Identifying parking location using single camera reverse projection
A system and method are disclosed, in which two or more points of a target object from a two-dimensional (2D) image may be “reverse projected” from a 2D plane to a 3D space. The method may comprise receiving a 2D image from a 2D camera, the 2D image including a representation of a physical topography and a target object of a plurality of target objects, the plurality of target objects being associated with a common physical characteristic, the common physical characteristic of the target object being detectable within the 2D image. The method may further comprise determining a position relative to the physical topography of the target object based upon a position of the target object in the image and the common physical characteristic of the target object.
1. A computer-implemented method, comprising:
receiving a two-dimensional image from a two-dimensional camera, the two-dimensional image including a representation of a physical topography and a target object of a plurality of target objects, the plurality of target objects being associated with a common physical characteristic, the common physical characteristic of the target object being detectable within the two-dimensional image; and
determining, utilizing reverse projection, a three-dimensional position relative to the physical topography of the target object based upon a two-dimensional position of the target object in the image and the common physical characteristic of the target object, the common physical characteristic used to determine at least a dimension of the three-dimensional position, the physical topography represented in a three-dimensional space, the three-dimensional position determined relative to the three-dimensional coordinate space, wherein the three-dimensional coordinate space forms a plurality of cuboids, and wherein determining the three-dimensional position in the three-dimensional coordinate space comprises mapping the target object to one or more cuboids of the plurality of cuboids.
2. A computer-implemented system, comprising:
one or more processors; and
one or more computer-readable storage media storing instructions that, upon execution with the one or more processors, cause the computer-implemented system to, at least:
receive an image from a camera, the image corresponding to a representation of a physical topography, the physical topography being mapped to a three-dimensional coordinate space;
determine, based at least in part on a target object recognition model, that a target object is present in the image, a target object type of the target object being a predetermined physical object type, wherein each object of the predetermined physical object type corresponds to at least two points, the at least two points each being detectable within the image and being mappable onto the three-dimensional coordinate space, and wherein each object of the predetermined physical object type has one or more physical characteristics, each physical characteristic corresponding to an aspect of mapping a point of the at least two points onto the three-dimensional coordinate space;
determine a position of the target object in the three-dimensional coordinate space based at least in part on the determination that the target object type of the target object matches the predetermined physical object type and the one or more physical characteristics of the predetermined physical object type; and
output a position of the target object relative to the physical topography, based at least in part on the position of the target object in the three-dimensional coordinate space.
3. The computer-implemented method of claim 1 , wherein determining the three-dimensional position further comprises:
based upon the one or more cuboids of the plurality of cuboids and one or more discrete location identifiers, associating the target object position with a discrete location of a plurality of discrete locations, wherein each of the plurality of discrete locations respectively corresponds to the one or more discrete location identifiers.
4. The computer-implemented method of claim 3 wherein the plurality of discrete locations are vehicle parking spaces of a vehicle parking lot.
5. The computer-implemented method of claim 1 wherein the common physical characteristic comprises two points of the plurality of target objects being spaced substantially the same.
6. The computer-implemented method of claim 5 wherein determining the three-dimensional position comprises setting a scaling factor corresponding to a first point of the two points of the plurality of target objects to be substantially the same as a scaling factor corresponding to a second point of the two points of the plurality of target objects.
7. The computer-implemented method of claim 6 wherein the target object is a vehicle license plate, and the first and second points are corners of the vehicle license plate.
8. The computer-implemented method of claim 1 ,
wherein determining a three-dimensional position in three-dimensional space of the target object comprises determining the three-dimensional position based at least in part on parameters of the camera, the parameters of the camera comprising at least one or more intrinsic parameters, one or more extrinsic parameters, and one or more distortion coefficients.
9. A computer-implemented system, comprising:
one or more processors; and
one or more computer-readable storage media storing instructions that, upon execution with the one or more processors, cause the computer-implemented system to, at least:
receive an image from a camera, the image corresponding to a representation of a physical topography, the physical topography being mapped to a three-dimensional coordinate space;
determine, based at least in part on a target object recognition model, that a target object is present in the image, a target object type of the target object being a predetermined physical object type, wherein each object of the predetermined physical object type corresponds to at least two points, the at least two points each being detectable within the image and being mappable onto the three-dimensional coordinate space, and wherein each object of the predetermined physical object type has one or more physical characteristics, each physical characteristic corresponding to an aspect of mapping a point of the at least two points onto the three-dimensional coordinate space;
determine a position of the target object in the three-dimensional coordinate space based at least in part on the determination that the target object type of the at least one target object matches the predetermined physical object type and the one or more physical characteristics of the predetermined physical object type; and
output a position of the target object relative to the physical topography, based at least in part on the position of the target object in the three-dimensional coordinate space.
10. The computer-implemented system of claim 9 wherein the predetermined physical object type is a vehicle license plate type.
11. The computer-implemented system of claim 10 wherein the one or more physical characteristics of the object of the vehicle license plate type comprises the object being a rectangular shape, wherein a length of the object is substantially double a width of the object.
12. The computer-implemented system of claim 11 wherein outputting information comprises converting the rectangular shape to a square shape, based in part on the determination that the target object type is the vehicle license plate type.