Non-contact automatic mapping method based on deep learning

Abstract

The invention discloses a non-contact automatic mapping method based on deep learning. The method comprises the following steps: calibrating a camera by using a color test card through a Zhang Zhengyou calibration method to obtain camera internal parameters; collecting a plurality of original images of the object and recording focal length parameters; correcting the color of the original image byusing the color test card picture to obtain an image A; and establishing and training a camera external parameter estimation network; inputting the image A, the camera internal parameters and the object three-dimensional model into a trained camera external parameter estimation network to obtain camera external parameters; inputting the external and internal parameters of the camera, the three-dimensional model of the object and the image A into a rendering pipeline for processing to obtain a rendering result, and performing rendering alignment; and enabling the rendering pipeline to store thecolor value of the image of the area determined to be used in the image A in a texture mapping file according to the UV coordinates. According to the invention, the automatic mapping process is realized, a large number of manual steps are reduced, parameters transmitted by a user can be received for fine adjustment, and the high-quality mapping speed is accelerated.

Description

technical field [0001] The present invention mainly relates to a mapping method and process for a three-dimensional model in image processing and computer graphics, as well as a method for estimating internal and external parameters of a camera, and relates to a deep learning method in the field of artificial intelligence. Background technique [0002] With the rapid development of computer technology, the digital needs of all walks of life are also increasing day by day. For some industries, the digitization process is to digitize real-world objects into computers using computer graphics technology and store and display them, and the mapping process is an important part of digitization. The process of mapping means that after certain processing of the existing original texture image, the color of the texture image is mapped to the coordinates of the 3D model, so that the rendering of the 3D model is closer to the appearance of the real world. [0003] The existing texture ...

AI Technical Summary

Problems solved by technology

The most time-consuming step in the process of manual mapping is to accurately align the required parts of the captured image with the 3D model. The problem in this process is that the professional software and the camera parameters used to capture the image are not the same, so The image is not distorted to the same degree as the viewing angle in the professional software, resulting in misalignment

Therefore, the disadvantage of manual mapping is that it takes a long time and the quality is low

[0004] The method of automatic mapping can better solve the above problems. The important step in the process of automatic mapping is to obtain the internal and external parameters of the camera, but the existing automatic mapping method is mainly to manually calibrate the feature points of the model and the screen coordinates in the corresponding texture map. Then the external parameters of the camera are obtained through the coordinate pair. When manually calibrating the model, many methods add sharp objects to the model as feature points in the scanning process for the convenience of calibration. However, this method is not suitable for some industries such as the cultural relics protection industry. Applicable because these industries require non-contact operation during the scanning process, so adding sharp objects is not feasible

And it is not suitable for industries that require the quality of texture maps, because a model may have dozens of texture maps, and manually selecting coordinate pairs for each picture is still a very time-consuming task.

Images