In order to achieve the detection of multiple regions in the product surface rapidly and accurately, we select variable step mechanism to obtain the circumferential image sequence of the products to be tested in the optimal recognition interval quickly so that some incomplete data could be utilized to realize the rapid detection of multiple areas in the limited orientation. First of all, we need to determine the optimal identification range of the area and the rotation step through relativity calculation and projection method. Secondly, the scale invariant feature transform (SIFT) algorithm and binary search method are selected to find the optimal location information of product in the standard image library. Finally, we distinguish there are some defects or not by relativity calculation. The experiment results demonstrate that the variable step size method based on optimal identification interval can save an average time of 6.37 s than the traditional full-circumferential fixed step method while ensuring the detection accuracy.