Abstract: Ratio of intact kernel among rice sample is an important index when grading rice in market. Crack in brown rice kernel is the major factor to produc broken kernel during rice milling. A new method using wavelet transform to detect crack in brown rice kernel was described. Images of brown rice kernels were taken with new developed illumination style. The wavelet transform was applied to rice kernel image on binary scale. Edge feature of brown rice kernel was extracted while noise of image was smoothed. The partial maximum of wavelet transform was extracted and subsequently crack in brown rice kernel was detected effectively. Experiments showed that accuracy of crack identification is more than 92% and the new method is more satisfactory than classical differential operators.