Computer Vision Image Precision Measurement on Cigarette Filter Stick

摘要

Cigarette filter stick is a deformable and irregular shape parts, it could not be measured by traditional instruments. This paper put forward a novel computer vision measurement: relative measurement based on image analysis. In this method a standard part and a target part were put in the same imaging fields, measured image was collected by CCD camera. Many methods for edge detection are compared, and a new edge detection method based on data fusion is given. It improves the precision of image location. The standard part's roundness error was evaluated by minimum region Law. The cigarette filter stick' circumference and roundness errors have been calculated by comparing with the standard's dimensions (relative method). An algorithm for edge sub pixel location based on Zernike moment mask of square was used to calculate measured valuables. The results show that the relative method accuracy based on the single pixel edge image was good enough to the results based on the sub-pixel edge image.