Method for Fog Detection

• US 20140247968A1
• Filed: 05/15/2014
• Published: 09/04/2014
• Est. Priority Date: 11/16/2011
• Status: Active Grant

• Current Assignee: Bayerische Motoren Werke AG
• Original Assignee: Bayerische Motoren Werke AG
• Inventors: PAVLIC, Mario
• Granted Patent: US 9,384,401 B2
• Time in Patent Office: Days
• US Class Current: 382/104
• CPC Class Codes: G06F 18/24 Classification techniques
• CPC Class Codes: G06T 2207/10004 Still image; Photographic image
• CPC Class Codes: G06T 2207/10016 Video; Image sequence
• CPC Class Codes: G06T 2207/10024 Color image
• CPC Class Codes: G06T 2207/20024 Filtering details
• CPC Class Codes: G06T 2207/20056 Discrete and fast Fourier transform, [DFT, FFT]
• CPC Class Codes: G06T 2207/20081 Training; Learning
• CPC Class Codes: G06T 2207/30252 Vehicle exterior; Vicinity of vehicle
• CPC Class Codes: G06T 7/90 Determination of colour characteristics
• CPC Class Codes: G06V 20/56 exterior to a vehicle by using sensors mounted on the vehicle

Abstract

A method for detecting fog by use of a camera image or a video image includes the taking of a two-dimensional image with at least one color channel or several color channels as a function of two independent location coordinates. The method determines a two-dimensional gray-scale function for the at least one color channel or for each of the several color channels, which defines the value of the gray scale as a function of the two independent location coordinates of the two-dimensional image. Then, the method performs a two-dimensional Fourier transformation of the two-dimensional gray-scale function depending on two independent frequency coordinates.

Claims

1. A method for detecting fog by use of a camera image or a video image, the method comprising the acts of:

taking a two-dimensional image with at least one color channel or several color channels as a function of two independent location coordinates;

determining a two-dimensional gray-scale function for the at least one color channel or for each of the several color channels, which defines a value of the gray scale as a function of the two independent location coordinates of the two-dimensional image; and

performing a two-dimensional Fourier transformation of the two-dimensional gray-scale function depending on two independent frequency coordinates.

2. The method according to claim 1, wherein:

the two-dimensional gray-scale image is scaled, and

the scaling compensates intensity gradients of the gray scales by use of a low-pass filter and by use of a high-pass filter.

3. The method according to claim 2, wherein:

the square of the absolute value of the Fourier transform, which is called a power spectrum, is calculated, and

the power spectrum is analyzed by use of digital image processing.

4. The method according to claim 3, wherein:

the power spectrum is analyzed by use of a family of Gabor filters, and

the power spectrum is filtered by use of each Gabor filter of the family of Gabor filters over the two-dimensional frequency range, and the filtering result is called a Gabor characteristic for this Gabor filter.

5. The method according to claim 4, wherein:

the Gabor characteristics are characteristics-reduced, and

the reduction of characteristics is carried out by means of a main component analysis to reduced Gabor characteristics.

6. The method according to claim 5, wherein:

the Gabor characteristics or the characteristics-reduced Gabor characteristics are multiplied by a predefined weighting function in order to compute a fog indicator,

the fog indicator is compared with a predefined threshold value, and

the method assigns a value for “

fog”

or another value for “

no fog”

with a confidence measurement to a classification variable.

7. The method according to claim 6, wherein:

the predefined weighting function and the predefined threshold value are predefined by an evaluation of the process by use of training data.

8. The method according to claim 3, wherein:

the power spectrum is characteristics-reduced, and

the reduction of characteristics is carried out by use of a main component analysis in the frequency domain to reduced characteristics.

9. The method according to claim 3, wherein:

the power spectrum is characteristics-reduced, and

the reduction of characteristics is carried out by use of a support vector method in the frequency domain to reduced characteristics.

10. The method according to claim 9, wherein:

the reduced characteristics are classified,

the classification is carried out by use of a linear classifier or a non-linear classifier,

the method assigns a value to a classification variable,

a confidence measurement is assigned to the classification variable, and

the value of the classification variable scales with the occurrence of fog or the density of fog with the confidence measurement.

11. A vehicle, comprising:

a control unit,

a camera- or video system, and

at least one driver assistance system, wherein the control unit is operatively configured to execute processing in real-time that;

receives a two-dimensional image, taken via the camera or video system, with at least one color channel or several color channels as a function of two independent location coordinates;

determine a two-dimensional gray-scale function for the at least one color channel or for each of the several color channels, which defines a value of the gray scale as a function of the two independent location coordinates of the two-dimensional image;

perform a two-dimensional Fourier transformation of the two-dimensional gray-scale function depending on two independent frequency coordinates, wherein

the two-dimensional gray-scale image is scaled,

the scaling compensates intensity gradients of the gray scales by use of a low-pass filter and by use of a high-pass filter

the square of the absolute value of the Fourier transform, which is called a power spectrum, is calculated,

the power spectrum is analyzed by use of digital image processing.

the power spectrum is analyzed by use of a family of Gabor filters,

the power spectrum is filtered by use of each Gabor filter of the family of Gabor filters over the two-dimensional frequency range, and the filtering result is called a Gabor characteristic for this Gabor filter,

the Gabor characteristics are characteristics-reduced,

the reduction of characteristics is carried out by means of a main component analysis to reduced Gabor characteristics,

the Gabor characteristics or the characteristics-reduced Gabor characteristics are multiplied by a predefined weighting function in order to compute a fog indicator,

the fog indicator is compared with a predefined threshold value,

the method assigns a value for “

fog”

or another value for “

no fog”

with a confidence measurement to a classification variable, and

transmits the value of the classification variable to the at least one driver assistance system, wherein

the at least one driver assistance system outputs a warning to the driver in the event of occurring fog, and/or the driver assistance system can be operated in a configuration specified for occurring fog.