design aspects in building up situation awareness in a human-robot team, developing systems using a user-centric methodology involving end users throughout the entire R&D cycle, and regularly deploying implemented systems under real-life circumstances for experimentation and testing. This has yielded substantial scientific advances in the state-of-the-art in robot mapping, robot autonomy for operating in harsh terrain, collaborative planning, and human-robot interaction. NIFTi deployed its system in actual disaster response activities in Northern Italy, in July 2012, aiding in structure damage assessment.

The paper describes experience with applying a user-centric design methodology in developing systems for human-robot teaming in Urban Search & Rescue. A human-robot team consists of several semi-autonomous robots (rovers/UGVs, microcopter/ UAVs), several humans at an off-site command post (mission commander, UGV operators) and one on-site human (UAV operator). This system has been developed in close cooperation with several rescue organizations, and has been deployed in a real-life tunnel accident use case. The human-robot team jointly explores an accident site, communicating using a multi-modal team interface, and spoken dialogue. The paper describes the development of this complex socio-technical system per se, as well as recent experience in evaluating the performance of this system.

This paper shows that the successively evaluated features used in a sliding window detection process to decide about object presence/absence also contain knowledge about object deformation. We exploit these detection features to estimate the object deformation. Estimated deformation is then immediately applied to not yet evaluated features to align them with the observed image data. In our approach, the alignment estimators are jointly learned with the detector. The joint process allows for the learning of each detection stage from less deformed training samples than in the previous stage. For the alignment estimation we propose regressors that approximate non-linear regression functions and compute the alignment parameters extremely fast.

We overview different approaches to safety in (semi)autonomous robotics. Part icularly, we focus on how to achieve safe behavior of a robot if it is requested to perform ex ploration of unknown states. Presented methods are studied from the viewpoint of reinforcement learning, a partially-supervised machine learning method. To collect training data for this a lgorithm, the robot is required to freely explore the state space - which can lead to possibly dangerous situations. The role of safe exploration is to provide a framework allowing explora tion while preserving safety. The examined methods range from simple algorithms to sophisticat ed methods based on previous experience or state prediction. Our overview also addresses the i ssues of how to define safety in the real-world applications (apparently absolute safety is un achievable in the continuous and random real world). In the conclusion we also suggest several ways that are worth researching more thoroughly.

We exploit image features multiple times in order to make sequential decision process faster and better performing. In the decision process features providing knowledge about the object presence or absence in a given detection window are successively evaluated. We show that these features also provide information about object position within the evaluated window. The classification process is sequentially interleaved with estimating the correct position. The position estimate is used for steering the features yet to be evaluated. This locally interleaved sequential alignment (LISA) allows to run an object detector on sparser grid which speeds up the process. The position alignment is jointly learned with the detector. We achieve a better detection rate since the method allows for training the detector on perfectly aligned image samples. For estimation of the alignment we propose a learnable regressor that approximates a non-linear regression function and runs in ne2076-1465gligible time.

This paper addresses object detection and tracking in high-resolution omnidirectional images. The foreseen application is a visual subsystem of a rescue robot equipped with an omnidirectional camera, which demands real-time efficiency and robustness against changing viewpoint. Object detectors typically do not guarantee specific frame rate. The detection time may vastly depend on a scene complexity and image resolution. The adapted tracker can often help to overcome the situation, where the appearance of the object is far from the training set. On the other hand, once a tracker is lost, it almost never finds the object again. We propose a combined solution where a very efficient tracker (based on sequential linear predictors) incrementally accommodates varying appearance and speeds up the whole process. Next we propose to incrementally update the detector with examples collected by the tracker. We experimentally show that the performance of the combined algorithm, measured by a ratio between false positives and false negatives, outperforms both individual algorithms. The tracker allows to run the expensive detector only sparsely enabling the combined solution to run in real-time on 12 MPx images from a high resolution omnidirectional camera (Ladybug3).

This paper presents a method for feature-based 3D object recognition in cluttered scenes. It deals with the problem of non-uniform sampling density which is inherent in typical range sensing methods. We suggest a method operating on polygonal meshes which overcomes the problem by exploiting surface area in both establishing local frames and creating feature descriptors. The method is able to recognize even highly occluded objects and outperforms state of the art in terms of recognition rate on a standard publicly available dataset.

We present an original integration of high level planning and execution with incoming perceptual information from vision, SLAM, topological map segmentation and dialogue. The task of the robot system, implementing the integrated model, is to explore unknown areas and report detected objects to an operator, by speaking loudly. The knowledge base of the planner maintains a graph-based representation of the metric map that is dynamically constructed via an unsupervised topological segmentation method, and augmented with information about the type and position of detected objects, within the map, such as cars or containers. According to this knowledge the cognitive robot can infer strategies in so generating parametric plans that are instantiated from the perceptual processes. Finally, a model-based approach for the execution and control of the robot system is proposed.

Linear predictors (LPs) are being used for tracking because of their computational efficiency which is better than steepest descent methods (e.g. Lucas-Kanade). The only disadvantage of LPs is the necessary learning phase which hinders the predictors applicability as a general patch tracker. We address this limitation and propose to learn a bank of LPs off-line and develop an on-line detector which selects image regions that could be tracked by some predictor from the bank. The proposed detector differs significantly from the usual solutions that attempt to find the closest match between a candidate patch and a database of exemplars. We construct the detector directly from the learned linear predictor. The detector positively detects the learned patches, but also many other image patches, which were not used in LP learning phase. This means, that the LP is able to track also previously unseen image patches, the appearances of which are often significantly diverse from the patches used.

This paper addresses object detection and tracking in high-resolution omnidirectional images. The foreseen application is a visual subsystem of a rescue robot equipped with an omnidirectional camera, which demands real time efficiency and robustness against changing viewpoint. Object detectors typically do not guarantee specific frame rate. The detection time may vastly depend on a scene complexity and image resolution. The adapted tracker can often help to overcome the situation, where the appearance of the object is far from the training set. On the other hand, once a tracker is lost, it almost never finds the object again. We propose a combined solution where a very efficient tracker (based on sequential linear predictors) incrementally accommodates varying appearance and speeds up the whole process. We experimentally show that the performance of the combined algorithm, measured by a ratio between false positives and false negatives, outperforms both individual algorithms.

We improve an object detector based on cascade of classifiers by a local alignment of the sliding window. The detector needs to operate on a relatively sparse grid in order to achieve a real time performance on high-resolution images. The proposed local alignment in the middle of the cascade improves its recognition performance whilst retaining the necessary speed. We show that the moment of the alignment matters and discuss the performance in terms of false negatives and false positives. The proposed method is tested on a car detection problem.

Loss-of-track detection (tracking validation) and automatic tracker adaptation to new object appearances are attractive topics in computer vision. We apply very efficient learnable sequential predictors in order to address both issues. Validation is done by clustering of the sequential predictor responses. No aditional object model for validation is needed. The paper also proposes an incremental learning procedure that accommodates changing object appearance, which mainly improves the recall of the tracker/detector. Exemplars for the incremental learning are collected automatically, no user interaction is required. The aditional training examples are selected automatically using the tracker stability computed for each potential aditional training example. Coupled with a sparsely applied SIFT or SURF based detector the method is employed for object localization in videos. Our Matlab implementation scans videosequences up to eight times faster than the actual frame rate. A standard-lengt

This paper summarizes results of face association experiments on real low resolution data from airport and the Labeled faces in the Wild (LFW) database. The objective of experiments is to evaluate different face alignment methods and their contribution to face association as such. The first alignment method used is Sequential Learnable Linear Predictor (SLLiP), originally developed for object tracking. The second method is well known face alignment method Active Shape Model (ASM). Both methods are compared versus face association without alignment. In case of high resolution LFW database the ASM rapidly increases the association results, on the other hand for real low resolution airport data the SLLiP method brought more improvement than ASM.

Anytime learning for the Sequence of Learned Linear Predictors (SLLiP) tracker is proposed. Learning might be time consuming for large problems, we present an anytime learning algorithm which, after a very short initialization period, provides a solution with defined precision. As SLLiP tracking requires only a fraction of the processing power of an ordinary PC, the learning can continue in a parallel background thread continuously delivering improved SLLiPs, ie. faster, with lower computational complexity, with the same pre-defined precision. The proposed approach is verified on publicly-available sequences with approximately 12000 ground truthed frames. The learning time is shown to be twenty times smaller than learning based on linear programming proposed in the paper that introduced the SLLiP tracker [TR]. Its robustness and accuracy is similar. Superiority in frame-rate and robustness with respect to the SIFT detector, Lucas-Kanade tracker and Jurie's tracker is also demonstrated.

We propose a learning approach to tracking explicitly minimizing the computational complexity of the tracking process subject to user-defined probability of failure (loss-of-lock) and precision. The tracker is formed by a Number of Sequences of Learned Linear Predictors (NoSLLiP). Robustness of NoSLLiP is achieved by modeling the object as a collection of local motion predictors --- object motion is estimated by the outlier-tolerant Ransac algorithm from local predictions. Efficiency of the NoSLLiP tracker stems from (i) the simplicity of the local predictors and (ii) from the fact that all design decisions - the number of local predictors used by the tracker, their computational complexity (ie the number of observations the prediction is based on), locations as well as the number of Ransac iterations are all subject to the optimization (learning) process. All time-consuming operations are performed during the learning stage - t.

The digital video processing becomes more and more important area of the computer vision. Between the quite developed methods for static images processing and video processing there are many clear differences, for example the lower overall image quality of the video, the higher volume of the video data and the real-time processing requirement. In this work we focus on the task of 3D tracking of the human head for the application in automated indexing of the feature-length movies. One of the most successful real-time tracking algorithms is the CONDENSATION algorithm and a well known approach to face detection is the Viola-Jones detector, based on the AdaBoost learning algorithm. We combine the two approaches and design a 3D head tracking algorithm, which is able to automatically learn the head appearance and track the full-angle head turnaround.

A new learning method for motion estimation of objects with significantly varying appearance is proposed. Varying object appearance is represented by a low dimensional space of appearance parameters. The appearance mapping and motion estimation method are optimized simultaneously. Appearance parameters are estimated by unsupervised learning. The method is experimentally verified by a tracking application on sequences which exhibit strong variable illumination, non-rigid deformations and self-occlusions.

Adaptation of a tracking procedure combined in a common way with a Kalman filter is formulated as an constrained optimization problem, where a trade-off between precision and loss-of-lock probability is explicitly taken into account. While the tracker is learned in order to minimize computational complexity during a learning stage, in a tracking stage the precision is maximized online under a constraint imposed by the loss-of-lock probability resulting in an optimal setting of the tracking procedure. We experimentally show that the proposed method converges to a steady solution in all variables. In contrast to a common Kalman filter based tracking, we achieve a significantly lower state covariance matrix. We also show, that if the covariance matrix is continuously updated, the method is able to adapt to a different situations. If a dynamic model is precise enough the tracker is allowed to spend a longer time with a fine motion estimation, however, if the motion gets saccadic, i.e. unpr

We consider a single-camera model-based 3D head tracking problem. The existing methods usually estimate only limited out of plane rotations, because the back of the head contains no good features to track. To solve this problem, we propose an adaptive tracking algorithm with a simple 3D ellipsoidal textured head model. The rotational and translational parts of the motion are estimated separately. The translation is estimated by a particle filter contour tracking. The remaining relative rotation is then estimated by a combination of feature tracker with \sc RANSAC-based outlier detection.

Comprehensive textbook about Matlab programming in image processing and computer vision

A novel object representation for tracking is proposed. The tracked object is represented as a constellation of spatially localised linear predictors which are learned on a single training image. In the learning stage, sets of pixels whose intensities allow for optimal least square predictions of the transformations are selected as a support of the linear predictor. The approach comprises three contributions: learning object specific linear predictors, explicitly dealing with the predictor precision - computational complexity trade-off and selecting a view-specific set of predictors suitable for global object motion estimate. Robustness to occlusion is achieved by RANSAC procedure. The learned tracker is very efficient, achieving frame rate generally higher than 30 frames per second despite the Matlab implementation.

A novel object representation for tracking is proposed. The tracked object is represented as a constellation of spatially localised linear predictors which are learned on a single training image. In the learning stage, sets of pixels whose intensities allow for optimal least square predictions of the transformations are selected as a support of the linear predictor. The approach comprises three contributions: learning object specific linear predictors, explicitly dealing with the predictor precision - computational complexity trade-off and selecting a view-specific set of predictors suitable for global object motion estimate. Robustness to occlusion is achieved by RANSAC procedure. The learned tracker is very efficient, achieving frame rate generally higher than 30 frames per second despite the Matlab implementation.

Virtual immersive environments or telepresence setups often consist of multiple cameras which have to be calibrated. We present a convenient method for doing this. The minimum is three cameras, but there is no upper limit. The method is fully automatic and a freely moving bright spot is the only calibration object. A set of virtual 3D points is made by waving the bright spot through the working volume. Its projections are found with sub-pixel precision and verified by a robust RANSAC analysis. The cameras do not have to see all points, only reasonable overlap between camera subgroups is necessary. Projective structures are computed via rank-4 factorization and the Euclidean stratification is done by imposing geometric constraints. This linear estimate initializes a post-processing computation of non-linear distortion which is also fully automatic. We suggest a trick on how to use a very ordinary laser pointer as the calibration object. We show that it is possible to calibrate an immers

We present a software for complete metric calibration of synchronized multicamera setups. The minimum is three cameras, but there is no upper limit. The method is fully automatic and a freely moving bright spot is the only calibration object. No camera pre-calibration is required. The software computes complete set of intrinsic camera parameters including the parameters of non-linear distortion as well as external orientation of the cameras in one common coordinate system. The software is written in Matlab, runs in non-interactive mode and produces both textual and intuitive graphical output. It is very robust and may be run with the same set of parameters in different setups. The software is free.

An optimization algorithm and statistical description of articulated body model estimation is proposed. The optimization algorithm fits the model into segmented multiview images. The input of our algorithm is a sequence of segmented images captured by several cameras and a structure of the articulated model. The output of the optimization procedure is shape and motion of the articulated model. The optimization runs over all cameras and all images in the sequence. We focus on description and optimization of probability distribution of the model parameters given segmented multiview sequence. We demonstrate the performance of the algorithm on real sequences of walking human.

We present a camera network system consisting of several modules of 2-3 low end cameras attached to one computer. It is not possible for a human to observe all the information coming from such a network simultaneously. Our system is designed to select the best viewpoint for each part of the video sequence, thus automatically creating one real-time video stream that contains the most important data. It acts as a combination of a director and a cameraman. Cinematography developed its own terminology, techniques and rules, how to make a good movie. We illustrate here some of these techniques and how they can be applied to a camera network, to solve the best viewpoint selection problem. Our system consists of only fixed cameras, but the output is not constrained to already existing views. A virtual zoom can be applied to select only a part of the view. We propose a view interpolation algorithm which makes it possible to create new intermediate views from the existing camera i

Central catadioptric cameras are cameras which combine lenses and mirrors to capture a very wide field of view with a central projection. In this paper we extend the classical epipolar geometry of perspective cameras to all central catadioptric cameras. The epipolar geometry is formulated as the geometry of corresponding rays in a three-dimensional space. Then, using the model of image formation of central catadioptric cameras, the constraint on corresponding image points is derived. It is shown that the corresponding points lie on epipolar conics. In addition, the shape of the conics for all types of central catadioptric cameras is classified. Finally, the theory is verified by experiments with real and simulated central catadioptric cameras.

This paper proposes a scheme for the reliable reconstruction of indoor scenes from few catadioptric images. A set of hand-detected correspondences are established across (not necessarily all) images. Our improved method is used for the estimation of the essential matrix from points which have appropriately normalized coordinates. Hartley's decomposition is used for estimation of motion parameters. A heuristic is suggested for selecting the point pairs which are {\em most reliable\/} for 3D reconstruction. The known mid-point method is applied for computing the 3D model of a real scene. The parameters of the catradioptric sensor are approximately known but no precise self-calibration method is performed. The experiments show that a reliable 3D reconstruction is possible even without complicated non-linear self-calibration and/or recontruction methods.

Epipolární geometrie pro panoramatické kamery.

Nepřesně kalibrovaná kamera při odhadování vlastního pohybu, propagace chyby