Direct method unsupervised monocular image scene depth estimation method

Abstract

The invention discloses a direct method unsupervised monocular image scene depth estimation method, which belongs to the field of computer vision and depth estimation, and comprises the following steps of: constructing a neural network; calculating an image re-projection error; and calculating an image mask and updating the pose of the camera. According to the method, the defects that monocular image depth estimation is high in environment requirement, prone to being interfered by a low texture area, poor in camera pose estimation precision and the like are overcome. A traditional monocular depth estimation problem is combined with a visual odometer, so that the depth estimation precision is remarkably improved, and the method has an auxiliary effect on positioning and navigation of a moving vehicle. The method has the advantages of being high in precision, high in flexibility, wide in application range and the like, can be used for surrounding environment perception, collision prevention and positioning navigation of equipment such as automatic driving vehicles and mobile robots, and has various application scenes.

Description

technical field [0001] The invention belongs to the field of computer vision, neural network, and depth estimation, and more specifically, relates to a direct-method unsupervised monocular image scene depth estimation method. Background technique [0002] At present, depth estimation has been developed rapidly driven by related technologies such as neural networks and sensors, and has been widely used in the fields of intelligent robots, pedestrian recognition, face unlocking, VR applications, and automatic driving. The primary task of depth estimation is to estimate the distance from the front object to the camera based on a single color image collected by the camera. [0003] There are two main ways to obtain the 3D information corresponding to the scene from the real scene: one is to use a sensor that can perceive the 3D depth information of the scene to collect the depth information in the scene, and the other is to recover the 3D information from the 2D image correspond...

AI Technical Summary

Problems solved by technology

The traditional depth estimation method is mainly based on the geometric constraints of feature point matching and environmental assumptions. The binocular method and the multi-eye depth estimation method require precise camera extrinsic parameters, and errors caused by extrinsic parameter changes cannot be eliminated.

Images