基于傅立叶描述子的中长跑自动计圈技术研究

Study on Auto-counting for Middle-distance Race Based on Fourier Descriptor

为了识别号码布图像中的三个数字,基于傅立叶描述子建立数字0～9的Bayes线性判别模型,从号码布图像中提取数字二值图像。用8-连接跟踪法获取数字区域边界轮廓的坐标序列,经归一化、重采样后获取的傅立叶描述子具有平移、缩放和旋转不变性,其低频分量对数字的区分度较好。Bayes判别分析结果表明:所建立的数字0～9的判别模型的内部验证的识别率达100%。基于图像处理技术识别号码布数字是可行的,为研制中长跑自动计圈系统提供了方法。

In order to distinguish three digits from a number cloth image, Bayes-criterion-based linear discrimination models of 0 through 9 were developed based on Fourier descriptors. A digit binary image was extracted from a number cloth image based on image processing. A sequence of coordinates of digit region boundary was calculated by using 8-link tracing arithmetic, Fourier descriptors are shift, scaling and rotation invariant, which were calculated from normalized and re-sampling of the sequence of coordinates, and their low-frequency components were separable to the digits from 0 to 9. Bayes discrimination results： showed that discrimination models of 0 through 9 were developed with an accuracy of 100% based on re-substitution. It is feasible to recognize digits of number cloth by using image processing, which helps to develop auto-counting system for middle-distance race.