为了解决数字图像亚像素级高精度尺寸检测问题, 采用了基于人工神经网络的线阵相机镜头畸变矫正方法, 比传统的多项式拟合法矫正精度高且操作简便; 提出了将光栅尺融入线阵相机扫描机构的新方法, 利用高精度光栅尺测位移信号触发线阵CCD图像采集动作, 精确获取线阵相机扫描目标的亚像素级图像。实验证明, 新的测量技术可在对运动系统要求不高的情况下获取高精度图像, 图像检测精度可以达到1μm。

In order to solve the problem of high-precision dimension detection at sub-pixel level in digital images, a linear array camera lens distortion correction method based on artificial neural network is adopted, which is higher in accuracy and easier to operate than the traditional polynomial fitting method. A new method of integrating the grating ruler into the scanning mechanism of the linear array camera is proposed, which uses the displacement signal measured by the high-precision grating ruler to trigger the image acquisition action of the linear array CCD to accurately acquire the sub-pixel image of the scanning target of the linear array camera. Experiments show that the new measurement technology can obtain high-precision images with low requirements on the motion system, and the image detection accuracy can reach 1μm.