RECONFIGURABLE CLEAR PATH DETECTION SYSTEM

• US 20130265424A1
• Filed: 04/09/2012
• Published: 10/10/2013
• Est. Priority Date: 04/09/2012
• Status: Active Grant

• Current Assignee: GM Global Technology Operations LLC (General Motors Company)
• Original Assignee: GM Global Technology Operations LLC (General Motors Company)
• Inventors: ZHANG, WENDE, WANG, JINSONG
• Granted Patent: US 9,626,599 B2
• Time in Patent Office: Days
• US Class Current: 348/148
• CPC Class Codes: B60W 40/072 Curvature of the road
• CPC Class Codes: G06F 18/254 of classification results, e.g. of results related to same input data
• CPC Class Codes: G06V 10/809 of classification results, e.g. where the classifiers operate on the same input data
• 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 reconfigurable clear path detection system includes an image capture device and a primary clear path detection module for determining corresponding probability values of identified patches within a captured image representing a likelihood of whether a respective patch is a clear path of the road. A plurality of secondary clear path detection modules each are used to assist in identifying a respective clear path of the traveled road in the input image. One or more of the secondary clear path detection modules are selectively enabled for identifying the clear path. The selectively enabled secondary clear path detection modules are used to identify the clear path of the road of travel in the input image. A fusion module collectively analyzes the clear path detection results of the primary clear path detection module and the selectively enabled secondary clear path detection modules for identifying the clear path in the input image.

Claims

1. A reconfigurable clear path detection system for a vehicle comprising:

an image capture device for capturing images of a scene in a path of travel;

a primary clear path detection module determining a clear path in an input image captured from the image capture device;

a plurality of secondary clear path detection modules, each secondary clear path detection module configured to independently assist in identifying a respective clear path of the road of travel in the input image, wherein one or more of the secondary clear path detection modules are selectively enabled for assisting in identifying the clear path of the road of travel, wherein only the selectively enabled secondary clear path detection modules are used to identify the clear path of the road of travel in the input image; and

a fusion module for collectively analyzing the clear path detection results of the primary clear path detection module and the selectively enabled secondary clear path detection modules for identifying the clear path in the input image.

2. The reconfigurable clear path detection system of claim 1 wherein the plurality of secondary clear path detection modules includes a temporal coherence module, wherein a clear path detection probability from a previous time step frame is utilized to update a clear path detection probability in a current time step frame, wherein a probability update is based on a weighted inter-frame matching technique which is derived from vehicle motion and an on-ground plane assumption.

3. The reconfigurable clear path detection system of claim 1 wherein the plurality of secondary clear path detection modules includes a top-down view classification module, wherein a top-down view image is generated from the input image, and wherein features are extracted from the top-down view image for identifying the clear path.

4. The reconfigurable clear path detection system of claim 1 wherein the plurality of secondary clear path detection modules includes a road structure detection module, wherein vanishing point and pass-vanishing point line tracking is performed and learning-based road edge detection is applied for identifying road edges and road structure in the input image.

5. The reconfigurable clear path detection system of claim 1 wherein the plurality of secondary clear path detection modules includes an on-line similar scene module, wherein the on-line similar scene module utilizes an image captured in real-time and compares the captured image to various images retrieved from a database that are previously labeled or identified.

6. The reconfigurable clear path detection system of claim 1 wherein a combination of the primary clear path detection module and the selectively enabled clear path detection modules are collectively analyzed by weighting each of the results of the primary clear path detection module and the selectively enabled secondary clear path detection modules.

7. The reconfigurable clear path detection system of claim 6 wherein weighting each of the results of the primary clear path detection module and the selectively enabled secondary clear path detection modules are represented by the formula as follows:

8. The reconfigurable clear path detection system of claim 7 wherein the clear path is further determined as a function of the identified road structure and is represented by the following formula:

P<sub>rdstr—

fuse</sub>(c)=Conf<sub>rd—

edge</sub>(c)·

P_fuse(c)

where—

P<sub>rdstr—

fuse</sub>(c) is a probability of the clear path as a function of the identified road structure, P_fuse(c) is the probability of the clear path for a respective segmented patch generated by a first fusion module, and Conf<sub>rd—

edge</sub> is a confidence map with regions inside and outside of the identified road edges being assigned different confidence values.

9. The reconfigurable clear path detection system of claim 1 wherein the selectively enabled secondary clear path detection modules are selected based on a trade-off between a processing time and accuracy of identifying the clear path in the input image.

10. The reconfigurable clear path detection system of claim 1 wherein the selectively enabled secondary clear path detection modules utilize time-displaced captured images of the traveled road for identifying the clear path in the input image.

11. The reconfigurable clear path detection system of claim 10 wherein the time-displaced captured images of the traveled road are captured within a predetermined frame rate of capturing the input image.

12. The reconfigurable clear path detection system of claim 1 further comprising an output device for displaying the clear path of road to a driver of a vehicle.

13. The reconfigurable clear path detection system of claim 1 further comprising an object detection system, the object detection system utilizing the identified clear path for detecting objects in the input image.

14. A method for detecting a clear path of a road of travel for a vehicle, the method comprising the steps of:

capturing images of a scene in a path of travel by an image capture device;

determining a clear path in an input image by a primary clear path detection module, the primary clear path detection module analyzing the input image from the captured image device, the primary clear path detection module segmenting the input image into a plurality of patches, determining probability values for each of the patches representing a likelihood of whether a respective patch is a clear path of travel, associating feature data of the input image with the segmented patches, and applying a trained classifier for identifying a clear path of travel in the input image based on the feature data and the corresponding probability values;

providing a plurality of secondary clear path detection modules for independently identifying a respective clear path of the travel in the input image;

selectively enabling one or more of the secondary clear path detection modules for identifying the clear path of the travel, wherein only the selectively enabled secondary clear path detection modules are used for identifying the clear path of the road of travel in the input image;

a fusion module collectively analyzing the clear path detection results of the primary clear path detection module and the selectively enabled secondary clear path detection modules for identifying the clear path in the input image.

15. The method of claim 14 wherein the plurality of secondary clear path detection modules includes a temporal coherence module, wherein a clear path detection probability from a previous time step frame is utilized to update a clear path detection probability in a current time step frame, wherein a probability update is based on a weighted inter-frame matching technique which is derived from vehicle motion and an on-ground plane assumption.

16. The method of claim 14 wherein the plurality of secondary clear path detection modules includes a top-down view classification module, wherein a top-down view image of the captured image is generated from one or more captured input images exterior of the vehicle, and wherein features are extracted from the top-down view image, and wherein extracted features from the top-down view are applied to a classifier for identifying the clear path in the input image.

17. The method of claim 14 wherein the plurality of secondary clear path detection modules includes a road structure detection module, wherein vanishing point and pass-vanishing point line tracking is performed and learning-based road edge detection is applied for identifying road edges and road structure in the input image.

18. The method of claim 14 wherein the plurality of secondary clear path detection modules includes an on-line similar scene module, wherein the on-line similar scene module utilizes an image captured in real-time and compares the captured image to various images retrieved from a database that are previously labeled or identified.

19. The method of claim 14 wherein the fusion module weights the result of the primary clear path detection module and the results of the selectively enabled secondary clear path detection modules for identifying the clear path in the input image.

20. The method of claim 19 wherein weighting each of the results of the primary clear path detection module and the selectively enabled secondary clear path detection modules are represented by the formula as follows:

21. The method of claim 20 wherein the clear path is further determined as a function of the identified road structure and is represented by the following formula:

P_rdstrfuse(c)=Conf<sub>rd—

edge</sub>(c)·

P_fuse(c)

where—

P<sub>rdstr—

fuse</sub>(c) is a probability of the clear path as a function of the identified road structure, P_fuse(c) is the probability of the clear path for a respective segmented patch generated by a first fusion module, and Conf<sub>rd—

edge</sub> is a confidence map with regions inside and outside of the identified road edges being assigned different confidence values.

22. The method of claim 14 wherein the enabled secondary clear path detection modules are selected based on trade-off between a processing time and accuracy of identifying the clear path in the input image.

23. The method of claim 14 wherein the selectively enabled secondary clear path detection modules utilize time-displaced captured images of the path of travel for identification of the clear path in the input image.

24. The method of claim 14 wherein the time-displaced captured images of the path of travel include images captured within a previous predetermined time period.

25. The method of claim 14 further comprising the step of displaying the clear path of travel to a driver of the vehicle using an output display device.

26. The method of claim 14 further comprising the step of providing the identified clear path of travel to an object detection system for detecting objects in the input image.