在线阵 CCD检测系统中, 图像的边缘检测是关键技术之一。传统阈值比较法容易受到随机噪声的干扰, 为了减少这种干扰, 提出了一种基于 Sigmoidal函数最小二乘拟合的一维图像边缘检测方法。利用 Sigmoidal函数对图像边缘区域内所有点的灰度信息进行最小二乘拟合, 能够有效避免单个信号点上的噪声对测量结果的影响。对拟合函数一阶求导, 并寻找其一阶导数最大值对应点, 该点即为图像的边缘位置。实验分别采用阈值比较法和基于 Sigmoidal函数最小二乘拟合法, 对采集图像的边缘位置检测结果进行比较, 实验结果表明, 基于 Sigmoidal函数拟合方法稳定性好, 具有较高重复精度, 能够有效抑制系统随机噪声的影响, 具有较好的检测效果。

The edge detection of one-dimension image is the key component in linear CCD detection system. In order to reduce the effect of random noise in the traditional threshold comparison method, a new method is proposed based on the Sigmoidal function fitting. In this method, the Sigmoidal function is regarded as the fitting model of the image edge, the gray value of the image edge area is fitted to the model in the principle of least-squares. The location of the edge point in the image is obtained by calculating the maximum point of the ordinary derivative of the fitted function. An experiment is designed and the image edge point is detected with the threshold comparison method and the Sigmoidal function fitting method respectively. The experiment results show that the Sigmoidal function fitting method has the advantages of high repeat accuracy, high stability, and suppressing the influence of random noise.