3:00 PM - 3:30 PM
[15p-A32-5] Spatio-wavelength fusional perception based on hyperspectral imaging
Keywords:Spectral Imaging, Object detection
Although the pattern recognition technology is significantly advanced in imaging applications, there are still limitations. For example, in the disaster rescue or surveillance camera for security applications, small object detection from wide area is needed. However, the detection of small object is difficult because it does not provide enough shape features. Moreover, the training data is unavailable for machine learning if the detection of unspecific objects is required. Motion detection is used in such cases, but serious difficulty exists when background is moving or the target object is standing still. In this presentation, the possibility of advanced image perception system using the technology of spectral imaging. Specifically, small object detection from wide-area hyperspectral image and lesion detection and classification in dermatological image analysis are introduced. Applying the methods of anomaly detection, spectral unmixing, and spectral classification, it will realize highly-functional imaging system that breakthrough the limitation of the image recognition based on conventional RGB images.