Monocular vision scene depth estimation method based on deep learning

Abstract

The invention relates to a monocular vision scene depth estimation method based on deep learning, and the method comprises the steps of employing a VGG-13 network model, employing a depth separable convolution layer to replace a standard convolution layer so as to reduce the model parameter quantity, and obtaining a network model which can be used for obtaining a parallax image; inputting the monocular image into the trained network model to generate the disparity maps of multiple scales, and generating a single disparity map consistent with the input image in scale by combining multi-scale fusion and disparity map smoothing; and generating a corresponding depth image according to the geometric transformation relationship between the disparity map and the depth map in the multi-view geometry. The method has the beneficial effects that the simple and easily available binocular visible light image is used for training the network model without acquiring the high-cost real depth data, andthe standard convolution is replaced with the depth separable convolution, so that the parameter quantity of the network model can be reduced to one seventh of the previous parameter quantity, and the reasoning speed of the model is increased.

Description

technical field [0001] The invention belongs to the field of three-dimensional reconstruction of computer vision, and relates to a method for estimating the depth of a monocular vision scene based on deep learning. Background technique [0002] Three-dimensional structural information is indispensable information for human beings to observe and understand the environment, understand and analyze the scene. Correctly judging and recognizing the three-dimensional structure of the scene allows the computer to perform tasks such as target positioning and path planning more reasonably and accurately. Therefore, how to reconstruct the three-dimensional structure of the scene from two-dimensional images or video sequences is a major research topic in the field of computer vision. Key points and difficult points. 3D scene reconstruction mainly depends on obtaining the depth map corresponding to the image. At present, there are mainly two ways to obtain depth information. One is to...

AI Technical Summary

Problems solved by technology

[0004] However, the method of scene depth estimation based on deep learning for monocular images has the following disadvantages: First, the general method of depth estimation for monocular images based on deep learning regards depth restoration as a classification task of image pixels, which requires real Depth data is used to train the network, and this kind of data usually needs to be obtained by lidar, which is not only expensive to obtain but also sparse in data, and poor in practicability; second, the previous method has a single network structure and a relatively simple model, which ultimately leads to low accuracy of the results. Low, and the transferability of the model is poor; third, in order to improve the accuracy of the model, the previous method usually uses a deeper network structure to extract the feature information of the image, resulting in a huge amount of parameters and slow inference speed in practical applications. Poor real-time performance

Images