Research Of The Auto-Focus For Textile Fibers And Multi-Layer Image Processing

Till now, textile fibers inspections are manually performed in our country. The inspection is time consuming and costly, moreover, the result is affected by the operators. In order to overcome such shortcomings, the Ministry of Education of China sponsored the project for fiber automatic image analyzing. The research consists of: the auto-focus of microscope, the pattern recognition of digital image.In this thesis, we study the methods for improving quality for fibers image, a practicable solution is discussed to deal with the effect of microscope auto-focus. Because of the effect of illumination, diffraction and photo-electricity transition, noise is induced in image acquiring system. The image pre-processing methods are discussed in the paper, the median filtering is applied to eliminate the noise and it is verified by the experiment results. Also, the real time processing of the detecting system requires fast and veracious auto-focus algorithm. Several auto-focus algorithms are discussed and compared in the paper. The SMD algorithm is applied as the basic algorithm in our research. At the same time, an optimized focus-searching scenario is developed here. Finally, the characteristics of fiber slices limit the effective application of ordinaryauto-focus algorithms. In order to solve the multi-focus problem for fibers, we study the algorithms with slice swatches of cotton, hemp, and some other fibers. In the thesis, a images is divided into several sub-images, then the optimized focus for each sub-image is calculated. The value of pixels near the dividing line is recalculated to eliminate the artifacts caused by image dividing. This method for multi-focus is applied in the auto analyzing and detecting system.In the automatic fiber image analyzing and detecting system, the detailed discussion about the embedded system and the controlling system is included in the Appendix.