OCR signage recognition with skew & slant correction for visually impaired people

It is a challenge for visually impaired people (VIPs) to navigate independently whenever they attempt to find their way in unfamiliar buildings searching for amenities (i.e. exits, ladies/gents toilets) even with a walking stick or a guide dog. Camera-based computer vision systems have the potential to assist VIPs in independent navigation or way finding in unfamiliar places. To leverage on previous research of Signage Recognition Framework which could only recognize public signage with slanted angle less than30°, an improved OCR signage recognition model with skew and slant correction in public signage is presented. The proposed OCR method consists of Canny edge detection algorithm, Hough Transformation and Shearing Transformation were used to detect and correct skewed and slanted images. The proposed model would capture a public signage image, compare the image in the database using template matching algorithm and convert to machine readable text in a text file. The text will then be processed by Microsoft Speech Application Program Interface (SAPI) speech synthesizer and translated to voice as output. Experiments were conducted on 5 blind folded subjects to test the performance of the model. The proposed OCR recognition model has achieved satisfactory recognition rate of 82.7%.