Abstract: It is difficuh to extract crack defects front a defect image by traditional detection techniques, for the image of surface crack defects on a medium plate is influenced greatly by background nolo- and oxidized scale. Morphological wavelet, which combines the advantages of wavelet decomposition with mathemarital morphology, was introduced to decompose the defect image and detect calastrophe signals. According to the similar linear character of crack defects irn shape, the crack defects can he accurately extracted from the image. Experimental results sit, show that the morphological wavelet transform presented here is more effective than other techniques for the detection of surface crack defects of a medium heavy plate.