Technical Abstract: It is important to find an appropriate pattern-recognition method for in-field plant identification based on spectral measurement in order to classify the crop and weeds accurately. In this study, the method of Support Vector Machine (SVM) was evaluated and compared with two other methods, Decision Tree (DT), and Radial Basis Function Neural Network (RBF-NN). The three methods were used to model and classify the spectral data measured on plants in the corn fields. The spectral measurement was conducted over the fields covered by seedling corns, and two kinds of weeds, Dchinochloa crasgalli, and Echinochloa crusgalli weeds. Among the measured data, respectively 48 corn samples (category 1), 33 Echinochloa crusgalli samples (category 2), and 33 Dchinochloa crasgalli samples (category 3) were randomly selected as the sample data. The sample data were divided into a training data set (2/3 of the data) and a testing data set (1/3 of the data). The SVM method provided an 81.58% correct classification rate of the corn and two weeds with the testing data set while the methods of DT and RBF-NN provided the rates of 63.16% and 52.63%, respectively. This indicates that among the three methods, SVM has the highest accuracy in identification of corn and weeds in the fields in the case of limited samples. SVM could provide a method to build a real time tool to identify crop and weeds with high accuracy in practice.