How to analyze large sonar data sets with machine learning
Together with Norwegian Geological Survey and Lundin Norway AS, TerraNor has developed routines to do map and analyse large datasets from camera and sonar. In this webinar they show how to process large sonar datasets with a machine learning solution in eCognition Developer. The data from sonar sensors are bathymetry and backscatter. Sonar is very similar to radar and the input data can be sidescan or multibeam sonar. Based on sonar, we create several derivatives: slope, aspect, northness, bathymetric position index, roughness and rugosity. The total number of layers used in the analysis can be 20 or more. eCognition provides the tools required to fuse and communicate between these complex datasets.
eCognition goes through 3 general steps to do the analysis:
- Create and control the derivatives from bathymetry and backscatter
- Train the machine learning algorithm based on samples and input data
- Classify based on the trained dataset. Classification is user defined and often focuses on geology, but can also consider fauna.
We will show how the approach works and additional products like UXO maps and coral maps.
Was this article helpful?
Articles in this section
- Data Fusion Approaches to Tree Canopy Change Detection
- Welcome to eCognition 10
- Timber Cruising with eCognition
- eCognition Developer for Beginners
- What’s New in eCognition 9.5
- Go Big with eCognition Server
- Agricultural Field Boundary Delineation with Multi Temporal Sentinel 2 Imagery
- Deep Learning, UAVs and Precision Agriculture
- Counting Features with Template Matching
- The Future of Vegetation Analyses in Feeder Corridors with YellowScan & Trimble