Design of Radiometric Thermography System for Object Recognition
The paper presents a custom-made radiometric thermography system which provides a full radiometric IR inspection. It contains heterogeneous stereo-vision system with RGB and thermal camera, as well as the processor unit with developed embedded software modules. The software units were realized using C++ programming language within ROS development environment. In order to obtain the better representation of points in the space as well as their projections on the cameras planes, the stereo calibration was performed. The correspondence between RGB and thermal images is represented with homography. The object detection is performed using OpenCV feature detectors, while the recognition is carried out by Hu moments computation and K-means clustering. For the showcase, incandescent light bulbs, LED light bulbs and quartz heaters are detected and recognized as such. The developed system is capable of providing a heterogeneous inspection of heating systems, power lines, etc.