Open Access
Classification of south american camelid and goat fiber samples based on fourier transform infrared spectroscopy and machine learning
(Taylor and Francis Group, 2024) Quispe Bonilla, Max David; Trigo Vilaseca, Jesús Daniel; Serrano Arriezu, Luis Javier; Huere, Jorge; Quispe Peña, Edgar; Beruete Díaz, Miguel; Institute of Smart Cities - ISC; Universidad Pública de Navarra / Nafarroako Unibertsitate Publikoa.
Some animal fibers are considerably cheaper than others. Hence the existence of counterfeit products, which are detrimental to legitimate producers and consumers alike. Fourier-Transform Infrared (FTIR) spectroscopy can extract a characteristic waveform from fibers that can be later used for classification. However, visually inspecting such waveforms is imprecise. Previous research has complemented FTIR with other mathematical or physical methods to improve accuracy. In parallel, Artificial Intelligence (AI) is an emerging field that could be helpful in this domain. The objective of this work is therefore to develop and validate two machine learning models, namely, Deep Neural Networks (DNN) and Support Vector Machine (SVM) to classify spectra of fibers by species. The spectra are acquired using an FTIR spectrometer in Attenuated Total Reflectance (ATR) mode (FTIR-ATR). Camelid (alpaca: n = 51, llama: n = 50, vicuña: n = 50) and goat (mohair: n = 35 and cashmere: n = 20) samples were evaluated, from which 1236 FTIR-ATR spectra were obtained. Some visual differences were observed between the spectra of the different species. Accuracies up to 96.75% and 95.12% were obtained when evaluating the DNN and SVM models. Furthermore, an accuracy of 97.8% was obtained when evaluating the FTIR-ATR spectra of South American Camelids (SAC) fibers with DNN, and 97.2% when evaluating them with SVM. A 100% accuracy was obtained when evaluating the FTIR-ATR spectra of vicuña fibers with both models. No significant differences were found (p-value = 0.368) by comparing the number of hits against the total number of alpaca, llama, vicuña, mohair and cashmere fibers using DNN. As per the results, it seems that DNN is more accurate than SVM. In conclusion, FTIR-ATR spectrometry techniques combined with machine learning models are a reliable alternative for the identification of SAC and goats through the spectrum of their fibers.
Open Access
Application of artificial intelligence and digital images analysis to automatically determine the percentage of fiber medullation in alpaca fleece samples
(Elsevier Science, 2022) Quispe Bonilla, Max David; Serrano Arriezu, Luis Javier; Trigo Vilaseca, Jesús Daniel; Quispe Bonilla, Christian; Poma Gutiérrez, Adolfo; Quispe Peña, Edgar; Ingeniaritza Elektrikoa, Elektronikoaren eta Telekomunikazio Ingeniaritzaren; Institute of Smart Cities - ISC; Ingeniería Eléctrica, Electrónica y de Comunicación; Universidad Pública de Navarra / Nafarroako Unibertsitate Publikoa
The aim of this research is to develop and validate two computer programs based on artificial intelligence (AI) and digital image analysis (DIA) in order to determine the incidence of medullation in white alpaca fibers. Two data sets were analyzed: 76 samples of Huacaya alpaca fibers obtained from Huancavelica, Peru, and 200 samples of white alpacas of two genotypes (Huacaya, n =100; Suri, n = 100), obtained from Arequipa, Peru. The preparation of each sample followed the procedure described in IWTO-8-2011. The Pytorch framework was used to generate several training models based on the You Only Look at Once (YOLO) architecture. Circa 4000 pictures of fibers were taken and 661 of them were selected as representative. Using the LabelImg software, the fibers present in each representative picture (approximately 10 fibers/picture) were labeled as one of these two classes: either medullated or non-medullated. Subsequently, the data augmentation technique was applied to expand the data set to 3966 photographs. Thus, 90 of them were used as initial validation data, while the reaming 3876 pictures (containing a total of 23,964 labeled fibers) were used as training data. Matlab was used to develop the DIA-based software. More specifically, algorithms of pre-processing, segmentation, smoothing, skeletonization and Hough transform were implemented to detect medullated and non-medullated fibers. Correlation and linear regression analyses were used to evaluate the models. The medullation percentage results show that there is no statistically significant difference between the AI-based method and the projection microscope method (p-value = 0.668 and 0.672 for the t-student and Wilcoxon tests, respectively). Moreover, the correlation of each of the developed computer methods with the projection microscope method is very strong (r = 0.99 and 0.97). This confirms the software ability to perform the recognition of fibers with and without medullation. Similar results (p-value = 0.357) were obtained when comparing the projection microscope method and DIA-based software method. Finally, using the proposed framework, the average time required to analyze a sample was 19.44 s. As a result, this software allows the implementation of practical, precise, and efficient methodologies to determine the incidence of medullation of alpaca fibers.

Quispe Bonilla, Max David

Email Address

person.page.identifierURI

Birth Date

Job Title

Last Name

First Name

person.page.departamento

person.page.instituteName

ORCID

person.page.observainves

person.page.upna

Name

Filters

Author

Subject

Date

Has files

Item Type

Type

Settings

Sort By

Results per page

Search Results