Video understanding method and device

Abstract

The invention proposes a video understanding method and device, and the method comprises the steps: obtaining video data; carrying out the scene depth estimation of the obtained video data; and carrying out the understanding of scene contents based on the scene depth estimation result. According to the embodiment of the invention, the method achieves the analysis and understanding of the scene contents according to the depth information of the scene through the depth estimation of the scene, and can achieve the efficient and accurate recognition and location of an object in the video contents.

Description

technical field [0001] The present invention relates to the technical field of video processing, and more particularly to a video understanding method and device. Background technique [0002] The structuring of video data, through the intelligent understanding of video content, is of great value for video advertising, intelligent security, etc. For example, for video content on the Internet (such as TV programs), through the intelligent processing of video, a collection of objects in some videos can be obtained (for example, a car begins to appear at a certain position in the tth frame of the video until t+ end of frame k, where t and k are natural numbers). Links to advertisements can be made through the understanding of video content. On the other hand, in security scenarios, such as a bicycle being lost, the bicycle can be found through the structured data of the video, and then locked at a certain moment when it was stolen by someone. The applications of intelligent ...

AI Technical Summary

Problems solved by technology

However, when the amount of video data increases rapidly with the explosive growth of the number of cameras, it is no longer feasible to understand and mine video data by manpower

Existing intelligent algorithms often analyze and understand video content directly, but when the video scene is noisy, it is often difficult to accurately locate objects in the video content

Images