基于BP神经网络的数字图像相关非迭代灰度梯度算法

为了提高非迭代灰度梯度算法的亚像素位移求解精度与效率,提出一种基于神经网络的非迭代灰度梯度改进算法。该算法应用BP神经网络直接建立变形前后散斑图子图像的灰度及灰度梯度与亚像素位移之间的非线性映射关系,无需对相关系数进行最小二乘优化求解。应用模拟散斑图像对算法的亚像素位移求解精度与效率进行了验证。结果表明,作者提出的算法的计算精度与效率较传统的非迭代灰度梯度算法均有提高。最后进行了真实的刚体平移实验,其结果进一步证明了作者提出的算法的有效性。

Non-iterative Gray-gradient Algorithm Based on BP Artificial Neural Network in Digital Image Correlation

In order to improve the accuracy and efficiency of solution of the subpixel displacement measurement of non-iterative gray-gradient algorithm in digital image correlation, an improved algorithm based on artificial neural network is proposed. The Back Propagation (BP) artificial neural network was employed to build directly the nonlinear mapping relationship between the gray and gray-gradient in the subimages of speckle images before and after deformation and subpixel displacements, without the need for least square analytical optimal solution of correleation coefficient. Speckle images were simulated and applied to verify the accuracy and efficiency of solution of the Subpixel displacement measurement of the algorithm. The results reveal that the proposed algorithm has higher computation accuracy and efficiency than traditional non-iterative gray-gradient algorithm. The validity of the proposed algorithm is verified further by rigid body translation experiment finally.