Abstract
Research indicates that deaths due to fall incidents are the second leading cause of unintentional injury deaths in the world. Death by fall due to a person texting or talking on mobile phones while walking, impaired vision, unexpected terrain changes, low balance, weakness, and chronic conditions has increased drastically over the past few decades. Particularly, unexpected terrain changes would many times lead to severe injuries and sometimes death even in healthy individuals. To tackle this problem, a warning system to alert the person of the imminent danger of a fall can be developed. This paper describes a solution for such a warning system used in our bio-inspired wearable pet robot, KiliRo. It is a terrain perception system used to classify the terrain based on visual features obtained from processing the images captured by a camera and notify the wearer of terrain changes while walking. The parrot-inspired KiliRo robot can twist its head and the camera up to 180 degrees to obtain visual feedback for classification. Feature extraction is followed by K-nearest neighbor for terrain classification. Experiments were conducted to establish the efficacy and validity of the proposed approach in classifying terrain changes. The results indicate an accuracy of over 95% across five terrain types, namely pedestrian pathway, road, grass, interior, and staircase.