Due to the diversity and complexity of image background, it is difficult to obtain the prominent image edge based on single traditional algorithms. In this paper, an improved adaptive edge detection method focusing on defective steel plate images is proposed, which can be applied to extract the target areas by filtering out background features. Firstly, the bilateral filter is adopted in the Canny detector instead of the traditional Gaussian filter to remove the noise in the image. Secondly, the improved Otsu method is applied to obtain the double threshold of the Canny detector. Finally, the characteristic indicators with prior knowledge are constructed to filter out the background features and achieve the extraction of target area. The feasibility and effectiveness of this method has been verified by applying the edge detection and extraction of defective steel plate images. The experimental results indicate that the proposed edge detector, which can filter out the background features effectively and the performance is better than traditional methods in terms of detection accuracy and robustness.

Edge detection, Steel plate image, Canny detector, OTSU, Target extraction