Method and a System For Detecting a Road at Night

• US 20080199045A1
• Filed: 05/16/2006
• Published: 08/21/2008
• Est. Priority Date: 05/31/2005
• Status: Active Grant

• Current Assignee: Koninklijke Philips N.V.
• Original Assignee: Koninklijke Philips Electronics N.V. (Koninklijke Philips N.V.)
• Inventors: Ekin, Ahmet
• Granted Patent: US 7,936,903 B2
• Time in Patent Office: Days
• US Class Current: 382/104
• CPC Class Codes: G06T 2207/10048 Infrared image
• CPC Class Codes: G06T 2207/30252 Vehicle exterior; Vicinity of vehicle
• CPC Class Codes: G06T 7/11 Region-based segmentation
• CPC Class Codes: G06T 7/136 involving thresholding
• CPC Class Codes: G06V 10/44 Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
• CPC Class Codes: G06V 20/56 exterior to a vehicle by using sensors mounted on the vehicle

Abstract

A method of detecting a road feature in an image signal derived from an infrared-sensitive camera. The method, in overview, comprises processing an image frame by assigning binary values to pixels in the frame in response to their representative temperature, and then to analyze spatially the binary mask to identify regions of pixels having mutually similar assigned binary values. The road feature is subsequently found from the analysis of the identified regions of mutually similar binary values and a visual indication of the road feature in relation to the image frame provided to the user.

Claims

1. A method for detecting a road at night or in poor visual conditions, said method being based on an image frame and being executed by a computer, said image frame, I, being divided into pixels, each having an intensity value assigned corresponding to a temperature of a point of an environment that the image frame represents, said method comprising the steps of

calculating a mask, R, (230) by detecting in said image frame pixels candidating to represent a road;

selecting (250) from the road candidate pixels an initial cluster of road candidate pixels having a lower boundary coinciding with or being located in the vicinity of the lower part of the image frame as an estimate of the road;

determining an initial horizon estimate (240) as a first horizontally oriented boundary between horizontal regions characterized by different densities of road candidate pixels;

defining a final road cluster (260) by correcting the selected initial cluster of road candidate pixels (250) having an upper boundary, said correction being based on the location of said upper boundary relative to the initially estimated horizon and a predetermined upper limit for the location of the horizon.

2. A method according to claim 1, wherein said detection of road candidate pixels (230) is based on a calculation of a first threshold, Th, from the pixel intensity values of said image frame, and wherein the road candidate pixels (230) are detected by comparing each pixel intensity value to said first threshold value, Th.

3. A method according to claim 2, wherein said first threshold value, Th, is calculated as the mean, Ī

, of the pixel values in said image frame.

4. A method according to claim 3, wherein said first threshold value, Th, is weighed with the standard deviation, σ

₁, of the pixel intensity values in said image frame.

5. A method according to claim 1, wherein said initial horizon estimate (240) is calculated from a vertical projection, P_v, of the road pixel candidates of said mask, R, and the derivative, P_Dv, of said projection, P_v.

6. A method according to claim 5, wherein said initial horizon estimate (240) is based on:

a second horizon estimate, H₂ , determined as the uppermost row of the mask, R, having an absolute value of a derivative, P_Dv, larger than a predetermined constant, T_Large;

or

an average of said second horizon estimate H2 and a first horizon estimate H₁, wherein said first horizon estimate, H₁, is determined as the uppermost row of the mask, R, having a vertical projection, P_v, larger than the sum of a vertical projection, P_v(i_up) of a predetermined row of the mask, R, and said a predetermined constant, T_Large.

7. A method according to claim 1, wherein said selection of the road candidate pixels (250) comprises identifying connected clusters of equal candidate pixel values in the mask, R, and selecting the largest coherent cluster, that is larger than a predetermined percentage of all pixels, and has a lower boundary located below a certain row, RC_max, of the mask, R.

8. A method according to claim 7, wherein said identification of connected clusters of equal pixel values in the mask, R, is performed by applying a connected component labelling algorithm to the mask, R.

9. A method according to claim 1, wherein said selection of a final road cluster (260) comprises:

correcting (640) the initial road cluster (250), if the upper boundary of the initial road cluster is above a preselected second threshold (N_sky) by redefining the upper boundary of the initial road cluster with the initial horizon estimate; and

selecting (630) the initial road cluster (250), and adjusting the horizon estimate to be coinciding with the upper boundary of the initial road cluster, if the upper boundary of the initial road cluster is located below or at said predetermined second threshold, N_sky.

15. A data carrier storing instructions for a data processor which when loaded into a computer system performs a method according to claim 1.

10. A system for detecting a road at night or in poor visual conditions, the system comprising one or more data processors and one or more data memories, the system is storing data in the one or more memories representing an image frame, I, being divided into pixels, each having an intensity value assigned corresponding to a temperature of a point of an environment that the image frame represents, the system is provided with instructions which upon execution by the one or more processors adapt the system to:

calculating a mask, R, (230) by detecting in said image frame pixels candidating to represent a road;

selecting (250) from the road candidate pixels an initial cluster of road candidate pixels having a lower boundary coinciding with or being located in the vicinity of the lower part of the image frame as an estimate of the road;

determining an initial horizon estimate (240) as a first horizontally oriented boundary between horizontal regions characterized by different densities of road candidate pixels;

defining a final road cluster (260) by correcting the selected initial cluster of road candidate pixels (250) having an upper boundary, said correction being based on the location of said upper boundary relative to the initially estimated horizon and a predetermined upper limit for the location of the horizon.

11. A system according to claim 10, further comprising a single infrared camera for obtaining an image frame, I.

12. A system according to claim 10, comprising multiple infrared cameras for obtaining image frames, I.

13. A system according to claim 10, wherein the system is further adapted to displaying an image of the final road cluster (260) to a user.

14. Use of a system according to claim 12 to calibrate the multiple cameras.