Customizable interactive video production method and device based on UCL

Abstract

The invention discloses a customizable interactive video production method and device based on UCL. The method comprises the following steps: firstly, improving a Faster R-CNN target detection networkmodel to achieve an application effect of performing fine-grained category detection and accurate recognition on a target appearing in a video; segmenting the video, extracting a plurality of features of a video segment, and extracting semantic information of each video segment by adopting a video semantic extraction model based on LSTM; and finally, semantic indexing is performed on the video byusing a unified content label UCL, and a knowledge base based on the UCL is constructed, so that a customizable interactive video application can be supported to perform automatic retrieval and the like on detailed information of a target appearing in the video. According to the invention, automatic production of a customizable interactive video can be realized, and timeliness and flexibility ofinformation embedded in the video and semantic relevance and on-demand customizability of video target information in an interactive video application can be ensured.

Description

technical field [0001] The invention relates to a UCL (Uniform Content Label, uniform content label)-based method and device for making a customizable interactive video, belonging to the field of video image processing. Background technique [0002] The vigorous development of the network video industry has promoted the continuous innovation of video multimedia applications. Compared with the traditional sense, where users passively accept all video content, people seek a more active and interactive way to understand and discover the content information in the video, and at the same time, can make redundant information in the video or information that users do not need. to filter and filter. Interactive video is one such new application of video. At present, interactive videos can be divided into three categories: the first category is customizable, which allows users to adjust some variable elements of the video while watching, but the video itself is linear. For example...

AI Technical Summary

Problems solved by technology

However, currently many interactive video production tools can be used to produce customizable videos, but video producers need to manually embed hyperlinks of items or person information into the video, which is time-consuming and laborious

And once the information is embedded in the video, it should not be changed, and the timeliness and flexibility are greatly reduced

[0003] In terms of technology, in recent years, domestic and foreign scholars have made some progress in target detection and knowledge base construction, but there are still some problems in practical application. First, the target detection algorithm of the candidate area and the target detection algorithm based on regression, Although it can achieve high detection accuracy and fast detection speed when detecting the target, it cannot detect the detailed category information of the target. For example, it can be detected that the object is a bottle of Coke through target detection, but it cannot be detected. specific brand of cola

Images