PASSENGER VEHICLE MAKE AND MODEL RECOGNITION SYSTEM

• US 20170097944A1
• Filed: 10/03/2016
• Published: 04/06/2017
• Est. Priority Date: 10/02/2015
• Status: Active Grant

• Current Assignee: Regents of the University of California (University of California)
• Original Assignee: Regents of the University of California (University of California)
• Inventors: THAKOOR, Ninad, BHANU, Bir
• Granted Patent: US 10,223,609 B2
• Time in Patent Office: Days
• CPC Class Codes: G06V 10/32 Normalisation of the pattern dimensions
• CPC Class Codes: G06V 10/462 Salient features, e.g. scale invariant feature transforms [SIFT]
• CPC Class Codes: G06V 10/754 involving a deformation of the sample pattern or of the reference pattern; Elastic matching
• CPC Class Codes: G06V 20/54 of traffic, e.g. cars on the road, trains or boats

Abstract

A system, a method, and a non-transitory computer readable recording medium for recognizing a passenger vehicle make and model are disclosed, which includes a make and model recognition (MMR) database, which is constructed from rear view videos of passenger vehicles.

Claims

1. A system for recognizing a passenger vehicle make and model, the system comprising:

a make and model recognition (MMR) database, which is constructed from rear view videos of passenger vehicles;

a universal view, which is generated from the make and model recognition database by averaging images from the make and model recognition database, and wherein images generated are identified as “

UVaMMRDB”

;

a showcase view, which represent visually distinct types of vehicles, and wherein copies of the images aligned to their type specific showcase view and stored in the MMR database, and wherein the images are identified as “

SVaMMRDB”

;

a moving object detection system, which is configured to detect region of interest (ROI) of passenger vehicles for classification, and the ROI is registered with the universal view, which aligns landmarks in the ROI, the aligned ROI is identified as the universal view aligned ROI (“

UVaROI”

);

after illumination normalization, the UVaROI is compared to UVaMMRDB to find the nearest neighbors in the MMR database, and based on the nearest neighbors a type is assigned to the ROI, and the ROI is registered with the showcase view of the type to which it is assigned to and the registration process aligns landmarks in the ROI and the showcase view, and the showcase view aligned ROI is identified as “

SVaROI”

; and

wherein after illumination normalization, the SVaROI is compared to SVaMMRDB to find the nearest neighbors in the MMR database, and based on these neighbors a make and model is assigned to the ROI.

2. The system of claim 1, wherein the make and model recognition database holds rear view images of passenger vehicles and make and model information.

3. The system of claim 1, wherein the universal view is iteratively improved by aligning the MMR database to the current universal view and averaging the aligned images.

4. The system of claim 1, wherein copies of the images aligned to the universal view and stored in the MMR database.

5. The system of claim 1, wherein the showcase view for a vehicle type is generated by averaging images of the particular vehicle type in the MMR database.

6. The system of claim 1, wherein the showcase view is iteratively improved by aligning the images of the particular vehicle type with current showcase view and averaging the aligned images.

7. The system of claim 1, wherein the registration process aligns landmarks in the ROI and the universal view, which comprises a two-step alignment process of aligning the rows and the aligning the columns;

and

wherein the alignment steps are interchangeable.

8. The system of claim 1, wherein the system can be used for:

identifying make and model of vehicles on the road;

parking garage or parking lot allocation can be carried out based on vehicle make and model;

vehicle fingerprinting and security related applications ensuring limited/controlled access to vehicles; and

/or

in conjunction with automatic license plate recognition and vehicle registration database, vehicle make and model verification can be carried out to avoid license plate spoofing.

9. The system of claim 1, wherein

the system can be implemented for finding the make and model of the vehicles from front and/or side views;

only the type classification part of the system can be implemented;

only Make recognition can be carried out by ignoring the model information; and

/or

types of vehicles used in the system can be changed by either removing some of the types or by adding more vehicle types.

10. A method for recognizing a passenger vehicle make and model, the method comprising:

constructing a make and model recognition (MMR) database, which is constructed from rear view videos of passenger vehicles;

generating a universal view (MMRDB) from the make and model recognition database by averaging images from the make and model recognition database, and wherein images generated are identified as “

UVaMMRDB”

;

representing visually distinct types of vehicles, and wherein copies of the images aligned to their type specific showcase view are stored in the MMR database and identified as “

SVaMMRDB”

;

providing a moving object detection system, which is configured to detect region of interest (ROI) of passenger vehicles for classification, and the ROI is registered with the universal view, which aligns landmarks in the ROI, and wherein the aligned ROI is identified as the universal view aligned ROI (“

UVaROI”

);

after illumination normalization, comparing the UVaROI to UVaMMRDB to find the nearest neighbors in the MMRDB, and based on the nearest neighbors, assigning a type to the ROI, and registering the ROI with the showcase view of the type to which it is assigned to and the registration process aligns landmarks in the ROI and the showcase view, and wherein the showcase view aligned ROI is identified as “

SVaROI”

; and

wherein after illumination normalization, comparing the SVaROI to SVaMMRDB to find the nearest neighbors in the MMRDB, and based on these neighbors assigning a make and model to the ROI.

11. The method of claim 10, comprising:

storing rear view images of passenger vehicles and make and model information in the MMR database.

12. The method of claim 10, further comprising:

aligning the database to the current universal view and averaging the aligned images to iteratively improve the universal view.

13. The method of claim 10, further comprising:

aligning copies of the images to the universal view; and

storing the copies of the images in the MMR database.

14. The method of claim 10, comprising:

generating the showcase view for a vehicle type by averaging images of the particular vehicle type in the database.

15. The method of claim 10, comprising:

iteratively improving the showcase view by aligning the images of the particular vehicle type with current showcase view and averaging the aligned images.

16. The method of claim 10, wherein the registration process aligns landmarks in the ROI and the universal view, which comprises a two-step alignment process of aligning the rows and the aligning the columns.

17. A non-transitory computer readable recording medium stored with a computer readable program code for recognizing a passenger vehicle make and model, the computer readable program code configured to execute a process comprising:

constructing a make and model recognition (MMR) database, which is constructed from rear view videos of passenger vehicles;

generating a universal view (MMRDB) from the make and model recognition database by averaging images from the make and model recognition database, and wherein images generated are identified as “

UVaMMRDB”

;

representing visually distinct types of vehicles, and wherein copies of the images aligned to their type specific showcase view are stored in the MMR database and identified as “

SVaMMRDB”

;

providing a moving object detection system, which is configured to detect region of interest (ROI) of passenger vehicles for classification, and the ROI is registered with the universal view, which aligns landmarks in the ROI, and wherein the aligned ROI is identified as the universal view aligned ROI (“

UVaROI”

);

after illumination normalization, comparing the UVaROI to UVaMMRDB to find the nearest neighbors in the MMRDB, and based on the nearest neighbors, assigning a type to the ROI, and registering the ROI with the showcase view of the type to which it is assigned to and the registration process aligns landmarks in the ROI and the showcase view, and wherein the showcase view aligned ROI is identified as “

SVaROI”

; and

wherein after illumination normalization, comparing the SVaROI to SVaMMRDB to find the nearest neighbors in the MMRDB, and based on these neighbors assigning a make and model to the ROI.

18. The computer readable recording medium of claim 17, comprising:

storing rear view images of passenger vehicles and make and model information in the MMR database.

19. The computer readable recording medium of claim 17, further comprising:

aligning the database to the current universal view and averaging the aligned images to iteratively improve the universal view.

20. The computer readable recording medium of claim 17, further comprising:

generating the showcase view for a vehicle type by averaging images of the particular vehicle type in the database; and

iteratively improving the showcase view by aligning the images of the particular vehicle type with current showcase view and averaging the aligned images.