A visual traffic data anomaly identification and position filling method

Abstract

This invention provides a method for visual flow data anomaly identification and supplementation, comprising: labeling stations with topological tags based on watershed topology; acquiring and preprocessing basic hydrological data of the stations; constructing candidate supplementation paths based on the preprocessed basic hydrological data, wherein the candidate supplementation paths include at least a path derived based on the water level-flow relationship and a predicted path based on historical water level-flow time series data; constructing an anomaly hierarchical identification mechanism to identify abnormal time series change characteristics of online flow data and scene characteristics of video monitoring images, generating corresponding abnormal time series feature labels and anomaly cause labels; selecting a matching supplementation method from the candidate supplementation paths based on the topological tags, abnormal time series feature labels, and anomaly cause labels, and outputting the corresponding flow supplementation data and anomaly cause. This invention can automate and intelligently identify and supplement visual flow data anomalies, improving the continuity and reliability of flow data in complex scenarios.

Images