Satellite-Altimeter Bathymetry Mapping With Neural Depth Prediction
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing bathymetric survey methods face challenges in accurately determining sea floor depth, especially in intricate underwater topographies, lacking a scalable and economically viable alternative.
Innovation Solution
A deep learning framework integrates high-resolution satellite imagery with precise altimeter data using a neural network to generate detailed and accurate bathymetric maps, achieving an RMSE of less than 30 cm.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If conventional bathymetric survey methods are used, then measurement precision can be achieved, but device complexity and cost increase significantly
Solution Approach 1:
The patent replaces traditional mechanical bathymetric survey systems (sonar, echo sounders) with a computational approach using satellite imagery and machine learning algorithms. The neural network model processes satellite image data to predict bathymetric values, eliminating the need for complex physical survey equipment while achieving comparable or superior measurement precision.
Solution Approach 2:
The patent introduces satellite imagery as an intermediary medium between the observation platform and the target object (sea floor). Instead of direct mechanical measurement, the system uses satellite images as intermediate data that are then processed through machine learning models to infer bathymetric information, simplifying the overall measurement system.
2Measurement precision
If traditional bathymetric survey methods are used, then measurement precision can be achieved, but loss of time and productivity decrease
Solution Approach 1:
The patent performs preliminary actions by pre-training the neural network model with extensive bathymetric data before actual survey operations. This pre-processing of computational models allows the system to rapidly process satellite imagery during actual surveys without requiring time-consuming real-time calculations, significantly improving productivity while maintaining precision.
Solution Approach 2:
The replacement of mechanical survey methods with satellite imagery processing and machine learning enables parallel processing of large areas, dramatically increasing the rate of data collection while maintaining measurement accuracy through the trained neural network model.
3Measurement precision
If extensive data collection is performed to improve accuracy, then measurement precision increases, but loss of time and cost increase
Solution Approach 1:
The patent uses a trained neural network model that has been excessively trained on comprehensive datasets during the training phase. This allows the model to achieve high measurement precision when processing actual survey data with minimal input requirements, reducing the need for extensive real-time data collection while maintaining accuracy.
4Reliability
If conventional survey methods are used, then reliability of bathymetric data can be ensured, but ease of operation and accessibility decrease
Solution Approach 1:
The patent implements a self-service system where the neural network model automatically processes satellite imagery and generates bathymetric maps without requiring manual survey operations. The system performs self-calibration and self-processing, eliminating the need for skilled operators to conduct complex field surveys while ensuring data reliability through the trained model.
Data Source
AI summary
Method, system, and non-transitory computer readable storage medium are provided for creating a shelf bathymetric map. In some embodiments, a satellite image with a shelf area and related altimeter data for at least a number of points of the shelf area are obtained. The shelf area in the satellite image and depth values for the shelf area from said altimeter data are selected. The selected depth values are associated with points of the shelf area based on geo-coordinates. A training dataset and a testing dataset are created from the associated points. A neural network is trained on the training dataset. The neural network is tested on the testing dataset. Inference of the neural network is adjusted based on testing results. A depth value for each satellite image pixel of the shelf area is predicted by the neural network.