一种运用自聚类方法优化椭圆图像识别的方法

A method of optimizing ellipse images recognition by automatic clustering

在高分辨率图像中,为获得较高的椭圆目标识别效率与拟合精度,提出了一种利用自动聚类技术优化椭圆识别和拟合的方法。本方法先对原始图像进行降采样;在缩减图的颜色空间中进行kmeans自动聚类并进行原始图像的预分割;经图像连通性分析后,分离出若干目标区域;再利用各目标区域的边界信息拟合椭圆;最后依次经边界长度约束、椭圆方程参数约束、椭圆拟合精度误差指标正确识别出椭圆目标物,并获得较高精度的椭圆方程。实验表明,该方法能够自适应图像质量的较大范围波动,具有较高的识别正确率和拟合精度,同时也兼顾了识别速度,在工业视觉测量领域具有一定的理论和实用价值。

For high-resolution images,in order to get high efficiency and fitting error of ellipse recognition,a method of optimizing ellipse recognition and fitting by automatic clustering is proposed. Down sampling of the original image is the first step ;then kmeans automatic clustering and pre-segmentation of the original image in the color space of reduced image is conducted ;after image connectivity analysis,target areas are separated and the boundary information of each target area are used to fit the ellipse ;the ellipse object is finally recognized correctly by boundary length constraint, elliptic equation parameter constraints and the indicators of ellipse fitting precision error,with a high accuracy elliptic equation obtained.It is proved by experiments that this method could adapt to the wide fluctuation of the quality of images,and has a high recognition correct rate and fitting accuracy.The speed of recognition is taken into account too.This method has theoretical and practical value in the field of industrial vision measurement.