Technique for Detection of Straight Lines on the Images of the Stationary Land Scenes

Abstract

Topic of this article is the problem of processing of the images of the stationary land scenes for the purpose of detection of the straight lines. Straight lines are considered as one of the most typical and robust components of the geometrical content. Straight lines detection on an image can be used in machine vision systems for solving of a number of information processing problems and control of the autonomous robotic mobile objects, such as: elongated landmarks detection in navigation problems; merging of several images of an observed scene with different resolution characteristics. The detected straight lines can be used directly as typical characteristic features in the recognition algorithms and object selection algorithms. At the same time, the detected straight lines could be used to form a different geometrical content, such as points of intersection and convergence of the straight lines, etc. For solution of the task of the straight lines detection an algorithm was proposed based on the use of a directed filtration and modified Hough transform. The directed filtration allows us to get a set of the contour points, while preserving information about direction of the brightness gradient. The offered modification of the Hough transform uses the aprioristic information about the prevailing directions of the brightness gradients, and also a condition that one contour point belongs only to one of the detected straight lines. The reviewed examples show, that even in case of a poor quality of the initial image (low contrasts and high noise level) the offered algorithm allows us to select information about the extended objects in the form of the contour points set, belonging to the straight lines, which determine the scene geometry.