A highway abnormal event detection method, device, equipment, medium and product

Abstract

The application discloses a highway abnormal event detection method, device, equipment, medium and product, relates to the technical field of data processing, and comprises the following steps: collecting video data and radar data of a target section of a highway; aligning the video data and the radar data by time stamp to obtain radar-video data; extracting multiple key frames from the video data based on a preset frequency; inputting the multiple key frames and the radar-video data into a pre-constructed abnormal event identification model to obtain an abnormal event detection result output by the abnormal event identification model; and inputting the abnormal event detection result and a preset dynamic instruction into a pre-constructed multi-modal large language model to obtain abnormal event work order information output by the multi-modal large language model; wherein the abnormal event work order information at least comprises an event number, an event time, an event location, an event type, an event description and an event object appearance feature. The application can improve the accuracy of detecting abnormal events on the highway.

Images