Crop Row Vision Guidance for Robots Without GNSS Maps
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing agricultural navigation systems for autonomous robots, particularly in early-stage crop fields, face challenges such as reliance on GNSS which can be unreliable due to signal blockage, multipath errors, interference, and the need for precise maps, and conventional perception systems fail to accurately detect crop rows in early growth stages, especially when weeds are present.
Innovation Solution
A method and system using synchronized RGB and ToF cameras to identify crop plants with object detection, transform points to an absolute reference frame, cluster and merge points, and estimate guiding coordinates, allowing precise navigation without relying on GNSS or pre-stored maps, even in early growth stages with variable crop row alignment.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Reliability
If GNSS is used for navigation, then positioning can be achieved, but signal blockage and multipath errors reduce reliability and accuracy
Solution Approach 1:
The patent introduces an intermediary perception system (cameras and processors) that detects crop rows and physical markers in the environment to serve as a mediator between the robot and its destination. This intermediary system provides alternative positioning cues when GNSS signals are blocked or inaccurate, resolving the contradiction by not relying solely on the problematic GNSS system.
Solution Approach 2:
The patent replaces the electromagnetic-based GNSS system with a vision-based mechanical/optical detection system. By using cameras to detect crop rows and physical markers, and processors to calculate positions based on visual information, the system substitutes the unreliable electromagnetic signal-based navigation with a visual-mechanical approach that works reliably in agricultural environments.
2Ease of operation
If conventional perception systems are used, then navigation can be implemented, but they fail to accurately detect crop rows in early growth stages
Solution Approach 1:
The patent applies preliminary action by detecting and tracking physical markers planted at known positions before the crop rows become visible. The system uses these pre-placed markers as reference points to establish the coordinate system and guide the robot from the beginning, rather than waiting for crops to grow to a detectable size. This resolves the contradiction by enabling navigation implementation while maintaining detection accuracy through the use of pre-deployed reference markers.
Solution Approach 2:
The patent changes the detection parameter from relying on crop size and visual characteristics to detecting fixed physical markers with known parameters. By shifting the detection target from variable crop parameters to constant marker parameters, the system maintains measurement precision regardless of crop growth stage, while still enabling ease of operation through automated marker detection.
3Productivity
If laser weeding is applied, then weed control can be achieved, but heavy equipment is required resulting in complex systems
Solution Approach 1:
The patent segments the navigation system into independent modular components: perception modules (cameras), processing modules (position calculation), and control modules (robot guidance). This segmentation allows the complex laser weeding system to be managed through a simplified modular navigation architecture, where each module performs a specific function. The navigation system itself is kept relatively simple by using visual detection and marker-based positioning, compensating for the overall system complexity through modular design.
Solution Approach 2:
The patent creates a universal navigation system that can serve multiple functions: positioning the robot, detecting crop rows, guiding laser application, and tracking field boundaries. By designing a multi-functional perception and navigation system, the patent reduces overall device complexity by eliminating the need for separate specialized systems for each function, while still achieving effective weed control through integrated laser guidance.
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 accurate and reliable navigation of autonomous robots in early-stage crop fields, effectively guiding them along crop rows and detecting entry and exit points, enhancing precision and reducing environmental impact by minimizing the need for herbicides.
Implementation Method 1
Obtaining a point cloud of crop locations, PC vf, by associating points of the point cloud from the at least one ToF camera image with the pixels identified as belonging to a crop plant
Data Source
Figure 1
Figure 2
Figure 3
AI summary
It is proposed a method and system for detecting crop rows (during early growth stage of the crop) and guide autonomous robots in agricultural tasks by following said detected crop rows. The proposed solution enables the use of mobile autonomous robots in crop fields, without requiring excessive modification of the work environment (such as altering the crop field to accommodate the robots), and without relying on precise maps or GNSS navigation (that must be constantly updated due to the continuous change of the field conditions).The solution proposed in this invention comprises three main tasks or sub-procedures (in other words, this procedure can be divided into 3 fundamental parts or phases): 1) 4D vision for crop identification using object detection, 2) crop row identification, and 3) Guiding of the robot (by Entry-Exit points detection).