Software

The web-based software developed within the BASTA project is intended for offshore survey projects, especially for munitions detection and monitoring campaigns. The AI-based analysis and visualization concept regarding UXO identification will be implemented into the AmuCad.org (Amunition Cadastre Sea) platform (www.amucad.org) developed by EGEOS GmbH. For the underlying data processing the technologies developed by the sensor data platform and maritime big data specialist TrueOcean (www.trueocean.io) will be used. The software performs four main tasks:

  • it provides a web-based management, storage and sharing platform for huge data sets
  • it evaluates the data quality in terms of the data quality parameters that are developed within the project
  • it offers a visualization platform for the data quality parameters and results that should support decision making for the user
  • it implements AI modules that should improve the detection rate of munition objects in the sea

A general overview of data processing steps that are planned for the software is shown in Figure 1. The main components for the BASTA project are the data management (data and metadata), storage, computation of data quality parameters, AI testing and the visualization (generation of map/image). These components are based on the following technologies. The storage and management will be handled using graph databases and cloud-based object storage. The processing will be performed using a Hadoop [1] cluster that is interfaced using Spark [2]. The visualization will probably be performed using GeoSpark that is now converted to Apache Sedona [3]. These are all open source software components that are supplied by the Apache software foundation [4].


Figure 1: General workflow for the data processing (Ⓒ EGEOS).

The data quality parameters can be important to compare different data sets or to evaluate whether these data are suitable to fulfill a specific goal, e.g. does the data allow detecting an object with a specific size. Therefore, the data quality parameters could be a direct output to the user for data assessment or might be used as additional information for the evaluation of the AI results. The quality parameters are defined in collaboration with the project stakeholders, through a questionnaire (here), in review of existing guidelines [5, 6] and based on theoretical values. An example of computed data quality parameters is illustrated in Figure 2 for multibeam and magnetic data acquired in the same area. The quality parameters add another dimension of information to the data that can be used for further analysis. An optimal solution for visualization and usage of this additional information will be developed within the project.


Figure 2: Example of layered data information computed for multibeam data (left) and magnetic data (right) in the same survey area. In both stacks the four bottom layers are exemplary data quality parameters. The second layer from top (water depth, residual field) are the actual data and the top layer (curvature, analytical signal) are postprocessing results of the data. Please note that the color scales are arbitrary (Ⓒ EGEOS).

The testing of AI is composed of three steps which are the preparation of input data, the AI algorithm (e.g. convolutional neural networks) and the output. The input is a crucial and important step, as there is the saying “garbage in, garbage out” in computer science. With increasing complexity, as for multi-sensor data (e.g. Multibeam, Side Scan Sonar, Magnetics, Subbottom Profiler), the verification of a “correct” input gets difficult and hence the data quality parameters could be a supportive tool for this matter.


[1] www.hadoop.apache.org

[4] www.apache.org

[5] Offshore Wind Accelerator (OWA), 2020. Guidance for geophysical surveying for UXO and boulders supporting cable installation

[6] IHO, 2020. International Hydrographic Organization Standards for Hydrographic Surveys, 6th Edition