It can be difficult to guide yourself across a crosswalk when your visual capabilities are limited, which can be an everyday issue for someone with impaired vision. This paper aims to alleviate that issue for zebra stripe crosswalks by proposing an algorithm that incorporates multiple properties of zebra stripe crosswalks with a neural network to assist in quickly and accurately identifying a crosswalk in video and pictures taken from a smartphone camera.
This method improves the accuracy of zebra crosswalk detection in images. In a large dataset, it correctly identified 76.5% of zebra crosswalks, while reducing the false discovery rate (q-value) from 20% without using neural networks to 2.21% using this neural network method. Only 2.04% of non-crosswalk images as crosswalks using the neural network method.
| Identifer | oai:union.ndltd.org:CALPOLY/oai:digitalcommons.calpoly.edu:theses-2726 |
| Date | 01 June 2016 |
| Creators | Banich, Jason David |
| Publisher | DigitalCommons@CalPoly |
| Source Sets | California Polytechnic State University |
| Detected Language | English |
| Type | text |
| Format | application/pdf |
| Source | Master's Theses |
Page generated in 0.0017 seconds