Vehicle headlight detecting method and apparatus, and region-of-interest segmenting method and apparatus

• US 8,155,381 B2
• Filed: 12/22/2008
• Issued: 04/10/2012
• Est. Priority Date: 12/27/2007
• Status: Active Grant

• Current Assignee: Alpine Electronics Incorporated (Alps Alpine Co., Ltd.)
• Original Assignee: Alpine Electronics Incorporated (Alps Alpine Co., Ltd.)
• Inventors: Song, Chunyan, Liu, Wei, Yuan, Huai
• Primary Examiner(s): Wu, Jingge
• Application Number:
  US12/341,193
• Publication Number:
  US 20090169105A1
• Time in Patent Office: 1,205 Days
• Field of Search: 382103-107, 348148-149
• US Class Current: 382/103
• CPC Class Codes: G06V 10/25 Determination of region of interest [ROI] or a volume of interest [VOI]
• CPC Class Codes: G06V 20/54 of traffic, e.g. cars on the road, trains or boats
• CPC Class Codes: G06V 20/584 of vehicle lights or traffic lights

Abstract

A method and apparatus for detecting vehicle headlights and a region-of-interest (ROI) segmenting method and apparatus are disclosed. The ROI segmenting method includes: performing an edge extracting operation on a captured image to obtain edges of the captured image; selecting edges meeting predetermined criteria from the obtained edges, the predetermined criteria being the similarity between the region surrounded by the selected edges and the pattern formed by a vehicle headlight in physical reality at a position of the selected edges; determining the region surrounded by the selected edges within the captured image as a vehicle headlight pattern; and segmenting the ROI which potentially includes the vehicle pattern from the captured image based on the determined vehicle headlight pattern. With such a method and apparatus, the ROI of a vehicle may be acquired from the image without using the vehicle'"'"'s bottom shadow.

Claims

1. A method for segmenting a region of interest (ROI) from an image, comprising:

performing an edge extracting operation on a captured image to obtain edges of the captured image;

selecting edges meeting predetermined criteria from the obtained edges, the predetermined criteria being the similarity between the region surrounded by the selected edges and the pattern formed by a vehicle headlight in physical reality at a position of the selected edges;

determining the region surrounded by the selected edges within the captured image as a vehicle headlight pattern; and

segmenting the ROI potentially including the vehicle pattern from the captured image based on the determined vehicle headlight pattern.

2. The method according to claim 1, wherein the act of segmenting further comprises:

moving the left side and the right side of the determined vehicle headlight pattern respectively to the left and to the right by a predetermined width to obtain a left extension side and a right extension side, and extending the bottom side of the determined vehicle headlight pattern downwards by a predetermined depth to obtain a bottom extension side;

extracting edges from a left and right sides candidate region formed by the bottom side, the left extension side, the right extension side, and the bottom extension side of the vehicle headlight pattern;

computing the number of edge points contained in each column within a left side candidate region and a right side candidate region segmented from the left and right sides candidate region along a vertical line through the center of the motorcycle headlight pattern; and

determining the columns containing the most edge points in the left side candidate region and in the right side candidate region as the left side and the right side of the ROI, respectively.

3. The method according to claim 2, further comprising:

if no edges are extractable in either the left side candidate region or the right side candidate region, the column in one candidate region, being symmetrical with the column containing the most edge points in the other candidate region with respect to the center of the motorcycle headlight, is selected as the column containing the most edge points in the one candidate region.

4. The method according to claim 2, further comprising:

if no edges are extractable from both of the left side candidate region and the right side candidate region, determining the left extension side and the right extension side as the left side and the right side of the ROI respectively.

5. The method according to claim 2, wherein the act of segmenting further comprises:

determining a bottom side candidate region of the ROI based on the captured image;

extracting edges from the bottom side candidate region;

computing the number of edge points contained in each row within the bottom side candidate region; and

determining the row containing the most edge points in the bottom side candidate region as the bottom side of the ROI.

6. The method according to claim 5, wherein the act of determining the bottom side candidate region further comprises:

computing the actual height in physical reality corresponding to the pixels'"'"' height from each row within the captured image to the bottom side of the vehicle headlight pattern;

finding a first particular row whose corresponding actual height equals the maximum height from the vehicle headlight to the ground in physical reality, and a second particular row whose corresponding actual height equals the minimum height from the vehicle headlight to the ground in physical reality; and

determining the region formed by the first particular row, the second particular row, the left side, and the right side of the ROI within the captured image as the bottom side candidate region.

7. The method according to claim 5, wherein the act of segmenting further comprises:

computing the pixels'"'"' height of the vehicle pattern formed by the vehicle in physical reality at a position of the bottom side of the ROI within the captured image; and

determining a top side of the ROI based on the bottom side of the ROI and the pixels'"'"' height of the vehicle pattern.

8. A method for detecting a vehicle headlight pattern from an image, comprising:

performing an edge extracting operation on a captured image to obtain edges of the captured image;

selecting edges meeting predetermined criteria from the obtained edges, the predetermined criteria being the similarity between the region surrounded by the selected edges and the pattern formed by the vehicle headlight in physical reality at a position of the selected edges; and

determining the region surrounded by the selected edges within the captured image as a vehicle headlight pattern.

9. The method according to claim 8, the predetermined criteria comprising:

the selected edges featuring a closed contour, with its perimeter within a range between a perimeter of the smallest headlight and a perimeter of the largest headlight and an aspect ratio of its boundary rectangle within a predetermined range.

10. The method according to claim 8, further comprising the following act before the selecting act:

removing edges without a closed contour and/or edges featuring a closed contour with its perimeter less than the corresponding perimeter within the captured image of the smallest headlight in physical reality from the obtained edges.

11. The method according to claim 10, wherein the act of removing further comprises:

performing a filling operation on the edges featuring a closed contour among the obtained edges;

performing an eroding operation on the captured image by adopting a predetermined template; and

performing a dilating operation on the captured image by adopting the predetermined template.

12. The method according to claim 8, wherein the edge extracting operation is performed on an image portion between the vicinity of a road vanishing line within the captured image and the bottom side of the captured image.

13. An apparatus for segmenting a region of interest (ROI) from an image, comprising:

an edge extracting module for performing an edge extracting operation on at least a portion of a captured image to obtain edges of the captured image;

a selecting module for selecting edges meeting predetermined criteria from the obtained edges, the predetermined criteria being the similarity between the region surrounded by the selected edges and the pattern formed by a vehicle headlight in physical reality at a position of the selected edges;

a determining module for determining the region surrounded by the selected edges within the captured image as a vehicle headlight pattern; and

a segmenting module for segmenting the ROI potentially including the vehicle pattern from the captured image based on the determined vehicle headlight pattern.

14. The apparatus according to claim 13, wherein the segmenting module further comprises:

an extending unit, for moving the left side and the right side of the determined vehicle headlight pattern respectively to the left and the right by a predetermined width to obtain a left extension side and a right extension side, and moving the bottom side of the determined vehicle headlight pattern downwards by a predetermined depth to obtain a bottom extension side;

a candidate region edge extracting unit for extracting edges from a left and right sides candidate region formed by the bottom side, the left extension side, the right extension side, and the bottom extension side of the vehicle headlight pattern;

a computing unit for computing the number of edge points contained in each column within a left side candidate region and a right side candidate region segmented from the left and right sides candidate region along a vertical line through the center of the vehicle headlight pattern; and

a determining unit for determining the columns containing the most edge points in the left side candidate region and the right side candidate region as the left side and right side of the ROI respectively.

15. The apparatus according to claim 14, wherein:

if no edges are extractable from either the left side candidate region or the right side candidate region, the computing unit computes the column in one candidate region, being symmetrical with the column containing the most edge points in the other candidate region with respect to the vertical line, as the column containing the most edge points in the one candidate region.

16. The apparatus according to claim 14, wherein:

if no edges are extractable from both of the left side candidate region and the right side candidate region, the determining unit determines the left extension side and the right extension side as the left side and the right side of the ROI respectively.

17. The apparatus according to claim 14, wherein the segmenting module further comprises a bottom side candidate region determining unit for determining a bottom side candidate region of the ROI based on the captured image, wherein:

the candidate region edge extracting unit extracts edges from the bottom side candidate region;

the computing unit computes the number of edge points contained in each row within the bottom side candidate region; and

the determining unit determines the row containing the most edge points in the bottom side candidate region as the bottom side of the ROI.

18. The apparatus according to claim 17, wherein the bottom side candidate region determining unit further comprises:

an actual height computing unit for computing the actual height in physical reality corresponding to the pixels'"'"' height from each row within the captured image to the bottom side of the vehicle headlight pattern;

a determining unit, for determining a first particular row whose corresponding actual height equals the maximum height from the vehicle headlight to the ground in physical reality, and a second particular row whose corresponding actual height equals the minimum height from the vehicle headlight to the ground in physical reality; and

a candidate region determining unit for determining the region formed by the first particular row, the second particular row, the left side, and right side of the ROI within the captured image as the bottom side candidate region.

19. The apparatus according to claim 17, wherein the segmenting module further comprises:

a pixel height computing unit for computing the pixels'"'"' height of the vehicle pattern formed by the vehicle in physical reality at a position of the bottom side of the ROI within the captured image; and

a top side determining unit, for determining a top side of the ROI based on the bottom side of the ROI and the pixels'"'"' height of the vehicle pattern.