Scribd
Upload
85 views20 pages

Iris Pi

This document describes an assistive technology called Irisπ that helps blind people navigate city environments. It has two modules: (1) a crosswalk module that detects crosswalks using computer vision and guides users across streets, and (2) an elevator module that detects elevator floor numbers using templates and informs users. The system aims to make daily tasks like crossing streets and using elevators independently possible for visually impaired individuals. It analyzes images using techniques like thresholding, contour detection, and template matching to locate helpful navigational elements.

Uploaded by

locorecto007
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
85 views20 pages

Iris Pi

This document describes an assistive technology called Irisπ that helps blind people navigate city environments. It has two modules: (1) a crosswalk module that detects crosswalks using computer vision and guides users across streets, and (2) an elevator module that detects elevator floor numbers using templates and informs users. The system aims to make daily tasks like crossing streets and using elevators independently possible for visually impaired individuals. It analyzes images using techniques like thresholding, contour detection, and template matching to locate helpful navigational elements.

Uploaded by

locorecto007
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 20

Iris

An ultra low-cost solution for the blind


in the city
Capstone Students: Luis Maurad, Jean Sanchez, Jason Kwong
Advisor: Professor Zhigang Zhu
Department of Computer Science, City College of New York
Introduction
More than 285 million visually impair people in the world. That is
more than 4.2% the world population.
50 million daily elevator ridesamong58,000 elevators in NYC.
4 crosswalk for every street interception
Very little attention research on helping blind people cross the
streets and use the elevators
It would be nice to have a system to help blind people, specially
in the city, in their daily lives.
Motivation
Walking across a street is a simple process:
1. Find a crosswalk
2. Determine if the sign is in walk mode
3. Check for any oncoming cars
4. Walk across before the sign changes

This process becomes a blind person's travel nightmare:
5. Where is the crosswalk?
6. Can I cross the street or is the sign in 'Don't walk' mode?
7. Am I within the crosswalk or have I crossed its bounds (bump into a car,
walk into intersection)?
8. Is the timer about to run out?
Motivation
Elevator Floor Reading Process:
1. Walk into elevator
2. Press Floor Button
3. Look at panel to see current floor
4. Get off at required floor
A blind Persons Process:
5. Enter the elevator
6. Ask someone to press the floor number or press button if
there is no one available(Braille buttons)
7. Count the floor beeps
8. beep count == floor level ? get out of elevator
Overview
Iris

1. Laptop Computer (running Linux) 4. Sun Glasses
2. USB Cameras X2 5. Small Speaker
3. Numeric Keypad
Overview
Iris

Overview
Iris
The Iris system currently consists of two modules:
1.Crosswalk Module

Detect crosswalks when a user walks near one


Guide its user to the crosswalk location
Guide its user as he or she crosses the street

2.Elevator Module
Detect elevator floor indicator
Detect the number
Inform the user of the floor through speech
Design
Crosswalk Module
The crosswalk detection module works as follows:
1. Capture image/frame from the camera live feed
2. Pre-process frame
3. Contour Detection
4. Convex Hull Detection
5. Postprocessing
6. Crosswalk Location Detection
Frames & Preprocessing
Frames are captured from the cameras' live video feed which
are then processed to remove unneeded data. A thresholding
filter is applied to accomplish this.
Thresholding
Frames & Preprocessing
After thresholding the resulting image will contain only white
elements/objects on a black background. However this image is not
ready for contour detection. The image will be dilated to merge any
broken components that may have resulted from the thresholding
process.
Dilate
Contour Detection
After preprocessing the frame and before being able to determine the
presence of a crosswalk we must analyze each component in the
frame, to do this we apply Contour Detection to the image which
results in a outline map of every object in the frame.
Contour
Detection
Convex Hull & Defects Detection
Once we have a contour map of the image we the procede to
determine the convex hull and defects of every contour to determine if
its a crosswalk lane. If a contour has too many defects its then
discarded.
Convex Hull: The smallest convex region that contains the points in S
Post-Processing & Location
Finally the system will group contours together and create a bounding
box around the crosswalk, the sides of these bounding box will be
used to determine the vanishing point in the frame in relation to the
crosswalk, which in turn allow us to calculate the direction of the
crosswalk relative to the center of the camera.
Design
Elevator Module
The elevator module help user to reach a desire floor
Combines different computer vision and image processing
techniques together with statistical analysis to improve
the accuracy of the readings.
The goal is to provide the best possible readings of the
floor indicator.
The elevator module matches predefined number
templates from 1-9 and arrows going up and down.
The system then analyses the output of thematcherto
improve the results
Design
Elevator Module
Load Templates
Simple Preprocessing
Grayscale
Gaussian Filter
Canny Edge Detection
Contour Detection
Contour Extraction
Approximate Rectangle
Size and Area Filtering
Histogram Analysis+Color Filtering
HSV color space filtering
Template matching
Template Resizing + Matching
Return the most frequent output
Text to speech
Design
Elevator Module
Experiment Results
Elevator Module
Sample Size (1-2): 70 Images
Sample Size (3-5): 5 videos 60-90 seconds. Random samples every 1-3
secs.
Final Words

References
V. Ivanchenko, J. Coughlan and H. Shen. "Detecting and Locating
Crosswalks using a Camera Phone." Fourth IEEE Workshop on
Embedded Computer Vision, in conjunction with Computer Vision
and Pattern Recognition (CVPR 08). Anchorage, Alaska. June 2008.

Luis David Lpez and Olac Fuentes, Color-Based Road Sign
Detection and Tracking, International Conference on Image Analysis
and Recognition (ICIAR), Montreal, CA, August 2007
S. Wang, C. Yi, and Y. Tian, Signage Detection and Recognition for
Blind Persons to Access Unfamiliar Environments, Journal of
Computer Vision and Image Processing, Vol. 2, No. 2, 2012.
References
J. Coughlan and H. Shen. A Fast Algorithm for Finding Crosswalks
using Figure-Ground Segmentation. 2nd Workshop on Applications
of Computer Vision, in conjunction with ECCV 2006. Graz, Austria.
May 2006.

Ellen Klingbeil, Blake Carpenter, Olga Russakovsky, Andrew Y. Ng.
Autonomous Operation of Novel Elevators for Robot
Navigation. ICRA, 2010.

World Health Organization. Visual Impairment and blindness http:
//www.who.int/mediacentre/factsheets/fs282/en/ . October 2011.

Phys.org:http://phys.org/news/2011-07-bionic-glasses-poor-vision.
html. July 6, 2011

You might also like