珠海口岸常见鼠形动物外形的模式识别研究

目的探索鼠形动物鉴别较为客观、高效的方法,以作为专家经验的补充。方法对珠海口岸常见的4种成年鼠形动物的5个外形指标(体重、体长、尾长、耳长、后足长)进行分析。以统计学方法和神经网络方法进行模式识别的试探。结果分别以全模型法(强迫引入法)和逐步选择法建立分类函数,并建立BP神经网络来对种类进行分类鉴定,用训练样本数据进行回判时,分别得到96.0%、96.5%、95.5%的正确率,3种方法的正确率无显著性差异。结论模式识别方法能较好地鉴别珠海口岸的常见鼠形动物。

Pattern recognition of the common rat-shaped animals at Zhuhai port

Objective To find a good and effective method for the discrimination of the rat-shaped animals,which should be as an experimental supplement for our work.Methods The common rats at Zhuhai port were searched and analysed by methods of statistics and neural network about the body weight,body length,tail length,ear length and pes length.Results According to the different methods of enter independent together discriminant mothod,stepwise mothod and back propagation NN,96.0%,96.5% and 95.5% of the original grouped cases were correctly classified by emulating the recognition respectively.Conclusion The common rat-shaped animals at Zhuhai port could be discriminated correctly by pattern recognition methods.