Image Analysis - Artificial Intelligence
Most autonomous systems rely on images to understand complex environments and make decisions. As part of the RISE digitalization, we carry out image analysis studies that utilize both traditional and AI methods to solve practical problems in various application areas. Examples include recognition of human behaviour and tracking of traffic objects.
Image analysis deals with algorithms and processes to gain different levels of understandings from digital images or video sequences. The tasks include acquisition, preparation, processing, analysis and understanding of information from visual sensors.
Recent years have seen the emergence of deep learning techniques in the field. Artificial intelligence hade started to outperform traditional techniques in different processes, including feature extraction and various analyses. This opens up opportunities to solve existing problems and also creates challenges for researchs and innovations. Open research questions include: (i) how to deal with limited amount of training data, (ii) high labour cost of training data annotation, (iii) understandability of the system in collaborations with human and other systems, etc.
RISE has research experiences of using both traditional and artificial intelligence techniques in image analysis to align with industrial needs in the following areas:
- Traffic behavior modeling, recognition and prediction
- Segmentation, classification, clustering, anomaly detection
- Multimodal data analysis
- Human and AI interactions
- Visual data privacy protection