Capstone.

September 2016 – April 2017

I enrolled into a Capstone project in my last semester in University of Toronto. It was the closest project in school where I felt like I was build a company.

What is Capstone?

“An experience in engineering practice through a significant design project whereby student teams meet specific client needs through a creative, iterative, and open-ended design process.” It is a project where it simulate a professional contract work done according to the client desire. We worked with people with different disciplinary knowledge and skill. I was in a group of of student consist of electrical engineering, computer engineering and mechanical engineering/ computer engineering. It sounds like a nightmare to work together. I was so afraid that even before the semester started, I email our supervisor of what material should I study to catch up with the group. She told me not to worry too much. It turn out to be very good experience. What I learn from capstone is that you can’t finish the project by yourself like many other school project. You need to communicate well with the team and trust them with their solution. So this is what all capstone about, working together as a group to solve a problem and having fun while doing it.

My captone group presenting to out supervisor, Joyce Poon.

Problem

Defense Research and Development Canada (DRDC) is in need of a projectile detection sensor system for conducting marksmanship studies. The current system (SIUS LOMAH) is able to collect data within the accuracy requirements of the client. However, there are a few issues with this system. These include a time consuming setup and calibration process, lack of feedback about the calibration offset and operation status of the sensor system, the collected data is not easily exportable for further analysis, and the client-side user interface is lacking in its representation of the collected data and usability.

The focus of the proposed design was the accurate collection of marksmanship data, secure transmission of this data, and clear digital representation of the data for the soldiers and the researchers. Additionally, our team targeted portability, easy installation and calibration. The above mentioned considerations enable easier operation, simplified processing of data, and improved marksmanship.

The project was organized into three systems: the Detection System, Communication System, and User Interface and Data Store. The Detection System is powered by a series of micro-controllers that return real-time data to the local web server. The web server stores the collected data in local Data Store (Database). The web server serves the User Interface via a web browser to various user devices.

Engineering behind our detection system

Light sensor approach

The solution developed by the team was to use a set of precision linear sensor arrays developed by AMS, designated as TSL1402R.

There are four distinct circuits within the detection system The first is the sensor circuit, which outputs a digital signal if the voltage of the sensor is below a certain threshold during a read operation. The second is the pixel recording circuit, which stores which pixels were indicated by the sensor circuit to be below the voltage threshold. The third circuit is the control circuit, which ...

READ MORE »

TSL1402 sensor

The key advantage to using these devices is that majority of signal processing circuitry required to analyze the sensor output is internal to the chip. In addition, each 2cm device contains 256 sensors, allowing for very precise measurement of which light sensor experienced a reduction in light.

User Interface

I am the one who is responsible for the frontend of the detection system. The frontend received the data with REST API from the data base running on the server. The frontend is where the user would be interacting with the detection system. The user would be interacting with the web application on an ASUS Transformer Pad. Features include:

Displays the detected shot in real time.
A CSV file containing the all the shot data is available to download.
Users can set theme colour to their liking that set well to the surrounding environment.

The web application is build with ReactJS framework. ReactJS framework served pretty well for the features we needed to implement. There are field that need to be updated when the detection system is live. ReactJS is decided to handle the update on a efficient matter.

Admin login

Administrator account has more features than general user account..

User login

General user account have access to the dectection system display.

Lock screen

User can lock his screen while away.

Display detected shot

Detected shot will be display on the screen

Download data file

Administrator can download shot data into a csv file format.

Switch Theme

User can switch into theme with different constranst.

Challenges

When the shooter shoot through the detection system, the target will be display on a the screen. One of the challenges is to figure out how to display the shoots on the screen. One approach is to draw the canvas with every new shoot. The downside of this approach is that ReactJS is not build to draw on the html canvas. Which means I wouldn’t gain any advantage of using ReactJS. So to phrase the problem, how to display the shoot on the screen efficiently while still having ReactJS benefit.

While searching for a solution, I stumbled on a ReactJs scatter plot chart that has similar principal that I was trying to solve, to plot a dot on a plane. The chart library was build for ReactJS, it is very efficient as in that it will only display dot on the screen only when it received new data.

A plot chart are not design to have the coordinate (0,0) on the centre of the chart and the chart might have rectangular shape. So what I did is to put some heuristic rule on the plot chart. Limit the chart width and height to be always squared shape, set the (0,0) coordinate to the centre, set the y axis and x axis to have the same range, etc.

Scatter Plot

The display is run on a C3 scatter plot library.

Sentiment analysis with tweets

Developed a program that extract features from raw tweets to classify its topics and sentiments.
Utilized WEKA to classify and analysis data into positive and negative sentiment.
Reported classification accuracy on test data among, SVM, naïve Bayes and decision tress.
Concluded that SVM perform better that its counterpart if large amount of data is given.

Object Recognition in 2D Medical Images

Developed a visual based automated system to characterize histopathology images of cancerous cells.
Utilized convolutional neural network and color segmentation to detect patches with a nucleus at the center and to classify the detected nuclei patches in the different classes.
Preformed 0.65 weighted average F1 score on nucleus classification and 0.7 of nucleus detection.

Simple router with NAT

Developed a simple router with NAT that handles packets routing in a team of 2.
Implemented the functions that unwraps the incoming network packets, translates the packet to the desired destination and sends the packet to the correct port.
Passed all tests that was assign to the project.

SynergySpace

Developed an Airbnb like web application for co-working space rental with a team of 4.
Contributed the implementation of the listing and payment page of the application.
Resolved all features that are need for the application
Prepared a professional report that addressed all aspects of the project, such as database model, framework mode and user stories.

About Me

Student at the University of Toronto studying Computer Science. Fast learner and always improving myself. Also I'm a spicy food lover who coincidentally know Spanish.

Available for Hire!

Download Resume

Toronto, Canada
Phone: (647) 609-9289
Email: carloskb.lee at mail.utoronto.ca