基于几何形态特征的鼠类分科自动识别技术研究

    袁雄峰1,胡红东2,曾罗1,何龙凉1,闫正跃1,朱佩思1,马晓光3

　　1. 防城港出入境检验检疫局, 广西 防城港  538001;  2. 中国人民解放军防化学院;3.中国检验检疫科学研究院

　　摘要:目的   应用计算机技术提取鼠类头骨标本数字图片几何形态特征, 探讨计算机自动分科鉴定的方法。方法   以鼠科、仓鼠科和松鼠科为研究对象, 选取13个鼠种, 用数码相机获取头骨标本正面数字图片, 结合图像处理和特征提取技术,测量其几何形态特征作为分类变量, 用逐步判别法筛选形态特征, 非参数法进行判别分析, 全部数据分析采用SAS 8.0软件。结果 逐步判别法筛选出12个形态特征作为分类变量:面积(x1)、周长(x2)、短轴长(x4)、偏心率(x6)、紧凑性(x7)、球状性(x8)、叶状性(x9)、圆形性(x10)、凸周长(x12)、和不变距hu1(x16)、hu3(x18)、hu6(x21)。非参数判别分析能够有效的对13种鼠类分科,回判的正确率为100%, 交叉判别的正确率为98.97%。结论   以鼠类头骨标本几何形态特征为分类变量,用非参数判别分析可以实现鼠类计算机自动分科鉴定。

　　关键词:几何形态特征; 鼠类; 头骨; 数字鉴定

　　中图分类号:R184.38   文献标识码:B

　　Study on automatic recognition of rodent in family level

　　based on geometric morphology

　　YUAN Xiong-feng*, HU Hong-dong, ZENG Luo, HE Long-liang, YANZheng-yue, ZHU Pei-si, MA Xiao-guang

　　*Fangchenggang Entry-Exit Inspection and Quarantine Bureau,Fangchenggang, Guangxi 538001, China

　　Abstract:   Objective   Extracting the geometricmorphology characters of rodent skull with computer visiontechnology, to explore the way of automatical identification ofrodent in family level.   Methods   13 species were choseamong Muridae, Circetidae and Sciuridae. Combining digital imageprocessing and mathematic morphological character extractingtechnology, mathematic morphological characters were extracted fromdigital images of rodent skulls. Var SAS V8.0 for windows, the datawas explored for automatic recognition with nonparametric stepwisediscriminatory method.   Results     Beforenonparametric discriminant analysis, 12 morphological characterswere selected by stepwise discriminant method: area (x1), perimeter(x2), short axis (x4), eccentricity (x6), compactness (x7),sphericity (x8), lobation (x9), circulatity (x10), bump area (x10),convex perimeter (x12) and arc length constraints hu1 (x16), hu3(x18), hu6 (x21). The results showed that all species could beeffectively identified in family level with nonparametricdiscriminant. The discriminant accuracy rate was 100% andCross-validated accuracy rate was 98.97%.   Conclusion  With geometric morphology characters of rodent skull, the automaticrecognition in family level is possible as nonparametricdiscriminant method was used..

　　Key words:   Geometric morphological characters; Rodent;Skull; Automatic recognition《中国国境卫生检疫杂志》2014年2期