Plant Classification Using Temporal Tracking for Weed Detection
Find Innovative SolutionsGenerate Solutions
Solution Overview
Problem
Current methods for classifying plants in agricultural settings often incorrectly identify plants as weeds, leading to unnecessary treatment with herbicides, as they fail to reliably distinguish between crop plants and weeds, especially in edge areas and when plants are partially outside the evaluation window.
Innovation Solution
A method using an optical and/or infrared detection unit evaluates image data to reliably classify plants by considering their connection to a row of plants and distance from the evaluation window's limits, ensuring accurate assignment to either crop or weed species, and a device with a control system and spray nozzle for targeted treatment.
Engineering Contradictions & Design Principles
Engineering Contradiction Analysis
1Productivity
If plants in edge regions are classified based on their connection to plant rows within the evaluation window, then classification speed is improved, but classification accuracy deteriorates because plants protruding from the evaluation window are incorrectly identified as weeds
Solution Approach 1:
The patent extends the evaluation from a two-dimensional evaluation window to a three-dimensional space-time volume by incorporating temporal information from multiple evaluation times. Plants protruding from the evaluation window are tracked across time, and their connection to plant rows is verified through temporal persistence, adding a time dimension to the classification process.
Solution Approach 2:
The patent performs preliminary classification at the first evaluation time to identify plants that may belong to crop rows but are partially outside the evaluation window. These plants are flagged for further verification at a second evaluation time before final classification, allowing preliminary sorting without premature commitment to incorrect weed identification.
2Area of stationary object
If the evaluation window is moved continuously to cover more area, then detection coverage is improved, but the ability to track plants across evaluation windows deteriorates
Solution Approach 1:
The patent performs preliminary identification of plants at the first evaluation time that protrude from the evaluation window boundaries. These plants are marked for special tracking, and their positions are carried forward to the second evaluation time to verify continuity with plant rows, preserving connection information despite window movement.
Solution Approach 2:
The patent uses feedback from the first evaluation time to guide the second evaluation, specifically tracking the positions and movements of plants that were initially identified as potentially belonging to crop rows. This feedback mechanism ensures continuous monitoring of edge-region plants across moving evaluation windows.
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
This approach ensures safe and reliable differentiation of crop plants from weeds, preventing incorrect treatment by accurately assigning plants to their species and enabling timely and precise application of herbicides to weeds.
Implementation Method 1
an optical and/or infrared detection unit evaluates image data
Implementation Method 2
an optical and/or infrared detection unit evaluates image data
Data Source
Figure 1
Figure 2
Figure 3~5
AI summary
The invention relates to a method for classifying plants (P), wherein the plants (P) or the plant components are detected by means of an optical and/or infrared capturing unit (25) within an analysis region (30) at a first analysis time, the captured image data (BD) of the capturing unit (25) is analyzed using an algorithm (28), the analysis making it possible to distinguish at least one first type of plant (PA1) from at least one second type of plant (PA2), and the second type of plant (PA2) is treated in particular with a medium, preferably a liquid spray (1).