Cloud Detection in Single Panchromatic Images Using Feature-Space Thresholds

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current cloud detection methods in digital images, particularly in satellite imagery, are inadequate for accurate automated filtering, often requiring human intervention and relying on multiple images or complex processing techniques like stereo imaging and fractal dimensions, which are not always feasible or efficient.

Innovation Solution

A computer-implemented method for identifying clouds in a single panchromatic or monochromatic digital image using homogeneity and brightness features, without the need for stereo imaging, different angles, wavelengths, or thermal data, by applying feature-space transformations and thresholds to distinguish cloud pixels from background pixels.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If traditional cloud detection methods are used, then cloud detection capability is provided, but accuracy is insufficient and human intervention is required

Engineering Contradiction:
Improvecloud detection accuracyVSAvoidautomated filtering capability
Core Design Contradiction:
Measurement precisionVSExtent of automation

Solution Approach 1:

The patent transforms the input image into a new feature space using mathematical transformations (e.g., logarithmic transformation of normalized difference cloud index). This changes the parameter representation of cloud features, making them more distinguishable from background features in the transformed space, thereby improving detection accuracy and enabling automated filtering

Inventive Principle:
Principle #35Parameter changes

Solution Approach 2:

The patent replaces manual human inspection and filtering with an automated computer-implemented method. The system uses algorithmic processing of image features, automatic threshold determination, and computational classification to substitute the mechanical human operation, achieving both high accuracy and full automation

Inventive Principle:
Principle #28Mechanics substitution (Replace mechanical system)

2Measurement precision

If multiple images or complex processing techniques are used, then cloud detection capability is improved, but device complexity increases

Engineering Contradiction:
Improvecloud detection accuracyVSAvoidprocessing system complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The patent extracts only the essential cloud detection features (brightness and homogeneity) from the image data, transforming them into a simplified feature space. This extraction approach avoids the need for complex stereo imaging, multiple wavelengths, or thermal data, reducing device complexity while maintaining detection accuracy through focused feature analysis

Inventive Principle:
Principle #2Taking out (Extraction)

3Productivity

If traditional cloud detection methods are used, then cloud detection is provided, but processing efficiency is low and resource-intensive

Engineering Contradiction:
Improveprocessing efficiencyVSAvoidcomputational resource consumption
Core Design Contradiction:
ProductivityVSUse of energy by stationary object

Solution Approach 1:

The patent applies mathematical transformations to convert raw image parameters into a optimized feature space where cloud detection is more efficient. The transformation of normalized difference cloud index using logarithmic functions and the creation of new feature combinations streamline the detection process, reducing computational iterations and resource consumption while improving processing speed

Inventive Principle:
Principle #35Parameter changes

Data Source

PatentEP3408828B1Systems and methods for detecting imaged clouds
Publication Date: 2024.05.15 ISRAEL AEROSPACE IND LTD
  • EP3408828B1 patent drawingFigure 1
  • EP3408828B1 patent drawingFigure 2
  • EP3408828B1 patent drawingFigure 3

AI summary

A computer-implemented method for identifying clouds in a digital image, comprising pixels, of a scene, the method comprising quantifying pixel-level characteristic/s in each of a multiplicity of pixels within a digital image of a scene; comparing function/s of the pixel-level characteristic/s to threshold/s thereby to generate comparison result/s; and using a controller for generating an output identifying clouds in the digital image, including identifying presence of cloudiness at at least one first pixel in the digital image, at least partly because the at least one comparison result indicates that the first pixel falls below the threshold/s, and identifying absence of cloudiness at at least one second pixel in the digital image, at least partly because the at least one comparison result indicates that the second pixel exceeds the threshold/s.