Granted Patent B2

US 12,651,364 · App. 18/129,336 · Granted Jun 9, 2026

Inventor: Tadas Orentas (Jupiter, FL)

Assignee: Jenoptik Smart Mobility Solutions, LLC

G06T7/62G06T2207/30196G06T2207/30232

Abstract

Claims (89)

1 . A method for estimating at least a length of a vessel, the method comprising:

monitoring of a scene by a camera, wherein the camera provides a first plurality of two-dimensional images of the scene;

processing at least a portion of the two-dimensional images of the first plurality of two-dimensional images provided by the camera using an object detection algorithm in order to detect at least one reference object in at least one two-dimensional image of the first plurality of two-dimensional images wherein each reference object is assumed to have a same reference width;

using at least one two-dimensional image of the first plurality of two-dimensional images in which at least one reference object was detected as a training image and determining, for each training image, a pixel position and at least a pixel width of each reference object in the respective training image;

creating a width scaling model based on the determined pixel positions of the reference objects and the determined pixel widths of the reference objects as well as the reference width of the reference objects, wherein the width scaling model provides a functional relation between a pixel position and a pixel width of an object in a two-dimensional image of the scene on the one hand and a width of the object on the other hand;

monitoring of the scene by the camera, wherein the camera provides a second plurality of two-dimensional images of the scene;

processing at least a portion of the two-dimensional images of the second plurality of two-dimensional images provided by the camera using an object detection algorithm in order to detect at least one vessel in at least one of the two-dimensional images of the second plurality of two-dimensional images;

determining a pixel position and at least a pixel width of the vessel in the respective two-dimensional image; and

deriving a vessel length estimation for the vessel from the width scaling model based on the pixel position and pixel width of the vessel in the respective two-dimensional image,

wherein creating the width scaling model comprises:

creating a three-dimensional width data cloud based on at least the determined pixel positions and pixel widths of the reference objects, wherein the data cloud comprises a plurality of three-dimensional data points, wherein each of the data points represents a pixel position determined for a specific reference object in a specific training image and at least a pixel width or a width scaling value determined for the specific reference object based on the specific training image;

smoothing the three-dimensional width data cloud in at least one dimension; and

wherein the three-dimensional width data cloud is generated from reference objects detected at different pixel positions across a plurality of training images captured at different vessel-relative distances, such that the width scaling model compensates for perspective-induced non-linear scaling across the image field.

2 . The method of claim 1 ,

wherein at least the width scaling model is created based on at least 500 reference objects detected in at least one training image.

3 . A method for estimating at least a length of a vessel, the method comprising:

monitoring of a scene by a camera, wherein the camera provides a first plurality of two-dimensional images of the scene;

processing at least a portion of the two-dimensional images of the first plurality of two-dimensional images provided by the camera using an object detection algorithm in order to detect at least one reference object in at least one two-dimensional image of the first plurality of two-dimensional images wherein each reference object is assumed to have a same reference width;

using at least one two-dimensional image of the first plurality of two-dimensional images in which at least one reference object was detected as a training image and determining, for each training image, a pixel position and at least a pixel width of each reference object in the respective training image;

creating a width scaling model based on the determined pixel positions of the reference objects and the determined pixel widths of the reference objects as well as the reference width of the reference objects, wherein the width scaling model provides a functional relation between a pixel position and a pixel width of an object in a two-dimensional image of the scene on the one hand and a width of the object on the other hand;

monitoring of the scene by the camera, wherein the camera provides a second plurality of two-dimensional images of the scene;

processing at least a portion of the two-dimensional images of the second plurality of two-dimensional images provided by the camera using an object detection algorithm in order to detect at least one vessel in at least one of the two-dimensional images of the second plurality of two-dimensional images;

determining a pixel position and at least a pixel width of the vessel in the respective two-dimensional image; and

deriving a vessel length estimation for the vessel from the width scaling model based on the pixel position and pixel width of the vessel in the respective two dimensional image,

wherein creating the width scaling model comprises:

creating a three-dimensional width data cloud based on at least the determined pixel positions and pixel widths of the reference objects, wherein the data cloud comprises a plurality of three-dimensional data points, wherein each of the data points represents a pixel position determined for a specific reference object in a specific training image and at least a pixel width or a width scaling value determined for the specific reference object based on the specific training image;

smoothing the three-dimensional width data cloud in at least one dimension, and

wherein creating the width scaling model comprises:

fitting a width scaling surface to the smoothened three-dimensional width data cloud, wherein the width scaling surface provides a functional relation between a pixel position on the one hand and a pixel width of a reference object or a width scaling value on the other hand.

4 . A method for estimating at least a length of a vessel, the method comprising:

monitoring of a scene by a camera, wherein the camera provides a first plurality of two-dimensional images of the scene;

processing at least a portion of the two-dimensional images of the first plurality of two-dimensional images provided by the camera using an object detection algorithm in order to detect at least one reference object in at least one two-dimensional image of the first plurality of two-dimensional images wherein each reference object is assumed to have a same reference width;

using at least one two-dimensional image of the first plurality of two-dimensional images in which at least one reference object was detected as a training image and determining, for each training image, a pixel position and at least a pixel width of each reference object in the respective training image;

creating a width scaling model based on the determined pixel positions of the reference objects and the determined pixel widths of the reference objects as well as the reference width of the reference objects, wherein the width scaling model provides a functional relation between a pixel position and a pixel width of an object in a two-dimensional image of the scene on the one hand and a width of the object on the other hand;

monitoring of the scene by the camera, wherein the camera provides a second plurality of two-dimensional images of the scene;

processing at least a portion of the two-dimensional images of the second plurality of two-dimensional images provided by the camera using an object detection algorithm in order to detect at least one vessel in at least one of the two-dimensional images of the second plurality of two-dimensional images;

determining a pixel position and at least a pixel width of the vessel in the respective two-dimensional image; and

deriving a vessel length estimation for the vessel from the width scaling model based on the pixel position and pixel width of the vessel in the respective two-dimensional image,

wherein creating the width scaling model comprises:

creating a three-dimensional width data cloud based on at least the determined pixel positions and pixel widths of the reference objects, wherein the data cloud comprises a plurality of three-dimensional data points, wherein each of the data points represents a pixel position determined for a specific reference object in a specific training image and at least a pixel width or a width scaling value determined for the specific reference object based on the specific training image; and

smoothing the three-dimensional width data cloud in at least one dimension,

wherein deriving the vessel length estimation from the width scaling model comprises:

deriving a width scaling value from the width scaling model based on the pixel position of the vessel in the respective two-dimensional image and calculating the vessel length estimation based on the scaling value and the pixel width of the vessel,

wherein the three-dimensional width data cloud is generated from reference objects detected at different pixel positions across a plurality of training images captured at different vessel-relative distances, such that the width scaling model compensates for perspective-induced non-linear scaling across the image field.

5 . The method of claim 1 ,

wherein processing at least a portion of the first plurality of two-dimensional images comprises that each reference objects is assumed to have a same reference height; and

wherein the method further comprises:

determining, for each training image, at least a pixel height of each reference object in the respective two-dimensional training image;

creating a height scaling model based on the determined pixel positions of the reference objects and the determined pixel heights of the reference objects as well as the reference height of the reference objects, wherein the height scaling model provides a functional relation between a pixel position and a pixel height of an object in a two-dimensional image of the scene on the one hand and a height of the object on the other hand;

determining a pixel height of the vessel in the respective two-dimensional image;

deriving a vessel height estimation for the vessel from the height scaling model based on the pixel position and pixel height of the vessel in the respective two-dimensional image.

6 . The method of claim 1 ,

wherein detecting at least one reference object comprises detecting humans as reference objects.

7 . The method of claim 6 ,

wherein each human is assumed to have a width of 2 ft.

8 . The method of claim 1 ,

wherein monitoring the scene comprises monitoring at least a portion of a passageway for vessels.

9 . The method of claim 1 ,

wherein at least the pixel width of a reference object is determined as the pixel width of a bounding rectangle of the reference object, wherein the bounding rectangle represents the maximum extents of the reference objects in the respective two-dimensional training image or the pixel width of a vessel is determined as the pixel width of a bounding rectangle of the vessel, wherein the bounding rectangle represents the maximum extents of the vessel in the respective two-dimensional image.

10 . The method of claim 1 ,

wherein at least the pixel height of a reference object is determined as the pixel height of a bounding rectangle of the reference object, wherein the bounding rectangle represents the maximum extents of the reference objects in the respective two-dimensional training image or the pixel height of a vessel is determined as the pixel height of a bounding rectangle of the vessel, wherein the bounding rectangle represents the maximum extents of the vessel in the respective two-dimensional image.

11 . A system for estimating at least a length of a vessel, the system comprising:

a camera for monitoring a scene and obtaining two-dimensional images from said scene; and

an image processing device operatively coupled to the camera, the image processing device comprising a processor and a memory operatively coupled to the processor, the memory storing instructions to cause the processor to perform operations comprising:

processing at least a portion of a first plurality of two-dimensional images provided by the camera using an object detection algorithm in order to detect at least one reference object in at least one two-dimensional image of the first plurality of two-dimensional images wherein each reference object is assumed to have a same reference width;

using at least one two-dimensional image of the first plurality of two-dimensional images in which at least one reference object was detected as a training image and determining, for each training image, a pixel position and at least a pixel width of each reference object in the respective training image;

creating a width scaling model based on the determined pixel positions of the reference objects and the determined pixel widths of the reference objects as well as the reference width of the reference objects, wherein the width scaling model provides a functional relation between a pixel position and a pixel width of an object in a two-dimensional image of the scene on the one hand and a width of the object on the other hand;

processing at least a portion of a second plurality of two-dimensional images provided by the camera using an object detection algorithm in order to detect at least one vessel in at least one of the two-dimensional images of the second plurality of two-dimensional images;

determining a pixel position and at least a pixel width of the vessel in the respective two-dimensional image; and

deriving a vessel length estimation for the vessel from the width scaling model based on the pixel position and pixel width of the vessel in the respective two-dimensional image,

wherein creating the width scaling model comprises:

creating a three-dimensional width data cloud based on at least the determined pixel positions and pixel widths of the reference objects, wherein the data cloud comprises a plurality of three-dimensional data points, wherein each of the data points represents a pixel position determined for a specific reference object in a specific training image and at least a pixel width or a width scaling value determined for the specific reference object based on the specific training image;

smoothing the three-dimensional width data cloud in at least one dimension,

wherein the image processing device is configured to generate the three-dimensional width data cloud from reference objects detected at different pixel positions across a plurality of training images captured at different vessel-relative distances, such that the width scaling model compensates for perspective-induced non-linear scaling across the image field.

12 . The system of claim 11 ,

wherein the camera is a camera providing a resolution of at least 3840 pixels in width and at least 2160 pixels in height.

13 . A non-transitory computer readable storage medium with instructions stored thereon that, when executed by a processor, cause the processor to perform operations comprising:

processing at least a portion of a first plurality of two-dimensional images provided by the camera using an object detection algorithm in order to detect at least one reference object in at least one two-dimensional image of the first plurality of two-dimensional images wherein each reference object is assumed to have a same reference width;

using at least one two-dimensional image of the first plurality of two-dimensional images in which at least one reference object was detected as a training image and determining, for each training image, a pixel position and at least a pixel width of each reference object in the respective training image;

creating a width scaling model based on the determined pixel positions of the reference objects and the determined pixel widths of the reference objects as well as the reference width of the reference objects, wherein the width scaling model provides a functional relation between a pixel position and a pixel width of an object in a two-dimensional image of the scene on the one hand and a width of the object on the other hand;

processing at least a portion of a second plurality of two-dimensional images provided by the camera using an object detection algorithm in order to detect at least one vessel in at least one of the two-dimensional images of the second plurality of two-dimensional images;

determining a pixel position and at least a pixel width of the vessel in the respective two-dimensional image; and

deriving a vessel length estimation for the vessel from the width scaling model based on the pixel position and pixel width of the vessel in the respective two-dimensional image,

wherein creating the width scaling model comprises:

creating a three-dimensional width data cloud based on at least the determined pixel positions and pixel widths of the reference objects, wherein the data cloud comprises a plurality of three-dimensional data points, wherein each of the data points represents a pixel position determined for a specific reference object in a specific training image and at least a pixel width or a width scaling value determined for the specific reference object based on the specific training image;

smoothing the three-dimensional width data cloud in at least one dimension, and

wherein the instructions cause the processor to generate the three-dimensional width data cloud from reference objects detected at different pixel positions across a plurality of training images captured at different vessel-relative distances, such that the width scaling model compensates for perspective-induced non-linear scaling across the image field.

14 . The method of claim 1 , wherein deriving a vessel length estimation from the width scaling model comprises:

deriving a width scaling value from the width scaling model based on the pixel position of the vessel in the respective two-dimensional image and calculating the vessel length estimation based on the scaling value and the pixel width of the vessel.

Assignments (2)

Continuity (1)

References Cited (10)