With image processing, we can identify and track objects of interest within images or video feeds.
We can do this by making use of opencv's library for image processing. https://opencv.org/
The method we can use to do this is to classify by:
1. Shape (insulators, transformers, tree leaves, tree flowers/pods, branches, reef shapes)
2. Texture (rust surface, reef structures)
3. Color (reef colors, transformer colors)
1/25/2018, Sample Reef Images
https://drive.google.com/drive/folders/0B1qRGF93OqxQenBkR0ZzWHJKcHc?usp=sharing
========
1/21/2018
========
1. Our own data set: Haole Kao trees, reefs, etc.
We should take some pictures of Haole Koa tree asap, as we can do this ourselves. I am asking the reef images again from ARL.
We need to build our own image sets; Haole Koa trees, reefs, etc.
The issue with Google images is: they are not the same object. We need to build our own object imag set, such as Haole Kao tree leaves, flowers, branches.
2. Our own filters: Haar, CNN
We have our Haar filters working. We need to build our own CNN filters.
3. Tasks
1) training filters with images
2) detecting objects in the testing images
3) auto extracting images from a video stream
Evan's Fall 2017 report EE699 report Evan Law, Fall 2017