An Improved Canny Algorithm for Edge Detection in the Thickness Measurement of Reconstituted Tobacco by Modulated Line Laser

In the thickness measurement of reconstituted tobacco by modulated line laser, it is necessary to extract the edge of the modulated line laser to perform the straight line fitting. However, in the edge detection, the conventional Canny edge detection algorithm suffers from the problem of edge blur and manual double threshold setting caused by filtering. In this paper, a dynamic adaptive weighted median filtering scheme is proposed. Instead of the scheme of single Gaussian filtering in the conventional Canny algorithm, the proposed scheme uses an adaptive dual threshold selection algorithm based on the maximum entropy of the image to cope with the problem of dual threshold setting. With the above two improvement schemes, the filtering parameters can be dynamically adjusted to reduce the degree of smoothing to the edges while reducing the noise. In addition, the edge detection quality of Canny algorithm can be effectively improved by the adaptive threshold selection method. The results show that the improved Canny algorithm can effectively filter out the noise and has the advantages of high adaptability and edge clarity.