Guy Martin

Guy Martin

Greater Montreal Metropolitan Area
4K followers 500+ connections

Join to follow
R&CV: Robotics and Computer Vision

About

In Montreal CANADA +1 (514) 382-9392
We beat even NASA. The entire imaging system…

Services

Request proposal

Articles by Guy

See all articles

Activity

Experience

  • R&CV: Robotics and Computer Vision

    Montreal CANADA

Education

  • Polytechnique Montréal Graphic

    Polytechnique Montréal

    -

    Simulation of Antropomorphic Robot Manipulator and non Linear Control using Inverse Kinematics and Kinetics
    Research Cooperation with Université Catholique de Louvain-la-Neuve, Belgium

Licenses & Certifications

  • Galil Motion Control Integrator

    Since 1989, top 3 experts in Eastern Canada

    Issued

  • Medar Welding Seminar

    -

    Issued

Volunteer Experience

  • Baseball Coach

    Salaberry Sports Federation

    - 3 years 5 months

    Children

    Coaching, baseball, registered at Baseball Quebec

Publications

  • What is Wrong with Telemetry !

    White Paper

    An overview of the multiple bias sources in imaging and camera calibration that will offset telemetry: target shape, location, size, and range. We give our lab analysis of Bayer pattern sub pixel edge bias as the next limitation in sub pixel imagery. Our next patent will publish our RAW image calibration for a color camera to compensate chromatic distortion.

    See publication

  • Digital Zooming and Chromatic Distortion

    White Paper

    95% of digital cameras are affected by chromatic distortion causing a colored edge blurr and loss of image sharpness. The error finds its origin in the splitting of light in the optics, and wrong Bayer pattern interpolation to account for the phenomena. It can lead to several pixels edge and corner location shift in telemetry, and lower image quality in digital zooming.
    First of two papers on chromatic distortion, I introduce the chromatic color model and calibration procedure as derived…

    95% of digital cameras are affected by chromatic distortion causing a colored edge blurr and loss of image sharpness. The error finds its origin in the splitting of light in the optics, and wrong Bayer pattern interpolation to account for the phenomena. It can lead to several pixels edge and corner location shift in telemetry, and lower image quality in digital zooming.
    First of two papers on chromatic distortion, I introduce the chromatic color model and calibration procedure as derived from my 2015 camera model. In the following paper (scheduled for early 2017), I will deal with Bayer pattern recovery to account for chromatic edge shift, first from a calibrated camera, to conclude with approximate models and schemes to compensate.

    See publication

  • Foreign Comparative Testing (FCT) Product Summary as filed in the US Army database

    Something to read if you need a good overview of our published results. More to come

    See publication

  • The Zooming Lens Model

    White Paper

    Extension of our camera model disclosed in May 2105 for IEEE TePRA, this white paper shows how to handle a zooming lens. It discusses the calibration, distortion and pairing with a fixed lens camera for Intelligent Systems. ISO/TC 172 - Optics and photonics identifies the lack of a zooming lens model as the missing link for the widespread use of digital imaging at large.

    See publication

  • Accurate 3D Telemetry from Knowing the Image Scale

    White Paper

    Combining our camera model and our 3D true perspective approach in telemetry, we achieve 10:1 accuracy increase. Our camera model gives a single image scale, relating width, to height, to depth
    proportions in the image like no other model can, giving the exact solution to plane target tracking
    in the single camera case. We increased telemetry accuracy while cutting down
    computational time, avoiding chromatic distortion. We also do the extension to the classic stereo
    pair, and show…

    Combining our camera model and our 3D true perspective approach in telemetry, we achieve 10:1 accuracy increase. Our camera model gives a single image scale, relating width, to height, to depth
    proportions in the image like no other model can, giving the exact solution to plane target tracking
    in the single camera case. We increased telemetry accuracy while cutting down
    computational time, avoiding chromatic distortion. We also do the extension to the classic stereo
    pair, and show how to switch from single to stereo camera to cope with occlusion.

    See publication

  • Sub-Pixel Edge Extraction in the 2D Plane-White Paper

    In 1995 Frederic Devernay [1] proposed a method for sub pixel edge extraction from a quadratic interpolation of the NMS method. Starting from his work, we extend his approach into a surface model able deal with edge point displacement, either from zooming, or chromatic distortion.
    In target center extraction, we achieve 0,03 pixel accuracy

    See publication

  • System and Method for Imaging Device Modelling and Calibration

    WIPO

    PCT/CA2014/000534 published on WIPO January 2015, and the core knowledge of our imaging expertise.

    See publication

  • The Major Error in the Camera Model is the Tilted Axis Assumption

    White Paper

    The error introduced by the tilted axis assumption dominates the lens geometric distortion contribution.
    A camera has two distortion sources: the lens and the off squareness of the image plane

    See publication

  • Cybersickness: Caused by 3D bias and distortion, as well as image lag-latency

    White paper

    Users of augmented reality suffer headaches from 2 sources: Image lag, and image distortion. Over 0,250sec lag between natural vision and synthetic images creates nausea. Distortion causes headaches and could very well induce Alice in Wonderland Syndrome in AR users. A calibrated camera uses the tilted axis assumption to compensate image plane loss of squareness with the line of sight. We have proven it creates more distortion, and headaches.
    My conclusions show that before anything…

    Users of augmented reality suffer headaches from 2 sources: Image lag, and image distortion. Over 0,250sec lag between natural vision and synthetic images creates nausea. Distortion causes headaches and could very well induce Alice in Wonderland Syndrome in AR users. A calibrated camera uses the tilted axis assumption to compensate image plane loss of squareness with the line of sight. We have proven it creates more distortion, and headaches.
    My conclusions show that before anything else,

    1- Every single projective 3D inconsistency for monocular vision should be removed
    2- 3D stereo inconstistency has to be removed from the system to avoid depth perception errors by the brain
    3- Any image lag-latency should be removed to isolate that component in motion sickness

    and only then other human perception channels can be analyzed.

    Other authors
    See publication

  • A Compared Camera Calibration Result between Zhang’s Model and Ours

    Lab test results showing why we are 4 times faster and the only ones to achieve absolute positioning accuracy from a digital camera. Before us, Microsoft’s Zhang model was the fastest and most accurate. From COTS, we achieved focal length calibration at 10e-10mm giving the image true scale and perspective, and removing a 2 pixel bias on the image center.

    See publication

  • HIGH ACCURACY CAMERA MODELLING AND CALIBRATION FOR AUTOMATED IMAGERY

    NATO Symposium on Emerging and Disruptive Technologies-Madrid

    Tracking down bias and systematic errors in camera calibration.
    To achieve sub-pixel accuracy, we corrected systematic bias sources in
    1. The camera model and calibration procedure
    2. Sub pixel edge and corner extraction
    3. 3D telemetry from a stereo camera pair or moving camera
    From 1 000 000:1 accuracy gain on the camera focal distance, we proposed solutions to increase deblurring of atmospheric distortion.

    See publication

  • Software Solutions for NVG ENVG Integration

    ICee Webtool

    Necessary conditions to achieve sub pixel sensor fusion in night vision systems using low accuracy digital cameras.
    The presentation focuses on software sensor fusion of Infrared/Colour/Synthetic imaging to preserve sub pixel image accuracy for image analysis and 3D telemetry

    See publication

  • R&D Management in Small and Medium Business Environments

    Lecture at the Sherbrooke University post graduate job fair

    Insight on managerial/budget constraints and guidelines for Research and Development in a limited ressources business structure

  • Mathematical Modelling & Use of Digital Cameras in Advanced Imaging

    ICee Webtool

    We discovered a systematic bias in the modelling and calibration
    procedure for digital cameras. Using of the shelf industrial cameras and lenses, we demonstrate that a software image correction and corrected modelling approach gains up to
    10:1 higher accuracy (image or 3D telemetry)
    4:1 faster computation time
    30% added loss less video compression
    Even from low accuracy 1024x768 640x480 Bayer pattern colour
    cameras, we retrieve the camera focal distance…

    We discovered a systematic bias in the modelling and calibration
    procedure for digital cameras. Using of the shelf industrial cameras and lenses, we demonstrate that a software image correction and corrected modelling approach gains up to
    10:1 higher accuracy (image or 3D telemetry)
    4:1 faster computation time
    30% added loss less video compression
    Even from low accuracy 1024x768 640x480 Bayer pattern colour
    cameras, we retrieve the camera focal distance to an
    unmatched 10e-10mm…roughly the size of an hydrogen molecule
    The accuracy increase provides multiple integration trade offs between, computation speed, cost, accuracy, compression, …

  • Dynamic Behaviour or Robot Manipulators Applied to Simulation and Control

    Master Thesis, Memoir, 401 p

    Analysis of the mathematical representation dependency on the multi-bodied dynamic simulation of robotic arms.
    It serves as a starting point for the use of dynamic simulation in a non linear predictive control strategy for known robot trajectory tracking.
    The memoir introduces a pseudo non linear integration compensation strategy for modelling errors: friction, gear backlash and hysteresis...

Projects

  • Research Mission for the Quebec Ministry of Education

    -

    I was sent to Belgium at Université 'Catholique de Louvain-La-Neuve' to evaluate, learn and train on a Robotics Computer Simulation Software based on Jean Claude Samin's PhD work at Caltech. During my stay I was asked to give a lecture on my studies on the optimization of Robot's Dynamic Behaviour Model used for simulation and non linear real time control.

    Other creators
    • Jules OShea
    • Jean Claude Samin

  • SAE Formula Racer

    -

    Structural Analysis: Formula one racer using a 500cc engine, North American Engineering student competition. Overall strusture was evaluated using finite element analysis and weight was cut down in half from previous year version. First for fuel efficiency, and safety during the competition in Colorado.

    Other creators
    • SAE student chapter of Polytechnique

  • CAD-CAM-CAE Program Implementing at Polytechnique

    -

    As a representative of the Student's Organization, I was on the board to evaluate the curricula proposed by every Engineering department. I was reporting to the Student's Organization president.
    I was eventually offered a job by Quebec's Ministry of Education to recommend and evaluate computer equipment and software for the entire Education Network of the Québec Province.

    Other creators
    • Benoit Cazavant

Honors & Awards

  • Poeple's Choice Award

    M2D2: Massachusetts Medical Device Development

    Business new technology competition in Lowell Massachusetts.
    We won for the medical imaging performance enhancements our software technology provides:
    Increased Speed, Accuracy, even from lower cost COTS devices.

  • Coolest Technology

    CENE and the Canadian General Consulate in Boston

    We broke the accuracy limitiation in digital imaging. From 10e-4mm, we brought down the focal distance (image scale) calibration error to 10e-10mm, increased computational speed by 4:1, and removed a systematic bias in all camera parameters preventing accurate lens distortion removal, 3D scanning, zooming lens use, fast software fusion,... applicable in most related Intelligent System integrations. The camera accuracy is now limited only by pixel count. (filed for patent)
    We were selected by…

    We broke the accuracy limitiation in digital imaging. From 10e-4mm, we brought down the focal distance (image scale) calibration error to 10e-10mm, increased computational speed by 4:1, and removed a systematic bias in all camera parameters preventing accurate lens distortion removal, 3D scanning, zooming lens use, fast software fusion,... applicable in most related Intelligent System integrations. The camera accuracy is now limited only by pixel count. (filed for patent)
    We were selected by CENE: Canadian Entrepreneurs of New England as one of the 20 most promising startups in Canada. In the business pitch competition held at the Cambridge Innovation Centre on the MIT campus, we were awarded 'Coolest Tech'.
    The event took place during the '48 hours in the hub' event
    We were offered 3 months office space, full support of the Canadian Commercial Delegation, mentoring, access to VCs and Angels, networking by CENE ...

Languages

  • English

    Native or bilingual proficiency

  • French

    Native or bilingual proficiency

Recommendations received

6 people have recommended Guy

Join now to view

View Guy’s full profile

  • See who you know in common
  • Get introduced
  • Contact Guy directly
Join to view full profile

Other similar profiles

Explore more posts

Explore collaborative articles

We’re unlocking community knowledge in a new way. Experts add insights directly into each article, started with the help of AI.

Explore More

Others named Guy Martin in Canada

90 others named Guy Martin in Canada are on LinkedIn

See others named Guy Martin

Add new skills with these courses

See all courses