| Title |
Road Crossing Guidance Technique for the Visually Impaired through Crosswalk Detection using Multiple Region of Interest |
| Authors |
손인수(In Su Son) ; 강동욱(Dong Wook Kang) ; 정경훈(Kyeong Hoon Jung) |
| DOI |
https://doi.org/10.5573/ieie.2022.59.6.57 |
| Keywords |
Assistive technology; Visually impaired; Road crossing; Crosswalk detection; Multiple ROI |
| Abstract |
In this paper, we propose a method for the visually impaired to safely cross the road through crosswalk detection and its slope estimation techniques. By adopting multiple ROI(Region of Interest), it is possible to reduce the effect of obstacles such as vehicles or other pedestrians and to determine the relative position and direction of the crosswalk. A binary image is obtained by two-step binarization for each ROI, and statistical and morphological features are extracted to determine whether it is a part of crosswalk or not. The projection data of the binary image, which is used to calculate the number of pattern change, is also used to estimate the slope indicating the direction of the crosswalk. A frame-by-frame result is determined by combining the results of multiple ROI and it is finally stabilized through temporal filtering. The performance of the proposed method was evaluated for still images. The recall, precision, and F1 score of crosswalk detection were 0.935, 0.990 and 0.962, respectively. And the slope of the crosswalk was estimated within a small error. In addition, an experiment was carried out on video sequence assuming a real road crossing situation. It showed that the frame-by-frame result was successfully stabilized through the temporal filtering. |