VIPeR |
ABOUT THE PROJECT
The ability of robots to move and navigate autonomously in an unknown environment is directly related to their faculty to extract information surrounding them. Polarization information is used by many animals but remains unused in the fields of robotics. For instance, the sky produces a polarization pattern that is used by bees to orient themselves. The main goal of the VIPeR (Polarimetric Vision Applied to Robotics Navigation) project is to prove by developing an experimental platform that polarization can be efficiently used for navigation tasks. For this project, our Le2i UMR CNRS 6306 Laboratory will provide us the ability to use the robotic platform with a ground robot vehicle (Summit XL) and a UAV (ASTEC Pelican). The first application will consist in embedding a commercial three-CCD polarization sensor on a terrestrial vehicle to illustrate the ability to avoid puddles or ice hazards on the road. The polarization effect by reflection will be used since after reflection, natural light that is unpolarized becomes partially linearly polarized. Furthermore, using the same properties of polarization by reflection in urban environment, the vehicle will have the ability to distinguish semi-reflective surfaces such as glasses.
The second application will consist in embedding a polarimetric sensor on a UAV. In this case, many constraints regarding the weight of the sensor will arise. The previous technic could not be used in that sense, and two other solutions will be provided. The first option is the use a pixelated polarized sensor which is low resolution but provide polarization information in real time. The second option is a mechanical controlled polarizer placed in front a conventional sensor. This last option presents a more flexible solution but involves new algorithms to be developed in order to extract the polarization information. To illustrate the usefulness of such a sensor embedded on a UAV, we will exploit the polarization properties of the scattering light. The attitude estimation task will be developed by using the polarization pattern of the sky. Indeed, depending on the sun position, the sky provides a useful compass for orientation in space and can be an efficient tool to compensate the drift that frequently appear in inertial measurement unit after a long term running. An additional application will be proposed by extending algorithms usually used to solve the “blue effect” on natural images with high depth of field. Indeed, the “blue effect” is polarized and the polarization parameters are directly linked to the distance from the objects. From the polarization parameters measurement, we will be able to estimate roughly the distance of the objects.
The VIPeR project is a project over two years, which aims to start a new topic in the laboratory by combining Vision for Robotics and Non-conventional Imaging field. Once the contribution of polarimetric imaging will be concretely demonstrated through these two prototypes, this project will allow us to initiate larger projects including laboratories specialized in autonomous vehicles.
The second application will consist in embedding a polarimetric sensor on a UAV. In this case, many constraints regarding the weight of the sensor will arise. The previous technic could not be used in that sense, and two other solutions will be provided. The first option is the use a pixelated polarized sensor which is low resolution but provide polarization information in real time. The second option is a mechanical controlled polarizer placed in front a conventional sensor. This last option presents a more flexible solution but involves new algorithms to be developed in order to extract the polarization information. To illustrate the usefulness of such a sensor embedded on a UAV, we will exploit the polarization properties of the scattering light. The attitude estimation task will be developed by using the polarization pattern of the sky. Indeed, depending on the sun position, the sky provides a useful compass for orientation in space and can be an efficient tool to compensate the drift that frequently appear in inertial measurement unit after a long term running. An additional application will be proposed by extending algorithms usually used to solve the “blue effect” on natural images with high depth of field. Indeed, the “blue effect” is polarized and the polarization parameters are directly linked to the distance from the objects. From the polarization parameters measurement, we will be able to estimate roughly the distance of the objects.
The VIPeR project is a project over two years, which aims to start a new topic in the laboratory by combining Vision for Robotics and Non-conventional Imaging field. Once the contribution of polarimetric imaging will be concretely demonstrated through these two prototypes, this project will allow us to initiate larger projects including laboratories specialized in autonomous vehicles.