Paphiopedilum species are known as unique flower shape， gorgeous color， long ornamental period and have great ornamental and economic value. The classification of the genus is controversial. The diversity and correlation analysis of 27 phenotypic traits were carried out by investigating 29 species of Paphiopedilum， and cluster analysis was carried out based on the principal components of phenotypic traits. The results showed that the phenotypic traits of Paphiopedilum were rich in diversity. The variation range of genetic diversity index （H′） of 17 quantitative characters was 0.7834-2.0318. Among them， the strong variation characters with genetic diversity index greater than 2.0 were flower length and flower width， the weak variation character with genetic diversity index less than 1.0 was flower number， and the variation range of genetic diversity index of 10 quality characters was 0.5098-1.1241. The diversity indexes of petal shape and Lip main color were the highest， both exceeding 1.0， and the smallest were staminode bottom concave and anther type，from the results of diversity analysis， it can be seen that 29 species of Paphiopedilum have large differences in phenotypic traits among species， with rich diversity. The coefficients of variation （CV） of 17 quantitative traits ranged 18.22%-59.09%， the petal length/petal width and peduncle length exceeded 50%， the results showed that the interspecific phenotypic characters of Paphiopedilum palustratum were dispersed greatly.Correlation analysis showed that the phenotypic characters of Paphiopedilum were closely related and complex. Five principal components with eigenvalues greater than 1 were selected by principal component analysis， and the cumulative contribution rate was 84.176%， which could reflect most of the information of the phenotypic characters of Paphiopedilum， and the characters and indexes of flower organs had a great impact on the phenotypic diversity. Cluster analysis showed that 29 species of Paphiopedilum were divided into 6 groups， which was similar to other classification methods， but there were also some differences.