Peduncle Position Estimation for Occluded Harvesting
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Harvest robots face instability in determining the proper insertion direction of the end effector due to the inability to detect peduncle positions when they are obscured by fruits, leading to ineffective harvesting.
Innovation Solution
A peduncle position estimating apparatus and method using a light source, image acquiring section, and image processing section to emit IR light, capture RGB, IR, and distance images, and perform image processing to estimate peduncle positions even when obscured, employing bright spot detection, grouping, and main component analysis to determine accurate insertion directions.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If the end effector inserts into the center of the peduncle to harvest fruits, then the harvesting is effective, but the peduncle position cannot be detected when obscured by fruits leading to unstable harvesting
Solution Approach 1:
The patent introduces an estimation section as an intermediary component that calculates the peduncle position based on the detected fruit positions when the peduncle cannot be directly detected. This mediator resolves the contradiction by providing a reliable peduncle position estimate through mathematical calculation (centroid computation) even when direct observation is blocked by fruits.
Solution Approach 2:
The system creates a virtual model of the peduncle position based on the spatial distribution of fruits. By computing the centroid coordinates from multiple fruit positions, the system generates an accurate representation of the hidden peduncle location without needing to directly observe it, thus solving the detection problem.
2Measurement precision
If multiple image sensors are used to detect peduncle position from different angles, then detection accuracy improves, but device complexity increases
Solution Approach 1:
The patent transitions from spatial dimension (multiple physical sensors at different positions) to computational dimension (mathematical calculation based on single sensor data). By using coordinate geometry and centroid computation, the system achieves accurate peduncle position estimation without adding physical sensors, thus improving measurement precision while maintaining simple device structure.
Applied Scientific Principles
This section explains which scientific principles are used to turn an abstract innovation direction into a practical engineering solution.
Function Achieved in This Case
Enables stable harvesting motions by accurately estimating peduncle positions regardless of their visibility, ensuring consistent fruit picking regardless of the positional relationship between fruits and peduncle.
Implementation Method 1
a light source 101 configured to emit light
Implementation Method 2
an image acquiring section 102 configured to acquire an image of the bunch of fruits 404
Data Source
Figure 1
Figure 2
Figure 3A~3C
AI summary
A equipment of localization of peduncle and a method of localization of peduncle that can achieve a stable harvesting motion by estimating the position of a peduncle even in the case where the peduncle is located behind fruits and the like and cannot be recognized as viewed from the image sensor side. The peduncle position estimating apparatus includes image acquiring section 102 that acquires an image of fruits; and estimation section 104 that estimates a position of a peduncle that is coupled with the fruits based on positions of the fruits and a distance between the fruits in the image acquired by the image acquiring section.