Machine-Learned Crop Yield Estimation from Harvester Sensor Data
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing agricultural yield monitoring systems for harvesters, such as sugarcane harvesters, are costly and require significant maintenance, making them impractical for many users, and they rely on sensorized plates that accumulate dirt and debris, leading to inefficiencies.
Innovation Solution
Implementing a machine-learned yield estimation model using onboard sensors to estimate crop yield based on operation-related conditions, eliminating the need for expensive and high-maintenance sensor systems.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If sensorized plates are used to monitor crop yield, then measurement precision is improved, but device complexity and maintenance requirements increase
Solution Approach 1:
The patent replaces the mechanical sensorized plate system with a machine learning model that processes operational data from existing sensors. This substitution eliminates the need for complex mechanical load-sensing plates while maintaining yield estimation accuracy through computational algorithms that analyze motor current, rotational speed, and other operational parameters.
Solution Approach 2:
Instead of directly measuring yield with physical sensors, the system creates a virtual model (machine learning algorithm) that copies the functionality of the sensorized plate by inferring yield from operational parameters. This virtual copy achieves similar measurement objectives without the physical complexity and maintenance burden of actual load cells and sensorized plates.
2Measurement precision
If sensorized plates are used to monitor crop yield, then measurement precision is improved, but maintenance requirements increase
Solution Approach 1:
The machine learning model replaces the mechanical sensorized plate system, eliminating components that accumulate dirt and require cleaning. The software-based solution has no moving parts or surfaces that need maintenance, thereby significantly reducing maintenance requirements while preserving measurement capability.
Solution Approach 2:
The system uses existing operational sensors and data already collected during normal harvester operation to estimate yield, eliminating the need for separate maintenance activities. The machine learning model continuously processes available data without requiring manual intervention, cleaning, or calibration that would be necessary for physical sensorized plates.
3Measurement precision
If sensorized plates are used to monitor crop yield, then measurement precision is improved, but cost increases
Solution Approach 1:
The patent substitutes expensive mechanical sensorized plates with a computational approach using machine learning. This replacement dramatically reduces hardware costs while maintaining yield estimation accuracy, as the solution leverages existing onboard sensors and processing capabilities already present in modern harvesters.
Solution Approach 2:
The machine learning model serves multiple functions simultaneously: it estimates crop yield, monitors operational efficiency, and provides insights into harvester performance. This multi-functionality consolidates what would otherwise require separate specialized equipment, thereby reducing overall system cost while maintaining measurement precision.
Data Source
AI summary
In one aspect, a computing system for estimating crop yields for agricultural harvesters. The computing system includes one or more processors, and one or more non-transitory computer-readable media that collectively store a machine-learned yield estimation model configured to receive data associated with one or more operation-related conditions for an agricultural harvester and process the data to determine a yield-related parameter indicative of a crop yield for the agricultural harvester. In addition, the computer-readable media stores instructions that, when executed by the one or more processors, configure the computing system to perform operations, the operations comprising: obtaining the data associated with one or more operation-related conditions; inputting the data into the machine-learned yield estimation model; and receiving a value for the yield-related parameter as an output of the machine-learned yield estimation model.