19th Summer School on Image Processing Szeged - Hungary Project #5 and St ps Dávid Almási University of Debrecen Severino Gomes-Neto Federal University.

19th Summer School on Image ProcessingSzeged - Hungary

Project #5and St ps

DávidAlmásiUniversity

of Debrecen

SeverinoGomes-NetoFederal University ofRio Grande do Norte

MelindaKatonaUniversityof Szeged

Tamás Sámuel

TasiUniversityof Szeged

Project

Recognition of Doors and Steps

• Description– Appeal: Aid permanently or temporarily

blind people to know in advance the existence of doors and steps meanwhile they walk

– Input: Images from mobile phones (or webcams)

– Goal: recognize doors and steps– Tools: Matlab R2010a + Image Processing

Toolbox

Project Database Pipeline Accuracy Evaluation

Database

Actual PicturesProject Database Pipeline Accuracy Evaluation

StairsNon stairs

Actual PicturesProject Database Pipeline Accuracy Evaluation

DoorsNon doorsNon stairs

Pipeline

Recognizing StepsProject Database Pipeline Accuracy Evaluation

Input Image

Gray Scaled

Crop

Histogram Equalization

Sharpening

Fourier Transform

Output Steps!

Edge Detection

Recognizing DoorsProject Database Pipeline Accuracy Evaluation

Input Image

Gray Scaled

Crop

Histogram Equalization

Sharpening

Fourier Transform

Output Door!

Edge Detection

Accuracy Evaluation

Positives and Negatives

Positives

Negatives

StepsTrue 16 (89%) 8 (44%)*

False 2 (11%)10

(56%)*

DoorsTrue 10 (83%) 2 (50%)False 2 (17%) 2 (50%)

Project Database Pipeline Accuracy Evaluation

* Computing “out of the range” stairs as false positives. Actually, discarding those, the results are:8 true negatives (73%) and 3 false negatives (27%)