基于特征点匹配的甲骨文识别

Oracle Bone Script Recognit ion Basedjon Feature Points

传统的甲骨文识别方法在识别精度、识别速度和抗干扰能力等方面还有进一步的发展空间,尤其是现有甲骨文识别技术对专家的依赖性高而甲骨文信息共享普及率则相对较低。本文通过实验模拟现实场景,采用特征不变匹配算法和编码法,从甲骨文单字图片每个像素的灰度值入手,以像素之间的变化关系作为图片的特征点进行甲骨文识别。无论对甲骨文原始单字图片进行旋转、缩放、对比度调整,还是电子设备噪声处理,其识别测试结果都非常好,其识别准确率为99%,识别精度相当高,这说明了将特征点匹配算法和编码法结合起来使用,对不同失真图像的识别能力很高,对甲骨文识别具有非常重要的作用。

Traditional Oracle Bone Script Recogmition(OBSR)has been subject to abundant research,recognition accuracy,speed and reliability can still be improved.In particular,the existing OBSR technology relies heavily on experts while Oracle Bone Inscription adoption rate and information sharing is relatively low.In this paper,we use SIF T feature invariant matching algorithm to simulate the real scene,Starting from the gray value of each pixel of Oracle Bone Inscription single-character image,we use the cha nging relationship between the pixels as the feature points of the image to recognize oracle bone inscription.Whether the oracle bone inscription single character image is rotated,scaled,cont rast adjusted and electronic equipment noise processed,the recognition test results are very good.Its recognition accuracy is 99%,and the recognition accuracy is quite high.This shows that SIFT algorithm has a high recognition ability for different distorted images,that is to say,the A established by SIFT algorithm.SIFT feature point Library of bone inscriptions plays an important role in Oracle inscription recognition.