Apparatus and method for tracking object based on radar image reconstruction
View Patent ↗Disclosed is an apparatus for tracking an object based on radar image reconstruction, the apparatus including a two-dimensional image producing unit configured to produce a two-dimensional image by collecting multiple one-dimensional radar signals from which clutter is removed, an object detecting unit configured to determine the presence or absence of an object from the two-dimensional image and estimate a two-dimensional position of the present object, and an object tracking unit configured to track a movement path of the object based on the two-dimensional position estimated by the object detecting unit.
1. An apparatus for tracking an object based on radar image reconstruction, the apparatus comprising:
a two-dimensional image producing unit configured to produce a two-dimensional image by collecting multiple one-dimensional radar signals using sensors of a plurality of radars from each of which clutter is removed;
an object detecting unit coupled to the two-dimensional image producing unit and configured to determine whether an object is present or absent from the two-dimensional image and estimate, in response to determining that the object is present, a two-dimensional position of the object; and
an object tracking unit coupled to the object detecting unit and configured to track a movement path of the object based on the two-dimensional position estimated by the object detecting unit,
wherein the two-dimensional image producing unit calculates the two-dimensional image that minimizes an objective function expressed using a vector made by stacking the multiple one-dimensional radar signals from which the clutter is removed, and a system matrix that reflects characteristics of the sensors of the plurality of radars collecting the multiple one-dimensional radar signals.
2. The apparatus of claim 1 ,
wherein the objective function comprises a regularization term that reflects prior information with respect to the two-dimensional image to be calculated.
3. The apparatus of claim 2 ,
wherein the regularization term depends on at least one of:
prior information representing that the two-dimensional image to be restored comprises low-frequency elements and similarity between adjacent pixels has a predetermined reference value; and
prior information representing that similarity between two-dimensional images restored from two adjacent radar frames has a predetermined reference value due to a sampling speed of the two adjacent radar frames.
4. The apparatus of claim 1 ,
wherein the object detecting unit is configured to estimate, in response to determining that the object is present, the two-dimensional position of the object using a time of arrival (TOA) method.
5. The apparatus of claim 1 ,
wherein the objective function is expressed as follows:
J ( f k )=∥ g k −Hf k ∥,
where:
k represents an index of a frame acquired by a radar system;
f k represents a column stacked vector of the two-dimensional image to be obtained;
g k represents a column stacked vector of an observed value of a kth frame acquired by the radar system; and
H represents the system matrix that reflects characteristics of sensors of a plurality of radars.
6. A method of tracking an object based on radar image reconstruction, the method comprising:
producing a two-dimensional image by collecting multiple one-dimensional radar signals from each of which clutter is removed;
determining whether an object is present or absent from the two-dimensional image and estimating, in response to determining that the object is present, a two-dimensional position of the object; and
tracking a movement path of the object based on the two-dimensional position estimated by the object detecting unit,
wherein the producing of the two-dimensional image comprises calculating the two-dimensional image that minimizes an objective function expressed using a vector made by stacking the multiple one-dimensional radar signals from each of which the clutter is removed, and a system matrix that reflects characteristics of sensors of a plurality of radars.
7. The method of claim 6 ,
wherein the objective function comprises a regularization term that reflects prior information with respect to the two-dimensional image to be calculated.
8. The method of claim 7 ,
wherein the regularization term depends on at least one of:
prior information representing that the two-dimensional image to be restored comprises low-frequency elements and similarity between adjacent pixels has a predetermined reference value; and
prior information representing that similarity between two-dimensional images restored from two adjacent radar frames has a predetermined reference value due to a sampling speed of the two adjacent radar frames.