SVM based column-level approach for crosswalk detection in low-resolution images

Development of automated assistance systems for the visually impaired is still relatively uncommon research topic. However, the use of the cameras and image processing is widely spread and can be a good source of useful information for the visually impaired user. In this work, we are considering usage of SVM (Support Vector Machine) for detecting the crosswalk region in low-resolution images. The proposed approach uses SVM to make decisions on column level thereby providing the horizontal position of the crosswalk in image. Such information can potentially help the visually impaired person to cross the road safely. Preliminary testing was conducted and satisfying results were obtained especially in terms of performance.