Information processing device, information processing method, and information processing system
An information processing device includes a memory and a processor configured to obtain first point group information that represents three-dimensional positional information, by using image data captured by a first camera; output second point group information obtained by reducing noise of the first point group information, by using one or more filters; and set filter coefficients of the one or more filters, by using (i) a filter coefficient estimation model trained in advance by using (a) test data for learning that includes image data captured by a second camera and (b) training data based on third point group information that represents three-dimensional positional information obtained by a position sensor, and (ii) observation data that includes the image data captured by the first camera.
1 . An information processing device comprising:
a memory; and
a processor configured to:
extract feature values from a plurality of frames included in image data captured by a first camera, identify corresponding points between the plurality of frames and execute a visual simultaneous localization and mapping (Visual SLAM) process using the corresponding points to obtain first point group information that represents three-dimensional positional information;
output second point group information obtained by reducing noise of the first point group information, by using one or more filters; and
set filter coefficients of the one or more filters, by using (i) a filter coefficient estimation model trained in advance by using (a) test data for learning that includes image data captured by a second camera and (b) training data based on third point group information that represents three-dimensional positional information obtained by a position sensor, and (ii) observation data that includes the image data captured by the first camera, a feature count extracted in a feature value extraction process of the Visual SLAM, and a matching count of the corresponding points.
2 . The information processing device as claimed in claim 1 , wherein the filter coefficient estimation model is a neural network that is trained to learn filter coefficients of the one or more filters with which a difference between the second point group information and the training data is less than or equal to a threshold value.
3 . The information processing device as claimed in claim 1 , wherein the test data for learning and the observation data are obtained by using a vehicle, and include vehicle information on the vehicle.
4 . The information processing device as claimed in claim 3 , wherein the vehicle information includes vehicle speed information, gear information, or information on a parking mode selected by a user, on the vehicle.
5 . The information processing device as claimed in claim 1 , wherein the test data for learning and the training data are obtained by using a vehicle, and the position sensor includes a LIDAR installed on the vehicle.
6 . An information processing system, comprising:
a computer including a memory and a processor configured to:
extract feature values from a plurality of frames included in image data captured by a first camera, identify corresponding points between the plurality of frames and execute a visual simultaneous localization and mapping (Visual SLAM) process using the corresponding points to obtain first point group information that represents three-dimensional positional information, by using image data captured by a camera;
output second point group information obtained by reducing noise of the first point group information, by using one or more filters;
obtain third point group information that represents the three-dimensional positional information obtained by a position sensor; and
train a filter coefficient estimation model by using test data for learning that includes the image data and training data based on the third point group information, so as to output filter coefficients of the one or more filters with which a difference between the second point group information and the training data is less than or equal to a threshold value.
7 . An information processing method executed by a computer including a memory and a processor, the information processing method comprising:
extracting feature values from a plurality of frames included in image data captured by a first camera, identify corresponding points between the plurality of frames and execute a visual simultaneous localization and mapping (Visual SLAM) process using the corresponding points to obtain first point group information that represents three-dimensional positional information, by using image data captured by a first camera;
outputting second point group information obtained by reducing noise of the first point group information, by using one or more filters; and
setting filter coefficients of the one or more filters, by using (i) a filter coefficient estimation model trained in advance by using (a) test data for learning that includes image data captured by a second camera and (b) training data based on third point group information that represents three-dimensional positional information obtained by a position sensor, and (ii) observation data that includes the image data captured by the first camera, a feature count extracted in a feature value extraction process of the Visual SLAM, and a matching count of the corresponding points.