Distance detection method using images

• US 6,021,209 A
• Filed: 07/28/1997
• Issued: 02/01/2000
• Est. Priority Date: 08/06/1996
• Status: Expired due to Fees

• Current Assignee: Fuji Electric Company Limited (Fuji Electric Systems Company Limited)
• Original Assignee: Fuji Electric Company Limited (Fuji Electric Systems Company Limited)
• Inventors: Hirabayashi, Takehide, Shimizu, Hideo, Izumi, Akio
• Primary Examiner(s): Au, Amelia
• Assistant Examiner(s): Dastouri, Mehrdad
• Application Number:
  US08/901,066
• Time in Patent Office: 918 Days
• Field of Search: 382/104, 382/103, 382/106, 382/278, 340/937, 356/3.14, 348/113, 348/116, 348/135, 348/140, 348/170
• US Class Current: 382/103
• CPC Class Codes: G01S 11/12 using electromagnetic waves other than radio waves
• CPC Class Codes: G06T 7/593 from stereo images
• CPC Class Codes: G06V 20/58 Recognition of moving objects or obstacles, e.g. vehicles or pedestrians; Recognition of traffic objects, e.g. traffic signs, traffic lights or roads
• CPC Class Codes: G08G 1/165 for passive traffic, e.g. including static obstacles, trees
• CPC Class Codes: G08G 1/166 for active traffic, e.g. moving vehicles, pedestrians, bikes

Abstract

In a distance detection method, a pair of image data representing a pattern of images is obtained by using a pair of image sensors for receiving optical images. Then, a detection target extracted from each of the image data is captured while sequentially shifting positions of divisions in respective groups of divisions from window part data corresponding to a visual field to provide combinations, each having a pair of divisions, and a correlation value is calculated between each pair of the divisions. Also, a precision at a best correlation point as an indicator of reliability of the best correlation is obtained based on a changing condition of the correlation value near a shift value where the best correlation value has been calculated. Finally, a distance to the detection target is obtained by determining that the detection target is present in the visual field seen through the window corresponding to the window part data only when the precision exceeds a predetermined level.

Claims

1. A distance detection method comprising:

obtaining a pair of image data representing a pattern of images by using a pair of image sensors for receiving optical images via different optical paths;

capturing a detection target extracted from each of the image data while sequentially shifting positions of divisions in respective groups of the divisions from window part data corresponding to a visual field to provide combinations, each comprising a pair of the divisions, and calculating, respectively, a correlation value between each pair of the divisions;

selecting a best correlation value from the calculated correlation values to determine a best correlation point;

obtaining a precision at the best correlation point as an indicator of reliability of the best correlation point, based on a changing condition of the correlation values on both sides of the best correlation value, said precision being obtained by an average of slopes of a plurality of the correlation values before and after the best correlation point; and

detecting a distance to the detection target by determining that the detection target is present in the visual field seen through the window corresponding to the window part data only when the precision exceeds a predetermined level.

2. A distance detection method according to claim 1, wherein the precision as the indicator of the reliability of the best correlation is obtained by selecting a gentler slope in a plurality of the correlation values with slopes before and after the best correlation point.

3. A distance detection method according to claim 1, wherein the correlation value is calculated based on differential data for internal data in the window part data.

4. A distance detection method according to claim 1, further comprising specifying a direction and a width of an image capture visual field in connection with the correlation value, and selecting from the image data the window part data corresponding to the visual field to calculate the correlation value.

5. A distance detection method comprising:

obtaining a pair of image data representing a pattern of images by using a pair of image sensors for receiving optical images via different optical paths;

capturing a detection target extracted from each of the image data while sequentially shifting positions of divisions in respective groups of the divisions from window part data corresponding to a visual field to provide combinations, each comprising a pair of the divisions, and calculating, respectively, a correlation value between each pair of the divisions;

selecting a best correlation value from the calculated correlation values to determine a best correlation point;

obtaining a precision at the best correlation point as an indicator of reliability of the best correlation point, based on a changing condition of the correlation values on both sides of the best correlation value, said precision being changed based on an image contrast value in the window part in order to provide an effective precision, wherein as an indicator of the contrast value, sum of absolute values of differential values for internal data in the window part data is used; and

detecting a distance to the detection target by determining that the detection target is present in the visual field seen through the window corresponding to the window part data only when the precision exceeds a predetermined level.