Identification of Lines Based on Form Factor and Geometrical Descriptors

Abstract

The problems of releasing and identifying features of key elements and their identification in images are shown. The basic algorithms and methods for forming descriptors for key elements in the image are considered. The main disadvantages of existing methods in aerospace images are the loss of stability and the presence of a large number of false key elements when changing recording to the conditions (for example, brightness and contrast of the received images). To eliminate these drawbacks, it is proposed to use selected lines as key elements, and their geometrical characteristics to form a descriptor. To analyze the proposed algorithm and the SIFT method, fragments of aerospace images changed in brightness and contrast in a graphic editor from -50 to +50 percent of the original were used, which made it possible to evaluate the operation of these algorithms in conditions close to the real survey. It is shown that the proposed algorithm is more stable than the SIFT method with increasing contrast and decreasing the brightness of the test aerospace image. Also, the proposed algorithm is characterized by a smaller number of detected false key elements compared to the SIFT method.