A CNN Approach To Detect Vanishing Point By Considering Street Lane

Abstract

Linear perspective is widely used in highway road images to create the impression of depth on 2D/3D photos and detect vanishing points with the help of parallel lanes lines on road images. Automated understanding of linear perspective in landscape, road, street photo collections video has a number of real-world applications. This problem of automated understanding of linear perspective in images is addressed with the help of hough transform and CNN framework architecture to detect vanishing points. However, images of the road taken from street and highway video pose a very great technical challenge, because an insufficient number of parallel edges or lines intersect and lead to false vanishing points. To solve this problem, a state of art vanishing point detection method is proposed that exploits the vanishing point and intersection of parallel lines with the help of possible parallel lines to lane lines available in the image, the center of origin, and quadrant. The proposed strategy essentially performs best in detecting vanishing points on a public road image data set.