ROAD SHAPE RECOGNITION DEVICE

• US 20100299109A1
• Filed: 05/18/2010
• Published: 11/25/2010
• Est. Priority Date: 05/22/2009
• Status: Active Grant

• Current Assignee: Subaru Corp.
• Original Assignee: Fuji Heavy Industries Limited (Subaru Corp.)
• Inventors: Saito, Toru
• Granted Patent: US 8,340,896 B2
• Time in Patent Office: Days
• US Class Current: 703/1
• CPC Class Codes: G01S 11/12 using electromagnetic waves other than radio waves
• CPC Class Codes: G06T 2207/10012 Stereo images
• CPC Class Codes: G06T 2207/30252 Vehicle exterior; Vicinity of vehicle
• CPC Class Codes: G06T 7/12 Edge-based segmentation
• CPC Class Codes: G06V 20/588 Recognition of the road, e.g. of lane markings; Recognition of the vehicle driving pattern in relation to the road

Abstract

A road shape recognition device includes: distance and height detecting means for detecting distance data having a distance and height in real space regarding a road surface where a vehicle is traveling at multiple mutually different points; approximation line calculating means for dividing the plurality of distance data into near and far groups as viewed from the vehicle to calculate an approximation line of the distance data for each group each time the distance data of the boundary portion between the two groups is transferred from one of the groups to the other; statistics calculating means for calculating statistics from the corresponding approximation line for each group where the distance data is transferred; and road shape model generating means for selecting one out of combinations of the approximation lines to generate a road shape model using the selected combination.

Claims

1. A road shape recognition device comprising:

distance and height detecting means configured to detect distance data having at least information of distance and height in real space regarding a road surface where a vehicle is traveling at a plurality of mutually different points;

approximation line calculating means configured to divide said plurality of distance data into a near group and a far group as viewed from said vehicle based on said distance in real space to calculate approximation lines for approximating said distance data for said groups respectively each time said distance data of a boundary portion between said two groups is transferred from one of said groups to the other of said groups;

statistics calculating means configured to calculate statistics based on said approximation lines for each said group where said distance data is transferred; and

road shape model generating means configured to select one out of combinations of said approximation lines for said two groups to generate a road shape model using the selected combination of said approximation lines as at least a shape on a distance-height plane, based on said calculated statistics of said two groups.

2. The road shape recognition device according to claim 1,

wherein said approximation line calculating means uses the least square method to calculate said approximation lines for approximating said distance data for said groups respectively;

wherein said statistics calculating means calculates a variance or standard deviation as to said approximation lines of said distance data belonging to said groups respectively as said statistics; and

wherein said road shape model generating means selects a combination of said approximation lines that minimizes the total sum of said calculated statistics of said two groups.

3. The road shape recognition device according to claim 2,

wherein when the distance and height in real space of said distance data are represented as Z and Y respectively, the distance and height in real space of said distance data of said boundary portion to be transferred are represented as Za and Ya respectively, and the summation for each of said groups is represented using Σ

, said approximation line calculating means calculates said approximation lines for approximating said distance data for said groups respectively using the least square method by subtracting Za, Ya, Za² and ZaYa from Σ

Z, Σ

Y, Σ

Z² and EZY of one of said groups respectively, and adding Za, Ya, Za² and ZaYa to Σ

Z, Σ

Y, Σ

Z² and Σ

ZY of the other of said groups respectively each time the distance data of the boundary portion is transferred from one of said groups to the other of said groups.

4. The road shape recognition device according to claim 1,

wherein among of said plurality of distance data detected by said distance and height detecting means, said approximation line calculating means approximates, using said approximation lines, a plurality of distance data included in a predetermined range from the position of a road shape model at a current sampling cycle, which is estimated from a subsequent behavior of said vehicle based on said road shape models detected at past sampling cycles.

5. The road shape recognition device according to claim 1,

wherein said road shape model generating means replaces an intersection portion of said selected approximation lines of said two groups with an easement curve the tangent lines of which are said approximation lines to correct said generated road shape model.

6. The road shape recognition device according to claim 1, further comprising:

road marking detecting means configured to detect a road marking point corresponding to a marking on a road surface including a lane line marked lateral to said vehicle;

wherein said approximation line calculating means calculates said approximation lines said groups respectively using said plurality of distance data, which correspond to said road marking point detected by said road marking detecting means, and of which at least the distance and height in real space are detected by said distance and height detecting means.

7. The road shape recognition device according to claim 6,

wherein said distance and height detecting means uses stereo matching as to a pair of images captured by image capturing means to detect said distance data for each pixel of at least one image of said pair of images;

wherein said road marking detecting means detects said road marking point based on the luminance of pixels of said one image; and

wherein said approximation line calculating means corrects said distance data corresponding to said road marking point detected by said road marking detecting means based on the luminance of pixels near road marking points including said road marking point, and approximates a plurality of corrected distance data by said approximation lines.