A Computer Vision-based Method for Measuring Shape of Tobacco Strips

In order to describe the shape of tobacco strips accurately, a measuring method based on computer vision was proposed. Image segmentation method based on Mean-shift algorithm was used to extract the area, where tobacco strip was present from an image, then the profile of tobacco strip was precisely extracted by a morphological gradient algorithm. The shape of tobacco strip was quantitatively described by Fractal dimension and Fourier descriptors separately. The ability of shape description of Fractal dimension and Fourier descriptors was tested with tobacco strips of different shapes. The results showed that for tobacco strips of different shapes, Fractal dimension and Fourier descriptors took different values and the difference between the two increased with the widening of the discrepancy between shapes in a monotonic way. It was verified that both methods were reliable. By adopting the two methods together, the shape of tobacco strip can be more effectively evaluated.