🛠️ Hardware

Over the weekends, we designed and printed a slanted ‘chair’ for a stepper motor to rotate the gear attached to the objective lenses.

However, the gear we designed was too thin, and the struts were too weak. Therefore, we reprinted a stronger gear and tested it. During testing, we also realised that the locking mechanism introduced too much resistance to the gear system and we had to remove it.

Additionally, upon further testing, we found that the belt in the belt drive system continued to slip occasionally. After isolating the components, we realised that it was caused by the bending of the platform. As the movable chair moves along the platform, the weight distribution on the platform changes, causing the belt to lose tension. Therefore, it results in the belt slipping, which we observed. To resolve this, we increased the width of the bridge to reduce the bending of the platform.

To create less resistance for the gears, we removed the screws that formed the locking mechanism for holding the lens in place.

💻 Software

As part of the cell identification process, we needed to minimise the noise in the image and isolate each cell in a clean, uniformly framed image. To do so, our code trimmed a border of about 0.5% of the image dimension to remove any irregularities at its edges. It then converted the trimmed image to a grayscale and distinguished the foreground (cells) from the background via a binary threshold.

We then located the boundaries of the rectangle threshold region and cropped the original color image to that box, ensuring that there was only the area containing the cell left.

Next, we created a circular mask centered on the cropped region and merged this mask as an alpha channel with the cropped BGR image, thus producing a cleanly cropped image with no background. This could be applied across all the images taken.

This week was also the week we began putting everything together to begin automating Cooper. Due to issues with security when using website automation, I have decided to give up on that. Instead, Cooper will need a few pokes here and there to keep him going. With that, Cooper can successfully outsource the cell counting to GoogleColab, where images will be sent to a Google Drive to be accessed by SAM. The cell counts will then be written into a text file for Cooper to read.