Crop leaf spot image segmentation is the important steps in crop disease automatic recognition. To im-prove the leaf spot segmentation performance of the traditional threshold or clustering methods, a spot segmenta-tion based on SVM and morphology was proposed in the paper. The spot image segmentation was formulatedby SVM.The discontinuity edge, small noise, small hole and the small hole inside the lesion image were eliminated by the op-ening and closing algorithms of morphology. The experimental results showed that this approach outperformed othermethods and was effective for cucumber leaf disease segmentation.