Method and system for detecting road area

• US 6,549,642 B1
• Filed: 04/30/1999
• Issued: 04/15/2003
• Est. Priority Date: 05/01/1998
• Status: Expired due to Fees

• Current Assignee: NEC Corporation
• Original Assignee: NEC Corporation
• Inventors: Sakurai, Kazuyuki
• Primary Examiner(s): Johnson, Timothy M.
• Assistant Examiner(s): Tabatabai, Abolfazl
• Application Number:
  US09/302,451
• Time in Patent Office: 1,446 Days
• Field of Search: 382/104, 382/103, 382/199, 382/291, 382/153, 382/170, 701/300, 701/301, 340/901, 340/933, 340/905, 340/907, 340/910, 340/917, 340/918, 340/919, 340/937, 340/440
• US Class Current: 382/104
• CPC Class Codes: G06V 10/457 by analysing connectivity, e.g. edge linking, connected component analysis or slices

Abstract

In a system for recognizing the environment around a vehicle with an image, to provide a system and method of detecting a road area without lowering the accuracy over the entire of the road area including an upper portion of the road area. The system comprises a computer 100 which is operated under control of a program, an image input device 110, and means for outputting road area data 120. The computer 100 comprises means for calculating statistic for each horizontal line 101, means for fitting the statistic to a function 102, means for processing a threshold value of the road image 103 and means for detecting the road area 104.

Claims

1. In a system for recognizing the environment around a vehicle with an image, a system for detecting a road area wherein said system comprises

means for calculating a statistic for each horizontal line of a road image which is input from an image input device,

means for fitting said statistic to a function related to a vertical coordinate of the road image,

threshold value-processing means for processing a threshold value of said road image by using said function,

means for detecting the road area from the road image which has been subjected to threshold value-processing by said threshold value-processing means, and

wherein said function fitting means uses an inverse proportional function relating to the vertical coordinate of the road image as said function.

2. A system for detecting a road area as defined in claim 1 wherein said threshold value-processing means includes means for modifying the image based upon a position in the road image prior to said threshold value-processing.

3. A system for detecting a road area as defined in claim 1 wherein said threshold value-processing means includes means for presetting a threshold value based upon a statistic of an image which has not been subjected to the threshold value-processing.

4. A system for detecting a road area as defined in claim 3 wherein spatial differentiation is effected on said image which has not been subjected to the threshold value-processing and its statistic is used in said means for presetting the threshold value.

5. A system for detecting a road area as defined in claim 3 wherein spatial differentiation is effected on said image which has not been subjected to the threshold value-processing and a threshold value is preset as a linear function of an average value of pixel values in said means for presetting the threshold value.

6. A system for detecting a road area as defined in claim 1 wherein said means for calculating the statistic uses a brightness value at a peak in a histogram for each horizontal line as the statistic.

7. A method of detecting a road area comprising:

providing a road image comprising a road area therein, and

detecting the road area in said road image,

wherein detecting of the road area tracks changes in the vertical coordinate value of the brightness in the road area based upon a statistic which is calculated for each line in a given area in the road image which is input from image input means, and further comprising the step of;

fitting the statistic to an inverse proportional function relating to the vertical coordinate of the road image.

8. A method of detecting a road area comprising:

(a) calculating a statistic of each horizontal line of a road image which is input from an image input device,

(b) fitting the statistic of each horizontal line to a predetermined inverse proportional model function relating to the vertical coordinate of the image,

(c) approximating the statistic of each horizontal line by using said function and processing a threshold value of said road image based upon the difference between its characteristic value and a value of the model function on the vertical coordinate in each pixel of said road image, and

(d) detecting a road area from the road image which has been processed with said threshold value.

9. A method of detecting a road area wherein the method comprises the steps of:

(a) inputting a road image from an image input device,

(b) calculating a statistic of the brightness value as a characteristic value for each horizontal line in a predetermined area in which the road area occupies a predominant area in the image,

(c) fitting the statistic of each horizontal line to a predetermined inverse proportional model function on the vertical coordinate of the road image to approximate the characteristic value of each horizontal line of the road image using the function,

(d) calculating the difference between the characteristic value and the value of the model function on said vertical coordinate,

(e) modifying the difference between the characteristic value and the value of the model function based upon the position in the image so that value of the difference becomes lower for a pixel having a higher probability that the pixel is in the road area,

(f) calculating the statistic of the modified difference over the entirety of said road image,

(g) presetting a threshold value for said difference value based upon said statistic,

(h) converting the difference between the characteristic value and the function value in each pixel into two-valued variable, and

(i) cutting out an area which is in a lower portion of the road image as a road area based upon the difference value which is represented by the two-valued variable.

10. In a system for recognizing the environment around a vehicle with an image, a computer usable medium storing thereon a program, which, when executing in a computer includes:

(a) means for calculating a statistic for each horizontal line of a road image which is input from an image input device,

(b) means for fitting said statistic to an inverse proportional function related to a vertical coordinate of the road image,

(c) means for processing a threshold value of said road image by using said function value, and

(d) means for detecting a road area from the road image which has been processed with said threshold value.

11. A medium storing thereon a program for executing in a computer, and when executing on said computer performs the steps of:

(a) inputting a road image from an image input device,

(b) calculating statistic of the brightness value as a characteristic value for each horizontal line in a predetermined area such as a lower area in which the road area occupies a predominant area in the image,

(c) fitting the statistic of each horizontal line to a predetermined inverse proportional model function in the vertical coordinate of the road image to approximate the characteristic value of each horizontal line of the road image using said function,

(d) calculating the difference between the characteristic value and value of the model function in said vertical coordinate,

(e) modifying the difference between the characteristic value and the value of the model function based upon the position in the image so that value of the difference becomes lower for a pixel having a higher probability that the pixel is in the road area,

(f) calculating the statistic of the modified difference over the entirety of said road image,

(g) presetting a threshold value for said difference value based upon said statistic,

(h) converting the difference between the characteristic value and the function value in each pixel into two-valved variable, and

(i) cutting out an area which is in a lower portion of the road image as road area based on upon the difference value which is represented by the two-valved variable.