The Algorithm Research Of Color Classification And Illumination Correction For Dyeing Products Based On Computer Vision

The living standard of people is increasing, which promotes the rapid development of textileprinting and dyeing industry. Therefore, the accurate color quality control seems to be veryimportant. Intelligent evaluation for textile color difference using computer vision instead of thehuman eye has become a research hot spot of related colleges and scientific research institutions.In the system of color-difference evaluation for dyeing products based on computer vision,the problems need to be solved mainly include: the measured image acquisition, the featureextraction and modeling, the automatic classification of color difference as well as the lightcorrection of the image.The main research work and results of this paper are as follows:(1) The research of the feature extraction and modeling in color-difference evaluation fordyeing products. How to establish a robust and effective color feature model is the key ofwhether the detection system can be used for the practical application. This paper studies thefusion model of the color and texture feature and elaborates the performance evaluation methodof this model: robustness and: adaptability. In this paper, two important inspection index are alsointroduce: the missing report rate and the misreporting rate which provides important data modelfor the practical application of color difference detection system.(2) The proposition of automatic classification algorithm of the Morlet wavelet kenalsupport vector machine (SVM), which is based on LBP-LAB-GRA multidimensionalcharacteristics. In order to make the collected sample data classificate automatically better, thepaper introduces a kind of SVM classification algorithm based on multi-feature fusion throughthe comparative analysis some classification algorithms of SVM, AdaBoost based on cascadeweak classifier and a combination of wavelet analysis and SVM (Morlet wavelet), and and theproposed method was successfully used to color difference classfication of dyeing products. The results showthat: the algorithm has good classification effect and achieve automated evaluation of the color difference, andhas better robustness in the case of the inconsistent light.(3) The introduction of illumination correction algorithm based on the wavelet theory intothe dyeing cloth uneven illumination. If the surface of dyeing product is influenced by factorssuch as uneven illumination, a fuzzy surface color and texture will show up and the test result isnot ideal. The illumination correction algorithm based on the stretched histogram linear cancorrect the overall brightness, but it is necessary to set up the linear mapping function, and theselection of parameters is very important. As for the illumination correction algorithm based onthe histogram equalization, although it is not necessary to set up the linear mapping function, thecorrection effects of the solid color cloth is poor and this algorithm will change similar colorconversion to color changes and human eyes will feel these differences. And the illuminationcorrection algorithm based on the homomorphic filtering has a better correction effect, for thesolid color cloth, but for non-solid color of the cloth, this algorithm will cause a certain extent aloss on its surface texture features and the image distortion. The dyeing product correction algorithm based on the wavelet theory is put forward in this paper, which can better detect thesurface texture feature and, in brightness correction premise, avoid the loss of textureinformation and significantly improve the image quality.