VEHICLE DETECTION METHOD BASED ON HYBRID IMAGE TEMPLATE

• US 20160259981A1
• Filed: 06/28/2013
• Published: 09/08/2016
• Est. Priority Date: 06/28/2013
• Status: Active Grant

• Current Assignee: Institute Of Automation Chinese Academy Of Sciences (Chinese Academy of Sciences)
• Original Assignee: Institute Of Automation Chinese Academy Of Sciences (Chinese Academy of Sciences)
• Inventors: WANG, Feiyue, LI, Ye
• Granted Patent: US 10,157,320 B2
• Time in Patent Office: Days
• US Class Current: 1/1
• CPC Class Codes: G06T 2207/30236 Traffic on road, railway or crossing
• CPC Class Codes: G06T 2207/30252 Vehicle exterior; Vicinity of vehicle
• CPC Class Codes: G06T 7/74 involving reference images or patches
• CPC Class Codes: G06V 10/7515 Shifting the patterns to accommodate for positional errors
• CPC Class Codes: G06V 20/54 of traffic, e.g. cars on the road, trains or boats

Abstract

The invention discloses a vehicle detection method based on hybrid image template. This method consists of the three steps. Firstly, use no less than one vehicle image for template learning. Secondly, use information projection algorithm to learn a hybrid image template from the training images for vehicle object. The hybrid image template consists of no one less than image patch. Meanwhile, calculate the likelihood probability distribution of this template. Thirdly, use the learned HIT template to detect vehicle objects from testing images. The invention is suitable to detect vehicles with various vehicle shapes, vehicle poses, time-of-day and weather conditions. Besides vehicle localization, this method can also provide the detailed description of vehicle object.

Claims

1. A vehicle detection method based on hybrid image template HIT, comprising the following steps:

Step S1;

no less than one vehicle image is collected as a training image;

Step S2;

an information projection algorithm is utilized to learn all of image patches in the HIT for representing vehicle object from the training images and to compute image likelihood probability distribution of this hybrid image template;

Step S3;

the HIT learned from the step S2 is applied to detect vehicles from the input testing image and then to acquire the position of vehicles in the testing images;

the Step S3 further comprising the following sub-steps;

Step S31;

based on the HIT, the summation-maximization SUM-MAX operation is used to detect vehicle candidates with the maximum score from the input testing image, the Step S31 further comprising the following sub-steps;

Step S311;

a Gabor wavelets with no less than one orientation are utilized to filter the testing image, and then the sketch patches with these orientations are acquired;

Step S312;

the local maximization operation is applied to the sketch image to get a revised sketch image

Step S313;

the image patches in the HIT is used to filter the testing image and to detect vehicle patch candidates;

Step S3-1-4;

the local maximization operation is applied to the obtained vehicle patch candidates to get the revised vehicle patch candidates;

Step S3-1-5;

the revised vehicle patch candidates are merged according to their relative positions and scales in the HIT, and one or more vehicle candidate regions are generated from the testing image;

Step 3-1-6;

the patches in the HIT and the image likelihood probability are used to compute vehicle detection scores of the vehicle candidates region;

Step S3-1-7;

the vehicle candidate regions with the maximum vehicle detection scores is selected from all of the vehicle candidate regions; and

Step S32;

the maximum vehicle detection score is compared with a predefined vehicle detection threshold for detecting a vehicle object, and an iterative method is utilized to get all of vehicle objects in the testing image.

2. The method according to claim 1, wherein the HIT consists of one or more types of image patches, containing sketch patch, texture patch, color patch and flatness patch.

3. The method according to claim 2, wherein the sketch patch is modeled by a Gabor wavelet with one specific orientation, the texture patch is modeled by gradient histogram of local rectangular image region for the corresponding training image region, the color patch is modeled by color histogram of local rectangular image region for the corresponding training image region, and the flatness patch is modeled by superposition value of the responsive value of the Gabor filter for in one or more orientations in a local rectangular image region in the corresponding training image.

4. The method according to claim 3, wherein the gradient histogram for the texture patch is acquired by computing statistics of Gabor wavelet filter responses of local rectangular image region in the corresponding training image region;

The color histogram for the color patch is acquired by computing statistics of pixel values of the three color channels in color space of local rectangular image region; and

the flatness patch is acquired by summing Gabor filter responses of local rectangular image region with one or more orientations.

5. The method according to claim 1, wherein the likelihood probability of the HIT computed by the information project algorithm is

6. The method according to claim 1, wherein If the HIT contains the sketch patch, the texture patch, the color patch and the flatness patch, the step S313 further contains the following sub-steps:

Step S3131;

the color patches in the HIT are used to filter the testing image to detect and obtain candidate color patches. Local image regions in the testing image are detected as color patch candidates and these local image regions have the similar color feature with the color patches in the HIT; and

Step S3132;

the sketch patches, texture patches, and flatness patches in the HIT are utilized to filter the revised sketch image to detect and obtain sketch patch candidate, texture patch candidate and flatness patch candidates.

7. The method according to claim 1, wherein the number of Gabor orientation used in step S3-1-1 is no less than the number of the Gabor orientation applied to describing sketch patch in step S2, the number of filter orientation applied to compute texture patch in step S2, the number of Gabor orientation applied to compute flatness patch in step S2.

8. The method according to claim 1, wherein the vehicle detection score of vehicle candidate regions are calculated by using the following formula:

9. The method according to claim 1, wherein The step of S32 is particularly shown as follows:

the maximum vehicle detection score is compared with the predefined vehicle detection threshold;

if the maximum vehicle detection score is no less than the vehicle detection threshold, the corresponding vehicle candidate region is detected as a vehicle object;

then, the detected vehicle object is removed from the testing image and the remaining image is used to detect the remaining vehicle objects by performing the step S31 iteratively until the maximum vehicle detection score is less than the vehicle detection threshold and the iterative process is ended so that all of the vehicle objects in the testing image is obtained.

10. The method according to claim 9, the vehicle detection threshold is computed by the following steps:

Step S321, the vehicle region is detected from all of the training images according to the step S31, and the detection scores for the corresponding vehicle regions are calculated;

Step S322, the vehicle detection threshold is estimated by utilizing the detection scores of vehicle regions for all of the training images.