Computer Vision Security System


The use of security systems is becoming more prevalent in this modern age. With the advent of machine learning algorithms, there is a unique opportunity to integrate machine learning algorithms into modern camera-based security systems. We present an inexpensive, light-weight computer vision security system with an integrated deep neural network. The security system uses active image classification to detect intruders and provide a notification and a video to the owner over email. In addition, the system maintains an integrated graphical user interface controller as well an android application controller.

The project code is available on GitHub here under an MIT License

E-Puck Robot Control And A Star Navigation

This is a part of a lab I designed for undergraduate students at Vanderbilt University


A common problem in autonomous route planning and controls to implement efficient motor control, sensor feedback, and route planning algorithms to allow a robot to successfully navigate and traverse a line grid maze from a start point to a different destination point in an efficient manner. We present an E-Puck controller that utilized ground sensor feedback to perform a variety of functions; detect nodes (cross-sections), detect lines just past nodes, provideBraitenberg line tracking functionality, and allow for left and right turns on a line grid. In addition, we present an AStar algorithm that is utilized to efficiently solve for a route from a start node to an end node in a line grid maze. Finally, we present tuned parameters for node detection, line just passed node detection, line following, turning logic, and ground sensor threshold parameters resulting in a stable system.

The Webots project is available on GitHub here under an MIT License