A Method for Recognizing and Locating Geometric Marks Using Generalized Features

Abstract

The invention belongs to the technical field of photogrammetry and computer vision, and discloses a method for identifying and locating geometric marks by using generalized features. A three-dimensional rendering engine is used to generate a training data set for corresponding types of geometric objects, and a deep learning network is used to generate training data sets. The graphic is trained to realize the recognition of the geometric logo in the image; the identified graphic object is processed with common features such as contour extraction, and then the dimensionality reduction ICP algorithm is used to realize the registration in the discretized state of the logo and obtain its transformation parameters , so as to obtain the precise positioning of the geometric marker. The present invention realizes the quantitative description of the deformation of the geometric identification primitive in the image, and achieves a relatively satisfactory positioning accuracy; it can realize that for the commonly used geometric identification, there is no need to customize a specific algorithm, but directly through the above process, the realization of the Accurate positioning of markings, thereby simplifying the extraction and positioning process of manual markings, and forming a general method.

Description

Technical field [0001] The present invention belongs to the field of photographic measurement and computer visual technology, and more particularly to a method of identifying and positioning geometric identity using generalized features. Background technique [0002] At present, the existing technology in the industry is like this: [0003] In photographic measurement and computer visual field, geometric identification identification and positioning are commonly used in calibration, control point extraction, coding identification, and other algorithms. In terms of geometric identity of a certain type of shape, it is necessary to design a targeted algorithm, and the program is identified, or the geometric identity is positioned directly by manually. [0004] In summary, the problems existing in the prior art are: [0005] (1) Currently, the method of deep learning is not used for the identification of geometric identity, and is still a specific identification algorithm for specifi...

AI Technical Summary

Problems solved by technology

[0005] (1) At present, the method of deep learning has not been used for the recognition of geometric marks, and it is still to design specific recognition algorithms for specific geometric marks

[0006] (2) In the positioning process of the geometric mark, its common features have not been effectively utilized, and it is still necessary to rely on the unique features of the specific geometric mark

[0007] (3) The identification and positioning process of the entire geometric mark still needs manual intervention, and the efficiency needs to be improved

[0011] The main difficulty in solving the problem in the prior art lies in the inability to use the generalization method to accurately locate the geometric mark

Images