1 / 25

Team 14 Vivian Liu Eric Tsai

Driving Alarm System. Team 14 Vivian Liu Eric Tsai. Driving Alarm System. Purpose – To design an alarm system that will assist drivers who cannot focus on the road. Motivation – Everyone has experienced near accidents because of a few seconds of negligence. Initial Ideas.

hila
Download Presentation

Team 14 Vivian Liu Eric Tsai

An Image/Link below is provided (as is) to download presentation Download Policy: Content on the Website is provided to you AS IS for your information and personal use and may not be sold / licensed / shared on other websites without getting consent from its author. Content is provided to you AS IS for your information and personal use only. Download presentation by click this link. While downloading, if for some reason you are not able to download a presentation, the publisher may have deleted the file from their server. During download, if you can't get a presentation, the file might be deleted by the publisher.

E N D

Presentation Transcript

  1. Driving Alarm System Team 14 Vivian Liu Eric Tsai

  2. Driving Alarm System Purpose – To design an alarm system that will assist drivers who cannot focus on the road Motivation – Everyone has experienced near accidents because of a few seconds of negligence

  3. Initial Ideas • Measure reflected light from the retina • Determine if eye is closed or not focused on the road • Have 2 cameras operate independently • Capture 2 different images of the head • Hao Tang suggested 3D reconstruction • Use Active Appearance Model to detect facial features • Reconstruct 3D coordinates for each facial point

  4. Objectives • Develop an algorithm that can isolate the ear and nostrils • Needs to be fast and reliable • Cannot hinder the driver • Track the facial points to monitor head position • Design a circuit to generate a beeping noise • Only start beeping when the software sends signal

  5. Equipment

  6. Canny Algorithm • Developed by John F. Canny • Algorithm to detect a wide range of edges in an image • Should mark as many real edges as possible • The marked edges should be as close as possible to the real edge • Two threshold parameters used to generate the set of edges • Higher of the two is used to find the initial segments of strong edges • Lower one used to link the edges • Too high can cause the algorithm to miss important information • Too low and irrelevant information will be tagged

  7. Canny Algorithm Cont.

  8. Detecting Nostrils • Use Canny Algorithm to convert the frame into a binary image of edges • Use higher threshold to filter out background images • Attempt to fit ellipses around each set of edges • Approximate the ellipse with least squares method given a set of 2D points • Parse through the different ellipses to determine which 2 cover the nostrils • Determine based on the length of the major and minor axis of the ellipses • Take advantage of the face’s symmetry to filter out bad points

  9. Detecting Nostrils Cont.

  10. Detecting Nostrils Cont.

  11. Detecting the Ear • Apply the same method as detecting the nostrils • Attempt to isolate the ear canal though any part of the ear will do • Take advantage of the small elliptical shape of the canal • Restrain the ear to be near the center of the screen to aide in detection

  12. Detecting the Ear Cont.

  13. Detecting the Ear Cont.

  14. Lucas-Kanade Method • Two frame differential method for optical flow estimation • Optical flow : pattern of apparent motion of objects, surfaces, and edges • Still very popular • Attempts to find correlations between adjacent frames • Generate vector field showing where each pixel has moved from time t to t+δt • Use partial derivatives with respect to spatial and temporal coordinates • Assume the pixels locally flow at a constant rate

  15. Lucas-Kanade Method Cont.

  16. Lucas-Kanade Method Cont.

  17. Alarm Circuit • Only generate sound when a +12V signal is received from the serial port • Voltage regulator will force the output voltage to be 5V so the PIC won’t burn out • The PIC is programmed to play a repetitive note (C)

  18. Communicating with the PC • 9-pin serial port allows the PC to send signals to the PIC • Output signal is either at ±12V • Software will send a high signal when the timer reaches 3 seconds • Continue sending high signal until head position is within acceptable range • Diode will block the -12V output

  19. Testing Procedures • Make sure program can isolate facial features • Test the conditions for head tilting and nodding • Check if PIC can generate sound • Measure output voltage of serial port, try to switch to high through software

  20. Successes • Software was able to isolate key facial points • Tracking the facial points was mostly satisfactory • Good constraints for determining alertness • Capable of communicating with the alarm circuit • Alarm circuit generated annoying noises

  21. Improvements • Need to increase the frequency of a successful isolation • Determine good conditions on when to automatically recalculate the position of the facial points • Filter out unnecessary background images to allow for use in any environment • Depth perception • Alertness conditions will vary based on the distance of the face from the camera

  22. Special Thanks • Professor G. Swenson • Professor T. Huang • Professor J. Coleman • Tomasz Wojtaszek • Hao Tang • Yuxiao Hu • Xingdou Fu

  23. Questions?

More Related