Target identification is one of the main research content and also a key point in precision crop protection. The main purpose of the study is to choose the characteristic wavelengths (CW for short) to classify the cabbages and the weeds at their seedling stage using different data analysis methods. Using a handheld full-spectrum FieldSpec-FR, the canopies of the seedling plants, cabbage ‘8398, cabbage ‘zhonggan’, Barnyard grass, green foxtail, goosegrass, crabgrass, and small quinoa,, at three- & four-week growth were measured in the range of wavelength 350 ~ 2500nm.  In Unscrambler Data Analysis software system, the Principal Component Analysis (PCA) was applied respectively to extract CWs. Then plants were classified by means of Bayes discriminant analysis method with the chosen CWs as variable. The results showed that (1) According to the load factors and its changing rate of PCs corresponding to the spectral wavelengths, the CWs which were sensitive to plant identification were extracted respectively as 567, 667, 715, 1345, 1402, 1725, 1925, and 2015 nm for the first stage and 567, 667, 745, 1345, 1402, 1545, 1725, and 1925nm for the second stage. among the each 8 CWs of two stages, just two of the CWs were different, which indicated that the changes of spectral characteristics at different growth stages of cabbages have little influence on identification of cabbages and weeds. (2) The corresponding spectral data of the 8 CWs extracted from the data at the first stage were taken as the input variables of the model which was built up using Bayes discriminant analysis to classify two varieties of cabbages and five kinds of weeds. The correct classification rates for the training and testing sets were respectively 90.7% and 84.3%. When the two varieties of cabbages were regarded as the same category, using the analysis method the correct classification rates of the training and testing sets were respectively 95% and 100%, which indicated that different varieties of cabbages owned similar the spectral features.