Search

Visual object recognition and tracking require a good resolution of the object to accurately model its appearance. In addition, tracking systems must be able to robustly recover moving target trajectory, and possibly cope with fast motion and large displacements. Wide angle static cameras capture a global view of the scene but they suffer from a lack of resolution in the case of a large distance between the objects and the sensor. On the contrary, dynamic sensors such as Pan-Tilt-Zoom (PTZ) cameras are controlled to focus on a 3D point in the scene and give access to high resolution images by adapting their zoom level. However, when a PTZ camera focuses on a target, its very limited field of view makes the tracking difficult. To overcome these limitations, hybrid sensor systems composed of a wide angle static camera and a dynamic camera can be used. Coupling these two types of sensors enables the exploitation of their complementary desired properties while limiting their respective drawbacks. Calibration is required to enable information exchange between the two sensors to produce collaborative algorithms. The calibration of our system is difficult because of changes of both intrinsic (focal length, central point, distortion) and extrinsic (position, orientation) parameters of the dynamic sensor during system exploitation. Two approaches for dynamic stereo sensor calibration are possible: • Strong calibration involves a complete modeling of the system. Intrinsic parameters of each camera and extrinsic parameters are estimated. This approach enables the projection of 3D points, expressed in the world frame, in 2D points expressed in each image frame. • Weak calibration does not target to estimate intrinsic or extrinsic parameters. The objective is only to estimate the direct relation between pixels of the different sensors. From a pixel in the first camera, which is the projection of a given 3D point, the calibration gives the projection of the same 3D point to the second camera. In this approach, the recovery of 3D point coordinates is not more difficult.

Nowadays Rolling shutter CMOS cameras are embedded on a lot of devices. This type of cameras does not have its retina exposed simultaneously but line by line. The resulting distortions affect structure from motion methods developed for global shutter, like CCD cameras. The bundle adjustment method presented in this paper deals with rolling shutter cameras. We use a projection model which considers pose and velocity and needs 6 more parameters for one view in comparison to the global shutter model. We propose a simplified model which only considers distortions due to rotational speed. We compare it to the global shutter model and the full rolling shutter one. The model does not need any condition on the inter-frame motion so it can be applied to fully independent views, even with global shutter images equivalent to a null velocity. We also propose a way to handle epipolar geometry for rolling shutter. It is shown that constraint using essential matrix becomes non linear, and we show how to use it to recover poses and speeds from matched points. Results with both synthetic and real images shows that the simplified model can be considered as a good compromise between a correct geometrical modelling of rolling shutter effects and the reduction of the number of extra parameters.

International audience; This paper presents a new real-time localization system for a mobile robot. We show that autonomous navigation is possible in outdoor situation with the use of a single camera and natural landmarks. To do that, we use a three step approach. In a learning step, the robot is manually guided on a path and a video sequence is recorded with a front looking camera. Then a structure from motion algorithm is used to build a 3D map from this learning sequence. Finally in the navigation step, the robot uses this map to compute its localization in real-time and it follows the learning path or a slightly different path if desired. The vision algorithms used for map building and localization are first detailed. Then a large part of the paper is dedicated to the experimental evaluation of the accuracy and robustness of our algorithms based on experimental data collected during two years in various environments.

International audience; Object-Oriented Programming (OOP) paradigm is one of the most common paradigm in introductory programming courses. However, novices often have difficulties to understand the basic concepts which are of a high level of abstraction. Either tangible and virtual constructive games provide the students with a more familiar way for learning programming. This paper applies a construction game metaphor approach for learning OOP concepts and C++ syntax. After introducing some tangible and virtual constructive games for learning programming, we present an experimental prototype of a new 3D virtual game for learning OOP called PrOgO as well as the results of an experiment conducted with beginner student using PrOgO.

Is it possible to calibrate a camera in the air and then use the calibration results to infer a new calibration corresponding to the embedding of the camera in another fluid (possibly water)? This problem is dealt within the paper. It is important to avoid direct underwater calibration, as it is much more inconvenient for experiments than the usual (air) calibration by human workers. Optical laws that must be considered when using underwater cameras are investigated. Both theoretical and experimental point of views are described, and it is shown that relationships can be found between results of air and water (or any other isotropic fluid in which the camera can be submerged) calibration.

This paper presents a new method for inferring 3D information using a static camera equipped with a zoom lens. The modeling algorithm does not require any explicit calibration model and the calculations involved are straightforward. This approach uses several images of accurate regular grids placed on a micrometric table, as the calibration process. The basic idea is to compute a local transformation that allows the establishment of a relationship between a distorted grid detected on the CCD matrix and the real one located in front of the camera. This relationship takes automatically into account all distortion phenomena and allows one to obtain reconstruction results much more accurate than those of previous works in the same field. A complete experiment on real data is provided and shows that it is possible to compute 3D information from a zooming image set even if data are close to the optical axis.

This paper presents a system for modeling solids of revolution from a set of images taken with a zoom-lens. Using a zoom in monocular vision has interesting optical properties. The displacement of the optical center along the optical axis, when the focal length is modified, permits one to implement a triangulation process. Modeling a solid of revolution requires the estimation of the 3D location of its revolution axis. First, the authors describe an original method, using zoom properties, to compute the 3D pose of the revolution axis. It requires the detection in each image of the object axis projection. Then, the modeling algorithm is presented. It is based on the resolution of the inverse perspective problem for points detected in the images. An experimental result of reconstruction, from a real images set is finally given. >

Pan Tilt Zoom cameras have the ability to cover wide areas with an adapted resolution. Since the logical downside of high resolution is a limited field of view, a guard tour can be used to monitor a large scene of interest. However, this greatly increases the duration between frames associated to a specific location. This constraint makes most background algorithms ineffective. In this article we propose a background subtraction algorithm suitable to cameras with very low frame rate. Its main interest consists in the resulting robustness to sudden illumination changes. The background model which describes a wide scene of interest consisting of a collection of images can thus be successfully maintained. This algorithm is compared with the state of the art and a discussion regarding its properties follows.

It is shown that it is possible to infer 3-D information from a set of images taken with a zoom lens. A precise study of the optical properties of such a lens gives two major results: The intersection between the optical axis and the image plane can be independently and accurately determined, and the pin-hole model cannot be used directly for the approximation of such a complex lens system. To explain the optical phenomena occurring during a zoom-lens focal-length change, it is shown that a thick optical model must be considered. Experimental reconstruction results for a set of real images, are given. >

Constant Guillot1 constant.guillot@cea.fr Maxime Taron1 maxime.taron@cea.fr Patrick Sayd1 patrick.sayd@cea.fr Quoc-Cuong Pham1 quoc-cuong.pham@cea.fr Christophe Tilmant2 christophe.tilmant@lasmea.univ-bpclermont.fr Jean-Marc Lavest2 J-Marc.Lavest@u-clermont1.fr 1 CEA LIST Laboratoire Vision et Ingenierie des Contenus, BP 94, Gif-sur-Yvette, F-91191 France 2 LASMEA UMR 6602, PRES Clermont Universite/CNRS, 63177 Aubiere cedex, France

The production machine evolutions impose a better control and consequently we need more accurate models. To ensure this accuracy, models have to take into account phenomena neglected previously such as elasticity and friction. This paper proposes an adapted metrological device to measure the displacement of machine points and identify its dynamic behavior. It is based on the merging of computer vision and accelerometers. This hybrid device performs dynamic measurements which impose a very short exposure time (50 mus). Despite that, the accuracy reached today is equals to 30 mum for a measured displacement equals to 300 mm. Two application examples are then described in the paper, elastic behavior identification and friction parameters identification.

Video surveillance becomes more and more extended in industry and often involves automatic calibration system to remain efficient. In this paper, a video-surveillance system that uses stationary-dynamic cameras devices is presented. The static camera is used to monitor the largest scene but with a small resolution : master-slave system. When it detects a moving object, the Pan-Tilt-Zoom (PTZ) camera is controlled to be centered on this object with a high resolution. We use a method of camera-to-camera calibration, integrating zoom calibration in order to command the angles and the zoom of the PTZ camera. This method enables to take into account the intrinsic camera parameters, the 3D scene geometry and the fact that the mechanism of inexpensive camera does not fit the classical geometrical model. Finally, the results of precision and accuracy of this algorithm are highlighted on real cases to qualify subjectively the value of theses errors for more complex surveillance applications like pattern recognition or tracking.

We propose an adaptive multi-resolution mean field annealing (MFA) algorithm for motion estimation in medical image sequences. The proposed algorithm can be easily parallelized and therefore is suitable to be implemented on dedicated hardware. The algorithm ensures a trade-off between precision and computational time, i.e. it ensures a good efficiency compared to other deterministic and stochastic algorithms. Results are reported for CT medical image sequences. The results are compared in terms of precision and computational time with those of other algorithms, such as the basic block-matching and Horn & Schunck algorithm, as well as with the mono-resolution and homogeneous multi-resolution MFA algorithm.

Video surveillance becomes more and more extended in industry and often involves automatic calibration system to remain efficient. In this paper, a video-surveillance system that uses stationary-dynamic cameras devices is presented. The static camera is used to monitor a global scene. When it detects a moving object, the Pan-Tilt-Zoom (PTZ) camera is controlled to be centered on this object. We describe a method of camera-to-camera calibration, integrating zoom calibration in order to command the angles and the zoom of the PTZ camera. This method enables to take into account the intrinsic camera parameters, the 3D scene geometry and the fact that the mechanism of inexpensive camera does not fit the classical geometrical model. Finally, some experiment results attest the accuracy of the proposed solution.

Camera calibration is a necessary preliminary step to ensure high quality measurement in a section of a flow. The bundle adjustment technique is based on redundancy resulting from multi-viewing the calibration target. In this paper we proposed a well-adapted protocol for self-calibrating a Scheimpflug model. Such a protocol allows calibration by hand positioning a laser printed target. Experimental results indicate that with only 5 views we can accurately calibrate a Scheimpflug camera in Stereoscopic Particle Image Velocimetry condition.

International audience; An original method for computing instantaneous 3D pose and velocity of fast moving objects using a single view is presented. It exploits image deformations induced by rolling shutter in CMOS image sensors. First of all, a general perspective projection model of a moving 3D point is presented. A solution for the pose and velocity recovery problem is then described. The method is based on bundle adjustment and uses point correspondences. The resulting algorithm enables to transform a CMOS low cost and low power camera into an original velocity sensor. Finally, experimental results with real data confirm the relevance of the approach.

An original concept for computing instantaneous 3D pose and 3D velocity of fast moving objects using a single view is proposed, implemented and validated. It takes advantage of the image deformations induced by rolling shutter in CMOS image sensors. First of all, after analysing the rolling shutter phenomenon, we introduce an original model of the image formation when using such a camera, based on a general model of moving rigid sets of 3D points. Using 2D-3D point correspondences, we derive two complementary methods, compensating for the rolling shutter deformations to deliver an accurate 3D pose and exploiting them to also estimate the full 3D velocity. The first solution is a general one based on non-linear optimization and bundle adjustment, usable for any object, while the second one is a closed-form linear solution valid for planar objects. The resulting algorithms enable us to transform a CMOS low cost and low power camera into an innovative and powerful velocity sensor. Finally, experimental results with real data confirm the relevance and accuracy of the approach.

This paper deals with area detection and tracking using color-based image processing. The application addressed concerns human head and hands tracking in real time. The authors proposed an original method in order to locate colored areas, for a foreground image. Foreground image is provided by a probabilistic way using Gaussian mixture models (GMM) of probability density functions. The temporal tracking is achieved by a particle filter, which is well adapted to partial occlusions and non Gaussian models.

In high speed machining, tool accelerations need to be increasingly high. Parallel kinematics machines present the highest ones, but combined with slimmer structures, the structural deformations cannot be still ignored. To solve this problem, some studies improve the dynamic model by taking into account the flexibility of the structure, but the identification of such model is still a difficulty. This paper proposes a method to obtain a direct measurement of the structural deformations under dynamic loading conditions due to the action of the actuators. The main difficulties are the measurement volume(1 m3), the deformation size(a few 0.01 mm) and the movement velocity(a few 10 m/min). The principle of the method is to merge position measurement by vision and acceleration measurement by accelerometers. Indeed these two kinds of material are complementary. Accelerometers have a high frequency of acquisition, but a double integration is necessary to obtain the deformation information. This involves an accuracy problem. On the contrary, vision systems take the position information directly but the acquisition frequency is too low to complete the task. The aim is to calculate the positions with the acceleration measurements and to adjust them with the fewest direct position measurements.

The reconstruction of rigid scenes from multiple images is a central topic in computer vision. Approaches merging partial 3D models in a hierarchical manner have proven the most effective to deal with large image sequences. One of the key building blocks of these hierarchical approaches is the alignment of two partial 3D models, which requires to express them in the same 3D coordinate frame by computing a 3D transformation. This problem has been well-studied for the cases of 3D models obtained with calibrated or uncalibrated pinhole cameras. We tackle the problem of aligning 3D models – sets of 3D points – obtained using uncalibrated affine cameras. This requires to estimate 3D affine transformations between the models. We propose a factorization-based algorithm estimating simultaneously the aligning transformations and corrected points, exactly matching the estimated transformations, such that the reprojection error over all cameras is minimized. In the case of incomplete image data our algorithm uses an Expectation Maximization (EM) based scheme that alternates prediction of the missing data and estimation of the affine transformation. We experimentally compare our algorithm to other methods using simulated and real data.

The reconstruction of rigid scenes from multiple images is a central topic in computer vision. Approaches merging partial 3D models in a hierarchical manner have proven the most effective to deal with large image sequences. One of the key building blocks of these hierarchical approaches is the alignment of two partial 3D models by computing a 3D transformation. This problem has been well-studied for the cases of 3D models obtained with calibrated or uncalibrated pinhole cameras. We tackle the problem of aligning 3D models – sets of 3D points – obtained using uncalibrated affine cameras. This requires to estimate 3D affine transformations between the 3D models. We propose a factorization-based algorithm estimating simultaneously the aligning transformations and corrected points, exactly matching the estimated transformations, such that the reprojection error over all cameras is minimized. We experimentally compare our algorithm to other methods using simulated and real data.

In order to introduce more flexibility in calibration, the bundle adjustment approach is introduced for Stereo PIV systems in Scheimpflug condition. The calibration target is positionned by hand front of the video-cameras at different locations in space. From the multiple views of the calibration target , the parameters of the video-cameras with both the target locations and the 3D calibration points are estimated by non linear least mean squares. In the proposed calibration procedure, the laser plane equation is determined by searching for the homography between the image planes . The ability of the procedure for self-calibrating two f=50 mm video-cameras and recovering the equation of the measurement plane is experimentally tested.

This article deals with optical laws that must be considered when using underwater cameras. Both theoretical and experimental point of views are described, and it is shown that relationships between air and water calibration can be found.

This article raises the problem of errors caused by the metrology of a calibration pattern to the accurate estimation of the intrinsic and extrinsic calibration parameters modeling the video system. In order to take into account these errors a new approach is proposed that enables us to compute in the same time the traditional calibration parameters and the 3D geometry of the calibration pattern using a multi-images calibration algorithm. Experimental results shows that the proposed algorithm leads to reliable calibration results and proves that calibration errors no longer depend on the accuracy of calibration point measurement, but on the accuracy of calibration point detection in the image plane.

Reconstruction by zooming is not an unachievable task. As it has been previously demonstrated, axial stereovision technics allows to infer 3D information, but involves very small triangulation angles. Accurate calibration, data matching and reconstruction have to be performed to obtain satisfactory modelling results. In this paper, a new approach is proposed to realize dense reconstruction using a static camera equipped with a zoom lens.

This article deals with videogrammetric reconstruction of volcanic structures. As a first step, the method is tested in laboratory. The objective is to reconstruct small sand and plaster cones, analogous to volcanoes, that deform with time. The initial stage consists in modelling the sensor (internal parameters) and calculating its orientation and position in space, using a multi-view calibration method. In practice two sets of views are taken: a first one around a calibration target and a second one around the studied object. Both sets are combined in the calibration software to simultaneously compute the internal parameters modelling the sensor, and the external parameters giving the spatial location of each view around the cone. Following this first stage, a N-view reconstruction process is carried out. The principle is as follows: an initial 3D model of the cone is created and then iteratively deformed to fit the real object. The deformation of the meshed model is based on a texture coherence criterion. At present, this reconstruction method and its precision are being validated at laboratory scale. The objective will be then to follow analogue model deformation with time using successive reconstructions. In the future, the method will be applied to real volcanic structures. Modifications of the initial code will certainly be required, however excellent reconstruction accuracy, valuable simplicity and flexibility of the technique are expected, compared to classic stereophotogrammetric techniques used in volcanology.

This paper presents a new method that permits to estimate, in the viewer coordinate system, the spatial attitude of an articulated object from a single perspective image. Its principle is based on the interpretation of some image lines as the perspective projection of linear ridges of the object model, and on an iterative search of the model attitude consistent with these projections. The presented method doesn't locate separately the different parts of the object by using for each of them a technics devoted to the localization of rigid object but computes a global attitude which respects the mechanical articulations of the objet. In fact, the geometrical transformations applied to the model to bring it into the correct attitude are obtained in two steps. The first one is devoted to the estimation of the attitude parameters corresponding to a rotation and involves an iterative process. The second step permits by the resolution of a linear system to estimate the translation parameters. The presented experiments correspond to the localization of robot arms from synthetical and real images. The former case permits to appreciate the accuracy of the method since the final result of the pose estimation can be compared with the attitude parameters used to create the synthetical image. The latter case presents an experiment made in an industrial environment and involves the estimation of twelve paramaters since the observed robot arm owns six inner degrees of freedom. The presented method can be useful in some operation driven by remote control.