Stereo Vision Plant Feature Detection Using 3D Point Clouds
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing image recognition systems struggle to accurately identify and differentiate between plant features in close proximity, often mischaracterizing growths or miscounting the number of plants in a field due to overlapping foliage.
Innovation Solution
The system captures a sequence of stereo image pairs as a device travels along a row of crops, generating probability heatmaps and depth maps to create combined maps. Transformation matrices are used to align these maps, and clusters of points with high probability are identified to determine the presence and location of plant features.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If plants are planted close in proximity to maximize yield and minimize land use, then productivity increases, but measurement precision deteriorates due to overlapping foliage making it difficult to identify individual plant features
Solution Approach 1:
The patent transitions from two-dimensional image analysis to three-dimensional point cloud analysis. By using stereo vision to generate depth maps and creating 3D point clouds of plant structures, the system can distinguish overlapping foliage by their different spatial positions in three-dimensional space, thereby maintaining measurement precision while allowing close plant spacing for high productivity
Solution Approach 2:
The patent segments the continuous plant canopy into discrete individual plant instances through 3D point cloud processing. By clustering points in three-dimensional space and separating them by vertical position and spatial distribution, the system can identify and count individual plants even when their foliage overlaps in 2D images, resolving the contradiction between close planting and accurate plant identification
2Device complexity
If existing image recognition systems are used to identify plant features, then device complexity remains low, but measurement precision deteriorates due to mischaracterization of plant growths and incorrect plant counting
Solution Approach 1:
The patent introduces depth maps as an intermediary data structure between the captured images and the final plant identification results. The depth maps provide spatial information that mediates the interpretation of overlapping features, allowing the system to distinguish between foliage from different plants by their depth positions, thereby improving measurement precision without requiring overly complex device architecture
Solution Approach 2:
The patent replaces traditional mechanical/optical image recognition methods with computational point cloud analysis. Instead of relying on complex optical systems or manual inspection to distinguish overlapping plants, the system uses algorithmic processing of 3D point clouds to automatically identify and separate individual plant structures, improving precision while keeping the physical device relatively simple
Data Source
AI summary
Described are methods for identifying the in-field positions of plant features on a plant by plant basis. These positions are determined based on images captured as a vehicle (e.g., tractor, sprayer, etc.) including one or more cameras travels through the field along a row of crops. The in-field positions of the plant features are useful for a variety of purposes including, for example, generating three-dimensional data models of plants growing in the field, assessing plant growth and phenotypic features, determining what kinds of treatments to apply including both where to apply the treatments and how much, determining whether to remove weeds or other undesirable plants, and so on.