从电场角度引出的基于边缘特征的归一化形状识别算法研究

为了从一新的角度来解决图像形状识别问题，提出了一种从电场角度引出的基于边缘特征归一化的形状识别算法，即在图像处理过程中，像素将被赋予电量，各个图像的形状信息是用带电像素在三维空间产生的电势与电场强度的值来表征。该算法处理都是基于二值图像，并先通过边缘检测技术找出图像边缘信息，然后通过对边缘进行多边形逼近来得出边缘的角点信息，最后再对归一化的观测点进行电势与电场强度的计算。由于一个图形存在着无穷的观测点，因此在实际运用中合理选择观测点是重要且必要的，因为它能更加合理地反映图形的形状信息，故更加适合计算机对图形形状的判断。实验表明，当被识别的二值图像发生旋转、位移、变形时，该方法都能取得比较好的识别效果。

Educing a New Shape Identify Method Form Position of Electric Field

This paper will resolve shape identify from a new position. In course of image processing, the pixels are endowed with electric quantity. Every shape of image is indicated by electric field strength and electric potential in 3-D space around the image. In this paper, the processing is based on binary image. Firstly, edge of the image is got by edge detection techniques and then corner points by polygonal approximation. Lastly, electric field strength and electric potential are calculated for normalizing observation points. The calculated value will indicate different images. It exists infinite observation points for one image. Therefore it's important and necessary to choose appropriate observation points in practice. Because they indicate the shape information more reasonably and fit computer for judging. The algorithm derives from physics and easily be understood in physics. The paper give necessary deduction and experiments, when the images are revolved,moved and distorted ,the method also could show a better result in recognition. The key of following work is to resolve how to choose observation points, so that recognition result will be more precise and robust and can be competent for complex images.