Monday

The Harris corner detector was not accurate enough to match the right images so we had to ask for help in other ways to address the problem. We had a long talk in which he explained what Harris corner detector does and why it fails compared to other algorithms out there.

Tuesday

We had our lab meeting as usual. However, this was quick since nobody had much to update since last week. I took the opportunity to ask around if people knew of better ways to match images but no luck. Nonetheless, I met with Professor Crandall after the meeting to show him how poorly the SIFT detector did. He advised us to continue trying and looking for better ways to improve accuracy.

Wednesday

We started off our day brainstorming another matching which we could use, which didn’t involve Harris corner detection. After long hours of research and reading, we found a SIFT detector method which involved Flann Based Matching. We also attended an IARPA meeting where members of the US government’s main intelligence research agency gave talks on projects which were geared towards improving cybersecurity and national intelligence.

Thursday

We finally had a working algorithm that was able to match identical features in similar images and vice versa in unidentical images. We met with our mentor at 10:30 AM to show him our working algorithm and get instructions on what next we needed to do. This was indeed a big step for our project; the next task was for us to get a bigger dataset of microchips and test our prototype algorithm with.

Friday

My main task for the day was to increase the dataset, which will put our prototype algorithm to the test. I took pictures of 60 different microchip – 10 groups comprising of 6 microchips each. This dataset varies in size in order to create difficult tasks for the program to match. This method would allow us to determine the accuracy and how effective this program may work depending on the circumstance.