Awesome

Automatic-Defect-Inspection-of-Solar-Farm

To replicate work follow https://github.com/RonakDedhiya/Automatic-Defect-Inspection-of-Solar-Farm/blob/master/Tensorflow%20Object%20Detection%20API/README.md

<img src="./Dataset/ref/images1.PNG"> <img src="./Dataset/ref/images2.PNG">

Below Image shows different types of faults which occur in solar panel and by their thermal images, It can be visually classified!!!

Image1: From set of panels, one panel is brighter than rest one of them!! It's Over Heated - Overheated Components Fault(OHC Fault)

Image2: In a single panel, Middle cells are having different heat variation, its because of short circuit diode - Bypass Diode Fault

Image3: There is a bright spot visible which is called Hotspot, It occurs due to dirt, scratches or cracks

Image4: It is Zoomed out Image where each small rectangle is panel and we can see bottom panels are yellow in colour compared to upper panels, Its bacause of interconnection Fault - Here bottom set of panels are disconnected from power supply.(FI- Faulty Interconnection)

Image5: It is a combination of Hotspot and known as Cluster of Thermal Anomalies(CTA Fault)

<img src="./Dataset/ref/images3.PNG">

Below is the process we have used to do fault classification and localization using deep learning technique called object detection(one-shot learning) & Transfer Learning

<img src="./Dataset/ref/workflow-for-object-detection.png">

As we can see,we have provided a test image and our trained model is able to classify what type of fault it is and put a bounding box around it. Here, It is detecting Diode Fault and certain Hotspots

<img src="./Dataset/ref/images5.PNG">

This is the extension part where we are showing causes of fault from our knowledge base

<img src="./Dataset/ref/images6.PNG">

This is the precautions of fault detected

<img src="./Dataset/ref/images7.PNG"> <img src="./Dataset/ref/images8.PNG"> <img src="./Dataset/ref/images9.PNG"> <img src="./Dataset/ref/images10.PNG"> <img src="./Dataset/ref/images11.PNG">

Misclassification:

1.Diode Fault is misclassified as Faulty InterConnections

2.Over Heated Components Fault is misclassified as Hotspot

Zoomed Images creates confusion for deep learning model

Solution: Train it on large Image set

<img src="./Dataset/ref/images13.PNG"> <img src="./Dataset/ref/images14.PNG">

Conclusion:

This automatic defect inspection application for solar farms demonstrates that deep learning technology can be applied to solve real-world problems, such as unmanned inspection in harsh or dangerous environments. These architecture can learn the sophisticated features from the input images for classification and detection tasks. This is a general solution for numerous inspection services in the markets, which can be used for oil and gas inspection, such as pipeline seepage and leakage; utilities inspection, like transmission lines and substations; and even for crisis response to emergencies. The UAVs can fly high up for close-up inspections without putting workers in danger. And the automatic defect inspection system can greatly improve the efficiency of mass data analysis without the help of skilled workers.