Scribd
Upload
23 views11 pages

Solar Vision Project Report

Uploaded by

durgesh.vns24
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
23 views11 pages

Solar Vision Project Report

Uploaded by

durgesh.vns24
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PPTX, PDF, TXT or read online on Scribd
You are on page 1/ 11

Solar Vision: Leveraging

Automated Detection and


Quantification of Solar Panels in
Urban Areas to Promote
Renewable Energy Adoption and
Achieve
A Machine Sustainable
Learning-Based Approach
for Solar Panel Detection and
Development Goals
Quantification
Presented by: Durgesh Kumar Singh
Date: [Insert Date]
Project Overview
• Objective: Develop an automated system to
detect and quantify solar panels in urban
areas using machine learning techniques to
support sustainable energy goals.
• Goals:
• 1. Accurate detection of solar panels from
satellite imagery.
• 2. Quantification of solar panel installations to
estimate energy production potential.
• 3. Contributing to renewable energy adoption
Dataset Overview
• Origin of Data: Satellite imagery sourced from
public and private datasets.
• Uniqueness of Project:
• - Focuses on urban solar panel detection,
presenting challenges due to occlusions and
high building density.
• - Combines detection and quantification for
comprehensive analysis.
Data Preprocessing
• Bounding Box Creation: Precise manual
annotation of solar panels.
• Data Annotation: Multi-label annotation,
including panel orientation and shading
factors.
• Data Augmentation: Rotation, Scaling,
Flipping, Contrast Adjustments to improve
model robustness.
Data Visualization
• Heatmap Visualization: Displays solar panel
distribution in urban regions.
• Training Metrics: Precision, Recall, F1 Score,
mAP shown through graphs.
• Graphical Analysis: Accuracy and loss curves
during training and validation.
Model Architecture
• Model Selection: Koko, YOLO v8, YOLO v8 OBB,
YOLO v8 OBBN, YOLO v8 S, YOLO v8 M.
• YOLO models were chosen for real-time object
detection, while OBB variants capture rotated
solar panels for higher accuracy.
Panel Counting Methodology
• Holistic Edge Detection (HED): Fine-grained
edge detection to outline solar panels.
• Canny Edge Detection: Used for sharper
boundaries and accurate panel count.
Performance Evaluation
• Model Comparisons: Performance of Koko,
YOLO v8 variants, and edge detection
techniques.
• Key metrics: Precision, Recall, mAP, FPS
(Frames per Second).
• Best Model: YOLO v8 OBBN achieved the
highest accuracy and speed.
Visualization of Results
• Sample Detections: Satellite images with
bounding boxes showing detected solar
panels.
• Quantification Output: Estimated solar panel
counts and corresponding energy potential.
Future Work
• Scalability: Expand model to rural areas and
incorporate advanced satellite imagery.
• Improvement in Counting Accuracy: Explore
AI-powered edge refinement techniques.
• Integration with Policy: Support urban
planning with solar potential maps.
Conclusion
• Key Takeaways: Successfully automated solar
panel detection and quantification in urban
areas.
• This project contributes to renewable energy
adoption and supports sustainable urban
development.

You might also like