Alumni



Aymen Arfaoui, PhD
Alexandre Baril, MSc
Sébastien Bouchard, PhD
Xavier Dallaire, PhD
William Deschênes, MSc
Erik Kretschmer, PhD
Jonathan Laberge, MSc
Martin Larivière-Bastien, PhD
Hugo Lemieux, MSc
Julie Mandar, MSc
Stephane Melanson, MSc
Jocelyn Parent, PhD
Charles Pichette, MSc
Anne-Sophie Poulin-Girard, PhD



Aymen Arfaoui, PhD

Revenu Québec

PhD Project: Calibration and modelling of a hybrid panoramic stereoscopic vision system (2015)
This thesis aims to present our contributions to the resolution of two problems encountered in the field of computer vision and photogrammetry, which are camera calibration and stereovision. These two problems have been extensively studied in the last years. Different camera calibration techniques have been developed in the literature depending on the type of camera (classical or panoramic, with zoom lens or fixed lens..). Our first contribution is a compact and accurate calibration setup, based on diffractive optical elements, which is suitable for different kind of cameras. The technique is very robust and optimal results were achieved for different types of cameras. With the multiplication of camera types and the diversity of the projection models, a generic model has become very interesting. Our second contribution is a generic model, which is suitable for conventional and panoramic cameras. In this model, composed cubic splines functions provide more realistic model of both radial and tangential distortions. Such an approach allows to model either hybrid or panoramic stereovision system and to convert panoramic image to classical image. Consequently, the processing challenges of a hybrid stereovision system or a panoramic stereovision system are turned into simple classical stereovision problems. Keywords: Calibration, panoramic vision, distortions, fisheye, zoom, panomorph, epipolar geometry, three-dimensional reconstruction, hybrid stereovision, panoramic stereovision.




Alexandre Baril, MSc

Gentec-EO

Master's Project: LED smart lighting prototype using light modulator (2017)
Since the massive invasion of LED lighting over the illumination market, a clear trend of need appeared for a more efficient and targeted lighting. The project leads this trend by developing an evaluation board to test smart lighting applications with a new liquid crystal light modulator recently developed for broadening LED light beams. These modulator are controlled by electricals signals and they are characterised by a very linear working zone. This feature allows the implementation of a closed loop control with a sensor feedback. We show that the use of computer vision is a promising opportunity for closed loop control. The developed evaluation board integrates the liquid crystal modulator, a camera, a LED light source and all the required electronics to implement a closed loop control with a computer vision algorithm.




Sébastien Bouchard, PhD

Creaform

PhD Project: Design and implementation of a novel solar concentrator reducing the need for solar tracking (2016)
Since a long time, the Sun is seen as an interesting source of renewable energy. However, the prohibitive cost of solar energy has made it mostly unused. To increase it acceptance, it is necessary to bring the cost down and make it competitive with fossil fuels. To achieve that, many possibilities exist. One method is the use of an optical system to concentrate sunlight. This thesis presents a solution based on the design of a new type of solar concentrator of low-medium concentration level made of polymer to be able to achieve low-cost fabrication. The performances of this condensed optical system which uses microlenses were then improved by the introduction of a refractive index gradient in the polymer matrix. A short analysis of the cost of the energy produced by such solar concentrators permitted to conclude that they are of commercial interest for the solar community. For this reason, the most promising design based on the use of a graded-index to increase the efficiency has been protected by a patent.

Master's Project: Study of the effet of current and temperature on high-intensity white LED degradation (2011)
Light-emitting diodes have the potential to significantly reduce electrical consumption needed for lighting. Their long life gives hope for light sources that do not need frequent replacement. However, many problems still need to be solved if lightemitting diodes want to attain the promises of energy saving and long lifetime of the manufacturers. So, there is a need for independent testing of the diodes to confirm claims written on the datasheet.

So, three different accelerated life tests were carried on two different types of lightemitting diodes. The results were compared to a reference sample used under the guidance of the manufacturers. The objective of these tests is to verify the effect on the diodes of high temperature storage and high current injection with or without heat control. This allows extraction of the influence of each specific parameter on the diode's degradation and, at the same time, to determine if the accelerated degradation is proportional to the increase in the operating conditions. It seems that the degradation pattern is different for each light-emitting diode model, which will cause problems for the development of a global degradation model.




Xavier Dallaire, PhD

Immervision

PhD Project: Wide-angle lenses miniaturisation (2018)
The miniaturization of optical systems, particularly wide-angle systems, is a subject of great importance. The reduction in size of optical components allows the integration of cameras in a wider range of applications. Even though continuous improvements in production techniques have led to great advances in the field of miniaturization, new techniques have to be developed in order to further miniaturize. The aim of this PhD project is to adapt and develop miniaturization techniques applicable to wide-angle optical systems. Through the study of various miniaturization techniques, the folded lens joined to foveated imaging and the correction of aberration via plenoptic imaging were retained as candidates allowing the miniaturization of wide-angle camera. Chapitre 3 gives an overall picture of the various avenues used in the literature for the miniaturization of optical systems. A short description of the techniques is presented as well as the reasons why some were discarded. An original miniature wide-angle endoscope design is presented in Chapitre 4, as well as the entire design and tolerancing process. The use of a fold in the system reduces the effective size of the system. Foveated imaging is used to control magnification in areas of interest. Two versions of the endoscope with different variations of their lfl are analyzed. It is shown that active control of distortion at during design can maintain the performance of an optical system in certain key regions of the field of view while reducing the number of elements that compose it. A reprojection algorithm for reconstructing an aberrated plenoptic image is presented in Chapitre 5. It is shown, through simulations, that it is possible to correct aberrations present in an optical system. Monochromatic, polychromatic and wide angle cases were successfully addressed and corrected. It was also demonstrated that the correction algorithm do not amplify the noise present in the original image. Finally, a simple prototype of a plenoptic camera was designed and tested in the laboratory.

Master's Project: Analysis and tolerancing of systems using a freeform frontal lens (2013)
In light of the new development of the production methods, free-form lens are used more frequently in optical design. However, their mathematical definitions as well as their specific shapes tend to make their tolerancing and reactions to perturbations hardly predictable. In consequence, the tools used by optical designers, like optical design programs, are not always able to help as they should. First, this document presents the recent developments and the theoretical concepts concerning these subjects. Thereafter, the analysis of a panomorph lens which has a free-form lens as a first surface reveals numbers of particular behavior showing relations between the field curvature and the footprint. Finally, a new tolerancing technique is presented. Using a perturbative analysis requiring a minimum of computing power, it is possible to reduce the allowed margins of error of certain variables while maximising the gain in image quality. This technique was particularly efficient in cases where the adjustment of the focus was spatially limited.




William Deschênes, MSc

Doric Lenses

Master's Project: Adaptive optics testing focal plane box for the Observatoire du Mont-Mégantic (2016)
New technologies for large scale telescope projects need to reach a high level of technological maturity before they can be integrated. The project involves the creation of an adaptive optics test bench for evaluating the onsky performance of adaptive optics related devices. The test bed was successfully installed on the Mont Mégantic observatory, and was used to evaluate the performance of a pyramid wavefront sensor. The system effectively halved point-spread function size. Several improvements to the system are possible to improve performance.




Erik Kretschmer, PhD

Karlsruher Institut für Technologie

PhD Project: Modelling of the instrument spectral response of conventionnal and imaging fourier transform spectrometers (2014)

Spectroscopy and the measurement of light spectrum have become essential tools in a large number of fields, from analytic laboratories to remote sensing field measurements. In these applications, the use of Fourier transform spectrometers (FTS) is widespread and, more recently, imaging Fourier transform spectrometers (iFTS) are becoming ever more popular. The iFTS instruments enable spatially resolved highresolution spectral analysis within a single measurement, thus allowing the study of fine spectral variations in observed scenes. Such measurements inherently include systematic errors which can be in large part described by the instrument spectral response function, often referred to as SRF. In the case of iFTS, each pixel of the instrument will sport a different spectral response, which opens a whole new dimension not only in the measurement itself, but also for error analysis and instrument design. Because of their unique imaging capacity, iFTS instruments are a prime choice for remote sensing applications from airborne or spaceborne platforms for the measurement of the Earth atmosphere, as well as the atmosphere of other celestial bodies. The requirements on the spectral accuracy demanded by such missions are very high. To achieve these requirements, an excellent knowledge of the instrument spectral response is essential. Modelling of the spectral response of iFTS instruments is a possible approach to achieve the desired knowledge of instrument performances. This thesis offers a review of the factors affecting the spectral performances of FTS instruments from which a numerical model of the spectral response specifically designed with imaging instrument in mind is proposed. This model integrates optical effects which were up to now only studied separately – if at all – and not integrated in a global performance model of the instrument. The necessity to consider these effects, such as those caused by the optical transfer function of the detector imaging optic or the architecture of the imaging focal plane array (FPA), is demonstrated. Using dedicated measurement of the airborne iFTS GLORIA (Gimballed Limb Observer for Radiance of the Atmosphere), the application of the model for performance analysis and review is demonstrated.




Jonathan Laberge, MSc

Cégep de Rimouski

Master's Project: Analysis of the telescope's optical aberrations at the Observatoire du Mont-Mégantic (2012)
The telescope at the Observatoire du Mont-Mégantic is in operation since late 70's. The wide range of instruments available makes it one of the most versatile telescopes in the world. However, very few data exist with respect to its optical aberrations. It is important to know these aberrations if we want to improve the optical performance ofthe telescope.

In order to evaluate thess aberrations, two methods of wavefront reconstruction were used: the Shack-Hartmann wavefront sensor and the wavefront curvature sensor. Analysis of the reconsctructed wavefronts identifies the dominant aberrations as well as several parameters influencing these aberrations. Furthermore, the methods are compared in order to know if they can be used to align the primary and secondary mirrors.




Martin Larivière-Bastien, PhD

Telops

PhD Project: Performance improvement for panoramic and Panomorph imaging systems (2014)

Because of their large field of view, panoramic and Panomorph imagers are modules of interest in optical design. An important field of view introduces many aberrations in a system. To obtain the desired image quality, aberrations are generally minimized by the optimization of certain optical components, but this practice has its limits. The goal of this project is hence to investigate other techniques that would allow breaking these limits. Three types of techniques are investigated: hardware, software and hybrid. In the hardware section, the curved sensor and the freeform field lens are analyzed. It is shown that they allow for a best focus effect on the whole image. The curved sensor can also be used to create a monocentric system which simplifies the optical design and gives increased performances compared to the traditional system. In the software section, the gains obtainable by image processing are highlighted. It is shown that even with the important variation of the PSF across the field of view, deconvolution still increases image quality as a whole. Lateral color correction gives additional improvement of the image. Both methods are sufficiently fast to consider a video application. The last section addresses wavefront coding. Two methods of optimization are presented: optimization by the MTF and optimization by the variance between object and image. The optimization by the variance requires more computation time but gives better results because it considers the effect of image reconstruction and noise. In this section, simulations are also used to compare the performances and the applicability of two phase mask types. It is shown that the cubic phase mask is more useful to correct defocus and astigmatism while the quartic phase mask is better for coma and Gaussian white noise. Using the information obtained with the simulations, a prototype based on the IMV1 ImmerVision lens was built. Unfortunately the cubic phase mask manufactured had a surface roughness that was too high and the resulting images are noisy.




Hugo Lemieux, MSc

Eddyfi

Master's Project: Study and comparison of LED lamp power supply strategies (2011)

As energy efficiency is becoming a priority, light-emitting diodes (LEDs) luminaires are a first choice light source. Although more expansive than other sources, they offer, in addition to a great efficiency, a long useful lifetime, which makes them a very interesting option in the long run. As all other light sources, they suffer from luminosity degradation, which reduces their luminous flux as well as their efficiency. In order to better control this luminosity degradation, power strategies can be used. Two of these strategies are current control and chain control, the first varying the current and the second the number of open LEDs. To study and compare those two strategies, a mathematical model has been developed and a simulation program has been designed. Results showed that in addition to keep a more constant luminous flux, power strategies can enhance energy efficiency and lifetime of a LEDs luminaire or reduce the required number of LEDs, reducing the initial cost.




Julie Mandar, MSc

ABB

Master's Project: Development of SITELLE's performance model (2012)

SITELLE is a new imaging Fourier transform spectrometer to be installed at the Canada-France- Hawaii Telescope. The development of its dedicated performance model drives the design of the instrument and the flow down of the science cases requirements into system requirements. First, the selected configuration with off-axis flat mirrors makes the achievement of a high efficiency in the near ultra violet easier. Secondly, servomechanism’s desirable performances were defined in order to design a photon noise limited instrument, based on a relevant scene. These performances should be maintained during a 4 hours data-cube acquisition, under operational vibrations and external effects such as wind gust hitting the telescope. Ultimately, this instrument performance model is the core of the signal to noise ratio simulator that will help observers to evaluate the potential benefits of SITELLE for their target.




Stephane Melanson, MSc

D-Pace

Master's Project: Colour control of quadrichromatic LEDs using a color detector (2014)

Light emitting diodes (LEDs) are good candidates to become the main light sources of the future given their good efficiency and long life. However, a problem associated to the use of LEDs is the color shift observed at different temperatures, currents and lifetime. The purpose of this work was to implement a control algorithm that uses a color sensor as feedback. Two algorithms were developed to calibrate the detectors. The algorithms were implemented to control RGBA and RGBW LEDs. The first algorithm provides a good calibration for a restricted range of color near the calibration. The second algorithm gives good performances for the entire range of color available to the lamp, but it is less precise than the first algorithm.




Jocelyn Parent, PhD

Immervision

PhD Project: Locally-variable magnification imaging system for active distortion control (2012)

This thesis presents a novel type of optical system, a locally magnifying imager. The idea behind it is to modify in real-time the distortion by using an active component placed far from pupils, allowing to modify each field individually. With the correct shape on the active surface, it is thus possible to create zones of increased magnification. Also, if the total field of view has to remain constant, it is possible to do so by combining theses zone of interest to zones of settling back where the magnification is decreased. To fully understand how systems with controlled distortion work, an analysis of surface errors is presented. After, from theses results of image plane displacement from a given error, the concept of the locally magnifying imager is developed. ZEMAX simulations are done, confirming the mathematically obtained parameters. The simulations finally lead to an experimental prototype using a ferrofluidic déformable mirror as the active surface. Experimental results with a system that can produce ratios of magnification larger than 3 in a zone of interest are presented and analysed. In short, if the fundamental limits are countered, this locally magnifying imager has the potential of changing some imaging domains thank to their novel capabilities in optical design.




Charles Pichette, MSc

Université Laval

Master's Project: Design and optimisation of detection channels using single-photon avalanche diodes (SPADs) for photon counting in optical diffuse tomography (2017)

This dissertation showcases the improvements suggested for the diffuse optical tomography scanner of the TomOptUS group at Université de Sherbrooke. Diffuse optical tomography is an imaging modality that uses near infrared light (650 - 950 nm) to image small animals such as mice in depth (> 1 cm). This technique is very interesting for pharmacology or oncology where it can be used to track medicine or the progress of pathology. It also decreases the number of necessary sacrifices since it is a non-invasive technique. The temporal progress of the object under consideration can, in that case, be acquired with ease. This technique can also be used with fluorescent agents to track different objects of interest in the animal. This scanner works in time domain where the time of flight of individual photons that propagated through the subject is registered to reconstruct the laser pulse via time-correlated single photon counting with an ultra-fast laser source. The current scanner uses 7 no-contact detection channels positioned in a ring around the subject. This number of channels is too low to obtain a satisfying acquisition time. It was determined that the limiting factor is the need to mechanically rotate the channels around the subject to obtain the necessary angular coverage. To reduce the acquisition time, it was suggested to increase the number of detection channels to 32 or even 64. However, the current channels use photomultiplier tubes which are too bulky to be used with such a high density of detectors. Single photon avalanche diodes have been considered to replace them because of their relative small size, excellent temporal resolution, and better quantum efficiency. These characteristics make them especially efficient for photon counting. These photodiodes, however, have a photosensitive surface with a very small diameter compared to the photomultiplier tubes (25 - 100 μm compared to ≈ 1 cm for photomultiplier tubes). This reduces the photon count rate, lowers the signal to noise ratio, and makes the alignment difficult. The goal of this project is to optimize the design of new detection channels that use these single photon avalanche diodes. A primary analysis of the parameters and constraints on the system was first conducted to pinpoint the optimal parameters. Several optical designs are then presented and analyzed. These channels can achieved a maximum photon count rate with the use of immersion lenses. These immersion lenses act as optical concentrators and achieve a concentration ratio of ≈ n² which is ≈ 4 in our case. This translates to an increase in the photon count rate and signal to noise ratio of the same ratio. The affixing of the immersion lens on a custom photodiode module was then performed and the experimental confirmation of the increase in photon count was obtained with intrinsic and fluorescence measurements. This experimental increase is supported with Zemax simulations which are in good agreement with the experiments. Finally, it was experimentally confirmed that those immersion lenses do not affect the excellent temporal resolution of the single photon avalanche diodes.




Anne-Sophie Poulin-Girard, PhD

Université Laval

PhD Project: Stereoscopic Panomorph pair for 3D reconstruction of objects of interest in a scene (2016)

A wide variety of panoramic lenses are available on the market. Exhibiting interesting characteristics, the Panomorph lens is a panoramic anamorphic optical system. Its highly non-uniform distortion profile creates areas of enhanced magnification across the field of view. For mobile robotic applications, a stereoscopic system for 3D reconstruction of objects of interest could greatly benefit from the unique features of these special lenses. Such a stereoscopic system would provide general information describing the environment surrounding its navigation. Moreover, the areas of enhanced magnification give access to smaller details. However, the downfall is that Panomorph lenses are difficult to calibrate, and this is the main reason why no research has been carried out on this topic. The main goal of this thesis is the design and development of Panomorph stereoscopic systems as well as the evaluation of their performance. The calibration of the lenses was performed using plane targets and a well-established calibration toolbox. In addition, new mathematical techniques aiming to restore the symmetry of revolution in the image and to make the focal length uniform over the field of view were developed to simplify the calibration process. First, the field of view was divided into zones exhibiting a small variation of the focal length and the calibration was performed for each zone. Then, the general calibration was performed for the entire field of view. The results showed that the calibration of each zone does not lead to a better 3D reconstruction than the general calibration method. However, this new approach allowed a study of the quality of the reconstruction over the entire field of view. Indeed, it showed that is it possible to achieve good reconstruction for all the zones of the field of view. In addition, the results for the mathematical techniques used to restore the symmetry of revolution were similar to the results obtained with the original data. These techniques could therefore be used to calibrate Panomorph lenses with calibration toolboxes that do not have two degrees of freedom relating to the focal length. The study of the performance of stereoscopic Panomorph systems also highlighted important factors that could influence the choice of lenses and configuration for similar systems. The challenge met during the calibration of Panomorph lenses led to the development of a virtual calibration technique that used an optical design software and a calibration toolbox. With this technique, simulations reproducing the operating conditions were made to evaluate their impact on the calibration parameters. The quality of 3D reconstruction of a volume was also evaluated for various calibration conditions. Similar experiments would be extremely tedious to perform in the laboratory but the results are quite meaningful for the user. The virtual calibration of a traditional lens also showed that the mean reprojection error, often used to judge the quality of the calibration process, does not represent the quality of the 3D reconstruction. It is then essential to have access to more information in order to asses the quality of a lens calibration.

Master's Project: Optical testing setup for panoramic lenses (2011)

Panoramic lenses' characteristics are unique and their characterization can be a challenge. For vision applications like security or inspection, a precise knowledge of the distortion introduced by panoramic lenses is essential to produce natural unwrapped views. Also, the image quality must be uniformed over the field of view because all directions are important. For these reasons, we have developed a dedicated testing setup for panoramic lenses and a quick and easy measuring process. Using referenced equally-spaced targets, we obtained the radial image mapping curves for various azimuthal angles, allowing us to calculate the instant full-field resolution map. Also, transition targets were used to find field-dependent spatial frequency where the MTF is 50%. We tested two panomorph lenses and two fisheyes with two different cameras. For each lens, we discussed the experimental resolution and MTF curves and compared some of those results to theoretical design data.