A method of recognition weeds during cotton emergence based on principal component analysis (PCA) and support vector machine (SVM) was developed. For the effective classification of the variety of weeds in cotton field, the dimension of feature variable space was reduced by extracting color, shape, texture characteristics and principal component analysis. The experiment of classification for 120 samples of cottons and weeds showed that it was able to reduce training time and increase classification accuracy effectively by the first three principal components obtained by PCA dimensionality reduction. It was found by comparison that the best classification and recognition result was obtained by using the combination of the first three principal components and RBF kernel function SVM. The training time is 91ms and the average correct classification rate is 98.33%.