Dynamic environment offline visual odometer expanding method

Abstract

The invention discloses a dynamic environment offline visual odometer expanding method. The method comprises the following specific steps: firstly, extracting point features from an image sequence tocalculate an initial pose matrix, and obtaining a predicted current frame image according to the initial pose matrix; and carrying out optical flow and residual value calculation through two adjacentframes of images and the frame of image before prediction, and eliminating the dynamic feature points; carrying out static linear expansion according to the collinear relationship of the remaining static feature points, and constructing a collinear matching matrix; and performing BA optimization through the remaining static feature points and static linear features, and adjusting the pose of the camera. Experimental results in a dynamic environment show that compared with other methods, the method has the advantages that tracking errors can be effectively reduced, the precision and robustnessof the visual odometer are improved, the static straight line recognition and extraction time and the static straight line matching time are shortened, and the requirements of real-time application are met.

Description

technical field [0001] The invention belongs to the field of autonomous navigation of mobile robots, in particular to a method for offline expansion of a visual odometer in a dynamic environment. Background technique [0002] Visual odometry uses single or multiple visual sensors to estimate the pose of the carrier by relying on the matching relationship between images at adjacent moments in the image sequence. Vision sensors have become an important research direction and research hotspot in the field of robotics due to the advantages of rich environmental information obtained by laser sensors, easy access, and low cost. [0003] However, in order to simplify the positioning and mapping problems, most of the current visual odometry methods assume that the environment is static. When the proportion of dynamic objects in the image is small, it can be treated as noise. When the proportion of dynamic objects in the image is large, this method Not applicable, limits the use of ...

AI Technical Summary

Problems solved by technology

However, since the straight line feature is more advanced than the feature point structure, and its descriptor is more complex, the amount of calculation when matching a straight line is much larger than that of point matching, which is a challenge to the real-time performance of the system.

At the same time, in a dynamic environment, it is inevitable to identify and eliminate dynamic straight line features, making it more difficult for the system to meet real-time requirements

Images