分类算法最好的评价指标是分类误差,在综合常用舌色分类算法和在不同的颜色空间中提取舌色特征的基础上,基于360张舌图片,考查了常用的9个3维颜色空间和18个2维色度平面,采用交叉验证的方法对舌色分类器的性能进行评价,形成了一个系统的舌色色度空间的分类性能评价体系,为选择合适的颜色空间来提取特征和选择分类性能较强的分类算法对舌色分类提供了依据。结果表明,在知觉颜色空间、感知均匀空间内和包含亮度信息的2维色度平面上提取舌色特征对于舌色分类具有较好的分类性能。
Classification error is the best evaluated performance: Using 360 images and the performance of tongue classification is researched by cross validation method compared 9-3D color-spaces and 18-2D chrominance planes widely used in methods of tongue classification. The system of tongue-color classification performance is built to choose proper color-space to detect character and choose better classify arithmetic. The result shows that the better classify performance of tongue-color is in aesthesia color-space, apperceive color-space and 2D chrominance planes include luminance.