The present invention relates to a stereo vision image calibration method and a related image capturing device, and more particularly, to a stereo vision image calibration method and a related image capturing device capable of automatically repairing deviation of its image capturing unit.
With the advanced technology, the stereo vision camera device is widespread applied in daily life. The stereo vision camera device may have two image capturing units disposed on different position, and the image capturing units provide two images having different visual angles for a stereo vision image computation, such as a depth map. In an installing process of the stereo vision camera device, the two image capturing units may have position offset due to unpredictable impact, operational error, unstable installation of the base, or thermal expansion and contraction resulted by temperature variation. As position of the image capturing units are deviated, the stereo vision image computation based on the images captured by the image capturing units becomes abnormal. Thus, design of a stereo vision camera device and a related stereo vision image calibration method capable of automatically repairing deviation is an important issue in the monitoring industry.
The present invention provides a stereo vision image calibration method and a related image capturing device capable of automatically repairing deviation of its image capturing unit for solving above drawbacks.
According to the claimed invention, a stereo vision image calibration method is applied to an image capturing device having a first image capturing unit, a second image capturing unit and an image rectify unit, the first image capturing unit and the second image capturing unit are arranged along a first direction and utilized to respectively capture a first image and a second image, and the image rectify unit is adapted to rectify the first image and the second image by a stereo vision parameter built in production to generate a rectified first image and a rectified second image. The stereo vision image calibration method includes setting the rectified first image as a reference image and setting the rectified second image as an adjusting image, computing a global disparity vector between the adjusting image and the reference image, and processing a stereo vision calibration according to a vector component of a second direction different from the first direction of the global disparity vector while the vector component has a non-zero value.
According to the claimed invention, an image capturing device with a stereo vision image calibration function includes a first image capturing unit, a second image capturing unit, an image rectify unit and an operational processor. The first image capturing unit is adapted to capture a first image. The second image capturing unit is adapted to capture a second image having a visual angle different from the first image. The image rectify unit is electrically connected to the first image capturing unit and the second image capturing unit, and adapted to rectify the first image and the second image as a rectified first image and a rectified second image. The operational processor is electrically connected to the image rectify unit and adapted to set the rectified first image as a reference image and set the rectified second image as an adjusting image, to compute a global disparity vector between the adjusting image and the reference image, and to process a stereo vision calibration according to a vector component of a second direction different from the first direction of the global disparity vector while the vector component has a non-zero value.
The stereo vision image calibration method of the present invention utilizes the first image and the second image respectively captured by the first image capturing unit and the second image capturing unit for rectification to compute the disparity motion vector and the similar value so as to find out the global disparity vector, and then the cropping range of the image capturing units or the stereo vision parameter can be varied to calibrate distortion of the stereo vision image resulted by position deviation about the two image capturing units of the image capturing device, to ensure that the image capturing device can provide the high quality stereo vision image since the two images captured by the two image capturing units from different visual angles after calibration are highly overlapped. Comparing to the prior art, the present invention can effectively solve the distortion of the stereo vision image resulted by an installation fault of the image capturing device, so as to automatically repair the said fault and to prevent unpredictable risk such as environmental weather, so the image capturing device can be kept in best state at any time.
These and other objectives of the present invention will no doubt become obvious to those of ordinary skill in the art after reading the following detailed description of the preferred embodiment that is illustrated in the various figures and drawings.
Please refer to
As the first image capturing unit 12 and the second image capturing unit 14 are arranged along the second direction D2, which is not shown in figures, and arrangement of the first image capturing unit 12 and the second image capturing unit 14 along the first direction D1 is slightly deviated relative to the original production design, the operational processor 16 can execute the stereo vision image calibration procedure of the present invention to repair and acquire the correct stereo vision image computation result. The image rectify unit 15 rectifies the first image I1 and the second image I2 via the stereo vision parameter built in the production, and the rectified first image I1′ and the rectified second image I2′ can be located at the same coordinate plane and aligned with the second direction D2. For example, assume the second direction D2 is the vertical direction, the rectified first image I1′ contains an object (such as the table) having a corner point (such as at one of the four table corners), coordinates of the said corner point is (X1, Y1), coordinates of the same corner point (which means at the same table corner) of the object (which means the same table shown in the rectified first image I1′) inside the rectified second image I2′ is (X2, Y2), then X1 equals X2.
Please refer to
Following statement has an example of the disparity motion vectors (m1, n1) and (m2, n2), as shown in
It should be mentioned that a part of pixels on the adjusting image (which means the rectified second image I2′) may be not overlapped with the reference image (which means the rectified first image I1′) while the adjusting image is moved relative to the reference image within the searching range M. Please refer to
The present invention sets the searching range and the similar value computation to determine the global disparity vector, but the present invention is not limited by it. The present invention can determine the global disparity vector based on image correlation measurement or characteristic point computation. The said image correlation measurement can be referred to the webpage: https://siddhantahuja.wordpress.com/2010/04/11/correlation-based-similarity-measures-summary/. The said characteristic point computation about the global disparity vector can be referred to the technical literature: Liu, K. H., Liu, T. J., and Liu, H. H. A SIFT descriptor based method for global disparity vector estimation in multiview video coding, 2010 IEEE International Conference on Multimedia and Expo (ICME), IEEE, pp. 1214-1218.
In step 404, the stereo vision image calibration method can vary a cropping range of the first image capturing unit 12 or the second image capturing unit 14 to overcome the deviation of ΔD; for example, while the stereo vision image calibration is functioned according to the vertical vector component (ny) of the global disparity vector (mx, ny) and the said vertical vector component (ny) is equal to 3, the cropping range of the first image capturing unit 12 or the second image capturing unit 14 can be varied to vertically shift S pixels, and the amount of S is the amount of deviated pixels of the first image capturing unit 12 or the second image capturing unit 14 corresponds to 3 pixels in the rectify unit 15; further, the stereo vision parameter of the image rectify unit 15 can be varied to overcome the deviation of ΔD. In addition, as shown in
The above-mentioned embodiment sets the rectified first image I1′ as the reference image and sets the rectified second image I2′ as the adjusting image; however, the present invention can alternatively set the rectified first image I1′ as the adjusting image and sets the rectified second image I2′ as the reference image.
In conclusion, the stereo vision image calibration method of the present invention utilizes the first image and the second image respectively captured by the first image capturing unit and the second image capturing unit for rectification to compute the disparity motion vector and the similar value so as to find out the global disparity vector, and then the cropping range of the image capturing units or the stereo vision parameter can be varied to calibrate distortion of the stereo vision image resulted by position deviation about the two image capturing units of the image capturing device, to ensure that the image capturing device can provide the high quality stereo vision image since the two images captured by the two image capturing units from different visual angles after calibration are highly overlapped. Comparing to the prior art, the present invention can effectively solve the distortion of the stereo vision image resulted by an installation fault of the image capturing device, so as to automatically repair the said fault and to prevent unpredictable risk such as environmental weather, so the image capturing device can be kept in best state at any time.
Those skilled in the art will readily observe that numerous modifications and alterations of the device and method may be made while retaining the teachings of the invention. Accordingly, the above disclosure should be construed as limited only by the metes and bounds of the appended claims.
Number | Date | Country | Kind |
---|---|---|---|
105132338 | Oct 2016 | TW | national |