Machine Vision Plant Tracking for Precision Weed Control
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Traditional agricultural implements struggle to precisely cultivate between commodity plants and weeds, often damaging plants or missing weeds due to static tools and manual intervention, leading to inefficiencies in weed removal and treatment application.
Innovation Solution
A machine vision-enabled control system with a convolutional neural network, imaging device, and sensors is used to detect and track plants, allowing for precise control of agricultural tools to avoid damaging commodity plants and accurately target weeds, using post-processing techniques to improve object tracking and classification accuracy.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Manufacturing precision
If static tools are used for cultivation, then device complexity is reduced, but manufacturing precision and reliability deteriorate due to inability to precisely target weeds without damaging plants
Solution Approach 1:
The patent replaces traditional mechanical cultivation systems with a vision-based control system that uses imaging devices, convolutional neural networks, and actuators to dynamically control tool positioning. This substitution enables precise identification and targeting of weeds while avoiding commodity plants, achieving high manufacturing precision through intelligent control rather than mechanical precision alone
Solution Approach 2:
The system employs self-service through autonomous operation where the imaging device continuously captures images, the neural network automatically processes and identifies plants and weeds, and the actuators autonomously adjust tool positions. This self-service capability eliminates the need for manual intervention while maintaining high precision in weed removal operations
2Productivity
If manual intervention is used, then ease of operation is maintained, but productivity deteriorates due to inefficiency in weed removal and treatment application
Solution Approach 1:
The system achieves self-service by automating the entire weed identification and removal process. The imaging device captures images, the convolutional neural network identifies weeds and commodity plants, and the actuators automatically position tools to remove weeds. This autonomous operation dramatically improves productivity by processing multiple plants continuously without manual intervention while maintaining ease of operation through simple system activation
Solution Approach 2:
The system ensures continuity of useful action by continuously capturing images along the travel path, processing them in real-time, and continuously adjusting tool positions. This continuous operation eliminates interruptions associated with manual intervention, maintaining high productivity throughout the cultivation process while the system operates autonomously
3Reliability
If static tools are used, then device complexity is reduced, but reliability deteriorates due to damage to commodity plants and missed weeds
Solution Approach 1:
The patent replaces simple mechanical positioning systems with a vision-based detection and control system. The imaging device captures detailed images of plants, the convolutional neural network accurately identifies and classifies commodity plants versus weeds, and the actuators precisely position tools. This substitution dramatically improves reliability in plant identification and tool positioning while managing device complexity through integrated intelligent control
Solution Approach 2:
The system implements continuous feedback by capturing images, processing them through the neural network to identify plants and weeds, and using this information to adjust tool positions in real-time. This closed-loop feedback mechanism ensures high reliability in distinguishing commodity plants from weeds and prevents damage to valuable plants while effectively targeting weeds for removal
Data Source
Figure 1
Figure 2
Figure 3
AI summary
An illustrative control system for a precision agricultural implement includes a controller having a neural network, an imaging device, a plurality of sensors, and a plurality of actuators in communication with the controller, the controller configured for detecting and tracking objects of interest in a commodity field, such a commodity plants, and the plurality of actuators including a plurality of agricultural tool actuators the controller operates based on the detection and tracking of object of interest in the commodity field.