The invention discloses a segmentation method of overlapping leaf images. The method comprises the steps that colorful images containing overlapping crop leaves are obtained from an image collection device and normalized; green color features of the images are extracted, and non-green backgrounds are removed; by means of the Chan-Vese model, contour extraction is conducted on a target leaf; by means of a Sobel operator, edge detection is conducted on the target leaf; the contour extraction result is fused with the edge detection result, accurate segmentation of the overlapping leaves is achieved, and the target leaf is extracted. By means of the method, accurate and complete segmentation of the overlapping crop leaves can be achieved, a foundation is laid for subsequent recognition of leaf diseases, and a new method and technology are also provided for the field of the crop leaf segmentation.