STEREO VISION SYSTEM AND CONTROL METHOD THEREOF

• US 20100066811A1
• Filed: 01/29/2009
• Published: 03/18/2010
• Est. Priority Date: 08/11/2008
• Status: Active Grant

• Current Assignee: Electronics and Telecommunications Research Institute
• Original Assignee: Electronics and Telecommunications Research Institute
• Inventors: Choi, Jae Il, Hwang, Dae Hwan, Choi, Seung Min, Chang, Ji Ho
• Granted Patent: US 8,098,276 B2
• Time in Patent Office: Days
• US Class Current: 348/43
• CPC Class Codes: G06T 2207/10012 Stereo images
• CPC Class Codes: G06T 5/40 using histogram techniques
• CPC Class Codes: G06T 7/593 from stereo images
• CPC Class Codes: H04N 13/133 Equalising the characteristics of different image components, e.g. their average brightness or colour balance
• CPC Class Codes: H04N 13/239 using two 2D image sensors having a relative position equal to or related to the interocular distance H04N13/243 takes precedence
• CPC Class Codes: H04N 13/246 Calibration of cameras

Abstract

Provided are a stereo vision system and a control method thereof. A stereo vision system includes an image information extracting unit receiving left and right images of the left and right stereo cameras to extract color information for a brightness control of the images from the received images, an image preprocessing unit performing a process for reducing noises of the left and right images using the color information and a prestored calibration parameter, a stereo matching unit performing stereo matching of the left and right images processed by the image preprocessing unit through an algorithm to obtain a depth map, and a matching result measuring unit receiving the depth map obtained by the stereo matching unit to measure a matching degree, and changing the prestored calibration parameter according to a result of the measurement.

Claims

1. A stereo vision system comprising left and right stereo cameras, the stereo vision system comprising:

an image information extracting unit receiving left and right images of the left and right stereo cameras to extract color information for a brightness control of the images from the received images;

an image preprocessing unit performing a process for reducing noises of the left and right images using the color information and a prestored calibration parameter;

a stereo matching unit performing stereo matching of the left and right images processed by the image preprocessing unit through an algorithm to obtain a depth map; and

a matching result measuring unit receiving the depth map obtained by the stereo matching unit to measure a matching degree, and changing the prestored calibration parameter according to a result of the measurement.

2. The stereo vision system of claim 1, wherein the image preprocessing unit comprises:

a smoothing unit generating a histogram of the images through the extracted color information, smoothing the generated histogram, and equalizing average values of the left and right images; and

an image rectifying unit performing image rectification which makes an epipolar line of the left image in accordance with an epipolar of the right image using the prestored calibration parameter.

3. The stereo vision system of claim 2, wherein the smoothing unit calculates a normalized sum of the histogram generated upon the smoothing of the histogram, and again modifies the input images based on the calculated normalized sum to generate a result image.

4. The stereo vision system of claim 3, wherein a maximum brightness value of the color information is divided by the total number of pixels in the input images, and the normalized sum is obtained from a multiplication of the divided value and a cumulative sum of an input image histogram.

5. The stereo vision system of claim 1, wherein the matching result measuring unit comprises:

a Fourier transformer performing a Fourier transform of the depth map to change an entire image into a frequency domain;

a high pass filter detecting a high frequency signal from an output of the Fourier transformer, and outputting the detected signal; and

a result outputting unit outputting a coefficient value representing a matching degree using the high frequency signal and a prestored look up table.

6. The stereo vision system of claim 5, wherein the matching result measuring unit further comprises a parameter learning unit learning the calibration parameters for the coefficient value to converge to a setting value.

7. The stereo vision system of claim 6, wherein the parameter learning unit repeatedly performs learning for the coefficient value being a variable value to converge to the setting value.

8. The stereo vision system of claim 5, wherein the matching result measuring unit performs a learning process when the coefficient value is more than the setting value.

9. A method for controlling a stereo vision system comprising left and right stereo cameras, the method comprising:

extracting color information for a brightness control of an image when left and right images of the left and right stereo cameras are input;

generating a histogram of the images through the extracted color information, smoothing the generated histogram, and equalizing average values of the left and right images;

performing image rectification which makes an epipolar line of the left image in accordance with an epipolar of the right image using a prestored calibration parameter after the generating and smoothing of the histogram;

performing stereo matching of the left and right images processed in the generating and smoothing of the histogram and the performing of the image rectification through an algorithm to obtain a depth map; and

receiving the obtained depth map to measure a matching degree, and changing the prestored calibration parameter according to a result of the measurement.

10. The method of claim 9, wherein the generating and smoothing of the histogram comprises calculating a normalized sum of the histogram generated upon the smoothing of the histogram, and again modifying the input images based on the calculated normalized sum to generate a result image.

11. The method of claim 10, wherein a maximum brightness value of the color information is divided by the total number of pixels in the input images, and the normalized sum is obtained from a multiplication of the divided value and a cumulative sum of an input image histogram.

12. The method of claim 9, wherein the receiving of the depth map and the changing of the prestored calibration parameter comprises:

performing a Fourier transform of the depth map to change an entire image into a frequency domain;

detecting a high frequency signal from the frequency domain, and outputting the detected signal; and

outputting a coefficient value representing a matching degree using the high frequency signal and a prestored look up table.

13. The method of claim 12, wherein the receiving of the depth map and the changing of the prestored calibration parameter further comprise learning calibration parameters for the coefficient value to converge to a setting value.

14. The method of claim 13, wherein the learning of the calibration parameters comprises repeatedly performing learning for the coefficient value being a variable value to converge to the setting value.

15. The method of claim 13, wherein in the receiving of the depth map and the changing of the prestored calibration parameter, a learning process is performed when the coefficient value is more than the setting value.