Intra-row Weeding Knives with ML Crop Center Localization
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Existing intra-row weeding machines struggle to effectively remove weeds in the immediate vicinity of crops without damaging the crops, particularly in irregularly sown fields, due to limitations in image analysis and navigation precision.
Innovation Solution
A method and machine that utilize a camera and machine learning to precisely identify the center of each crop, position weeding knives below the ground surface, and loosen soil around the crop roots without damaging the above-ground parts by ensuring the knives are spaced far enough from the crop center, using a calibrated field of view and defined zones to guide the knives' movement.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If intra-row weeding machines use close vicinity navigation to remove weeds near crops, then weeding effectiveness is improved, but risk of crop damage increases
Solution Approach 1:
The system performs preliminary identification and localization of crop centers using image analysis before the weeding knives reach the crops. The evaluation software processes images to detect and localize crop centers, calculates remaining time/distance to contact, and prepares control commands in advance. This allows the knives to be precisely positioned and activated only when and where needed, eliminating crop damage while maintaining effective weeding.
Solution Approach 2:
The system continuously captures images, analyzes crop positions, and adjusts knife positioning and activation timing in real-time based on feedback from the evaluation software. The feedback loop ensures that the knives maintain optimal distance from crops and activate only when weeds are detected, preventing crop damage while maximizing weeding effectiveness.
2Measurement precision
If image analysis is used to identify crops and guide weeding, then navigation precision is improved, but system complexity increases
Solution Approach 1:
The system replaces complex mechanical navigation and positioning systems with an optical image analysis system. Instead of using multiple sensors, mechanical guides, or complex positioning mechanisms, the invention uses a camera to capture images and software to analyze crop positions, calculate distances, and control knife activation. This substitution achieves high navigation precision while keeping the physical hardware relatively simple.
Solution Approach 2:
The evaluation software automatically processes images, identifies crops, localizes centers, calculates timing and positioning parameters, and generates control commands without human intervention. The system is self-sufficient in navigating and controlling the weeding process, eliminating the need for complex external control systems or manual operation.
3Productivity
If weeding knives are positioned close to crops for effective weeding, then weed removal efficiency is improved, but crop damage risk increases
Solution Approach 1:
The system dynamically adjusts the positioning and activation of weeding knives based on real-time image analysis. The knives are not fixed at a constant distance but are continuously repositioned based on the calculated distance to crop centers and the detected presence of weeds. This dynamic adjustment allows the knives to operate close to crops for effective weeding while automatically maintaining a safe distance when crops are present.
Solution Approach 2:
The system applies different operational characteristics to different locations: knives are activated and positioned close to crops only in areas where weeds are detected, while maintaining distance from crops in areas where no weeds are present. This localized approach ensures effective weeding where needed while preventing crop damage in areas where crops are present.
Data Source
Figure 1A~2B
Figure 3A~3B.1
Figure 3B.2~3C
AI summary
The method of intra-row weeding of agricultural crops by means of a moving weeding machine is characterized in that the weeding machine is firstly calibrated in 3D space and, then, the weeding occurs in three lines: within the recording line, the camera continuously takes images of the field of view, which are stored in the evaluation software; within the location line, the evaluation software analyzes the location field where it detects and localizes agricultural crop centers using a machine learning model trained to recognize crop centers; within the contact line, a defined zone is defined and the evaluation software evaluates the time and/or distance up to contact of the knives with the detected crop centers detected in the location line with respect to the current travel rate of the weeding machine; when the knives enter the defined zones around the contact points, the evaluation software commands to withdrawn around the detected crop centers, and when the knives leave the defined zones around the contact points, the evaluation software commands to clamp the knives.