| The identification of fine wool and cashmere was still a difficult problem. The recognition difficulty was expanded with a lot of similarities on the morphological and physic-chemical properties. Many methods of identifying cashmere and wool were presented now, but accepted standard by industry was still the microscope method.Characteristic differences of the fiber scales were mainly based on the identification in microscope method, the accuracy of the results were closely related with the experience of the inspectors. So it was an objective fact the ratio of Cashmere and wool blend wasn’t measured accurately. Add that refinement and of wool was more serious, and the case of wool slenderizing and cashmere coarsening was more serious. So it was more difficult that cashmere and wool fibers were measured accurately. Along with the fast development of computer technology, it becomes a trend microscope detection was combined with computer technology to detect wool and cashmere.BEION F6fiber fineness instrument was used to collect images in this paper, and the characteristics of fiber scale was observed and acquired, then the images were stored by the computer, image processing tools were used for collecting, then pattern information was acquired, Such as:fineness, scale height, premiere, area, relative perimeter, relative area, thickness, the ratio of diameter and ratio, scale angle and so on, There was more focus on the front4or other indicators, but less concern about scales angle due to its difficulty of observation by microscope and accurate measurement. Based on the outline form of the scales, mathematical model of scales angle and thickness was established; then based on result of the model, the possibility and accuracy of identification of Australian wool and cashmere were studied. Statistics of Australian wool and cashmere scales angle and thickness were built, the distribution density functions and probability analysis of two kinds of errors were analyzed, and the conclusion was that the misjudgment ratio of the sum of the two errors was less than11.3%. A set of feature indicators were provided to distinguish the Australian wool and cashmere blend.There were10indicators that2new indicators were coupled with8original parameters to characterize scale patterns, then the10characteristic parameters were mathematically statistics analyzed and their fitting parameters were got, then their theoretical curves were compared and analyzed, the value of two types of error was got, then better specification was acquired by comparing. Correlation analysis of the10parameters was done, the indicator with smaller value for two types of errors and strong independence was selected, then their composite indicators were discussed. According to these indicators, then the main factor identification model was established; integrated identification model was established; clustering analysis model based on SPSS was established; and Bayesian model was established, and Bayesian models was with the highest accuracy. The model series identification model of wool and cashmere was established based on the level of accuracy of above models, concrete results were as follows:(1)Based on morphological features of the Australian wool and cashmere to establish angle model was with higher precision, scale angle and thickness were characteristic indictors to identify Australian wool cashmere, it was of great significance to distinguish Australian wool and cashmere.(2) It wasn’t better that parameters were more for Combination parameters and the two types of errors was up to21.46%in scale angle indictor, it was41.81%in thickness, and it was just11.3%in the ratio of scale and thickness. But in Combination parameters h2·Ad、(dh)2、,((dh)2)/Ad it was60.21%、62.11%、79.61%,the accuracy significantly reduce considerably. In combination parameter (dh)2/(Ad·Pd,it was30.62%, the accuracy significantly was increased.(3) because the scales angle of Australia wool and cashmere was seriously different, the indicators of scale angle and thickness were come up. according to the overlap between Australian wool and cashmere scales angle probability distribution of measured values, a was15.12%,β was6.34%， the absolute error of both was21.46%; scales thickness also has significant difference, here a was22.2%,β was19.61%, the absolute error of both was41.81%. Scales angle was better than scales thickness in accuracy. The established model of the Australian wool and cashmere scales angle had higher precision. Scales angle and scales thickness were characteristic indicators to identify Australian wool and cashmere, and they had important significance.θ1was over θ2,and t1was less than t2, the compound index(θ1)/(τ1) was greater than(θ2)/(τ2). With the ratio increasing, the overlapping portion of two types of index was small; the accuracy was improved, so it could be used as compound index to distinguish fibers.(4)10parameters were not completely independent in this paper after floristic analysis, in scale angle α was6.34%,β was15.12%, the absolute error of both was21.46%; in dilatometer α was22.33%,β was15.75%, the absolute error of both was38.08%; in relative scale interval α was 20.68%,β was21.13%, the absolute error of both was41.81%; in scale thickness a was22.2%,β was19.61%, the absolute error of both was41.81%;The independent4indicators have strong distinctiveness.(5) This paper uses4models to distinguish fibers. Recognition rate of the main factor identification model was93.4%; Recognition rate of clustering analysis model based on SPSS was94%; and Recognition rate of Bayesian model was97.8%, above4models were tandem compounded, the Recognition rate was99.9%.In order to search more visible parameters, to count the number of scale branched in the continuous20scales and scales based on the number to differentiate wool cashmere fiber, and better results were also obtained. |
PDF Full Text Request