Plant Classification Using Segmented Detection Zones

Resolve Bottlenecks,
Find Innovative Solutions
Generate Solutions

Solution Overview

Problem

Current methods for classifying plants in agricultural settings often fail to accurately distinguish crops from weeds, leading to a high rate of non-classification and inefficient herbicide application, particularly for plants with large components like leaves.

Innovation Solution

The method involves creating a neutral area and a far area around cultivated plant rows, evaluating plant components within these areas, and using geometric dimensions and overlap of image data to correctly assign plants as either crops or weeds, with the option to adapt area widths based on crop growth and season, and employing an optical and/or infrared detection unit for precise classification.

Engineering Contradictions & Design Principles

VSEngineering Contradiction Analysis

1Measurement precision

If conventional plant classification methods are used, then the classification process is simple, but the rate of nonclassification is high and accuracy is poor

Engineering Contradiction:
Improveplant classification accuracyVSAvoidclassification system complexity
Core Design Contradiction:
Measurement precisionVSDevice complexity

Solution Approach 1:

The detection area is segmented into three distinct zones: cultivation row tube (central area where crops grow), neutral area (intermediate transition zone), and far area (outer region). This segmentation allows different evaluation criteria to be applied to each zone, improving classification accuracy by considering the spatial context of plant components across multiple regions rather than treating all areas uniformly.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

The method transitions from traditional two-dimensional image analysis to three-dimensional spatial evaluation by defining zones with longitudinal, lateral, and depth dimensions. The cultivation row tube, neutral area, and far area create a volumetric classification space that captures the spatial distribution of plant components, enabling more accurate distinction between crops and weeds based on their positional relationships.

Inventive Principle:
Principle #17Another dimension (Dimensionality change)

2Reliability

If the detection area is expanded to improve classification accuracy, then more plants are detected, but the processing complexity and time increase

Engineering Contradiction:
Improveweed detection reliabilityVSAvoidclassification processing time
Core Design Contradiction:
ReliabilityVSLoss of time

Solution Approach 1:

By dividing the detection space into three zones, the system can apply different processing strategies to each segment. The cultivation row tube receives focused attention for crop identification, the neutral area serves as a transition zone for ambiguous cases, and the far area provides contextual information. This segmented approach allows reliable detection without uniformly processing the entire area with maximum complexity.

Inventive Principle:
Principle #1Segmentation

Solution Approach 2:

Different evaluation criteria and attention levels are applied to different spatial zones. The central cultivation row tube receives the most rigorous evaluation for crop identification, while the neutral and far areas provide supporting contextual information. This local differentiation of evaluation quality enables reliable classification without uniformly applying high-processing standards across all regions, thus reducing overall processing time.

Inventive Principle:
Principle #3Local quality

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 significantly reduces non-classification errors, ensuring that weeds are reliably identified and treated, thereby increasing the available space for crops by accurately differentiating and targeting weed areas for herbicide application.

Implementation Method 1

an optical and/or infrared detection unit for detecting image data of a region to be examined in the longitudinal direction of the cultivated area

Methodology Applied
Scientific EffectOptical detection: Reflection

Implementation Method 2

an optical and/or infrared detection unit for detecting image data

Methodology Applied
Scientific EffectInfrared radiation: Infrared Radiation

Data Source

PatentUS20230380404A1Method and device for classifying plants, and computer program product
Publication Date: 2023.11.30 ROBERT BOSCH GMBH
  • US20230380404A1 patent drawing
  • US20230380404A1 patent drawing
  • US20230380404A1 patent drawing

AI summary

A method for classifying plants. In the method, the plants or their plant components, in particular plant leaves, are detected in an evaluation region with the aid of an optical and/or infrared detection unit, and the detected image data of the detection unit are evaluated with the aid of an algorithm, via the evaluation a first plant type being distinguished from a second plant type, the second plant type being treated in particular with a medium, preferably a liquid spray, and the first plant type being planted in the form of a plant row in a tubular cultivation row tube.