| Abstract: | There are several Stream Processing (SP) and Complex Event Processing (CEP) frameworks for situation analysis and monitoring for sensor data. These frameworks are not appropriate for analyzing video situations since video data characteristics are quite different from traditional sensor data. This work proposes an approach to situation analysis of videos such as CCTV footage by extracting video contents at the frame level. As video content extraction (e.g., object detection) has achieved tremendous success with neural networks recently, it is possible to perform video situation analysis by leveraging SP and CEP frameworks. This work presents a principled approach for expressing different types of video situations using a non-procedural video query language (termed CQL-VA) based on continuous queries. The data representation, classes of queries to be supported and querying of situations are discussed in this work. We believe the approach presented will help extract meaningful situations flexibly and efficiently from the contents extracted from videos. This is a step to move from customized situation detection to non-procedural query-based situation detection with tremendous advantages |