Chemical industrial park hazard source sensing and monitoring method and system based on AI video identification

The invention relates to the technical field of image processing, and provides a chemical industrial park hazard source sensing and monitoring method and system based on AI video identification, and the method comprises the steps: firstly, extracting detailed features and context correlation information in an image through an image description mining technology, and providing comprehensive and accurate data support for subsequent hazard source sensing and discrimination; secondly, the property and the influence range of the abnormal state can be understood more deeply by identifying the abnormal state monitoring image block and acquiring focus description knowledge data; and finally, the information of the global image representation vector is comprehensively considered, so that the accuracy and the reliability of danger source perception judgment can be improved. Therefore, the method has important practical application value and popularization prospect in the aspect of hazard source perception and discrimin