Key point thermodynamic diagram guided self-supervised monocular visual odometer method

Abstract

The invention provides a self-supervision monocular vision odometer method guided by a key point thermodynamic diagram, and belongs to the field of computer vision. The method comprises the following steps: constructing a pose estimation network and a depth estimation network; inputting the video image sequence into a pose estimation network and a depth estimation network; extracting key points of each frame of image in the video image sequence and generating a key point thermodynamic diagram; multiplying the poses of each pair of adjacent frame images output by the pose estimation network to obtain the pose of a relatively long time period, and calculating a luminosity error loss function of the video image sequence pose consistency constraint based on the depth image output by the depth estimation network and the generated key point thermodynamic diagram; training the pose estimation network and the depth estimation network based on the obtained luminosity error loss function; and estimating the camera pose corresponding to each frame of image in the video image sequence with the pose to be estimated by using the trained pose estimation network. According to the invention, the precision of camera pose estimation can be improved.

Description

technical field [0001] The invention relates to the field of computer vision, in particular to a self-supervised monocular visual odometry method guided by a key point heat map. Background technique [0002] Visual odometry refers to a method of estimating the current position and attitude of a camera based on input video image frames. It can be widely used in robot navigation, autonomous driving, augmented reality, wearable computing and other fields. According to the type and number of sensors used, visual odometry can be divided into monocular visual odometry, binocular visual odometer and visual odometer fused with inertial information. Among them, the monocular visual odometry has the advantages of only needing one camera, less hardware requirements, and no need for correction. [0003] The traditional visual odometry method first extracts and matches image features, and then estimates the relative pose between two adjacent frames according to the geometric relationshi...

AI Technical Summary

Problems solved by technology

[0005] Existing self-supervised methods use too many redundant pixels during the training process, making the deep neural network unfocused during the learning process, resulting in cumulative errors in the estimated pose of the network

In addition, these methods only consider the pose consistency between adjacent frames, and do not consider the pose consistency of the video image sequence

Images