铸币图像的配准是基于图像识别的真伪币鉴别及瑕疵检测的重要步骤.针对铸币图像的特点,提出使用边缘纹理替代灰度信息作为图像配准的特征;配准算法使用极坐标系投影等多种积分投影方法,通过计算样本图像与模板图像投影曲线的互相关函数与最大互相关参数得到平移、旋转、缩放参数,并以多次迭代,实现铸币样本图像与模板图像的精确配准,为基于图像识别的真伪币鉴别及瑕疵检测提供良好基础.
Mint image registration is an important step of the authenticity identification and defect detection of valued coins based on image recognition. On account of the specific of mint image, the texture image formed by edges is proposed to substitute the grayscale as the characteristic for image registration. The proposed regis- tration algorithm utilizes various integral projection methods including polar coordinates projection. The translation, rotation and scale parameters are obtained by calculating the crosscorrelation functions and its maximum parameters of the projection profiles of the sample image and template image. This process is iterated multiple times to achieve accurate registration. The proposed registration algorithm lays a sound foundation for further identification of authenticity and defect detection of valued coins.