Related Experiment Video
Updated: May 21, 2025
08:27
Author Spotlight: Efficient Image Recognition Using Directional Gradient Histogram Technique and Support Vector Machines
Published on: January 5, 2024
933
Deep Learning-Based Recognition and Classification of Soiled Photovoltaic Modules Using HALCON Software for Solar
Shoaib Ahmed1, Haroon Rashid2, Zakria Qadir3
1Department of Electrical and Electronic Engineering, Faculty of Engineering, University of Ryukyus, 1 Subaru, Nishihara 903-0213, Nakagami, Okinawa, Japan.
Sensors (Basel, Switzerland)
|March 17, 2025
Summary
This study introduces a deep learning system for identifying soiled solar panels, improving cleaning robot efficiency. The advanced image recognition accurately classifies soiling, optimizing solar energy output and maintenance.
Area of Science:
- Renewable Energy Systems
- Artificial Intelligence in Engineering
- Computer Vision for Robotics
Background:
- Global solar photovoltaic (PV) capacity is rapidly expanding.
- Panel soiling significantly degrades PV system efficiency and energy output.
- Automated cleaning strategies are needed to maintain optimal performance.
Purpose of the Study:
- To develop a deep learning-based system for recognizing and classifying soiled PV images.
- To enhance the capabilities of solar cleaning robots using the HALCON software framework.
- To enable intelligent visual analysis for optimizing PV maintenance.
Main Methods:
- Utilized Extreme Learning Artificial Neural Network (EANN) and Convolutional Neural Network (CNN) architectures.
- Employed advanced image processing techniques within the HALCON framework for image acquisition, preprocessing, and segmentation.
- Integrated trained deep learning models for robotic control and soiling pattern classification.
Main Results:
- Achieved high precision in detecting and classifying soiling patterns: EANN at 99.87% and CNN at 99.91%.
- Demonstrated the system's effectiveness in precise soiling recognition and classification.
- Validated the potential for improving automated cleaning strategies and reducing unnecessary cleaning cycles.
Conclusions:
- The deep learning approach significantly enhances the accuracy of soiled PV image classification.
- The HALCON framework effectively supports the deployment of intelligent visual analysis for robotic solar panel cleaning.
- Intelligent visual analysis is crucial for optimizing maintenance and maximizing the performance of solar energy applications.
