Chinese Journal of Lasers, Vol. 44, Issue 8, 804003 (2017)
A Stereovision Measurement Method Using Epipolar Constraint to Correct Digital Image Correlation Matching
Baohua Shan1,2,3,*, Xiaoyang Huo1, and Yang Liu1
- 1[in Chinese]
- 2[in Chinese]
- 3[in Chinese]
Objective: As far as image matching, the widely used DIC method is essentially a kind of area matching algorithm. However, when the DIC method is used to perform stereo matching, the epipolar constraint which is a very important constraint isn’t adopted to restrict selection of the corresponding matching points on right image. To deal with this problem, a stereovision method considering epipolar constraint correction of DIC is proposed to improve measurement precision.
Methods: The epipolar constraint is introduced to correct the DIC matching result on right image. Firstly, the DIC method is used to obtain the image coordinates of initial matching point on right image at the initial time. Secondly, the fundamental matrix is calculated according to the calibration results of stereovision measurement system, and the equation of right epipolar line can be acquired according to the image coordinates of the corresponding point on the initial left image and the fundamental matrix. Then, the foot of the perpendicular from the matching point on the initial right image to the right epipolar line can be acquired, which is obtained by DIC method. This foot located on the right epipolar line satisfies the epipolar constraint, and is the nearest point from the matching point obtained by DIC on the initial right image. Therefore, this foot is regarded as the matching result through epipolar constraint correction. Finally, this correction method considering epipolar constraint is extended to sequential matching of right images.
Results: The uniaxial tension test of 304 stainless steel specimens was performed on universal testing machine. The measurement results of extensometer are regarded as the true values in the present paper, the displacement error curves obtained by two stereovision methods before and after the epipolar constraint are compared to testify the effectiveness of the proposed stereovision method. Experimental results give the following conclusions: 1. Two displacement curves measured by the stereovision methods have the same trends with that of the extensometer. In contrast to the traditional stereovision method, the displacement curves obtained by the proposed stereovision method considering the epipolar constraint is more close to that of the extensometer. 2. The displacement errors obtained by the proposed stereovision method are less than that of the traditional stereovision method which does not employ epipolar constraint. 3. The average error, the maximum error and the standard deviation of displacements obtained by the proposed stereovision method considering epipolar constraint are less than that of the traditional stereovision method which doesn’t consider epipolar constraint. Among of them, the amplitude of average error falls up to 42.1%, the maximum error decreases to 36.5%, and the standard deviation decreases to 8.6%.
Conclusions: The proposed stereovision measurement method employs the epipolar constraint to correct the DIC matching results, the nearest point located on the right epipolar line to DIC matching result is selected as the new matching point on the initial right image. Moreover, this correction method considering epipolar constraint is extended to sequential matching of right images. Results of the tension test show that displacement errors obtained by the proposed stereovision method can be reduced through epipolar constraint correction, this suggests that the measurement precision of the proposed stereovision method is improved in comparison with the traditional stereovision method. The proposed stereovision method combines the DIC method and the epipolar constraint, and corrects the DIC matching results by using the constraint of the minimum geometric distance. The proposed stereovision method has a clear concept and simple calculation, improves the measurement accuracy, and will have a practical value in vision measurement field.
Please Enter Your Email: