UAS Genotype Analysis Using Machine Learning Yield Forecasting
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current plant breeding programs face challenges in efficiently and timely measuring crop characteristics over large areas, such as plant height and yield potential, due to labor-intensive methods and limited experimental field sizes, which impede genotype analysis and selection.
Innovation Solution
The implementation of a computer-implemented system using unmanned aerial systems (UAS) to capture images and sensor data, which are processed using machine learning routines, like artificial neural networks, to forecast crop yields and compare genotype performance across different field conditions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If traditional manual methods are used to measure crop characteristics and determine yield, then measurement accuracy can be maintained, but the process becomes extremely time-consuming and labor-intensive
Solution Approach 1:
The system performs preliminary measurements of crop characteristics (height, color, volume) throughout the growing season using UAS imagery before harvest occurs. These early measurements are used by machine learning models to predict final yield, eliminating the need to wait until harvest time to determine genotype performance.
Solution Approach 2:
The patent replaces manual mechanical measurement methods (physical harvesting and weighing) with automated optical sensing systems (UAS with multispectral cameras) and computational analysis (machine learning models). This substitution enables rapid, non-contact measurement of crop characteristics across large areas.
2Adaptability or versatility
If experimental field size is increased to analyze more genotypes, then genotype analysis capability improves, but the ability to harvest all genotypes becomes limited by available resources
Solution Approach 1:
The system performs preliminary yield predictions during the growing season using UAS-collected data on plant characteristics. This allows breeders to identify top-performing genotypes before harvest, so they can focus harvesting resources on a smaller subset of promising genotypes rather than attempting to harvest and analyze all genotypes equally.
Solution Approach 2:
The system creates digital copies and models of crop performance through machine learning predictions. These virtual representations allow extensive genotype comparison without requiring physical harvest of all genotypes, effectively decoupling analysis capacity from harvest capacity.
3Ease of operation
If small sample measurements are taken to reduce labor, then labor requirements decrease, but the accuracy of extrapolating data across the entire field deteriorates
Solution Approach 1:
The system transitions from ground-based point measurements to aerial imaging that captures two-dimensional spatial distributions of crop characteristics across the entire field. Multispectral cameras measure vegetation indices, plant height, and canopy coverage over large areas simultaneously, providing comprehensive spatial coverage rather than isolated sample points.
Solution Approach 2:
The UAS system performs multiple measurement functions simultaneously: it captures visual imagery for plant health assessment, measures plant height through stereo photography, and collects spectral data for vegetation analysis. This multi-functional approach provides comprehensive field-wide data without requiring separate measurement campaigns for each parameter.
Data Source
AI summary
Various embodiments are disclosed for a machine learning system for automatic genotype selection and performance evaluation using multi-source and spatiotemporal remote sensing data collected from an unmanned aerial system (UAS). A computing device may be configured to access images of a field having a first genotype and a second genotype of at least one crop or plant planted therein. The computing device may apply an image processing routine to the images to analyze the images of the field and determine characteristics of the first genotype and the second genotype of the at least one crop or plant planted in the field. The computing device may then apply a machine learning routine to forecast a first estimated yield of the first genotype and a second estimated yield of the second genotype using the identified characteristics of the first genotype and the second genotype.