The invention relates to a method for detecting an insulator string breakage defect based on sparse representation. The method comprises the steps of firstly performing image segmentation on a shot original image containing an insulator string by using an image segmentation algorithm, removing background interferences such as accumulated snow and transmission lines, and segmenting the image into a plurality of regions which are not mutually connected; then performing ASIFT (Affine Scale-Invariant Feature Transform) matching on the regions which are not mutually connected and insulators in an insulator string library, wherein regions with the number of matching points being higher than a preset threshold are insulator string regions; then performing insulator single-disc segmentation on the recognized insulator string regions to acquire a plurality of insulator small discs; and finally, performing classified labeling on each insulator small disc by using a sparse representation classifier, recognizing whether each single small disc has a string breakage defect or not, and performing positioning and labeling in the original image. Compared with other methods, the method has higher recognition accuracy and is accurate in defect positioning.