基于平滑卷积定位的圆形印鉴提取方法研究

Extraction Method for Circular Seal Imprint Based on Smoothing Convolution Location

利用圆形印鉴中心的五角星所在区域红色像素分布集中的特点,提出了一种基于平滑卷积定位的圆形印鉴提取算法。该算法首先将彩色票据图像转换为灰度图像,接着利用平滑卷积定位五角星区域中的一个像素点,以这个像素点邻域的平均灰度值作为阈值二值化票据图像,然后搜索圆形印鉴边框,计算圆心坐标和半径,从而实现印鉴的提取。同时,为了提高提取速度,通过适当减小卷积过程中图像的分辨率,优化了整个提取过程中时间复杂度最高的平滑卷积算法。实验结果表明,这种方法能够在复杂背景下有效地提取出圆形印鉴,在经过算法优化后,提取速度提高了近12倍。

According to the characteristic of high concentration of red pixels in pentagram area, a new method of circular seal imprint extraction based on smoothing convolution location is proposed. Firstly, the colorful bill image is converted to grayscale image. Secondly, a pixel in pentagram area is located by smoothing convolution, and then the grayscale image is converted to binary image using the average of grayscale value of the neighborhood of the pixel as the threshold. Finally, the seal imprint is extracted after calculating the coordinate of the center and the radius of the circular seal imprint by searching its outmost frame. Meanwhile, in order to improve the extraction speed, the algorithm of smoothing convolution, with the highest time complexity in the whole extraction process, is optimized by reducing the image resolution appropriately during convolution. Experiments show that the method can effectively extract the circular seal imprint from complex background. And the computation speed increases by nearly 12 times after algorithm optimization.