Cloud Shadow Detection in Satellite Imagery Using ML
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Satellite imagery for agronomic fields is often compromised by clouds and cloud shadows, which can lead to inaccurate vegetative index values and hinder modeling accuracy, as existing methods struggle to accurately detect thinner clouds and their boundaries.
Innovation Solution
A machine learning system, such as a convolutional encoder-decoder, is trained to identify cloud and cloud shadow pixels by leveraging surrounding pixel information, with a second system using candidate shadow locations to enhance detection accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional classification techniques are used to identify clouds and cloud shadows, then detection speed is maintained, but detection accuracy deteriorates for thinner clouds and cloud boundaries
Solution Approach 1:
The patent segments the cloud detection task into multiple processing stages: initial cloud shadow candidate identification, refinement through machine learning classification, and boundary optimization. This segmentation allows the system to achieve high accuracy for thin clouds by progressively refining detections rather than attempting single-step classification.
Solution Approach 2:
The patent employs dynamic adjustment of detection parameters and iterative refinement processes. The machine learning model dynamically adjusts classification thresholds and the system iteratively improves boundary detection by considering surrounding pixel information, enabling adaptive optimization for varying cloud thicknesses and conditions.
2Measurement precision
If machine learning systems leverage surrounding pixel information to identify cloud pixels, then detection accuracy improves, but computational complexity increases
Solution Approach 1:
The patent applies local quality analysis by examining surrounding pixels in the vicinity of cloud shadow candidates. The machine learning model processes local neighborhoods of pixels to determine cloud presence, focusing computational resources on relevant local regions rather than processing entire images uniformly, thus improving accuracy while managing computational load.
3Measurement precision
If a second machine learning system is used to identify cloud shadow locations, then shadow detection accuracy improves, but system complexity and processing time increase
Solution Approach 1:
The patent implements preliminary action by first identifying cloud shadow candidate locations using geometric relationships and metadata before applying the second machine learning system. This pre-filtering step reduces the number of pixels requiring intensive ML processing, thereby improving shadow detection accuracy while minimizing the time penalty of using a complex two-system approach.
Data Source
Figure 1
Figure 2(a)~2(b)
Figure 3
AI summary
Systems and methods for identifying clouds and cloud shadows in satellite imagery are described herein. In an embodiment, a system receives a plurality of images of agronomic fields produced using one or more frequency bands. The system also receives corresponding data identifying cloud and cloud shadow locations in the images. The system trains. a machine learning system to identify at least cloud locations using the images as inputs and at least data identifying pixels as cloud pixels or non-cloud pixels as outputs. When the system receives one or more particular images of a particular agronomic field produced using the one or more frequency bands, the system uses the one or more particular images as inputs into the machine learning system to identify a plurality of pixels in the one or more particular images as particular cloud locations.