Plant-Part Count Normalization Using Range-to-Canopy Estimation
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for counting plant-parts-of-interest, such as flowers or pods, are limited by the visibility of vision sensors and inaccuracies in estimating spatial dimensions, particularly plant height, leading to under or over prediction of counts.
Innovation Solution
Utilizing depth-capable vision sensors like stereoscopic cameras to estimate range-to-canopy and normalize counts of visible plant-parts-of-interest, combined with machine learning models to extrapolate occluded parts, and time-series models for yield prediction.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Ease of operation
If manual counting of plant-parts-of-interest is used, then simplicity of the method is maintained, but productivity and accuracy deteriorate due to inability to count large numbers of plant parts efficiently
Solution Approach 1:
The patent replaces manual mechanical counting with automated computer vision systems using cameras and image processing algorithms. The vision sensor captures images of plant parts, and computational algorithms automatically detect, count, and track plant-parts-of-interest, eliminating the need for human operators to manually count while significantly improving productivity and counting capacity.
Solution Approach 2:
The patent creates digital copies of plant parts through image capture and processes these copies to extract counting information. By working with image data rather than physically counting actual plant parts, the system enables rapid, repeatable counting of large numbers of plant parts without the limitations of manual methods.
2Productivity
If vision sensors are used to count plant-parts-of-interest, then productivity is improved, but measurement precision deteriorates due to occlusions and varying plant heights affecting visibility
Solution Approach 1:
The patent transitions from two-dimensional image analysis to three-dimensional spatial understanding by incorporating depth information and perspective geometry. By modeling the three-dimensional arrangement of plant parts and applying perspective correction based on camera position and plant height, the system accurately counts plant parts even when they are occluded or at varying distances, significantly improving measurement precision.
Solution Approach 2:
The patent dynamically adjusts counting parameters based on detected plant characteristics such as height, orientation, and spatial position. The system modifies detection thresholds, field of view parameters, and counting algorithms according to the specific geometric parameters of each plant, enabling accurate counting across diverse plant configurations while maintaining high productivity.
3Ease of operation
If vision sensors count only visible plant-parts-of-interest, then ease of operation is maintained, but loss of information increases due to occluded plant parts not being counted
Solution Approach 1:
The patent performs preliminary geometric modeling and spatial reasoning to predict the locations and numbers of occluded plant parts before final counting is completed. By using depth information, perspective geometry, and plant structure models, the system anticipates where occluded parts should be and adjusts the count accordingly, preventing information loss while maintaining operational simplicity.
Solution Approach 2:
The patent implements feedback mechanisms where the system continuously monitors detection confidence and adjusts counting based on identified occlusions. When plant parts are detected to be occluded or partially visible, the system uses spatial reasoning and previously detected information to infer the presence of hidden parts, correcting the count and reducing information loss while keeping the user interface simple.
Data Source
Figure 1
Figure 2
Figure 3A~3B
AI summary
Implementations are described herein for normalizing counts of plant-parts-of-interest detected in digital imagery to account for differences in spatial dimensions of plants, particularly plant heights. In various implementations, one or more digital images depicting a top of a first plant may be processed. The one or more digital images may have been acquired by a vision sensor carried over top of the first plant by a ground-based vehicle. Based on the processing: a distance of the vision sensor to the first plant may be estimated, and a count of visible plant-parts-of-interest that were captured within a field of view of the vision sensor may be determined. Based on the estimated distance, the count of visible plant-parts-of-interest may be normalized with another count of visible plant-parts-of-interest determined from one or more digital images capturing a second plant.