| dc.description.abstract | There are plenty of blind people who blind since
birth or blinded by accident. Usually, they face a lot of
problems while walking because most of the surfaces are
uneven surfaces such as road bumps, and potholes. The goal
of this study is to find a vision-based method for helping
blind people to detect uneven surfaces. To identify uneven
surfaces, color differences, and shadows of the uneven
surfaces, techniques such as contour length, contour area,
and nonzero pixel ratio of an image are being used. The
image is initially captured and cropped to match the viewing
angle of a human. An HSV filter is applied to the cropped
image along with thresholding techniques for classifying the
image components such as road, grass, or concrete. Further
the HSV thresholds aid in obtaining more detailed
information from the image. Subsequently, the image is
divided into eight parts, and the nonzero pixel ratio, contour
area, and contour lengths are computed for each part. The
resulting data is stored in separate arrays, and maximum
values are determined. If the maximum values from two
arrays share the same indexes, it suggests the presence of
potholes. To test the effectiveness of the proposed method,
the test images of various surfaces were captured and test
results. From the test results, we found that the proposed
algorithm can identify things like potholes, and notify the
user in advance. The findings of this study contribute to
improving the mobility of visually impaired individuals by
assisting them in navigating uneven surfaces more
effectively | en_US |
