Three-dimensional reconstruction method and device for multi-stage unsupervised learning, and electronic equipment

Abstract

The invention provides a three-dimensional reconstruction method and device for multi-stage unsupervised learning and electronic equipment, and the method comprises the steps: for any viewpoint in a multi-viewpoint image of a to-be-reconstructed object, when a depth map synthesized by a scale feature map corresponding to a multi-viewpoint image in a previous stage does not meet a preset condition, updating the depth map by adopting a scale feature map corresponding to the multi-view image at the current stage, wherein the resolution ratio of the scale feature map corresponding to the next stage is higher than that of the scale feature map corresponding to the previous stage; and fusing the depth maps of all viewpoints to determine the three-dimensional point cloud of the object to be reconstructed. According to the method provided by the invention, the problem that a supervised deep learning mode excessively depends on data is avoided, the generalization of the model is enhanced, wide application is facilitated, and the generated high-precision depth map can ensure the accuracy and integrity of three-dimensional point cloud fusion, and the point cloud is dense.

Description

technical field [0001] The present invention relates to the technical field of three-dimensional reconstruction, in particular to a multi-stage unsupervised learning three-dimensional reconstruction method, device and electronic equipment. Background technique [0002] Traditional two-dimensional image acquisition and display technologies are difficult to meet the growing viewing needs of viewers. With the innovation of display technology and the improvement of computing power, methods for accurately and efficiently reconstructing 3D scene information have received a lot of attention. The mainstream 3D scene reconstruction method mainly includes two processes: solving the corresponding multi-view depth map based on the multi-view color image, and then aggregating the solved multi-view depth map to obtain a 3D point cloud model. With the rapid development of deep learning, 3D reconstruction methods based on deep learning have surpassed traditional methods on multiple evaluat...

AI Technical Summary

Problems solved by technology

[0005] The invention provides a three-dimensional reconstruction method, device and electronic equipment of multi-stage unsupervised learning, which are used to solve the problem that the supervised deep learning method in the existing three-dimensional point cloud fusion method is too dependent on data, and the depth map obtained by the fusion result can be distinguished Due to the defect that the depth results cannot be displayed and applied well due to the low efficiency, through the multi-stage unsupervised learning mode, in the process of using different scale feature maps at different stages to form a depth map, timely detect whether the fusion depth map results of each stage meet the requirements. Accuracy requirements, if not met, continue to use the higher-precision feature map of the next stage for fusion until the accuracy meets the requirements

Images