一种新的图像识别算法.pdfVIP

第31 卷 第9 期 计 算 机 工 程 2005 年5 月 Vol.31 № 9 Computer Engineering May 2005 ·人工智能及识别技术· 文章编号：1000—3428(2005)09—0173—03 文献标识码：A 中图分类号：TP391.41 一种新的图像识别算法 薛明东，郭 立，张国宣，刘士建 （中国科技大学电子科学与技术系，合肥 230026） 摘 要：提出了一种图像的识别算法，首先对图像进行Gabor 小波变换，为了解决图像中的平移、旋转、尺度不变性，以及局部遮挡问题, 对滤波结果计算其4 个不变矩作为反映整体形状特征的特征向量,再对提取的特征向量利用SVM 算法进行图像检测，检测结果表明该算法 具有较好的抗噪性并能解决目标检测中的遮挡问题。最后利用弹性网格的方法对其进行识别。文中的检测和识别结果表明，多通道的Gabor 滤波器对于纹理图像的特征描述比较充分，该识别算法有比较理想的鲁棒性和容错性,能得到较好的识别结果。 关键词：Gabor 小波变换；支持向量机；弹性匹配 A New Method of Image Recognition XUE Mingdong , GUO Li , ZHANG Guoxuan , LIU Shijian （Department of Electronic Science and Technology, University of Science and Technology of China, Hefei 230026） 【Abstract 】The paper introduces a system of image recognition, first abstract image characters based on Gabor transform, the gabor wavelet is a kind of multi-resolution representation of images. Using Gabor magnitude, Gabor frequency, and Gabor phase, the object image can be represented. Then using the SVM classification to check the plane images, the test shows that the arithmetic can solve the problem of image noise and image cover, and using SVM classification can be expanded simply with the target image.The image recognition result finally is got by using the flexible grid method. The recognition result shows that this Gabor wavelet transform can describe the image character more fully, and this recognition system has better robust and error-tolerated to a certain extend, and it can get the better classification and recognition results. 【Key words 】Gabor wavelet transform；Support vectors machines；Flexible marching 在图像处理与计算机视觉领域，图像识别是一项困