Agricultural Field Pattern Detection Using Fourier Masking
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing remote imagery technologies for agricultural fields face challenges in reliably detecting patterns due to non-stationary crop growth and noise, leading to reduced effectiveness in decision-making support tools, and existing methods require significant computational effort or suffer from bias.
Innovation Solution
A computer-implemented method using Fourier Transform to process remote image data, applying a mask function to identify pattern elements with sub-pixel precision, adjusting pixel values, and determining vegetation indices to filter out noise, allowing for precise pattern detection and improved agricultural management.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Difficulty of detecting and measuring
If traditional image processing methods (filtering, thresholding) are used for feature detection, then man-made structures can be detected, but agricultural fields with non-stationary crops cannot be reliably analyzed due to increasing complexity
Solution Approach 1:
The patent transforms the image data from spatial domain to frequency domain using Fourier Transform, changing the parameters of analysis from pixel intensity values to frequency components. This allows periodic patterns in agricultural fields to be detected as distinct frequency signatures, making non-stationary crop patterns analyzable in the same way as stationary man-made structures.
Solution Approach 2:
The patent replaces traditional mechanical image processing operations (filtering, thresholding) with a mathematical transformation approach (Fourier Transform). This substitution enables the detection of periodic patterns through frequency domain analysis, which is more effective for agricultural fields where crops exhibit periodic growth patterns and phenological stages.
2Reliability
If AI-based or machine learning approaches are used for pattern detection, then pattern recognition can be achieved, but computational effort increases significantly and bias problems arise
Solution Approach 1:
The patent replaces AI-based machine learning approaches with a mathematical physics-based method (Fourier Transform). This substitution eliminates the need for training data and complex neural networks, reducing computational effort significantly while maintaining reliability through deterministic mathematical operations that do not suffer from bias problems inherent in machine learning models.
Solution Approach 2:
The Fourier Transform method is self-service in nature, requiring no external training data or model adjustment. The method automatically adapts to any agricultural field pattern through the mathematical transformation, making it universally applicable without the computational overhead and bias issues of AI-based approaches that require extensive training.
3Measurement precision
If spatial resolution of satellite imagery is increased, then detail accuracy improves, but computational requirements and appearance of patterns/artifacts increase
Solution Approach 1:
The patent uses Fourier Transform to substitute traditional computational image processing. This mathematical transformation approach is more efficient than brute-force computational methods, allowing high-resolution image analysis without proportionally increasing computational requirements. The frequency domain representation enables pattern detection that scales more favorably with resolution.
Solution Approach 2:
By transforming image data to the frequency domain, the patent changes the parameters of analysis to expose periodic patterns that are more easily detected at high resolutions. This parameter transformation allows the system to maintain computational efficiency while leveraging high spatial resolution for more precise pattern detection and agricultural analysis.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
Enables reliable detection of patterns in agricultural fields with sub-pixel accuracy, reducing computational effort, and providing accurate vegetation indices to support informed agricultural practices such as fertilizer application and irrigation.
Implementation Method 1
applying a Fourier Transform to the pixel values of the subset of contiguous pixels
Data Source
AI summary
A computer-implemented method for detecting patterns in agricultural fields, the method including receiving remote image data of an agricultural field including a plurality of pixels, each pixel including at least one pixel value representative of the reflectance or emittance of at least one wavelength band; processing the pixel values for at least a subset of contiguous pixels in the received remote image data, including applying a Fourier Transform to the pixel values of the subset of contiguous pixels; processing the Fourier Transform output data to determine an offset value representing the distance of the center of a pixel to a nearest pattern element; generating a mask function, the mask function including set values for the processed pixels determined based on the offset value of the processed pixels; and determining the pixels containing pattern elements based on the mask function.