基于Hbase的海底监测视频大数据存储方法

Big Data Storage Method for Seabed Monitoring Video Based on Hbase

摘要: 海底观测会产生规模大、稀疏性强的视频数据，目前的存储方案效率较低。为此，本文提出一种基于Hbase的海底监测视频大数据存储方法。首先，采用运动目标检测方法对视频中的稀疏空白部分和内容密集部分进行分段，对稀疏空白段进行重度压缩，并对密集内容部分进行特征提取，为基于内容的视频检索做准备。接着，设计一种能统一存储上述两部分视频的大数据表，它能在插入稀疏视频段时动态添加压缩信息列，在插入密集内容视频时动态添加特征列，并用行键键尾对两种视频进行了区分，方便查询过滤。实验分析表明：本文方法相比传统方式能节约近75%的存储资源，并能实现基于特征的检索，极大的方便了视频数据的管理。

Abstract: Submarine observations produce large-scale, sparse video data, and current storage schemes are less efficient. To this end, a big data storage method for seabed monitoring video based on Hbase is proposed. First, the moving object detection method is used to segment the video to high-density and low-density parts, so the low-density parts can be greatly compressed, at the same time, feature extraction is performed on the dense content part to facilitate content-based video retrieval. Then, a big data table is designed to uniformly store the above two parts of the video, it can dynamically add compressed information columns when inserting low-density video parts and feature columns when inserting high-density video parts, also we use the trailing portion of the row-key to identify them, which facilitates query and filtering. Experiment results show that the proposed method can save nearly 75% of the storage resources compared with traditional methods. Besides, it can also provide feature-based retrieval, which greatly facilitates the management of video data.