Granted Patent B2

US 9,477,892 · App. 14/226,078 · Granted Oct 25, 2016

Abstract

Claims (66)

1. A method for training a vehicle detection system used in a street occupancy estimation of stationary vehicles, the method comprising:

defining first and second areas on an image plane of at least one image capture device associated with monitoring for detection of vehicles, wherein the first area includes an on-street parking lane and the second area includes a traffic lane adjacent to the on-street parking lane;

receiving video-data from a sequence of frames captured from the image capture device;

determining candidate frames that include sample objects relevant to a classification task in the second area;

extracting the sample objects from the candidate frames;

extracting features of each extracted sample object;

assigning labels to the each extracted sample object;

training at least one classifier using the extracted features of the labeled objects; and,

using the at least one trained classifier to classify a vehicle class of a stationary vehicle detected in the first area.

2. The method of claim 1 , further comprising:

assign a label of the each extracted object based on an attribute of the extracted object.

3. The method of claim 2 , wherein the attribute is selected from a group consisting:

size;

shape;

orientation;

aspect ratio; and,

a combination of the above.

4. The method of claim 1 , wherein the extracting the objects further comprise:

performing a geometric transformation on the objects detected in the second area to match an appearance of the objects in the first area.

5. The method of claim 1 , wherein the defining the second area includes:

deriving a geometric transformation suitable to map the first area coordinates to coordinates of the second area;

computing the coordinates of the second area by mapping the coordinates of the vertices of the first area through the geometric transformation.

6. The method of claim 4 , wherein the geometric transformation is selected from a group consisting:

a projective transformation performed on the objects;

an affine transformation performed on the objects;

a rotation of the objects;

a scaling of the objects; and,

a combination of the above.

7. The method of claim 1 , wherein the determining the candidate frames is performed using a foreground detection process via a background estimation and subtraction process.

8. The method of claim 1 , wherein the determining the candidate frames includes:

detecting motion in the second area between frames;

identifying a group of pixels associated with the detected motion;

comparing a size of the group of pixels to a predetermined threshold; and,

in response to the size meeting the threshold, classifying the frames as candidate frames.

9. The method of claim 1 further comprising:

performing blur correction on the extracted objects.

10. The method of claim 1 , wherein the extracting the objects includes cropping a group of pixels defining the objects out of the candidate frames.

11. A system for training a vehicle detection system used in a street occupancy estimation of stationary vehicles, the system comprising a computer including a memory and a processor adapted to execute:

a video buffering module operative to receive video-data from a sequence of frames captured from the image capture device;

a vehicle detection module operative to:

define first and second areas on an image plane of an image capture device associated with monitoring for detection of vehicles, the first area including an on-street parking lane and the second area including a traffic lane adjacent to the on-street parking lane;

determine candidate frames that include objects relevant to a classification task in the second area, and

extract the objects from the candidate frames;

a training module operative to:

extract features of each extracted object,

assign labels to the each extracted object, and

train at least one classifier using the extracted features of the labeled objects; and,

the classifier operative to classify a vehicle class of a stationary vehicle detected in the first area.

12. The system of claim 11 , wherein the training module is further operative to:

assign a label of the each extracted object based on an attribute of the extracted object.

13. The system of claim 11 , wherein the vehicle detection module is further operative to:

perform a geometric transformation on the objects detected in the second area to match an appearance of the objects in the first area.

14. The system of claim 11 , wherein the vehicle detection module is further operative to:

determine the candidate frames by deriving a geometric transformation suitable to map the first area coordinates to coordinates of the second area;

compute the coordinates of the second area by mapping the coordinates of the vertices of the first area through the geometric transformation.

15. The system of claim 11 , wherein the vehicle detection module is further operative to:

determine the candidate frames using a foreground detection process via a background estimation and subtraction process.

16. The system of claim 11 , wherein the vehicle detection module is further operative to:

detect motion in the second area between frames;

identify a group of pixels associated with the detected motion;

compare a size of the group of pixels to a predetermined threshold; and,

in response to the size meeting the threshold, classify the frames as belonging to candidate frames.

17. The system of claim 11 , wherein the training module is further operative to:

perform blur correction on the extracted objects.

18. The system of claim 11 , wherein the vehicle detection module is further operative to:

crop a group of pixels defining the objects out of the candidate frames.

Assignments (8)