Identifying plant pathogens for disease diagnosis and disease control strategy making is of great significance. In this study, based on near-infrared spectroscopy, a method for identifying three kinds of pathogens causing wheat smuts, including Tilletia foetida, Ustilago tritici, and Urocystis tritici, was investigated. Based on the acquired near-infrared spectral data of the teliospore samples of the three pathogens, pathogen identification models were built in different spectral regions using distinguished partial least squares (DPLS), backpropagation neural network (BPNN), and support vector machine (SVM). Satisfactory identification results were achieved using the DPLS, BPNN, and SVM models built in each of the 22 spectral regions. By contrast, the modeling effects of DPLS and SVM were better than those of BPNN. The modeling ratio of the training set to the testing set affected the identification results of the BPNN models more than those obtained using the DPLS and SVM models. In this study, a rapid, accurate, and nondestructive method was provided for plant pathogen identification, and some basis was provided for disease diagnosis, pathogen monitoring, and disease control. Moreover, some methodological references and supports were provided for identification of quarantine wheat smut fungi in plant quarantine.