The invention discloses a visible component classification method based on SVM. The method comprises the following steps of in a SVM training phase, A, acquiring a sample picture containing several kinds of visible components and taking the sample picture as a training sample; B, segmenting the sample picture into several visible component images; C, extracting an image characteristic of the visible component images, classifying the visible component images based on the image characteristic and constructing several cascaded visible component picture databases; D, constructing several cascaded SVM classifiers and using the corresponding visible component picture database to train; in a visible component identification classification phase, E, extracting the image characteristic of the visible component images to be identified and classified; F, based on the image characteristic, assigning the visible component images to the corresponding SVM classifiers to carry out identification and classification. The plurality of cascaded SVM classifiers are used to carry out multilevel classification on a visible component graph so that classification accuracy of the SVM classifiers to the visible components is effectively increased.