基于深度学习的短视频分类技术及监控预警系统

Short Video Classification Technology and Monitoring and Early Warning System Based on Deep Learning

针对目前互联网上传播的海量短视频,设计并实现了一套短视频文本检测预警系统。该系统通过检测短视频中文本的位置,利用OCR识别技术对文本内容进行识别,运用NLP技术对文本内容进行校正、理解和匹配,最后上传检测结果,实现了对短视频中文本信息的结构化。系统可以通过上传短视频来进行短视频文本内容的结构化以及分类,同时对含敏感信息的视频进行预警上报,对非敏感短视频确定相应的标签以进行后续搜索、推荐等附加服务。测试结果表明:短视频分类技术及监控预警系统的分类准确率为96.3%,召回率为93.7%,单帧图像响应时间快,软件成熟度评估达到CMMI3级,兼容性良好。

In view of the massive short videos currently spread on the Internet,a set of short vid-co text detection and early warning system was designed and implemented.The system detects the position of short video Chinese texts,uses OCR recognition technology to identify text con-tent,uses NLP technology to correct,understand and match text content,and finally uploads the detection results to realize the structuring of short video Chinese text information.The system can structure and classify the text content of the short video by uploading the short video,and at the same time report the video containing sensitive information,and determine the correspond-ing tag for the non-sensitive short vidco for subsequent search,recommendation and other addi-tional services.The test results show that the classification accuracy of short video classification technology and monitoring and carly warning system is 96.3%,the rccall rate is 93.7%,the rc-sponse time of single-frame image is fast,and the software maturity evaluation reaches CMMI3 level,and the compatibility is good.