Digital Terrain Model Correction Using Satellite Imagery and Machine Learning
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for creating digital terrain models (DTMs) under forest canopies are inaccurate and costly due to reliance on LiDAR technology, which is expensive and not readily available for remote areas, and conventional photo-interpretation methods result in high errors and low resolution.
Innovation Solution
A method that calculates a digital terrain model (DTM) using a digital elevation model (DEM) and digital surface model (DSM), correcting elevation errors caused by land cover and terrain curvature using imagery and ancillary data, and applying statistical or machine learning models to predict local elevation corrections at each target point.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If LiDAR is used to produce accurate DTMs under forest canopies, then measurement precision is improved, but cost and data availability worsen
Solution Approach 1:
The patent replaces expensive LiDAR systems with free or low-cost satellite imagery (Sentinel-2, Landsat) and open-source DEM data. This substitution uses inexpensive, readily available data sources to achieve comparable DTM accuracy without the prohibitive costs of LiDAR missions.
Solution Approach 2:
The patent replaces the active sensing mechanism of LiDAR (laser ranging) with passive optical sensing using satellite imagery. The system substitutes mechanical/optical ranging equipment with image processing and statistical modeling approaches, eliminating the need for specialized LiDAR hardware.
2Ease of manufacture
If photo-interpretation methods are used to create DTMs, then cost is reduced, but measurement precision and resolution worsen
Solution Approach 1:
The patent replaces manual photo-interpretation with automated computer-based image processing. Statistical models and machine learning algorithms automatically analyze satellite imagery to extract elevation information, eliminating manual labor while significantly improving accuracy through systematic computational methods.
Solution Approach 2:
The patent transforms the approach by changing from direct visual interpretation to statistical parameter analysis. Instead of manually reading contours, the system uses statistical relationships between image brightness, terrain curvature, and elevation to automatically predict DTM values with higher precision.
3Measurement precision
If LiDAR is flown at low altitude to achieve high accuracy, then measurement precision is improved, but productivity worsens
Solution Approach 1:
The patent makes the system universally applicable by using satellite imagery that covers entire regions simultaneously. Unlike low-altitude LiDAR that must fly multiple flight lines, the satellite-based approach processes entire territories in single passes, dramatically increasing coverage speed while maintaining accuracy through statistical modeling.
4Measurement precision
If manual classification of LiDAR returns is performed, then measurement precision is improved, but loss of time and cost worsen
Solution Approach 1:
The patent enables the system to automatically classify terrain features without human intervention. Statistical models and machine learning algorithms autonomously distinguish ground from non-ground points by analyzing image characteristics and terrain curvature, eliminating the need for manual technician review while maintaining classification accuracy.
Data Source
AI summary
A method and system for calculating a digital terrain model (DIM) for a target portion of the surface of the Earth. A digital elevation model (DEM) for the target portion specifies an elevation for target points on the Earth within the portion. A digital surface model (DSM) specifies the elevation above the target point of an obstructing surface. Elevation errors in the DEM are corrected. A curvature correction is done using the DSM. A model is calibrated using reference points using statistical techniques or machine learning. A model using reference points for predicting the amount of local elevation correction needed at each target point as a function of terrain curvature is employed. The models are applied at each target point of the DEM to produce the DTM.