Related Experiment Video
Updated: Oct 10, 2025
08:47
Author Spotlight: UAV Remote Sensing for Efficient Invasive Plant Biomass Estimation
Published on: February 9, 2024
1.7K
Machine Learning-Based Plant Detection Algorithms to Automate Counting Tasks Using 3D Canopy Scans
Serkan Kartal1, Sunita Choudhary2, Jan Masner3
1Department of Computer Engineering, Faculty of Engineering, Cukurova University, Adana 01330, Turkey.
Sensors (Basel, Switzerland)
|December 10, 2021
Summary
Machine learning (ML) accurately separates individual mung bean and chickpea plants from 3D scans using Convolutional Neural Networks (CNNs). This automated plant counting enhances phenotyping pipelines, improving efficiency and accuracy over manual methods.
Area of Science:
- Agricultural Science
- Computer Science
- Plant Science
Background:
- Manual plant counting in phenotyping is labor-intensive and prone to errors.
- 3D laser scanning offers detailed crop information but requires advanced processing.
- Integrating machine learning with 3D data is crucial for automated agricultural analysis.
Purpose of the Study:
- To evaluate machine learning (ML) for segmenting individual plants from 3D canopy laser scans.
- To develop an automated plant counting method for mung bean and chickpea crops.
- To assess the accuracy of Convolutional Neural Networks (CNNs) in 3D plant analysis.
Main Methods:
- 3D laser scanning of mung bean and chickpea canopies using PlantEye® scanners.
- Region Growing Segmentation for separating crop canopies from background.
- Dimensionality reduction to 2D, incorporating height as color, for CNN application.
Main Results:
- Achieved high accuracy in individual plant identification and counting: 93.18% for mung bean and 92.87% for chickpea.
- Successfully applied 2D CNNs after innovative dimensionality reduction of 3D data.
- Demonstrated the potential to replace inefficient manual counting in phenotyping pipelines.
Conclusions:
- ML, specifically CNNs with dimensionality reduction, effectively automates plant counting from 3D scans.
- This approach significantly enhances the efficiency and accuracy of crop phenotyping.
- Further research is needed to address the gap in ML applications for complex 3D plant feature extraction.
