Context adaptive approach in vehicle detection under various visibility conditions

• US 7,724,962 B2
• Filed: 07/05/2007
• Issued: 05/25/2010
• Est. Priority Date: 07/07/2006
• Status: Active Grant

• Current Assignee: Continental Automotive GmbH (Continental AG), Siemens Corporation USA (Siemens AG)
• Original Assignee: Siemens AG, Siemens Corp. (Siemens AG)
• Inventors: Baratoff, Gregory, Acunzo, David Jean Pascal, Zhu, Ying
• Primary Examiner(s): BHATNAGAR, ANAND P
• Application Number:
  US11/825,619
• Publication Number:
  US 20080069400A1
• Time in Patent Office: 1,055 Days
• Field of Search: 382/100, 382/103, 382/104, 382/159, 382/165, 382/169, 382/170, 382/181, 382/191, 382/192, 382/194, 382/195, 382/203, 382/224, 382/225, 382/226, 382/227, 707/6, 701/1, 701/2, 701/200, 701/300, 701/301, 701/302
• US Class Current: 382/225
• CPC Class Codes: G06V 10/255 Detecting or recognising potential candidate objects based on visual cues, e.g. shapes
• 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: G06V 20/584 of vehicle lights or traffic lights
• CPC Class Codes: G06V 2201/08 Detecting or categorising vehicles

Abstract

Adaptive vision-based vehicle detection methods, taking into account the lighting context of the images are disclosed. The methods categorize the scenes according to their lighting conditions and switch between specialized classifiers for different scene contexts. Four categories of lighting conditions have been identified using a clustering algorithm in the space of image histograms: Daylight, Low Light, Night, and Saturation. Trained classifiers are used for both Daylight and Low Light categories, and a tail-light detector is used for the Night category. Improved detection performance by using the provided context-adaptive methods is demonstrated. A night time detector is also disclosed.

Claims

1. A method for adaptive detection by a processor of an object in an image represented by image data, by using a plurality of data-driven clusters, each of the plurality of data-driven clusters being characterized by a range of values of one or more statistical parameters associated with a plurality of prior images, each data-driven cluster being part of a context category, a context category being part of a plurality of context categories, the plurality of data-driven clusters being greater than the plurality of context categories, comprising:

receiving the image;

determining a value for each of the one or more statistical parameters of a part of the image that contains the object by the processor;

the processor learning the data-driven clusters from the plurality of prior images, each of the prior images being acquired in a different lighting condition, a different traffic condition or a different camera setting;

assigning the image to one of the plurality of data-driven clusters according to the determined value of each of the one or more statistical parameters of the part of the image;

associating by the processor of the one of the plurality of data-driven clusters with one of at least three context categories, the at least three categories including a daylight, a lowlight and a nightlight category;

context adaptive learning of a classifier for detecting the object based on the one of at least three context categories associated with the assigned one of the plurality of data-driven clusters; and

detecting the object using the classifier.

2. The method as claimed in claim 1, wherein the classifier for detecting the object is learned based on the context category associated with the assigned one of the plurality of data-driven clusters.

3. The method as claimed in claim 1, wherein the object is a vehicle.

4. The method as claimed in claim 1, wherein the plurality of context categories includes at least a category of low light conditions, day light conditions, night conditions and saturation.

5. The method as claimed in claim 1, wherein the one or more statistical parameters of the image include a Bhattacharyya distance between an intensity histogram of the image and an intensity histogram associated with one of a plurality of cluster centers.

6. The method as claimed in claim 1, wherein a plurality of clusters is defined by clustering a plurality of images into k clusters by applying a k-mean algorithm in the space of intensity histograms, an intensity histogram is associated with a cluster center, and a Bhattacharyya distance between two intensity histograms is used in the k-mean algorithm.

7. The method as claimed in claim 1, wherein the classifier will detect the object by recognizing one or more features of a plurality of features of the object in an image, a success of recognizing a feature of the object depending on lighting conditions.

8. The method as claimed in claim 7, wherein the classifier is trained to look for one or more features to detect the object, the one or more features being optimal for detection under determined lighting conditions.

9. The method as claimed in claim 8, wherein the training of the classifier is assisted by an adaptive boosting algorithm.

10. The method as claimed in claim 9, wherein the boosting algorithm is AdaBoost.

11. The method as claimed in claim 7, wherein the object is a vehicle and the plurality of features includes at least one of a group of features consisting of features of edge, texture, contour and tail-lights of the vehicle.

12. A system for adaptive detection of an object in an image by using a plurality of data-driven clusters, each of the plurality of data-driven clusters being characterized by a range of values of one or more statistical parameters associated with a plurality of prior images, each data-driven cluster being part of a context category, a context category being part of a plurality of context categories, the plurality of data-driven clusters being greater than the plurality of context categories, comprising:

a processor;

software operable on the processor to;

receiving the image;

determining a value for each of the one or more statistical parameters of of a part the image that contains the object;

the processor learning the data-driven clusters from the plurality of prior images, each of the prior images being acquired in a different lighting condition, a different traffic condition or a different camera setting;

assigning the image to one of the plurality of data-driven clusters according to the determined value of each of the one or more statistical parameters of a part of the image that contains the object;

associating by the processor of the one of the plurality of data-driven clusters with one of at least three context categories, the at least three categories including a daylight, a lowlight and a nightlight category;

context adaptive learning of a classifier for detecting the object based on the one of at least three context categories associated with the assigned one of the plurality of data-driven clusters; and

detecting the object using the classifier.

13. The system as claimed in claim 12, wherein the object is a vehicle.

14. The system as claimed in claim 12, wherein the plurality of context categories includes at least a category of low light conditions, day light conditions, night conditions and saturation.

15. The system as claimed in claim 12, wherein the one or more statistical parameters of the image include a Bhattacharyya distance between an intensity histogram of the image and an intensity histogram associated with one of a plurality of cluster centers.

16. The system as claimed in claim 12, wherein a plurality of clusters is defined by clustering a plurality of images into k clusters by applying a k-mean algorithm in a space of intensity histograms, an intensity histogram is associated with a cluster center, and a Bhattacharyya distance between two intensity histograms is used in the k-mean algorithm.

17. The system as claimed in claim 12, wherein the classifier will detect the object by recognizing one or more features of a plurality of features of the object in an image, a success of recognizing a feature of the object depending on lighting conditions.

18. The system as claimed in claim 17, wherein the classifier is trained to look for one or more features to detect the object, the one or more features being optimal for detection under determined lighting conditions.

19. The system as claimed in claim 18, wherein the training of the classifier is assisted by an adaptive boosting algorithm.

20. The system as claimed in claim 19, wherein the boosting algorithm is AdaBoost.

21. The system as claimed in claim 17, wherein the object is a vehicle and the plurality of features includes at least one of a group of features consisting of edge, texture, contour and tail-lights of the vehicle.