基于傅里叶描述子的轮胎标识点形状识别

Shape Recognition of Tire Marking Points Based on Fourier Descriptors

针对人工识别轮胎标识点形状效率低、误差大等问题,提出了一种基于傅里叶描述子的轮胎标识点形状识别算法。首先对采集的轮胎标识点图像进行降噪、分割、轮廓提取等图像预处理操作;然后通过对预处理后的轮胎标识点图像进行傅里叶变换,提取轮胎标识点轮廓的傅里叶描述子系数;最后通过计算待识别轮胎标识点图像的傅里叶描述子系数,与轮胎标识点模板库中图像的傅里叶描述子系数的欧氏距离数值,其中与最小欧氏距离相对应的模板库形状即为待识别轮胎标识点轮廓的近似形状,从而实现轮胎标识点的形状识别。实验选取圆形、方形、菱形以及十字形,四种共计200幅轮胎标识点形状图像,进行标识点形状识别准确率测试。实验结果表明,该算法能准确地识别出轮胎标识点形状,四种轮胎标识点形状的平均识别准确率为97.25%,其中圆形和方形轮胎标识点的形状识别准确率达98%。

In order to solve the problem of low efficiency and high error rate in shape recognition of tire marking points manually,a method of shape recognition of tire marking points based on Fourier descriptors is proposed. Initially,the image preprocessing methods i. e image de-noising,segmentation,boundary extraction will be applied on tire marking points images before shape recognition of tire marking points; then the Fourier transform will be applied on those tire marking points,which have been disposed by methods of image preprocessing. In this way the Fourier descriptors of tire marking points can be extracted. Lastly,the Euclidean distance between the tire marking points of under checking and the ones in the standard template library will be calculated. The shape of tire marking points in template library with the minimum Euclidean distance is corresponding to the one which is under checking. In this way the shape of tire marking points can be recognised. In the experiment of shape recognition of tire marking points,200 tire marking points images is selected,which contain circular,square,rhombus and crisscross four different shapes. Experiment results show that the proposed method can recognize the shape of tire marking points effectively. The average shape recognition accuracy of four different tire marking points as 97. 25%. Especially,the accurate shape recognition rate of circular and square reach to 98%.