Ensuring the Invariance of Object Images to Linear Movements for Their Recognition

Abstract

In automated production, automated systems for recognizing two-dimensional images of these products are often used to control and diagnose the quality of manufactured products. However, in this case, difficulties arise with the reliability of recognition of these images due to the orthogonal displacement of object images, rotation of the image around the center of gravity, and rescaling of the image of the object. These destabilizing factors lead to significant methodological errors in estimating the proximity measure between the checked and standard objects. Therefore, the process of object image recognition must be invariant to changes in the position of the image. In the existing methods of pattern recognition, either invariant topological and geometric features are chosen for these changes (these features are less reliable), or non-invariant geometric features are selected with the subsequent provision of invariance to linear changes in images (they become more informative). Therefore, the problem is to ensure the invariance of the geometric features of images to linear changes. Existing methods cannot fully solve this problem. Therefore, an algorithm is proposed to solve this issue. The issue is solved by the fact that when an image with a random position appears, the angle of rotation of the image around the axis of inertia of the image is found. Then, the image is virtually returned to its original standard position. Thus, the problem of invariant recognition of two-dimensional images is solved. In the proposed algorithm, the contour points of two-dimensional images of objects are given by coordinates in the Cartesian coordinate plane. Therefore, for the correct recognition of such images, it is necessary to ensure that the image points are invariant to displacement and rotation. For the image to be invariant to orthogonal displacement, the coordinate system must be moved to the center of gravity of the image being defined. Then, by the moments of inertia about the coordinate axes of the image, the angle of its rotation relative to the initial position of the reference object is determined. After evaluating the angle of rotation of the image, the coordinates of the contour points are found by rotating the reference image in the computer memory by this angle. Then the coordinates of the contour points of the current image are compared with the coordinates of the contour points of the rotated reference image. Thus, the proposed algorithm makes it possible to invariantly recognize two-dimensional images of objects. The proposed algorithm was simulated on a computer and positive results were obtained.