Obstacle detection procedure for motor vehicle

• US 20110001615A1
• Filed: 07/02/2010
• Published: 01/06/2011
• Est. Priority Date: 07/06/2009
• Status: Active Grant

• Current Assignee: Valeo Vision (Valeo SE)
• Original Assignee: Valeo Vision (Valeo SE)
• Inventors: Fleury, Benoist, Reilhac, Patrice, Chanussot, Lowik, Kuoch, Siav Kuong, Rebut, Julien, Charpentier, Adrien, Dubois, Eric
• Granted Patent: US 8,379,928 B2
• Time in Patent Office: Days
• US Class Current: 340/436
• CPC Class Codes: B60R 21/0134 responsive to imminent contact with an obstacle , e.g. using radar systems
• CPC Class Codes: G01S 13/862 Combination of radar systems with sonar systems
• CPC Class Codes: G01S 13/867 Combination of radar systems with cameras
• CPC Class Codes: G01S 2013/9317 Driving backwards
• CPC Class Codes: G01S 2013/93272 in the back of the vehicles
• CPC Class Codes: G06T 2207/10021 Stereoscopic video; Stereoscopic image sequence
• CPC Class Codes: G06T 2207/20016 Hierarchical, coarse-to-fine, multiscale or multiresolution image processing; Pyramid transform
• CPC Class Codes: G06T 2207/30261 Obstacle
• CPC Class Codes: G06T 7/136 involving thresholding
• CPC Class Codes: G06T 7/70 Determining position or orientation of objects or cameras camera calibration G06T7/80
• 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

Abstract

The present invention concerns an obstacle detection procedure within the area surrounding a motor vehicle.

It features the fact that the procedure involves the following stages:

• • To define at least one first region of interest by carrying out a first detection of an obstacle by shape recognition in an acquired image of this environment;
  • To define at least one second region of interest by carrying out a second detection of an obstacle by movement detection in relation to the vehicle on a sequence of acquired images of this environment;
  • To classify the detected obstacle with, respectively, a first and second index of confidence applied to the first and second regions of interest in relation to given characteristics; and
  • To validate the classification of the object detected in relation to these indices of confidence and in relation to these regions of interest resulting in validated regions of interest.

Claims

1. An obstacle detection procedure (O) within an environment (E) of a motor vehicle, featuring the fact that it involves the following stages:

defining at least one first region of interest (ROI1) by carrying out a first detection of an obstacle (O) by shape recognition in an acquired image (I) of this environment (E);

defining at least one second region of interest (ROI2) by carrying out a second detection of an obstacle by movement detection in relation to the vehicle (V) on a sequence of acquired images (I) of this environment (E);

classifying the obstacle (O) detected with, respectively, a first (IC1) and second (IC2) index of confidence applied to the first (ROI1) and the second (ROI2) regions of interest in relation to given characteristics (C); and

validating the classification of the detected object (O) in relation to these indices of confidence (IC1, IC2) and in relation to these regions of interest (ROI1, ROI2) resulting in validated regions of interest (ROI1, ROI2).

2. The obstacle detection procedure according to claim 1, wherein said obstacle detection procedure also involves the further stages of:

carrying out a third obstacle detection (O) by sensor/s (ULS;

RAD) with detection range below a first threshold (S1) resulting in a determined position (POS);

projecting the determined position (POS) in a reference marker (COORef);

projecting the validated regions of interest (ROI1, ROI2) in this reference marker (COORef); and

aligning the two projections obtained (PJ1, PJ2) and attributing a determined position (POS) to the classified obstacle (O) in accordance with the alignment.

3. The obstacle detection procedure according to claim 2, wherein the reference marker is the image marker (XI, YI).

4. The obstacle detection procedure according to claim 1, wherein the first detection by shape recognition involves the following sub-stages:

constructing a pyramid of images (Isr) by recursive sub-sampling of an acquired image (I);

for each sub-sampled image (Isr);

this image (Isr) with a label (VIGN) representing a certain obstacle (O);

at each successive position of a label (VIGN) in this image (Isr), analyzing its content by means of a set of classifiers (Cls, Clw), these classifiers being determined during a phase of prior apprenticeship.

5. The obstacle detection procedure according to claim 4, wherein the first detection by shape recognition also involves a further sub-stage to determine a scanning zone (Zb) in a sub-sampled image (Isr).

6. The obstacle detection procedure according to claim 1, wherein a second movement detection involves the following sub-stages:

constructing a background image (Iref) recalculated for each image (I) of a sequence of images (SQ);

taking out the background image (Iref) to the current image (Ic) in the sequence of images (SQ) resulting in a resulting image (Ires);

discriminating certain movements in the resulting image (Ires) according to a threshold (Smax); and

extracting second regions of interest (ROI2) from the discrimination.

7. The obstacle detection procedure according to claim 1, wherein a second movement detection involves the following sub-stages:

determining points of interest (PtI) on an acquired image (I);

following these points of interest (PtI) on the next image (I+1) and determining their displacement (Vmi);

regrouping into second regions of interest (ROI2) these points of interest (PtI) having a similar displacement (Vmi);

determining the displacement (Vmv) of the vehicle (V); and

discriminating second regions of interest (ROI2) having a movement (Vmi) in the image which does not result from the displacement of the vehicle (Vmv).

8. The obstacle detection procedure according to claim 1, wherein the classification stage involves the following sub-stages:

creating a vector of characteristics (C) from the regions of interest (ROI);

classifying this vector of characteristics (C) by comparison with a border of decision (DG) determined previously during an apprenticeship stage; and

determining an associated index of confidence (IC) according to the distance of the vector of characteristics in relation to this border of decision (DG).

9. The obstacle detection procedure according to claim 1, wherein the vector of characteristics (C) is a histogram of the orientated gradients of a region of interest (ROI1, ROI2).

10. The obstacle detection procedure according to claim 1, wherein the validation stage involves the following sub-stages:

constructing a probability card (P_MAP) corresponding to an image (I) in which each classified region of interest (ROI1, ROI2) is represented by a distribution of probability;

accumulating these distributions of probability which tally in the probability card (P_MAP) in order to obtain at least one local maximum; and

validating the region of interest (ROI1, ROI2) which is closer to each local maximum of the probability card (P_MAP).

11. The obstacle detection procedure according to claim 2, wherein detection by sensor/s is carried out by means of ultrasound sensors (ULS).

12. The obstacle detection procedure according to claim 2, wherein detection by sensor/s is carried out by means of radar sensors (RAD).

13. The obstacle detection procedure according to claim 1, wherein it also involves a further stage of applying a change of perspective to an acquired image (I).

14. The obstacle detection procedure according to claim 1, wherein it also involves a further stage of follow-up of validated regions of interest on a sequence (SQ) of acquired images.

15. The obstacle detection device of an obstacle (O) within an environment (E) of a motor vehicle, arranged to commission a procedure according claim 1.

16. A computer program product (PG) including one or more sequences of instructions executable from an information processing unit, the execution of these sequences of instructions allowing the procedure to be commissioned according to claim 1.