A project whole-process data tracing method and system based on video semantic analysis

Abstract

The present application relates to the technical field of video information retrieval, in particular to a project whole-process data tracing method and system based on video semantic analysis, which obtains continuous video stream of engineering site through wearable collection equipment and fixed camera device, analyzes the synchronous relationship between engineering components and equipment of workers in video frames combined with timestamp information, forms a semantic video segment set by performing double constraint screening according to object boundary integrity and semantic time length, inputs image target detection and voice recognition model to extract work content and build a semantic event sequence, and then establishes event chain index and generates jump positioning structure according to the search request. In the present application, the unstructured video stream is analyzed into semantic event chain with time and behavior relationship, continuous tracking and accurate retrieval of project process are realized, automatic identification and high-consistency restoration of cross-stage work behavior are supported, and the data structuring level and process visualization ability are significantly improved.

Images