According to the influence on the low recognition rate of steel plate defect detection for uneven illumination, the algorithm of steel plate defect detection based on image processing is proposed. Based on image enhancement, image filtering and de-noising are performed, and image detailed features are retained. Image segmentation is realized by OSTU algorithm. The geometric features of steel plate defect images such as perimeter, area and width are effectively extracted, classified and recognized based on neural network. Experimental results show that the recognition accuracy of the proposed steel plate defect detection method is improved.