A Feature Point Tracking Method Based on Event Camera

Abstract

The invention discloses a feature point tracking method based on an event camera, aiming at improving the feature point tracking accuracy. The technical solution is to build an event camera-based feature point tracking system composed of a data acquisition module, an initialization module, an event set selection module, a matching module, a feature point update module, and a template edge update module. The initialization module extracts feature points and edge graphs from the image frame; the event set selection module selects the event set S of feature points from the event stream around the feature points; the matching module matches S with the template edges around the feature points to find t k Optical flow set G of n feature points at time k , the feature point update module according to G k calculate t k+1 The position set FD of n feature points at time k+1 , the template edge update module uses the IMU data to PBD k The position in is updated to get t k+1 The position set PBD of the template edge corresponding to n feature points at time k+1 . Adopting the present invention can not only improve the precision of tracking feature points on the event stream, but also prolong the average tracking time of feature points.

Description

technical field [0001] The invention relates to the field of computer image processing, in particular to using an event camera to complete the tracking of feature points in an image. Background technique [0002] As an important branch in the field of robotics, SLAM (full name Simultaneous Localization and Mapping) has been widely studied in recent years. SLAM attempts to solve the problem of how a robot moves in an unknown environment, how to determine its own trajectory by observing the environment, and construct a map of the environment at the same time. SLAM technology is the sum of many technologies involved in achieving this goal. A complete SLAM system mainly includes a front-end visual odometry part and a back-end optimization part. The visual odometry part estimates the robot state, and the estimation methods are mainly divided into two methods: feature point method and direct method. The feature point method is currently the mainstream method for robot state est...

AI Technical Summary

Problems solved by technology

Using this method, the position of the feature point is updated every time an event is received, which increases the amount of calculation, and this method will introduce more errors, resulting in poor tracking accuracy of feature points, and long-term tracking cannot be achieved.

[0006] Therefore, the existing methods of using event cameras for feature point tracking still have the disadvantages of low tracking accuracy and short tracking time.

Images