3D Orchard Reconstruction via Trunk Descriptor Alignment
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing methods for merging 3D reconstructions of fruit tree rows from both sides fail to align accurately without overlapping views or GPS coordinates, leading to inconsistent and costly solutions.
Innovation Solution
The method involves finding a rigid body transformation between occlusion boundaries using 2D shape matching and semantic constraints, and integrating tree morphology into the reconstruction system to output optimized morphological parameters, allowing for the alignment of front and back side reconstructions without the need for overlapping views or GPS coordinates.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Measurement precision
If traditional 3D reconstruction methods are used to merge front and back side models, then the reconstruction can be completed, but the alignment accuracy deteriorates without overlapping views or GPS coordinates
Solution Approach 1:
The patent introduces trunk descriptors as intermediary features to facilitate alignment between front and back side 3D models. These descriptors extract characteristic points from tree trunks in both reconstructions and use them as reference markers to compute the rigid body transformation, eliminating the need for overlapping views or GPS coordinates while maintaining alignment accuracy
Solution Approach 2:
The patent replaces the mechanical/GPS-based alignment system with a computational vision system. Instead of relying on physical overlapping views or GPS coordinates, the system uses 2D shape matching algorithms and semantic constraints to compute the spatial transformation between front and back side reconstructions, achieving accurate alignment through image processing rather than mechanical means
2Measurement precision
If labor-intensive manual measurement methods are used, then accurate morphological parameters can be obtained, but the productivity deteriorates
Solution Approach 1:
The patent implements a self-service measurement system where the 3D reconstruction and morphological parameter extraction are performed automatically by the computer vision system without human intervention. The system processes images, constructs 3D models, extracts trunk descriptors, aligns reconstructions, and computes morphological parameters (tree height, canopy volume, trunk diameter) autonomously, eliminating labor-intensive manual measurements while maintaining accuracy and significantly improving productivity
Solution Approach 2:
The patent replaces manual mechanical measurement tools (tapes, measuring devices) with an automated computer vision system. The system uses image processing algorithms to extract morphological parameters directly from 3D reconstructions, substituting physical measurement operations with computational methods that are both faster and equally accurate
Data Source
AI summary
A method includes constructing a three-dimensional model of a front side of a row of trees based on a plurality of images of the front side of the row of trees and constructing a three-dimensional model of a back side of the row of trees based on a plurality of images of the back side of the row of trees. The three-dimensional model of the front side of the row of trees is merged with the three-dimensional model of the back side of the row of trees by linking a trunk in the three-dimensional model of the front side to a trunk in the three-dimensional model of the back side to form a merged three-dimensional model of the row of trees. The merged three-dimensional model of the row of trees is used to determine a physical attribute of the row of trees.