<![CDATA[Electric Powered Wheelchairs or EPW has been used for many disable patients and still developing. EPW was developing to achieve flexibility to control its movement. EPW can be moved by eyes and head nowadays. One of the main accidents that often occurred to EPW's user is falling from EPW due to some obstacles that blocking the road or some descents such as stair descent which user did not see before. Hence, EPW needs a system to increase the user's safety. In previous research ultrasonic sensors were used to detect objects. But it needed a lot of sensors to detect obstacles on wide range and did not able to detect descents. Regarding that, researchers began to use camera to detect obstacles. This research use image processing methods to detect stair descent and generate warning sound through a speaker. HOG was used as a method to extract features from data and SVM algorithm as machine learning classifier. Pre-processing such as cropping, resizing, and blurring were used. Total features for each data were 3.780 features which generated from an image with 128x64 pixel size. This system had 80% accuracy of recognizing object and had 0,679672 second average computation.]]>
Copyrights © 2021
