Crop Yield Forecasting via Coarse-to-Fine Resolution Segmentation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Conventional crop yield forecasting methods are highly dependent on ground truth data, are computationally intensive, and costly, making them inefficient and scalable, especially in adverse weather conditions.
Innovation Solution
A processor-implemented method and system for high-resolution and scalable crop yield forecasting that uses satellite, weather, and soil data to develop two-stage crop yield forecasting models, employing preprocessing, feature selection, and stratified random sampling to generate coarse and high-resolution maps without requiring extensive ground data.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional remote sensing and modeling based methods use physical or process based models, then precision or accuracy of yield estimation is improved, but data dependency and scalability become major challenges
Solution Approach 1:
The patent segments the yield estimation process into two distinct stages: a first stage using physical/process-based models for coarse-resolution estimation, and a second stage using machine learning models for high-resolution refinement. This segmentation allows each stage to operate with different data requirements and complexity levels, reducing overall data dependency while maintaining accuracy.
Solution Approach 2:
The patent introduces coarse-resolution yield maps as an intermediary product between traditional modeling and final high-resolution yield estimation. These intermediate maps serve as training data for machine learning models, enabling the system to leverage both physical understanding and data-driven approaches without requiring extensive ground truth data at the highest resolution.
2Measurement precision
If conventional methods use ground truth data for model training, then forecasting accuracy is improved, but cost and difficulty of data collection increase significantly
Solution Approach 1:
The patent creates synthetic ground truth data by generating coarse-resolution yield maps through physical models and satellite observations. These synthesized maps serve as training data for machine learning models, replacing the need for extensive actual ground truth measurements while still providing realistic training examples for accurate forecasting.
Solution Approach 2:
The patent performs preliminary yield estimation at coarse resolution using physical models before conducting detailed high-resolution analysis. This preliminary action generates training data in advance, reducing the need for costly ground truth collection during the actual forecasting process while maintaining model accuracy.
3Manufacturing precision
If high resolution crop yield maps are generated directly, then spatial detail is improved, but computational intensity and processing time increase
Solution Approach 1:
The patent segments the resolution enhancement process into two stages: first generating coarse-resolution maps quickly using physical models, then applying machine learning-based super-resolution to achieve high spatial detail. This segmentation avoids the computational burden of directly generating high-resolution maps while maintaining the desired spatial precision.
Solution Approach 2:
The patent performs preliminary processing at lower resolution to generate training data and model parameters before conducting the final high-resolution yield estimation. This preliminary action at coarser scales reduces computational intensity while preserving the capability to produce detailed final results.
4Reliability
If extensive ground truth data is collected for model training, then model reliability in adverse conditions is improved, but cost and operational complexity increase
Solution Approach 1:
The patent extracts the essential training requirements by identifying that coarse-resolution physical model outputs and satellite observations are sufficient for training machine learning models. This extraction eliminates the need for extensive ground truth data collection operations while maintaining model reliability, particularly in adverse weather conditions where ground data collection is most difficult.
Data Source
AI summary
This disclosure relates to methods and systems for high resolution and scalable crop yield forecasting by first developing a first crop yield forecasting model to generate coarse resolution yield maps and further dynamically selecting a set of pixels from the coarse resolution yield maps. The coarse resolution yield maps, satellite, weather and soil related data are fed as input to a second crop yield forecasting to generate high resolution crop yield forecasting maps. Further, domain knowledge about crop growth stages, economically important crop growth stages and weather based triggers are identified to quantify extent of change in crop yield. This helps in crop yield forecasting during real time adverse weather conditions. Finally, an adjusted crop yield model is obtained after adjusting losses incurred due to the real time adverse weather conditions to obtain accurate high resolution crop yield forecasting maps. The method of present disclosure is inexpensive, light-weight, and scalable.