Webinar on Land Classification with Multispectral Imagery
Introduction
- Presented by CanDrone: Leading consultancy in aerial and ground-based remote sensing in Canada.
- Services: Drone-based solutions for forestry, agriculture, and urban development.
- Presenters: Cody Wildgust (Strategic Sales Expert) and Ian Perry (Solutions Specialist).
Agenda
- Introduction to technology
- Explanation of multispectral imagery
- Components of land cover classification
- Use cases of the technology
Multispectral Imagery
- Definition: Captures data from light in various bands of the electromagnetic spectrum.
- Visible Spectrum: 400-700 nanometers (red, green, blue).
- Multispectral Cameras: Can capture additional bands (e.g., infrared, red edge).
- Example: Micasense Altum camera with 5 lenses for various bands.
Data Processing
- Equipment Setup: Camera mounted on drones like DJI M300.
- Software: Photogrammetry software (e.g., Pix4D) for processing data into orthomosaics.
Multispectral Analysis
- Spectral Signature: Identifies objects by their interaction with light.
- NDVI (Normalized Difference Vegetation Index): Measures plant health by comparing near-infrared and red light reflectance.
Practical Applications
- Efficient Vegetation Monitoring: Identifying stressed or healthy crops from the air.
- Precision Agriculture: Reducing resource usage and improving field efficiency.
Land Cover Classification
- Process:
- Select training and testing data samples.
- Use image segmentation to group pixels.
- Apply machine learning for classification (e.g., Random Forest algorithm).
- Tools: QGIS for segmentation and analysis.
- Python Scripting: For automating large scale analysis and ensuring consistency.
Variations of NDVI
- Soil Adjusted Vegetation Index (SAVI): For sparse vegetation areas.
- Normalized Difference Red Edge: For complex canopies.
Fire Assessment
- Normalized Burn Ratio: Assesses fire impact using multispectral imagery.
Software and Resources
- QGIS: Open-source GIS platform for spatial data analysis.
- Python: For scripting workflows and processing large datasets.
Questions and Answers
- RGB cameras can isolate bands but lack near-infrared and red edge capabilities.
- Multispectral satellite images are available, e.g., Landsat.
- Discussion on using drones vs. satellites depending on resolution needs.
Conclusion
- Contact CanDrone: For multispectral services and consultation.
- Feedback: Suggestions for future webinar topics are welcome.
Overall, this webinar provided a comprehensive overview of using multispectral imagery for land classification, emphasizing the technology's efficiency and applications in various industries.