Inspection Of Cigarette Filter Rods Roundness With Machine Vision

Filter rod roundness is one of the important indexes for evaluating thequality of the cigarette. The traditional filter roundness detection methods requirekeeping contact with filter rods and are easily affected by environmental factorssuch as temperature disturbance that affects the efficiency and accuracy of thefilter rod detection. Machine vision technology is a kind of non-contact andonline sharp measurement and size detection method. Compared to the traditionalshape detection methods, it has the advantages of non-contact, high efficiencyand high precision in wide range. Therefore, we detect the degree of circular inthe filter rod and make the in-depth research on the detection methods in theroundness of the filter rod based on machine vision. Hough transform isintroduced to improve the initial contour of filter rod detection. For the blurringimage, snake module algorithm is used to detect the roundness of the filter rod.For some images with the adhesion and blur in the image acquisition forfilter rod, we adopt the gray variance operator and Radon transform algorithm todistinguish blurring image and adhesion. In order to improve the efficiency andprecision of the roundness measurement of filter rods, we propose an improvedHough transform algorithm, and an improved detection algorithm based on theactive contour model of Snake.In this paper, we make deep research on in the algorithm of circle detectionbased on the Hough transform. To overcome the high time complexity problem intraditional Hough transform circle detection, the improved Hough transformdetection algorithm of the filter rod roundness is put forward in this paper andverified through the theoretical analysis. Experimental result proves that thisalgorithm has good performance in detecting the image of clear and adjoinedfilter rod. The theoretical analysis and the actual experiment validate the effectiveness of this algorithm.The circle parameters in Hough spatial filter image processing are difficultto be determined in the traditional Hough transform. Combining with thecharacteristics of filter rod image, we introduce the active contour model ofSnake based on the theory of Snake algorithm and it has a good effect ofdetecting the blurring filter rod image. The paper proposes an improved Snakemodel for detecting blurring filter image. Experimental result shows that thismethod can detect the image of fuzzy filter with higher accuracy.Considering the precision and complexity of the algorithm, we fit thescattered points at the image edges by the least square method and evaluate theroundness of filter rod by the least square circle method. The simulation and realexperiment show that the algorithm has low time complexity and higher accuracyin detecting the roundness of filter rod.