Adaptive intelligent generation method of picture-text video thumbnails based on query terms

Abstract

The invention discloses a self-adaptive intelligent generation method of picture-text video thumbnails based on query words. The method comprises the following steps: acquiring an object video, extracting the object video, and obtaining audio and video information in the video; performing structured processing on the audio and video information to obtain structured video data and structured audiodata; selecting the structured video data to obtain a key frame corresponding to the query keyword semantics, namely a visual element; extracting and processing the structured audio data to obtain text elements related to query keyword semantics; obtaining visual elements and text elements for dynamic composition processing, and obtaining picture-text video thumbnails; the thumbnails are obtained,the semantic text of the thumbnails is extracted, the global color matching monitoring process is carried out, and the target video thumbnails related to the semantics of the query keywords are obtained. The embodiment of the invention can intelligently generate video thumbnails according to query keywords according to a system, thereby saving human resources and having more purposefulness than the existing automatic generation technology of video thumbnails.

Description

technical field [0001] The invention relates to the technical fields of algorithm optimization and analysis, computer image processing and deep learning, in particular to a method for adaptively and intelligently generating graphic and video thumbnails based on query words. Background technique [0002] Video thumbnails play a vital role in how users find videos and evaluate content relevance. The selection of thumbnails is not only related to the magnitude of video transmission in social networks, but also seriously affects the instant gratification of online viewers' information acquisition. However, selecting the most suitable thumbnail from thousands of video frames is a time-consuming and skillful manual process. Although there are currently some methods that can automatically or semi-automatically extract thumbnails from videos, there are still great limitations in specific scenarios, such as when the user needs to specify a video thumbnail related to a certain topic ...

AI Technical Summary

Problems solved by technology

[0004] The aforementioned technical shortcomings are that the design of the recognition module is complicated, and the definition of key frames is vague; key frames cannot be selected according to the wishes of end users, that is, there are discrepancies between the key frames understood by the machine and the key frames understood by the text. May not be what the end user wants; dynamic video thumbnails have disadvantages in terms of large-scale storage issues, taking up more storage space than a single frame

Images