Recently textured pavement blocks are utilized for visually impaired humans walking. However, it is still difficult for them to detect crosswalk. In this article, an image processing method is proposed to detect crosswalk and to notify the position and direction with voice guidance on a camera phone. First, lines are detected with Hough transform. Then through piecing short lines together and noise lines elimination, edges of crosswalk white lines are selected. Calculating the mean angle of the edges, the mean line which has the angle and crosses the median point of terminal points of the white lines is determined. Connected line of each median point of the terminal points in each side of the mean line shows crosswalk direction. To determine the start point, the near side median point is extended to the same distance between the overall median point. Finally, voice guidance informs the direction and start point.%視覚障碍者が独力で街頭を移動する際，点字ブロックなどを活用できるが横断歩道の存在位置を特定し渡る事は困難である．本研究では，携帯電話のカメラで撮像された画像から横断歩道の存在位置を特定し案内するための画像処理技術について検討する．ハフ変換により直線検出を行ない，抽出短線分の位置・傾き情報から線分統合・ノイズ除去により横断歩道白線線分を検出する．さらにその端点の重心位置を通り，平均的な傾きを持つ直線で白線線分を分け，それぞれの領域の線分端点の重心位置を結ぶ線分を横断歩道方向，手前側の重心位置と全重心位置との距離だけ手前に延ばした点を開始点とし，時計文字盤方向で音声ガイダンスを行なう．
展开▼
