Ludovico Carozza

Markerless Vision-Based Augmented Reality for Urban Planning

Augmented Reality (AR) is a rapidly develop- ing field with numerous potential applications. For ex- ample, building developers, public authorities, and other construction industry stakeholders need to visually as- sess potential new developments with regard to aesthet- ics, health and safety, and other criteria. Current state-of- the-art visualization technologies are mainly fully virtual, while AR has the potential to enhance those visualiza- tions by observing proposed designs directly within the real environment. A novel AR system is presented, that is most appropriate for urban applications. It is based on monocular vision, is markerless, and does not rely on beacon-based local- ization technologies (like GPS) or inertial sensors. Addi- tionally, the system automatically calculates occlusions of the built environment on the augmenting virtual objects. Three datasets from real environments presenting dif- ferent levels of complexity (geometrical complexity, tex- tures, occlusions) are used to demonstrate the perfor- mance of the proposed system. Videos augmented with our system are shown to provide realistic and valuable visualizations of proposed changes of the urban environ- ∗ To whom correspondence should be addressed. E-mail: f.n.bosche@

Mosaicing of optical microscope imagery based on visual information

Tools for high-throughput high-content image analysis can simplify and expedite different stages of biological experiments, by processing and combining different information taken at different time and in different areas of the culture. Among the most important in this field, image mosaicing methods provide the researcher with a global view of the biological sample in a unique image. Current approaches rely on known motorized x-y stage offsets and work in batch mode, thus jeopardizing the interaction between the microscopic system and the researcher during the investigation of the cell culture. In this work we present an approach for mosaicing of optical microscope imagery, based on local image registration and exploiting visual information only. To our knowledge, this is the first approach suitable to work on-line with non-motorized microscopes. To assess our method, the quality of resulting mosaics is quantitatively evaluated through on-purpose image metrics. Experimental results show the importance of model selection issues and confirm the soundness of our approach.

Automatic Perspective Camera Calibration Based on an Incomplete Set of Chessboard Markers

The measurement in metric units from a perspective view of a real world object is an important task in many computer vision applications. The goal of the perspective calibration is to map the reference coordinates of the 3D object into the 2D image coordinates. This is usually done by recovering the perspective transformation parameters by presenting to the camera a view of a known calibration pattern. Many of the available algorithms require to process all the points of the calibration pattern in order to recover the transformation parameters effectively. The approach we present for the computation of the inverse perspective transformation does not require to consider all the grid points. Rather, a grid matching between the 3D points projection and the model is performed exploiting a graph based approach. Extensive experiments have proved the efficacy of the algorithm in different application fields such as camera calibration and automotive headlight beam characterization.

An incremental method for mosaicing of optical microscope imagery

Digital imaging is nowadays widely employed in the field of optical microscopy. One of the most apparent benefits consists in the possibility for the researcher to see the whole biological sample in one image, achieved by collecting all the parts being inspected. Common approaches work in batch mode and rely on known motorized x–y stage offsets of the microscope holder. Or alternatively, the methods are conceived just to provide visually pleasant mosaics off-line, that are often built by altering the photometric values or the geometric properties of the original component images. This work presents an incremental mosaicing method for optical microscopy imagery, compliant with on-line requirements and suitable even for non-motorized microscopes. The resulting mosaics are very accurate and preserve the consistency of the original images so to be used for further global measurement steps. Nevertheless, the mosaics are visually pleasant so to be used for visual inspection as well. The experimental results obtained in different biological examinations confirm the efficacy of our approach.

A robust approach to reconstruct experimentally the camera response function

Many methods are known in literature to reconstruct the camera response function (RF), whether they use calibration chart or images of arbitrary scenes taken at different exposures. This works in case to have at ones disposal a camera with enough exposure steps to capture the whole dynamic range of the scene, representatively. However, in case of entry-level or low quality cameras, the resolution in the shutter time range of variation could yield just a reduced set of samples. Here we propose additional constraints to a well known method to reconstruct a RF even in the case that just a reduced set of exposures is available. Extensive experiments carried out using both a low quality and a professional camera commonly used in computer vision applications show the improvement achieved by our method in the reconstructed RF in case of few samples.

An Automatic System for the Real-Time Characterization of Vehicle Headlamp Beams Exploiting Image Analysis

A vehicles headlamp orientation and luminous and geometrical beam properties are a matter that is strictly ruled by the European Commission for transportation. To test the headlamps, a test system is usually manually aligned to the vehicle, and the human being has the definite opinion even on the beam-related measures. This paper presents a fully automatic system that exploits vision-based technology to extract the geometric parameters of the light profiles that are projected by vehicle headlamps in real time. The 3-D orientation of the longitudinal axis of the vehicle is recovered using stereoscopy. Furthermore, image analysis is used to automatically characterize the shadow-light border of a beam profile, as it would be perceived by an experienced human operator. A locally adaptive thresholding algorithm allows our system to automatically adjust to a wide range of light sources of different power. The alignment procedure and the beam characterization algorithm have been assessed through proper measuring apparatus that is capable of yielding accurate ground-truth data. In particular, the headlamp has been mounted on a special three-axis numerical control unit whose accuracy has been previously assessed, again using image analysis. Experimental results, which are carried out on a large number of different headlamps, show that our method is able to achieve accurate measurements in compliance with current regulations. Finally, it is worth remarking that our solution is fully automatic, and it just requires a simple setup procedure.

A Novel Vision-Based Approach for Autonomous Space Navigation Systems

Determining the attitude of a satellite is a crucial problem in autonomous space navigation. Actual systems exploit ensemble of sensors, also relying on star trackers, that recover the attitude of a satellite through matching part of the sky map with a star atlas stored on board. These are very complex systems that even suffer for Sun and Moon blinding. In this work we assess the feasibility to use a novel stand-alone system relying on a camera looking at the Earth . The satellite attitude is recovered through the parameters extracted from the registration of the image pairs acquired while orbiting. The experiments confirms that the accuracy of our approach is fully compliant with the application requirements.

An Image Registration Approach for Accurate Satellite Attitude Estimation

Satellites are controlled by an autonomous guidance system that corrects in real time their attitude according to information coming from ensemble of sensors and star trackers. The latter estimate the attitude by continuously comparing acquired image of the sky with a star atlas stored on board. Beside being expensive, star trackers undergo the problem of Sun and Moon blinding, thus requiring to work jointly with other sensors. The novel vision based system we are investigating is stand alone and based on an earth image registration approach, where the attitude is computed by recovering the geometric relation between couple of subsequent frames. This results in a very effective stand alone attitude estimation system. Also, the experiments carried out on images sampled by a satellite image database prove the high accuracy of the image registration approach for attitude estimation, consistent with application requirements.

High accuracy estimation of vehicle trajectory using a real time stereo tracking system

In this paper, we present the automatic real time stereo tracking algorithm we devised to derive the 3D orientation of the longitudinal axis of a vehicle by recovering its trajectory during a motion sequence. An accurate identification of vehicles longitudinal axis is required in automotive applications where measurements achieved by testing apparatus must be in compliance with regulations. Usually, these systems are made of ensembles of sensors or are invasive, needing fiducial markers placed on the vehicle being tracked. Our method is fully automatic, non invasive and employs common CCD technology to recover 3D vehicle axis information by exploiting patterns natively present on vehicles. The experiments carried out using different orientation angles show an extremely high accuracy that is even compliant with regulations. Accordingly, we can state that this is the first fully automatic system that uses stereo technology to achieve in real time such an accuracy regarding vehicles 3D orientation without exploiting any prior model.

An industrial vision-based technology system for the automatic test of vehicle beams

In this paper we describe the first industrial prototype to characterize automatically the headlamp beam properties using computer vision. The European commission for transportation provides strict regulations that have to be fulfilled as far as headlamp orientations, luminous and geometrical beam properties are concerned. To test the headlamps, the test system has to be properly aligned to the vehicle in order that the measures achieved on brightness and geometrical beam profile can be reliable. The system we present is composed of two integral subsystems. The first consists of a fixed stereo vision system capable of estimating automatically, in real time and with a very high accuracy, the longitudinal axis of the vehicle while it is approaching the stereo rig. The outcome is used to accurately align the second subsystem with respect to the vehicle. This subsystem is composed of a classic optical projection system endowed with a CCD camera used to perform automatically radiometric and geometric assessments of the beam projected by the headlamps. Experiments carried out for both the subsystems prove how the high accuracy achieved by our method makes the prototype compliant with current regulations. It is worth remarking that the technology employed is low cost, thus making our approach suitable for commercial headlight tester.