The Research And Development Of Fiber Detection System Based On Digital Image Processing

In recent years, many new kinds of profiled fiber showed up in chemical fiber industry. Theprofiled fiber is those fibers whose cross-section’s shape is not round. The characteristic of profiledfiber is close to the cross-section’s shape. At present, the chemical fiber cross-section’s shapeparameters detection still mainly relies on human, which is high labor-intensity and low repeatedmeasuring accuracy. Therefore, there is an urgent need for the auto-detection system in the productionof chemical fiber. The detection system should get the needed size parameters and shape parametersquickly and accurately. This paper focuses on the digital image processing of the profiled fiber,including the image segmentation of the touching fibers and the parameter measurementautomatically.The image processing of this paper includes three parts: the preprocessing part, the imagesegmentation part and the detection part. The preprocessing part is to transform the image into anblack-white image, which is to separate the target or foreground from the background. The imagesegmentation is not only to separate the target from the background, but also to separate touchingtargets into individual ones, which is the key to the paper. The detection part is to detect every singlefiber’s size parameters and shape parameters automatically. To solve the image segmentation, a newalgorithm is designed, which is a combination of the watershed algorithm based on distance transformand the marker-controlled wartershed segmention. First, transform the image into a black-white imageusing Niblack algorithm. Then, get the foreground marker using distance transform andmorphological reconstruction. Then, modify the distance image so that it only has minima at theforeground marker locations. Last, compute the watershed transform of the modified distance image.After the image segmentation, it is easy to get the fiber’s cross-section’s size parameters and shapeparameters like perimeter, area, the minimum circumscribed circle, the minimum area boundingrectangle.Based on these researches, a fiber detection system was developed using Visual C++. Thesoftware includes the preprocessing part, the image segmentation part and the detection part. Andsome experiments were done, and the results show this software can complete the image segmentationand the detection successfully.