نوع مقاله : مقاله پژوهشی
نویسندگان
1 استادیار مهندسی مکانیک بیوسیستم،دانشکده کشاورزی سنقر، دانشگاه رازی، کرمانشاه، ایران
2 دانشجوی دکترای مهندسی مکانیک بیوسیستم، دانشگاه بوعلی سینا، همدان ،ایران
چکیده
کلیدواژه‌ها
موضوعات
عنوان مقاله [English]
نویسندگان [English]
Wheat is an important grain product that constitutes about 20% of the calories consumed by the human population around the world. Due to the strong dependence of wheat growth and yield on its variety, the choosing the right variety for cultivation in different soil and water conditions is very important. In this research, the feasibility of using an e-nose system along with artificial intelligence based on metal oxide semiconductor sensors (MOS) as a non-destructive tool for the separation and identification of three varieties of wheat with the names: Salari dry wheat, Quds blue wheat and local red wheat is evaluated. Support vector machine (SVM), artificial neural network (ANN) and principal component analysis (PCA) were the methods used to achieve this goal. The obtained results showed that TGS822 and TGS2620 sensors play the most role and TGS813 and TGS2610 sensors play the least role in wheat variety detection. ANN analysis method with 91.7% accuracy showed better result than SVM method (75% accuracy) in identifying and classifying wheat varieties. In the meantime, the PCA method showed a relatively good performance in separating and identifying wheat varieties with 77% of the total variance of the total data. Also, the results showed that wheat of Salari dry wheat variety had different aromatic compounds with other two varieties of Quds blue wheat and local red wheat. The proper performance of the e-nose in the separation of wheat varieties can indicate the promising application of this technology in the separation and identification of wheat varieties.
کلیدواژه‌ها [English]
Wheat is an important cereal crop that accounts for about 20% of the calories consumed by the human population. Global wheat production reached 796 million tons in 2023. Given the great diversity in wheat varieties, choosing the right variety for cultivation in different soil and water conditions is of great importance, the growth and yield of wheat is strongly related to its variety. Old methods such as measuring yield and performance indicators in examining wheat varieties are a time-consuming method. On the other hand, recent and non-destructive methods for identifying wheat varieties are the hyperspectral imaging method and the electronic nose machine combined with artificial intelligence. Much research has been conducted in the field of using the hyperspectral imaging method in identifying wheat varieties. On the other hand, the e-nose machine, as a cheaper, faster and easier method, can be a more suitable method for identifying wheat varieties. Therefore, the aim of this research is to assess the feasibility of identifying three wheat varieties with the help of the smell machine and artificial intelligence.
In order to conduct this experiment, three wheat varieties were prepared: Salari dry wheat, Quds blue wheat and local red wheat cultivated in Sonqor County. An e-nose device equipped with 10 metal oxide semiconductor (MOS) sensors was used to conduct the experiments. To perform the olfactory tests, first 5 grams of the samples were placed in a closed container (sample chamber) for 30 minutes. The sample chamber was connected to the e-nose device and data was collected. 8 repetitions were considered for each sample. A fractional method was used to correct the baseline. The preprocessed data were analyzed using Support Vector Machine, Artificial Neural Network, and Principal Component Analysis using Unscrambler V 9.7 and Matlab 2015a software.
According to the results obtained from this research, it can be said that TGS822 and TGS2620 sensors showed the highest response and TGS813 and TGS2610 sensors showed the lowest response to the aromatic compounds of three wheat varieties. Which indicates the presence of organic vapor compounds in the aromatic compounds of wheat. The results of the classification of three wheat varieties, including Salari dry wheat, Quds blue wheat and local red wheat by SVM classification method, showed that this method was able to classify the three tested wheat cultivars well with an accuracy of 75%. Define and recognize. Also, according to the results of the confusion matrix, the SVM classification method was able to separate and classify all the samples into Salari dry wheat and Quds blue wheat with 100% accuracy. The disturbance matrix obtained from the neural network for three wheat varieties showed that the ANN method was able to classify and separate the three tested wheat varieties with an overall accuracy of 91.7%. to do. The artificial neural network analysis method was the best type of analysis with the identification accuracy of 91.7. According to the results obtained from this research, it can be said that e-nose in combination with artificial intelligence is a suitable method for detecting and identifying wheat cultivars.
In this research, an e-nose system based on ten metal oxide semiconductor (MOS) sensors was used to identify and distinguish three varieties of wheat, including Salari dry wheat, Quds blue wheat and local red. Support vector machine (SVM), artificial neural network (ANN) and principal component analysis (PCA) were used to identify and classify wheat varieties. The artificial neural network analysis method was the best type of analysis with the identification accuracy of 91.7. According to the results obtained from this research, it can be said that e-nose in combination with artificial intelligence is a suitable method for detecting and identifying wheat cultivars.
All authors contributed equally to the conceptualization of the article and writing of the original and subsequent drafts.
Data available on request from the authors.
This research was carried out in Razi University, Kermanshah-IRAN. Therefore, the authors are thankful to Razi University for their supporting.
The study was approved by the research Committee of the University of Razi. The authors avoided data fabrication, falsification, plagiarism, and misconduct.
The author declares no conflict of interest
)