Enhancing feature tracking with gyro regularization

作者：

Highlights：

• A 3-axis gyro mounted to a video camera can frequently predict the main component of optical flow.

• The gyro-predicted flow can be used to regularize feature tracking to help track ambiguous features through bad imagery.

• Gyro regularization does not require all features to belong to a rigid scene.

• Gyro regularization adds very little computational cost to feature tracking.

• The common practice of using gyros to initialize trackers offers no advantage over careful optical-only initialization.

摘要

•A 3-axis gyro mounted to a video camera can frequently predict the main component of optical flow.•The gyro-predicted flow can be used to regularize feature tracking to help track ambiguous features through bad imagery.•Gyro regularization does not require all features to belong to a rigid scene.•Gyro regularization adds very little computational cost to feature tracking.•The common practice of using gyros to initialize trackers offers no advantage over careful optical-only initialization.

论文关键词：Feature tracking,Optical flow,Inertial sensors,Gyroscopes

论文评审过程：Received 29 August 2014, Revised 3 November 2015, Accepted 30 January 2016, Available online 23 February 2016, Version of Record 30 May 2016.

论文官网地址：https://doi.org/10.1016/j.imavis.2016.01.004