The invention relates to the precision agriculture field and the unmanned aerial vehicle agricultural application field, especially to an aerial-image-based method for identification and counting of tobacco plants. The method comprises: S1, a tobacco planting area is shot by an unmanned aerial vehicle; S2, pretreatment is carried out on an aerial image and candidate areas for tobacco plants are obtained by segmentation; S3, color features and texture features of the candidate areas for tobacco plants are extracted and are used for classification by a classifier; S4, according to the extracted features of the candidate areas for tobacco plants, the candidate areas for tobacco plants are classifier by using the classifier; and S5, statistics of classification results is carried out and detected tobacco plants are marked in an original aerial image. Compared with the traditional method, the method provided by the invention has the following beneficial effects: the tobacco planting area is shot by the unmanned aerial vehicle and identification and counting of tobacco plants are realized by using an image identification method; the efficiency and accuracy are high; the data are reliable; and the operation is simple.