Related Experiment Video
Updated: Jul 9, 2025
11:49
Cereal Crop Ear Counting in Field Conditions Using Zenithal RGB Images
Published on: February 2, 2019
9.3K
Comparing CNNs and PLSr for estimating wheat organs biophysical variables using proximal sensing
Alexis Carlier1, Sébastien Dandrifosse1, Benjamin Dumont2
1Biosystems Dynamics and Exchanges, TERRA Teaching and Research Center, Gembloux Agro-Bio Tech, University of Liège, Gembloux, Belgium.
Frontiers in Plant Science
|December 6, 2023
Summary
Deep learning models, particularly convolutional neural networks (CNNs), effectively estimate crop biophysical variables using advanced training techniques like pseudo-labeling. This approach overcomes data scarcity in crop phenotyping for improved yield prediction.
Area of Science:
- Agricultural Science
- Remote Sensing
- Computer Science (Deep Learning)
Background:
- Accurate estimation of biophysical vegetation variables is crucial for crop monitoring and yield prediction.
- Traditional remote sensing methods face challenges due to complex plant structures and the need for extensive feature engineering.
- Limited labeled data hinders the application of deep learning, especially Convolutional Neural Networks (CNNs), in crop phenotyping for regression tasks.
Purpose of the Study:
- To evaluate the effectiveness of various CNN models for predicting wheat dry matter, nitrogen uptake, and nitrogen concentration.
- To address the challenge of limited labeled data in crop phenotyping using a novel training pipeline.
- To compare CNN performance against traditional machine learning approaches for biophysical variable estimation.
Main Methods:
- Utilized RGB and multispectral imagery from wheat tillering to maturity.
- Developed a training pipeline incorporating transfer learning, pseudo-labeling of unlabeled data, and temporal relationship correction.
- Compared the performance of different CNN architectures (EfficientNetB4, Resnet50) and a Partial Least Squares Regression (PLSr) model.
Main Results:
- CNN models significantly improved prediction accuracy when employing the pseudo-labeling method.
- EfficientNetB4 achieved the highest accuracy for above-ground biomass prediction (R² = 0.92).
- Resnet50 excelled in predicting Leaf Area Index (LAI), nitrogen uptake, and nitrogen concentration (R² = 0.82, 0.73, and 0.80, respectively).
Conclusions:
- CNNs, enhanced by pseudo-labeling, offer a promising and accessible solution for phenotyping quantitative crop biophysical variables.
- The developed training pipeline effectively overcomes data scarcity issues in deep learning for crop science applications.
- Further research is needed to fully realize the potential of CNNs in advanced crop phenotyping and management.
More Related Videos
Related Concept Videos
02:35Key Elements for Plant Nutrition
18.8K
Like all living organisms, plants require organic and inorganic nutrients to survive, reproduce, grow and maintain homeostasis. To identify nutrients that are essential for plant functioning, researchers have leveraged a technique called hydroponics. In hydroponic culture systems, plants are grown—without soil—in water-based solutions containing nutrients. At least 17 nutrients have been identified as essential elements required by plants. Plants acquire these elements from the...
18.8K
02:16Light Acquisition
8.5K
In order to produce glucose, plants need to capture sufficient light energy. Many modern plants have evolved leaves specialized for light acquisition. Leaves can be only millimeters in width or tens of meters wide, depending on the environment. Due to competition for sunlight, evolution has driven the evolution of increasingly larger leaves and taller plants, to avoid shading by their neighbors with contaminant elaboration of root architecture and mechanisms to transport water and nutrients.
8.5K