Method and system for tracking trajectory based on visual localization and odometry

1. A trajectory tracking method, executed in a computer system, for a mobile electronic device wherein:

the computer system comprises at least one processor configured to execute computer-readable instructions included in a memory, and

the trajectory tracking method comprises

estimating a relative pose of the electronic device using odometry, a relative pose value of the relative pose being estimated based on a relative pose value of a prior relative pose of the electronic device;

estimating an absolute pose of the electronic device using visual localization (VL) as camera pose information;

generating a pose graph including a node defined by the relative pose, a node defined by the absolute pose and an edge defining a difference between the relative pose and the absolute pose; and

optimizing the pose graph by correcting the relative pose value of the relative pose, estimated at a given timing in the pose graph, using an absolute pose value of the absolute pose estimated at the given timing, to minimize an error between the relative pose and the absolute pose at the given timing.

2. The trajectory tracking method of claim 1 , wherein tracking a trajectory comprises correcting an accumulated error, occurring when the relative pose is estimated using the odometry, based on the absolute pose value estimated using the VL.

3. The trajectory tracking method of claim 1 , wherein optimizing of the pose graph comprises:

defining a pose-graph error function by summing errors of all edges included in the pose graph; and

estimating the relative pose value and the absolute pose value of each of the nodes defined by the relative pose and absolute pose using a Gauss-Newton method based on the pose-graph error function.

4. The trajectory tracking method of claim 1 , further comprising estimating, by the at least one processor, the absolute pose by additionally using a visual object tracking (VOT) technology depending on whether the VL operates or based on the result of the absolute pose estimation using the VL.

5. A non-transitory computer-readable recording medium storing a program which, when executed by at least one processor, causes the at least one processor to execute the trajectory tracking method of claim 1 .

6. A computer system for tracking a trajectory of a mobile electronic device, comprising:

at least one processor configured to execute computer-readable instructions included in a memory,

wherein the at least one processor includes,

a relative pose estimation unit for estimating a relative pose of the electronic device using odometry;

an absolute pose estimation unit for estimating an absolute pose of the electronic device using visual localization (VL) as camera pose information; and

a trajectory optimization unit for generating a pose graph including a node defined by the relative pose, a node defined by the absolute pose and an edge defining a difference between the relative pose and the absolute pose, and optimizing the pose graph by correcting a relative pose value of the relative pose, estimated at a given timing in the pose graph, using an absolute pose value of the absolute pose estimated at the given timing based on a relative pose value of a prior relative pose of the electronic device, to minimize an error between the relative pose and the absolute pose at the given timing.

7. The computer system of claim 6 , wherein the at least one processor corrects an accumulated error, occurring when the relative pose is estimated using the odometry, based on the absolute pose value estimated using the VL.

8. The computer system of claim 6 , wherein the optimizing of the pose graph comprises,

defining a pose-graph error function by summing errors of all the edges included in the pose graph and estimating a pose value and the absolute pose value of each of the nodes defined by the relative pose and absolute pose using a Gauss-Newton method based on the pose-graph error function.

9. The computer system of claim 6 , wherein the at least one processor further estimates the absolute pose by additionally using a visual object tracking (VOT) technology depending on whether the VL operates or based on the result of the absolute pose estimation using the VL.