Detecting moving vehicles

Patent metadata

Field	Value
Publication number	US-9336446-B2
Application number	US-201314418473-A
Country	US
Kind code	B2
Filing date	Jul 25, 2013
Priority date	Jul 31, 2012
Publication date	May 10, 2016
Grant date	May 10, 2016

Abstract

A method of detecting at least one moving vehicle includes receiving ( 202 ) image data representing a sequence of image frames over time. The method further includes analyzing ( 204 - 206 ) the image data to identify potential moving vehicles, and comparing ( 208 - 212 ) at least one said potential moving vehicle with a vehicle movement model that defines a trajectory of a potential moving vehicle to determine whether the at least one potential moving vehicle conforms with the model.

First claim

The invention claimed is:

1. An autonomous method of detecting substantially in real-time at least one moving vehicle, the method comprising:
receiving image data, from an image capturing device, representing a sequence of image frames over time;
autonomously analysing, in a computer, substantially in real-time, the image data to identify potential moving vehicles; and
comparing, in said computer, at least one said potential moving vehicle with a vehicle movement model that defines a trajectory of a potential moving vehicle to determine whether the at least one potential moving vehicle conforms with the model; wherein
said vehicle movement model comprises analysing directional data relating to a said moving vehicle determined to conform with the vehicle movement model to determine if the moving vehicle is moving towards, away or across with respect to a location where the image data was captured.

2. A method according to claim 1 , wherein the directional data analysing includes analysing monocular visual depth cue relating to a said moving vehicle.

3. A method according to claim 2 , wherein the directional data analysing includes:
maintaining a count of pixel density of each said moving vehicle over a sequence of said image frames, and
computing a rate of change of pixel density count over time,
wherein if the rate is constant then the moving vehicle is determined to be moving across with respect to the location;
if the rate is positive then the moving vehicle is determined to be moving towards the location, or
if the rate is negative then the moving vehicle is determined to be moving away from the location.

4. A method according to claim 1 , wherein each said image frame in the sequence is defined by a set of image elements and the step of analysing the image data to identify the potential moving vehicles can include:
generating time-rate of intensity change estimations for corresponding pixels of the image frames;
extracting minima and maxima said time-rate of intensity change from the estimations;
identifying the minima and maxima as said potential moving vehicles.

5. A method according to claim 4 , wherein the step of generating time-rate of intensity change estimations comprises estimating first order temporal derivatives for the corresponding image elements.

6. A method according to claim 4 , wherein the step of comparing at least one said potential moving vehicle with the vehicle movement model includes defining a local region window within a said image frame based around a said pixel corresponding to a said potential moving vehicle.

7. A method according to claim 6 , wherein the local region window is used to support a best fitted model using linear least squares estimation.

8. A method according to claim 7 , wherein a size of the local region window is configurable by a user and sets an upper bound on proximity of multiple detections in the image frame in order for the pixels to be associated with a single said vehicle.

9. A method according to claim 6 , including generating the vehicle movement model that defines the trajectory of a said potential moving vehicle and generating a corresponding model fitting error for the local region window of the potential moving vehicle, wherein the trajectory provides an estimation of velocity and direction of the potential moving vehicle over a finite period of time, and the model fitting error gives an indication of confidence in accuracy of the detection.

10. A method according to claim 9 , where, for a said vehicle movement model, the method further predicts a future said trajectory of the moving vehicle.

11. A method according to claim 10 , wherein a constant velocity model is used for the future trajectory prediction.

12. A method according to claim 11 , wherein a 2D Gaussian is defined to weight any maxima or minima.

13. A non-transitory computer-readable storage medium storing a computer program element comprising:
computer code means to make the computer execute a method according to claim 1 .

14. Apparatus configured to autonomously detect at least one moving vehicle, substantially in real-time, the apparatus comprising:
a device configured to receive image data representing a sequence of image frames over time;
a device configured to autonomously analyse, substantially in real-time, the image data to identify potential moving vehicles, and
a device configured to compare at least one said potential moving vehicle with a vehicle movement model that defines a trajectory of a potential moving vehicle to determine whether the at least one potential moving vehicle conforms with the model; wherein
said vehicle movement model comprises analysing directional data relating to a said moving vehicle determined to conform with the vehicle movement model to determine if the moving vehicle is moving towards, away or across with respect to a location where the image data was captured.

Assignees

• Bae Systems Plc

Inventors

• Buchanan Roderick
• Revell James Duncan

Classifications

• G06V20/58 Primary

  Recognition of moving objects or obstacles, e.g. vehicles or pedestrians; Recognition of traffic objects, e.g. traffic signs, traffic lights or roads · CPC title

• G06V20/52

  Surveillance or monitoring of activities, e.g. for recognising suspicious objects (recognising microscopic objects G06V20/69) · CPC title

• G06V10/255

  Detecting or recognising potential candidate objects based on visual cues, e.g. shapes · CPC title

• G06T2207/30252

  Vehicle exterior; Vicinity of vehicle · CPC title

• G06K9/3241

  Physics · mapped topic