Drone Solar Panel Inspection Solution

Summary

Our client specializes in visual inspection data management, simplifying complex work processes, and making inspection analysis and reporting easy. To build an application that can detect anomalies in images of solar panel farms uploaded to our client's customer portal, our client needed an AI solution partner.

About the Customer

Specializing in Wind turbines, Solar PV, Transmission and distribution, Rooftop, and Building inspections, our client collaborates with industry leaders to efficiently analyze data. As leading industry experts they combine the strengths of Artificial Intelligence and expert industry analysis for optimal results in visual inspection.

Understanding the Challenge

Our client needed an AI solution partner to build an application that can detect the anomalies in the images of the solar panel farms that are uploaded on their customer portal.

Solution

Folio3 AI built an AI-powered application for the client that processes the images sent by our client’s customer portal. The AI model integrated by Folio3 AI is fine-tuned to detect and identify anomalies in uploaded thermal images according to the client’s use cases.

• Customized AI model: The algorithm was fine-tuned and trained on the client data set and customized according to solution needs to detect various brand logos in real time.
• Excluding Overlapping Images: The AI model can exclude overlapping images that are captured from different drone positions. Providing accurate detection results and anomaly counts.
• AI Anomaly Detection: The custom solar panel anomaly detection algorithm fine-tuned on client data will detect the following Diode Failure and Hotspot Failure on the images that are uploaded on the client portal.
• Maximized Accuracy: The model is capable of excluding any images detected outside the solar panel boundary to provide accurate results regardless of environmental factors that could compromise the model

Result

The Folio3 AI-built solar panel inspection solutions were deployed on the client’s existing platform with more than 90% model accuracy.