The invention discloses a method for detecting a tunnel defect, comprising the following steps: acquiring an image of a tunnel structure through an image pickup system; denoising the acquired image and then directly identifying the structure contour in the image, carrying out secondary image identification based on the identified structure contour to identify the defect phenomenon in the tunnel; and storing the portion having the defect phenomenon into a block file, calculating the area of the defect phenomenon in the block file according to the pixels in the block file, and inferring the category of the defect phenomenon. According to the invention, an emergency alarm can be carried out aiming at the defect phenomenon; the method is a substitute for traditional artificial analysis of detection data, and can rapidly, effectively, and comprehensively calculate and analyze the defect phenomenon in the subway shield tunnel.