AeroVision

Why I started

I want to study aerospace engineering, and I wanted to use my Python skills for something connected to aircraft. That led me to a simple question: could I compare two photos and highlight something that changed?

What AeroVision does

AeroVision is a small prototype that finds edges in an aircraft photo and highlights differences between images.

For this demonstration, I added an artificial mark to a copy of one photo. The program compares the copy with the original and draws a box around the change. I have not tested it on real aircraft damage.

How I built it

I used Python and OpenCV in a Kaggle notebook, with NumPy for working with the image data and Matplotlib for displaying the results.

The notebook saves the images and exports the test results as CSV and JSON files. I also created an AWS Lambda function that counts how many supplied results passed. I manually copy the results into Lambda; it does not process the photos.

What worked

The notebook passed ten tests with artificial changes and one comparison using an identical photo: 11 out of 11 controlled tests.

I was pleased to see the basic method working. These tests are only a starting point, though. They use one photo and do not show how well the program would work during a real inspection.

What was difficult

Getting the photo into the notebook took longer than I expected. I moved from Colab to Kaggle and changed the code to find the uploaded file. I also had to fix errors in my Lambda function and choose the correct test event.

Working through these problems gave me more practice with debugging and helped me understand how the different parts of my project work.

Where I am now

I worked on AeroVision individually. It is an early prototype with a working notebook, saved images, controlled test results, and a separate Lambda summary.

What I want to do next

I want to try real before-and-after photos and see how lighting and camera movement affect the comparison. I also need advice from people with aircraft inspection experience to understand what would actually be useful.

My longer-term goal is to help someone reviewing inspection photos find areas worth checking. There is still a lot to test before AeroVision could support that kind of work.

Built With

• amazon-web-services
• kaggle
• matplotlib
• numpy
• opencv
• python