Register now After registration you will be able to apply for this opportunity online.
This opportunity is not published. No applications will be accepted.
Learning Enhanced Infrared-Optical Imaging
The goal of this project is to establish an accurate pixel to pixel mapping between a thermal and optical camera.
Keywords: Deep learning, infrared camera, optical camera, camera synchronization, human detection, fixed-wing UAV
The goal of this project is to generate an accurate pixel to pixel mapping between a thermal and optical camera. In the first step, a PCB needs to be designed that uses a microcontroller to hardware-synchronize the RGB and thermal camera to ensure that both images are triggered at the same time instant. Next, the intrinsics and extrinsics of the cameras can be obtained with an existing calibration framework. However, the calibration is not accurate enough for a pixel per pixel mapping. Therefore, using deep learning methods, the calibration accuracy is to be improved such that the optical and thermal image can be overlayed.
The developed hardware and algorithm will be mounted on a fixed-wing UAV for human detection in the context of search-and-rescue missions.
The goal of this project is to generate an accurate pixel to pixel mapping between a thermal and optical camera. In the first step, a PCB needs to be designed that uses a microcontroller to hardware-synchronize the RGB and thermal camera to ensure that both images are triggered at the same time instant. Next, the intrinsics and extrinsics of the cameras can be obtained with an existing calibration framework. However, the calibration is not accurate enough for a pixel per pixel mapping. Therefore, using deep learning methods, the calibration accuracy is to be improved such that the optical and thermal image can be overlayed.
The developed hardware and algorithm will be mounted on a fixed-wing UAV for human detection in the context of search-and-rescue missions.
- Development of a PCB that hardware synchronizes a RGB camera and a thermal camera (FLIR Tau 2) using a microcontroller
- Rough calibration of RGB-Thermal camera intrinsics and extrinsics
- Improved pixel per pixel mapping using deep learning methods
- Development of a PCB that hardware synchronizes a RGB camera and a thermal camera (FLIR Tau 2) using a microcontroller - Rough calibration of RGB-Thermal camera intrinsics and extrinsics - Improved pixel per pixel mapping using deep learning methods
Mandatory:
- Interest in PCB design, microcontrollers, and camera synchronization
- Courses in deep learning and computer vision
Beneficial:
- Experience with PCB design, microcontrollers, and camera synchronization
- Hands-on experience in deep learning and computer vision
Mandatory:
- Interest in PCB design, microcontrollers, and camera synchronization - Courses in deep learning and computer vision
Beneficial:
- Experience with PCB design, microcontrollers, and camera synchronization - Hands-on experience in deep learning and computer vision