Improved Zernike Moment-Based Method for Subpixel Edge Detection of the Glass Bottle

As the accuracy of edge detection determines the accuracy of actual size measurement, in order to improve the measuring accuracy, this paper proposes that a faster algorithm should be adopted to detect the glass bottle dimension based on the Zernike moments. The traditional Zernike algorithm is combined with Otsu adaptive threshold algorithm to get a faster improved algorithm. The edge of glass bottle with subpixel level is detected by using the improved algorithm. The ellipses formed by the glass bottle's mouth and bottom are fitted by applying the least square method. At last, the system is calibrated by gauge block to get the actual measuring dimension. Experimental results show that the improved algorithm can not only make the edge detection reach the subpixel accuracy, but also avoid the edge misidentification and inefficient which will be caused by repeatedly manual adjustments to select the threshold value when we detect the edge.