Harvester Stereo Vision Yield Control on Sloped Crop Fields
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for controlling agricultural harvesters struggle to accurately estimate ground elevation and crop yield due to limited field of view and inability to distinguish unharvested crop from the ground, leading to unreliable terrain elevation data.
Innovation Solution
A method involving stereo vision data collection to create three-dimensional representations of crop canopies and ground surfaces, combining current and stored position data to estimate accurate ground elevation, and calculating yield differences between crop and ground surfaces within a grid-defined region.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If sensors are used to estimate ground elevation, then ground elevation data can be obtained, but the estimation becomes incorrect on slopes due to limited field of view and inability to distinguish unharvested crop from ground
Solution Approach 1:
The patent combines multiple data sources including stereo vision data, GNSS position data, attitude data from IMU, and stored ground elevation data to create a comprehensive ground elevation estimation system. This merging of multiple sensing modalities compensates for the limitations of individual sensors, particularly on slopes where single sensors fail to distinguish crop from ground.
Solution Approach 2:
The system uses stored ground elevation data from previous passes as an intermediary reference to compare against current stereo vision data. This intermediary dataset helps validate and correct current elevation estimates, providing a reliable baseline that compensates for limitations in real-time sensing on slopes.
2Measurement precision
If stereo vision data is collected to create three-dimensional representations, then crop yield estimation accuracy is improved, but data processing complexity increases
Solution Approach 1:
The patent segments the three-dimensional crop representation into discrete volumetric measurements within grid-defined regions. By dividing the field into manageable cells and calculating yield differences on a cell-by-cell basis, the system processes complex stereo vision data in smaller, more manageable units, reducing overall processing complexity while maintaining accuracy.
Solution Approach 2:
The system transforms two-dimensional stereo vision images into three-dimensional volumetric representations of crop canopies. This dimensional transformation enables accurate yield estimation by capturing crop volume, which correlates with biomass and yield, while the structured approach to 3D reconstruction manages the complexity through systematic coordinate transformations.
3Measurement precision
If current and stored position data are combined to estimate ground elevation, then elevation accuracy is improved, but alignment and processing requirements increase
Solution Approach 1:
The system uses stored ground elevation data from previous passes as feedback to validate and correct current elevation estimates. By comparing current stereo vision-based measurements with historical data at corresponding GNSS positions, the system iteratively refines elevation accuracy while using the feedback loop to identify and correct alignment errors.
Solution Approach 2:
The patent performs preliminary alignment and registration of stored ground elevation data with current position data before processing. By pre-aligning the coordinate systems and spatial references of historical and current datasets, the system reduces the processing complexity during real-time operation while maintaining high elevation accuracy through the aligned combined data.
Data Source
AI summary
An upper point cloud estimator is configured to estimate a three-dimensional representation of the crop canopy based on collected stereo vision image data. A lower point cloud estimator is configured to estimate a ground three-dimensional representation or lower point of the ground based on the determined average. The electronic data processor is configured to determine one or more differences between the upper point cloud (or upper surface) of the crop canopy and a lower point cloud of the ground, where each difference is associated with a cell within a grid defined by the front region. The electronic data processor is capable of providing the differences to a data processing system to estimate a yield or differential yield for the front region, among other things.