Document-Oriented Data Organization for Unmanned Aerial Vehicle Outputs

Abstract

Three-Dimensional (3D) point cloud is considered an important geospatial resource for a vast range of applications. This is due to the rapid technology on data acquisition, such as Unmanned Aerial Vehicle (UAV), Mobile Laser Scanning, and Terrestrial Laser Scanning (TLS). Yet efforts to exploit the use of these datasets are increasingly threatened by the massive dataset, data density, and data complexity. Traditional Relational Database Management System (RDBMS) existed years ago, but the capability of relational databases in handling these issues is questionable due to several drawbacks and limitations. To address these challenges, effective storage, querying, and organization is required. Document-oriented databases are becoming more prominent compared to relational database, since it is capable of handling petabytes of data emerging from the Big Data scheme. Thus, this study investigates the capability of the document-oriented database in organizing UAV outputs, such as images and point clouds, via a NoSQL database. There are 103,996,984-point clouds generated from UAV images and stored in the database. Several tests, such as time analysis, insert operation, and storage consumption, are performed and compared with the RDBMS. The results show that the document-oriented database outperforms the relational database during data retrieval, where the document-oriented database response is 37% faster than the relational database. Meanwhile for data updating, the document-oriented database response is 30% faster than a relational database. To retrieve the stored point cloud in the database, the Potree viewer is used to render the data on the web browser. Based on the result, the point cloud data is successfully rendered and can be manipulated for future applications.