1.

Differentiate Between Active and Passive Remote Sensing

Feature Active Remote Sensing Passive Remote Sensing

Source of Uses its own source of energy Relies on natural energy

Energy (emits signals) sources (usually sunlight)

Detects natural radiation

Emits radiation toward the
Function emitted or reflected by
target and measures reflection
objects

Works at Usually not (depends on

Yes
Night? sunlight)

- Radar (e.g., Synthetic

- Optical Sensors (e.g.,
Aperture Radar - SAR)
Examples Landsat, MODIS)
- LiDAR (Light Detection and
- Thermal Sensors
Ranging)

Weather Less affected by Affected by atmospheric

Dependency weather/clouds conditions (e.g., clouds, fog)

2. How AI is Used to Automate Land Cover Classification and Object

Detection

AI, especially Machine Learning (ML) and Deep Learning (DL), plays a major
role in analyzing large volumes of remote sensing data. Here's how it's used:

Land Cover Classification

 Goal: Categorize surface types (e.g., forests, water bodies, urban

areas) from satellite imagery.

 Methods:

o Supervised Learning: Uses labeled data to train models like

Random Forest, Support Vector Machines (SVM), or Convolutional
Neural Networks (CNN).

o Unsupervised Learning: Uses clustering algorithms (e.g., K-

means) to group similar pixels when labels are unavailable.

o Deep Learning Models (CNNs): Used for pixel-level

classification in high-resolution images.
  Example: Classifying agricultural land from Sentinel-2 imagery using a
trained CNN model.

Object Detection

 Goal: Identify and locate specific features like vehicles, buildings,

ships, etc.

 Methods:

o YOLO (You Only Look Once) and Faster R-CNN for detecting
and drawing bounding boxes around objects.

o Semantic Segmentation (e.g., U-Net): Identifies the precise

shape of objects (not just bounding boxes).

 Example: Using deep learning to detect illegal mining activities or

locate disaster-affected buildings in satellite images.

3. Impact and Future Scope of AI in Remote Sensing

Impact

 Improved Accuracy: AI models outperform traditional image

processing in precision and reliability.

 Efficiency: Automates tasks like classification and detection that were

once time-consuming and manual.

 Real-Time Monitoring: Helps in quick disaster response (e.g., flood or

wildfire mapping).

 Scalability: Can handle vast datasets (e.g., from satellites like

Sentinel or Landsat).

Future Scope

 Integration with IoT and Drones: Enhanced data collection and

processing in near real-time.

 Climate Change Monitoring: Predicting and mitigating

environmental impacts.

 Smart Agriculture: AI-driven analysis for crop health, yield

prediction, and irrigation management.
  Urban Planning & Smart Cities: Automated detection of urban
sprawl and infrastructure changes.

 Self-learning Models: Models that adapt and improve without

frequent retraining.

Unit 2
1. Define a Remote Sensing Sensor and Explain Its Role in Data
Acquisition

Remote Sensing Sensor: A remote sensing sensor is an instrument that

detects and records energy reflected or emitted from objects or surfaces on
Earth. It is the core device in remote sensing that gathers data from a
distance—commonly via satellites, drones, or aircraft.

Role in Data Acquisition:

 Measures electromagnetic radiation (EMR) from Earth's surface.

 Converts detected EMR into digital signals.

 Provides raw imagery or data for analysis of land, water, atmosphere,

vegetation, etc.

Active vs. Passive Remote Sensing Sensors

Aspect Active Sensor Passive Sensor

Emits its own signal (e.g., Relies on natural sunlight or

Energy Source
microwave, laser) thermal emissions

Works at Night? Yes No (needs sunlight)

Weather Works through clouds/rain Affected by weather and cloud

Independence (e.g., radar, LiDAR) cover

SAR (Synthetic Aperture Landsat TM/ETM+, MODIS,

Examples
Radar), LiDAR Sentinel-2 optical sensors

2. Three Main Types of Remote Sensing Platforms

a. Ground-Based Platforms
  Definition: Instruments located on the ground to observe nearby
features or atmosphere.

 Usage: Soil studies, vegetation monitoring, calibration of satellite

data.

 Example: Spectroradiometers placed in fields.

b. Airborne Platforms

 Definition: Sensors mounted on aircraft, drones, or balloons.

 Usage: High-resolution mapping, agriculture monitoring, mineral

exploration.

 Example: Hyperspectral sensors on drones.

c. Spaceborne Platforms

 Definition: Sensors mounted on satellites orbiting the Earth.

 Usage: Global-scale monitoring (e.g., weather, deforestation, climate).

 Example: Landsat, Sentinel-2, MODIS satellites.

3. Real-World Application: Deforestation Monitoring

Remote Sensing Application:

 Monitoring tropical forests (e.g., Amazon, Southeast Asia) for illegal

logging and environmental degradation.

Data Acquisition & Pre-processing:

 Satellites Used: Landsat, Sentinel-2, MODIS.

 Pre-processing Includes:

o Radiometric correction to normalize lighting differences.

o Geometric correction for aligning imagery with real-world

coordinates.

o Cloud masking and image enhancement.

 AI Tools: Used for change detection and forest cover classification

over time.
4. Radiometric and Geometric Corrections

Radiometric Correction:

Adjusts pixel values to remove errors due to sensor noise, atmosphere, or

lighting.

 Purpose: Ensure consistency in brightness and reflectance.

 Steps:

o Dark Object Subtraction

o Atmospheric correction (e.g., using models like FLAASH)

o Sensor calibration using known reflectance values

Geometric Correction:

Aligns the image to a geographic coordinate system to match the Earth's

surface.

 Purpose: Removes distortions due to Earth's curvature, sensor angle,

terrain variation.

 Steps:

o Identify Ground Control Points (GCPs)

o Apply image rectification or orthorectification

o Reprojection into desired coordinate system (e.g., UTM)

5. Define Data Fusion and Explain Its Benefits in Remote Sensing

Data Fusion: The process of integrating data from multiple sensors (or
platforms) to produce more accurate and comprehensive information.

Types of Fusion:

 Pixel-Level: Combines raw data (e.g., pansharpening for high-res

imagery)

 Feature-Level: Merges extracted features (edges, textures)

 Decision-Level: Combines analysis outcomes from multiple sources

Benefits in Remote Sensing:

  Enhances spatial, spectral, and temporal resolution.

 Provides more robust classification and change detection.

 Helps combine optical and radar imagery for better analysis in cloudy
areas.

 Supports multi-source decision-making (e.g., integrating LiDAR +

satellite data).

UNIT 3
1. What is Supervised Learning in the Context of Remote Sensing?

Supervised Learning in remote sensing refers to a machine learning

technique where algorithms are trained using labeled data—satellite or aerial
images with known land cover types or features.

Process:

1. Training Data: Collected samples (pixels or regions) with known

categories (e.g., water, forest, urban).

2. Model Training: The algorithm learns patterns in the data (spectral

values, textures, etc.).

3. Classification: The trained model is applied to new, unlabeled images

to classify each pixel or region.

Examples:

 Using Sentinel-2 data to classify agricultural crops.

 Land use mapping with Random Forest, Support Vector Machine (SVM),
or Decision Trees.

2. What is Deep Learning, and How Does It Differ from Traditional

Machine Learning in Remote Sensing?

Deep Learning (DL) is a subset of machine learning that uses multi-layered

neural networks (especially Convolutional Neural Networks – CNNs) to
automatically learn features from large volumes of data.
Key Differences:

Aspect Traditional ML Deep Learning (DL)

Manual (engineered
Feature Extraction Automatic (learned from data)
by experts)

Simpler (e.g., SVM, Complex (deep neural

Model Complexity
Decision Trees) networks)

Moderate-sized Requires large, labeled

Data Requirements
datasets datasets

Performance on High (especially in object

Moderate
High-Res Images detection & segmentation)

Land cover
Examples in Remote Cloud detection, building
classification with
Sensing segmentation with CNNs
SVM

3. Advantages and Challenges of Using AI-Driven Image

Classification in Remote Sensing

✅ Advantages:

 High Accuracy: Especially in high-resolution imagery and complex

environments.

 Automation: Speeds up large-scale data processing.

 Feature Learning: Learns spectral, spatial, and temporal features

without manual input.

 Real-Time Applications: Used in disaster management, smart

agriculture, and urban growth monitoring.

⚠️Challenges:

 Data Dependency: Requires large volumes of high-quality labeled

data.

 Computational Power: Demands high GPU/TPU resources, especially

for DL.

 Overfitting Risk: Especially when training data is limited or

unbalanced.
  Black Box Nature: Hard to interpret how DL models make decisions.

 Transferability: Models trained in one region may not perform well in

others.

4. Define Image Classification in Remote Sensing and Explain Its

Importance

Image Classification: It is the process of assigning a category or class label

to every pixel (or group of pixels) in a satellite or aerial image based on its
spectral or spatial features.

Types:

 Supervised Classification: Uses labeled training data (e.g.,

Maximum Likelihood, SVM).

 Unsupervised Classification: Clusters pixels without prior labels

(e.g., K-Means).

Importance:

 Land Use/Land Cover Mapping: Understand human activities and

natural resources.

 Environmental Monitoring: Detect deforestation, urban sprawl,

water quality changes.

 Disaster Management: Classify affected vs. unaffected areas in

flood/fire zones.

 Agriculture: Monitor crop types, health, and yield estimations.

 Urban Planning: Identify buildings, roads, vegetation for

infrastructure management.

UNIT 4
1. Object Detection in Remote Sensing Imagery

Object Detection:
It involves identifying and locating objects (like vehicles, buildings,
ships) within satellite or aerial images.

Common Algorithms:
 YOLO (You Only Look Once):

o Real-time object detection.

o Divides image into grids and predicts bounding boxes and class
probabilities.

o Fast but may compromise accuracy in dense scenes.

 Faster R-CNN:

o Two-stage detector: first generates region proposals, then

classifies them.

o Higher accuracy, slower than YOLO.

o Common in high-resolution remote sensing for detailed object

recognition.

 SSD (Single Shot Detector):

o Balances speed and accuracy; processes image in one pass.

o Efficient for detecting multiple objects at different scales.

2. Tracking Moving Objects in Remote Sensing

Tracking Vehicles/Ships:

 Uses time-series satellite/aerial images or video from drones.

Approaches:

 Frame-by-Frame Detection: Detect objects in each frame and match

them based on position and velocity.

 Optical Flow Methods: Track pixel movement between sequential

images.

 Kalman Filters: Predict future positions using current movement

trends.

 Deep Learning (e.g., DeepSORT + YOLO): Combines object

detection and tracking robustly.

Use Cases:

 Maritime surveillance.
 Monitoring illegal fishing or traffic flow in urban planning.

3. Change Detection in Remote Sensing

Change Detection:
Identifying differences in land use or cover over time using multi-
temporal imagery.

Methods:

 Image Differencing:

o Subtract pixel values from two dates.

o Highlights significant changes, e.g., NDVI before and after

deforestation.

 Post-Classification Comparison:

o Classify images from two times separately, then compare

classified maps.

o Useful for detailed thematic changes but error-prone if

classifications are poor.

 Pixel-Based Methods:

o Analyze individual pixel changes over time.

o Includes thresholding, change vector analysis, PCA-based

methods.

4. Significance of Time-Series Analysis in Remote Sensing

Time-Series Analysis: Analyzing sequences of remote sensing

images over time to observe trends, cycles, and anomalies.

Significance:

 Track seasonal crop growth.

 Detect gradual land degradation or urban expansion.

 Monitor ecosystem recovery post-disaster.

 Improve model predictions using temporal patterns.

 5. Change Detection Analysis (Example: Flood Impact)

Steps:

1. Collect Satellite Images: Before and after the flood (e.g., Sentinel-1
SAR for all-weather capability).

2. Preprocessing: Radiometric and geometric corrections, cloud

masking.

3. Water Classification: Apply threshold on NDWI or SAR backscatter.

4. Difference Mapping: Subtract pre- and post-flood classified maps.

5. Impact Assessment: Calculate area flooded, identify affected land

uses (cropland, urban zones).

Tools: Google Earth Engine, QGIS, Python (e.g., Rasterio,

Scikit-learn).

6. Advantages of Hyperspectral Imaging

Aspect Hyperspectral RGB/Multispectral

Spectral 3 (RGB) or ~10

100s of narrow bands
Bands (Multispectral)

Detail Identifies subtle material Broad-level

Level differences classification

Mineral exploration, crop General land use

Uses
stress detection classification

High precision, chemical Faster, less complex

Benefits
composition analysis processing

7. LiDAR Technology and DEM Generation

LiDAR (Light Detection and Ranging): Uses laser pulses from

airborne platforms to measure distances to Earth’s surface, creating 3D
point clouds.

Applications:
 Forest canopy analysis

 Infrastructure mapping

 Terrain modeling

DEMs from LiDAR:

1. Collect point cloud data (X, Y, Z coordinates).

2. Classify returns: Ground, vegetation, buildings.

3. Interpolate ground points to generate a Digital Elevation Model

(DEM).

Benefits:

 High vertical accuracy

 Penetrates vegetation to measure true ground surface

 Supports flood modeling, slope analysis, and urban planning

8. Multi-Modal Remote Sensing Data Fusion

Definition:
Combining data from different sensor types (e.g., hyperspectral +
LiDAR, optical + radar) to enhance analysis.

Why It’s Useful:

 Leverages complementary strengths:

o Hyperspectral: spectral detail

o LiDAR: elevation and structure

 Improves classification accuracy and contextual understanding

 Supports complex applications like precision agriculture and smart city

mapping

9. Case Study: Fusion of Hyperspectral + LiDAR for Vegetation

Health Monitoring

Application:
 Assessing forest health in a mountainous region.

Methodology:

 Hyperspectral Data: Captured reflectance patterns indicating plant

stress (chlorophyll, water content).

 LiDAR Data: Provided tree height, canopy structure, and terrain

elevation.

 Fusion Technique: Feature-level integration followed by Random

Forest classification.

Results:

 Achieved over 90% accuracy in classifying healthy vs. stressed trees.

 Enabled early detection of disease spread and areas prone to

landslides due to vegetation loss.

Implications:

 Supports proactive forest management.

 Offers insight into terrain-vegetation interaction.

UNIT 5
1. Importance of Mapping Land Cover and Land Use for
Sustainable Development

Definitions:

 Land Cover: Physical material on Earth's surface (e.g., forests, water,

buildings).

 Land Use: How people utilize land (e.g., agriculture, urban

development, recreation).

Importance:

✅ Environmental Management:

 Monitors deforestation, wetland degradation, desertification.

 Assesses biodiversity loss and natural resource depletion.

✅ Urban Planning & Infrastructure:

 Guides zoning regulations and expansion planning.

 Prevents urban sprawl and ensures green space conservation.

✅ Climate Change Mitigation:

 Evaluates carbon sinks (forests) and emission sources (urban zones).

 Supports reforestation and land restoration efforts.

✅ Disaster Management:

 Assesses floodplains, wildfire risk zones, and post-disaster land

changes.

✅ Sustainable Development Goals (SDGs):

 Helps monitor land use efficiency (SDG 11), protect ecosystems (SDG
15), and promote sustainable agriculture (SDG 2).

2. Methods for Land Cover Classification in Remote Sensing

Common
Method Description
Algorithms

Based on training data SVM, Random

Supervised (labeled samples). Forest,
Classification Human intervention Maximum
required. Likelihood

No labeled data needed.

Unsupervised K-Means,
Algorithm clusters pixels
Classification ISODATA
into classes.

Segments image into

Object-Based Decision Trees,
objects (not individual
Classification Rule-based
pixels). Useful for high-
(OBIA) approaches
res data.

Automatically extracts
Deep Learning-
features using CNNs or CNN, U-Net,
based
ResNets. Accurate but ResNet
Classification
data-hungry.

Tools Used:
 QGIS, ArcGIS, Google Earth Engine, Python (Scikit-learn, TensorFlow,
PyTorch)

3. Case Study: Urbanization in Delhi-NCR (India)

Context:

Delhi-NCR has experienced rapid urban growth due to population

pressure and industrialization.

Remote Sensing Role:

 Data Used: Landsat images (1990–2020)

 Techniques: Supervised classification with SVM and NDVI analysis.

 Findings:

o Urban area increased by ~120% in 30 years.

o Agricultural land and open land decreased significantly.

 Implications:

o Urban sprawl led to loss of vegetation and groundwater recharge

zones.

o Informed policymakers for green belt development and smart

city initiatives.

4. Role of Remote Sensing in Precision Agriculture

Precision agriculture uses high-resolution satellite/drone imagery and

sensor data to optimize farming operations.

Applications:

✅ Crop Health Monitoring:

 NDVI, NDRE, and other vegetation indices detect plant stress (disease,
drought).

✅ Soil Moisture & Fertility Mapping:

 Thermal and radar sensors assess water content and salinity.

✅ Variable Rate Application:

 Fertilizers, pesticides, and water are applied based on crop-specific
needs (reduces cost and runoff).

✅ Yield Prediction:

 Time-series analysis forecasts yield based on growth stages and

weather data.

✅ Field Zoning:

 Segments fields into management zones based on productivity.

Benefits:

 Higher crop yields with fewer resources.

 Environmentally sustainable farming.

 Early detection of threats (pests, droughts, diseases).