A real-time algorithm for aluminum surface defect extraction on non-uniform image from CCD camera

A novel real-time defect extraction framework is proposed for handling non-uniform images in high-speed aluminum strip surface inspection. The image is first preprocessed by Gaussian smoothing operator and Prewitt edge detection, which is robust to image non-uniformity. Afterwards, a fast adaptive segmentation algorithm is applied to further remove the effect of non-uniformity and enhance the edge detection. The final defect extraction image is achieved through morphological operations. The resultant method is computationally efficient and robust to non-uniformity. The proposed framework is evaluated on a large dataset of aluminum strip surface images obtained from the product line. The experimental results show that the proposed method achieves real-time defects extraction, and it outperforms the previous methods in accuracy.