Sessions
📽️ View PresentationDigital Image Processing
Band arithmetic, spectral indices, and image enhancement techniques.
Image Classification
Supervised, unsupervised, and object-based classification methods.
Lunch Break
Change Detection
Temporal compositing and multi-temporal change analysis.
AI and ML in EO
Modern applications of deep learning in Earth Observation.
📚 Required Readings
Band arithmetic, spectral indices, NDVI, and cloud masking.
Supervised, unsupervised, and object-based approaches.
Temporal compositing, multi-temporal analysis, and case studies.
📝 Key Concepts for the Exam
~10-20 discrete, wide bands (Landsat/Sentinel). Best for identifying land cover types like "forest" or "water".
100s-1000s of narrow, contiguous bands. Required for identifying specific minerals or chemical compositions.
- Radiometric Correction: Removing atmospheric "noise" (scattering/absorption) to reveal true surface reflectance. "Atmospheric modeling" is the key.
- Geometric Correction: Correcting for satellite tilt, earth curvature, and terrain. Essential for turning an "image" into a "map" that can be layered with other data.
- Bit Depth: 8-bit (256 values) vs. 16-bit (65,536 values). Higher bit depth allows for much subtler spectral distinctions.
Binary decision tree. Fast and interpretable, but prone to "overfitting" (being too specific to training data).
An "ensemble" of many trees. Highly accurate and stable. The "gold standard" for EO classification today.
Critique of "Environmental Determinism": Geography is not destiny. Human ingenuity and technology (like GIS/Remote Sensing) allow us to overcome environmental constraints.
🎓 Module 3 Knowledge Check
1. Which classification method requires the user to provide "training samples" of known land cover?
2. In Change Detection, what is "Post-Classification Comparison"?