Research And Implementation Of A Machine-Vision-Based Fiber Geometric Parameter Test System

With the rapid development of information technology, the fiber is needed more and more by the society. As a result, there are more and more enterprises which produce fiber. For most SMEs, the geometric parameter test of the fiber is manual. Because human beings are easily affected by physical and psychological abnormal, the process of the manual geometric parameter test is inefficient and the result is inaccurate. To increase the automotive and accuracy level of the fiber geometric parameter test, based on the theory of Machine Vision, we studied the principle of the fiber geometric parameter test and developed some algorithm to solve this problem in this thesis. Finally, a fiber geometric parameter test system is developed.Total technology scheme of the fiber geometric parameter test system is proposed according the analyses to principle of the fiber geometric parameter test and the technical requirement. Based on the conditions of the fiber image and classic image processing theories, the test process is divided into two parts: Digital Image Processing and Measurement.The preprocessing flow is designed. The format of fiber images is changed firstly. And then median filter is processed on the images. Finally, the images are enlarged. There is only one threshold to finish the image segmentation.To filter the noise and retention goals border, the contour tracking algorithm is developed. Based on the theories of least-squares curve fitting and plane analytic geometric parameter, we designed the geometric parameter parameters measurement algorithm. Two algorithms are also designed to remove points which are far from the fitting curve from the points set. These two algorithms are able to ensure the accuracy of curve fitting. Reference to the least-squares fitting method, an iterative algorithm of arc fitting is developed. And to evaluate the fitting accuracy, an fitting evaluation method is developed.A fiber geometric parameter test software is developed. Some experiments are designed to study the performance of the software. The advantage and disadvantage will be find out after these experiments, and the direction of improvement will be identified.