Computer Vision for those that need it

Posted by admin at November 24, 2015

Those who don’t wear prescription lens will never understand the discomfort of having to put on a device to enable you see well enough like others. Way down the sight spectrum are the terminally blind and then of course, the blind. These individuals will have to rely on methods and machines to be able to process their environment. How about a bit of technology to alleviate this problem, Computer Vision. Computer Vision is the tech used in processing images and relaying measurable reports, which includes techniques such as face/object detection, . This sub-discipline has been used extensively in robotics, but why not an extensive application in helping the blind. I am sure they are, so I decided to look into some research done in this field.

1. Research by specialists in the University of Lincoln, UK financed by the Google Faculty Research Award. This project will be combining computer vision and machine learning to develop applications that will help people with visual impediments. (http://www.lincoln.ac.uk/news/2015/06/1110.asp)

2. Arianna, a brilliant White Cane App from the University of Palermo in Italy: this app uses the smartphone camera to scan the floors for a colored tape and then vibrates when it find one. This colored tape runs across the premises guiding the user to locations of interest. (http://www.technologyreview.com/view/523401/app-turns-smartphone-into-virtual-cane-for-the-blind/)

3. VAVE, Roman Kozaks’ entry into the Google Science fair (https://www.youtube.com/watch?v=2e3LuM3pP_4)

Here is my own possible research project topic on the subject – Visually Impaired Friendly Environment VIFE:

This will be a piece of technology connected to a web service, that will process images or live video stream from a wearable camera, detect faces and objects in the users view, guiding the user via audio notes and speech. This gear can also include AR to detect familiar environments such as the users home, most visited spot (such as his/her favorite cafeteria or restaurant) or maybe popular location published by other users.

Possible implementations of this tech could be in the form of the following:
1. Camera enabled Mobile device or specialized Handheld scanner: This can be held out by the user to scan the vicinity for navigation data
2. Glasses or Head gear: worn by the user or the guiding animal to capture and relay images to a web server and also relay notifications from the web server to the user
3. Nav Wheels: Here imagine a voice command, obstruction-detection tech built into a two wheeler. Enough said.
4. Blind-Friendly Zones BFZ: This will involve the participation of office/complex owners installing specialized notification systems that will inform, warn and guide the visually impaired via secluded audio channels which are accessible to their tech.
5. VIFE Social Media: A social media add-on could help build a support network around the visually impaired person, helping with the generation of data on navigable places, tips and even auditory assistance.
6. Also, how about a specialized tech for the seeing-eye dog? Other than cameras and microphone for audio/visual data capture, the guiding animal can also wear a tracking device that will allow the user locate it without assistance or communicate with it even they both loose physical contact with each other.

Specifications of this tech will include the following:
1. Facial detection and Recognition: The user will know keep track of people in his/vicinity and know when familiar people are around
2. Crowd detection: It will be useful if the tech can detect areas of multiple movements as against areas will less movements and be able to calculate the probability of safeness. This can help the user decide where to go.
3. Position and Obstruction detection: The user will be able to detect peculiarities in the vicinity such as walls, stairs, guard rails, fountains, pools etc and be able to navigate them
4. Optical Character Recognition OCR for processing readable documents: This tech should be able to “read” documents and highlight key text of concern for the user. A simple text-to-speech tech can be integrated. Also, I dont know how blind people are able to tell a fifty from a five. This tech should be able to process notes and coins and detect the value of any piece of legal tender.

Finally, these specifications above can be broken down into sub-projects and built separately.

0 likes