External recognition device
An external recognition device that detects and tracks a moving object includes an object position detection unit that detects a position of the moving object as an observation value, based on an image; a region determination unit that determines a region to which the moving object belongs in the image, based on the observation value; an observation error setting unit that calculates an error to an observation value, based on a determination result; a state prediction unit that predicts a state of the moving object at a current time, based on the observation value up to a previous time that is a time earlier than the current time and the error; an association unit that associates the state of the moving object at the current time with the observation value; and a state update unit that updates the state of the moving object, based on a result of the association.
1 . An external recognition device configured to detect and track a moving object around a mobile body equipped with a camera, the external recognition device comprising:
a processor coupled to a memory, wherein the processor implements:
an object position detection unit that detects a position of the moving object based on an image captured by the camera and represents the position of the moving object as an observation value;
a region determination unit that determines a region in the image to which the moving object belongs based on the observation value detected by the object position detection unit, the image being separated into a plurality of regions in a vertical direction and the region determination unit determines the region in the image using the image separated into the plurality of regions;
an observation error setting unit that calculates an error associated with the observation value based on the region in the image to which the moving object belongs, a different error being calculated for respective regions of the plurality of regions;
a state prediction unit that predicts a state of the moving object at a current time, based on the observation value up to a previous time that is a time earlier than the current time and the error calculated by the observation error setting unit;
an association unit that associates the state of the moving object at the current time predicted by the state prediction unit with the observation value detected by the object position detection unit; and
a state update unit that updates the state of the moving object, based on a result of the association by the association unit,
wherein when the observation value in the image detected by the object position detection unit is associated with a region in which the moving object cannot be correctly detected among the plurality of regions, the moving object is detected by an optical flow or a median flow, and
wherein the observation value in the image detected by the object position detection unit is associated with a region having a high possibility of not including the entirety of the moving object among the plurality of regions, and at a time of the detection of the moving object by the alternative, detection is performed inside a region of the image calculated in accordance with a predictive value of the state prediction unit.
2 . The external recognition device according to claim 1 , wherein the region determination unit is configured to determine, by use of the image divided in a plurality of regions, which of the plurality of regions the observation value in the image detected by the object position detection unit belongs to, based on physical parameters that are a mounting position and a mounting orientation of the camera with respect to the mobile body, and lens distortion.
3 . The external recognition device according to claim 1 , wherein the region determination unit separates the plurality of regions into a region in which a relatively great error occurs due to a deviation of the observation value and a region in which a relatively small error occurs due to the deviation of the observation value in the vertical direction in the image.
4 . The external recognition device according to claim 1 , wherein the observation error setting unit is configured to
add an error of the observation value in the image detected by the object position detection unit to the observation value in the image in each of upward, downward, rightward, and leftward directions of the image,
project the observation value in the image to which the error is added on a mobile body coordinate system by use of a pinhole camera model, and
thereafter, set, as an observation error, an average of absolute values of differences in each direction relative to a position where the observation value in the image is projected on a mobile body coordinate system.
5 . The external recognition device according to claim 1 , wherein the association unit is configured to
associate, by use of a Mahalanobis distance, the state of the moving object at the current time predicted by the state prediction unit with the observation value detected by the object position detection unit, and
verify, by a chi-squared test, a result of the association by the association unit.
6 . The external recognition device according to claim 1 , wherein the state prediction unit is configured to predict the observation value by use of movement information of the mobile body, not on coordinates of the image but on a coordinate system in which a spot at system startup is an origin, and
the state update unit is configured to update the observation value on the coordinate system predicted by the state prediction unit, and tracks a target moving object.
7 . The external recognition device according to claim 1 , wherein in a state estimator including the state prediction unit and the state update unit, when a target of tracking has an internal state including a velocity, the state update unit is configured to calculate a movement velocity of the target of tracking, based on an observation value associated with each state estimator.
8 . The external recognition device according to claim 1 , wherein when the observation value is not obtained by a state estimator including the state prediction unit and the state update unit, an internal state and an error covariance matrix are updated by the state update unit, by use of a prediction result of the state prediction unit.
9 . The external recognition device according to claim 3 , wherein when the observation value in the image detected by the object position detection unit is a region above the region in which the relatively great error occurs due to the deviation of the observation value, and the observation value in the image detected by the object position detection unit is associated with a region higher than a vanishing point of the image among the plurality of regions, the position of the moving object is deleted.
10 . A method, performed by an external recognition device, for detecting and tracking a moving object around a mobile body equipped with a camera, the method comprising:
detecting a position of the moving object based on an image captured by the camera and representing the position of the moving object as an observation value;
determining a region in the image to which the moving object belongs based on the observation value detected, the image being separated into a plurality of regions in a vertical direction;
calculating an error associated with the observation value based on the region in the image to which the moving object belongs, a different error being calculated for respective regions of the plurality of regions;
predicting a state of the moving object at a current time, based on the observation value up to a previous time earlier than the current time and the error calculated;
associating the state of the moving object at the current time predicted with the observation value detected; and
updating the state of the moving object based on a result of the associating,
wherein when the observation value in the image is associated with a region in which the moving object cannot be correctly detected among the plurality of regions, the moving object is detected by an optical flow or a median flow, and
wherein the observation value in the image is associated with a region having a high possibility of not including the entirety of the moving object among the plurality of regions, and at a time of the detection of the moving object by the alternative, detection is performed inside a region of the image calculated in accordance with a predictive value of the state prediction unit.