Ground texture image-based navigation method and device, and storage medium

Patent metadata

Field	Value
Publication number	US-11644338-B2
Application number	US-202016920440-A
Country	US
Kind code	B2
Filing date	Jul 3, 2020
Priority date	Oct 19, 2018
Publication date	May 9, 2023
Grant date	May 9, 2023

Abstract

A navigation method based on ground texture images, an electronic device and storage medium. The method includes: performing transform domain based image registration on an acquired image of a current frame and an image of a previous frame, and determining a first pose of the image of the current frame; determining whether the image of the current frame meets a preset condition, and if so, inserting the image of the current frame as the key-frame image into a map, and performing loop closure detection and determining a loop key-frame image; performing transform domain based image registration on the image of the current frame and the loop key-frame image, and determining a second pose of the image of the current frame; and determining an accumulated error according to the first pose and the second pose, and correcting the map according to the accumulated error.

First claim

We claim:

1. A navigation method based on images of ground texture, performed by a robot provided with a collector, which is configured to collect the images of ground texture, the method comprising:
performing transform domain based image registration on an acquired image of a current frame and an image of a previous frame, and determining a first pose of the image of the current frame;
determining whether the image of the current frame meets a preset condition for creating a key-frame image, and in response to determining that the image of the current frame meets the preset condition, inserting the image of the current frame as the key-frame image into a map, and performing loop closure detection and determining a loop key-frame image;
performing transform domain based image registration on the image of the current frame and the loop key-frame image, and determining a second pose of the image of the current frame; and
determining an accumulated error according to the first pose and the second pose of the image of the current frame, and correcting the map according to the accumulated error, so as to perform navigation according to the corrected map;
wherein performing transform domain based image registration on the acquired image of the current frame and the image of the previous frame, and determining the first pose of the image of the current frame, comprises:
step 1: performing Fourier transformation on the image f 1 of the previous frame and the image f 2 of the current frame to obtain F 1 and F 2 ;
step 2: eliminating low-frequency noise from F 1 and F 2 by using a high-pass filter;
step 3: converting the filtered images from a rectangular coordinate system to a log-polar coordinate form;
step 4: performing Fourier transformation on the images F 1 and F 2 in log-polar coordinates;
step 5: determining a cross-power spectrum R 1 of the image f 1 of the previous frame and the image f 2 of the current frame in the log-polar coordinates based on a transformation result of step 4 and a calculation formula of the cross-power spectrum;
step 6: performing inverse Fourier transformation on the obtained power spectrum R 1 to obtain an inverse Fourier transformation result IR 1 ;
step 7: determining coordinates corresponding to a maximum peak value of the inverse Fourier transformation result IR 1 , and obtaining a scaling factor and a rotation factor according to the coordinates corresponding to the maximum peak value of the inverse Fourier transformation result IR 1 ;
step 8: performing inverse transformation on the image f 2 of the current frame according to the obtained scaling factor and rotation factor to obtain a new image f 3 ;
step 9: performing fast Fourier transformation on the images f 1 and f 3 to obtain F 1 and F 3 , calculating a power spectrum R 2 of F 1 and F 3 based on a power spectrum calculation formula, and performing inverse Fourier transformation on R 2 to obtain an inverse Fourier transformation result IR 2 ; and
step 10: determining coordinates corresponding to a maximum peak value of the inverse Fourier transformation result IR 2 as a translation parameter, and determining the first pose of the image of the current frame according to the translation parameter;

if the image of the previous frame is not an image of ground texture, selecting a preset number of key-frame images, matching feature points of each of the preset number of the key-frame images with feature points of the image of the current frame, and determining at least one candidate matched frame image according to a matching result;
wherein the image of the previous frame is not a ground texture image based upon determining whether the camera is blocked or has sudden movement;
calculating at least one third pose of the image of the current frame respectively according to the at least one candidate matched frame image; and
based on the at least one third pose of the image of the current frame, determining whether a distance difference between any of the at least one candidate matched frame image and the image of the current frame in a global coordinate system is less than a fifth threshold, and in response to determining that the distance difference between any of the at least one candidate matched frame image and the image of the current frame in the global coordinate system is less than the fifth threshold, calculating a pose of a new image of a next frame according to the image of the current frame.

2. The method according to claim 1 , wherein the preset condition for creating a key-frame image comprises:
map building is in an idle state, and a quantity of image frames between the image of the current frame and a previous key-frame image is greater than a preset first threshold; and
a distance difference between the image of the current frame and the previous key-frame image in a global coordinate system is greater than a preset second threshold.

3. The method according to claim 1 , wherein said that inserting the image of the current frame as the key-frame image into the map, and performing loop closure detection and determining the loop key-frame image comprises:
inserting the image of the current frame as the key-frame image into the map, and calculating similarities between the image of the current frame and all other key-frame images in the map respectively, and adding key-frame images with similarities greater than a third threshold to a candidate set;
selecting, from the candidate set, at least three key-frame images that meet a loop condition, sorting the at least three key-frames in an order of from high to low similarities between the at least three key-frame images and the image of the current frame, and using a key-frame image ranking first as a candidate key-frame image; and
determining whether a distance difference between the candidate key-frame image and the image of the current frame in a global coordinate system is less than a fourth threshold, and in response to determining that the distance difference between the candidate key-frame image and the image of the current frame in the global coordinate system is less than the fourth threshold, using the candidate key-frame image as the loop key-frame image.

4. The method according to claim 1 , wherein the first pose, the second pose or the third pose of the image of the current frame comprises a rotation and a translation of a movement of the image of the current frame.

5. The method according to claim 1 , further comprising: determining initial coordinate values in a global coordinate system at an initial position by identifying collected two-dimensional code information.

6. An electronic device, comprising:
at least one processor; and
a memory configured to store at least one program,
wherein the at least one program, when executed by the at least one processor, causes the at least one processor to implement operations of:
performing transform domain based image registration on an acquired image of a current frame and an image of a previous frame, and determining a first pose of the image of the current frame;
determining whether the image of the current frame meets a preset condition for creating a key-frame image, and in response to determining that the image of the current frame meets the preset condition, inserting the image of the current frame as the key-frame image into a map, and performing loop closure detection and determining a loop key-frame image;
performing transform domain based image registration on the image of the current frame and the loop key-frame image, and determining a second pose of the image of the current frame; and
determining an accumulated error according to the first pose and the second pose of the image of the current frame, and correcting the map according to the accu

Assignees

• Beijing Geekplus Tech Co Ltd

Inventors

• Wang Yingchun
• Ji Bin

Classifications

• G06T2207/20056

  Discrete and fast Fourier transform, [DFT, FFT] · CPC title

• G01C21/3837

  Data obtained from a single source · CPC title

• G06T7/37 Primary

  using transform domain methods · CPC title

• G06T7/74

  involving reference images or patches · CPC title

• G06T1/0014

  Image feed-back for automatic industrial control, e.g. robot with camera (robots B25J19/023) · CPC title