音视频大数据样本库入库规范

Entry Specification of Audio and Video Big Data Sample Library

当前各种音视频数据呈现几何级数的增长,有效组织庞大的音视频数据的难度增大,为了合理组织音视频大数据,提出了一种音视频数据入库规范。首先对原始音视频数据文件按照不同类别进行切分存储,将拆分数据段的类别、起止时间等信息存储在数据库表中。根据数据表信息,有针对性的调用某一类的音视频段,实现灵活的拼接拆分。建立大规模图片样本库,支持视频图像处理时的目标检测等任务。通过大量的试验证明,提出的样本库规范能实现音视频数据的正常入库和灵活调用。

At present,various audio and video data show a geometric increase,and it is more difficult to effectively organize huge audio and video data.In order to organize audio and video big data reasonably,we propose a standard for audio and video data storage.First,the original audio and video data files are segmented and stored according to different categories,and information such as the category,start and end time of the split data segment is stored in a database table.According to the information in the data table,a certain type of audio and video segments are called in a targeted manner to realize flexible splicing and splitting.A large-scale picture sample library should be established to support tasks such as target detection in video image processing.A large number of experiments have completed to prove that the sample library specification proposed in this paper can realize the normal storage and flexible calling of audio and video data.