|
|
Detection of People in Room Using Low-Cost Thermal Imaging Camera
Charvát, Michal ; Kempter, Guido (oponent) ; Drahanský, Martin (vedoucí práce)
As we are approaching times of Industry 4.0 and smart homes, there is a need for automatic reliable mechanisms for detecting living beings. There are many use cases---from aiding elderly or disabled people with their everyday life to increasing safety in dangerous workplaces by guarding hazardous areas. We propose a method for counting, detecting and locating people using a single low-cost thermal camera module FLIR Lepton 3 and Orange Pi computer. This paper describes the process of necessary hardware configuration of the whole system and a software solution to the problem of detecting and locating people. The method, we utilize in order to detect and locate people, is based on applying human temperature filter, image processing with object detection using OpenCV library, and 3D scene reconstruction with known environment parameters. This approach, even in its simplest form, provides accuracy around 90 % on our data set with various possibilities for improvement.
|
|
|
Detection of People in Room Using Low-Cost Thermal Imaging Camera
Charvát, Michal ; Kempter, Guido (oponent) ; Drahanský, Martin (vedoucí práce)
As we are approaching times of Industry 4.0 and smart homes, there is a need for automatic reliable mechanisms for detecting living beings. There are many use cases---from aiding elderly or disabled people with their everyday life to increasing safety in dangerous workplaces by guarding hazardous areas. We propose a method for counting, detecting and locating people using a single low-cost thermal camera module FLIR Lepton 3 and Orange Pi computer. This paper describes the process of necessary hardware configuration of the whole system and a software solution to the problem of detecting and locating people. The method, we utilize in order to detect and locate people, is based on applying human temperature filter, image processing with object detection using OpenCV library, and 3D scene reconstruction with known environment parameters. This approach, even in its simplest form, provides accuracy around 90 % on our data set with various possibilities for improvement.
|
host ::
RSS.