Abstract: As the traditional algorithms of image detection are applied to the inspection of strip surface, the treatment effect is not very ideal.According to this problem, a new strip surface defects detection technology is proposed negative selection algorithm based on B tree.According to factors of generation of mature detectors, the candidate detector are produced variation through low frequency and high frequency, representation of detector is changed to value-based detector instead of binary-based detector, and 3 properties such as self-number, detector-number and some antigen-number are extracted to construct indexes using B tree based negative selection algorithm to achieve quality of steel strip surface defect detection.Experiments show that the method is significantly superior to the traditional image detection algorithms in detection efficient.