Theme
Energy

Location
C-120

Title
Aware Adjustable Car Mirror

Team Members
Cristian Llerena, Michael Mogilevsky, Adarsh Narayanan, Muralii Krishnan Thirumalai, Hazem Zaky

Advisor(s)
Dr. Kristin Dana, Dr. Maria Striki, Dr. Zhao Zhang

Abstract
In today’s automotive industry, driver safety and user convenience remain top priorities. Yet mirrors are often overlooked, despite their essential role in maintaining situational awareness. Misaligned mirrors can increase blind spots and increase the risk of accidents. This Capstone project addresses this issue by developing a self-adjusting rear-view mirror, powered by computer vision and mechatronics. The computer vision part involves the usage of stereo vision cameras, OpenCV & Dlib’s 68-point facial landmarks model on a Raspberry Pi 5. The model detects key facial landmarks of the driver, to estimate their head orientation and position. These features are then triangulated into a 3D coordinate system via a Direct Linear Transform (DLT), from which the optimal angles for the rear-view mirror are computed and sent to the micro-controller through serial communication. An arduino micro-controller is used to implement a control system algorithm (PID Controller), which guarantees proper motion and control of the DC motors to place the mirror on the computed coordinates. The design for this dual-axis mirror device includes custom 3D-printed parts, and magnetic encoders as a feedback sensor. In parallel, a mobile application was developed to provide the user control over the position of the mirror. The app includes features for saving and loading presets, with a database hosted by Google Firebase. Ngrok was utilized to establish a secure HTTP tunnel between devices, aiming to avoid the leakage of the Raspberry Pi’s public IP. Together, all these modules provide an user-friendly and intelligent device for rear-view mirror adjustments.

Poster
[LINK]

Video
[LINK]