cultivar recognition

Nectarine is a plant that is cultivated as an important commercial product in some countries and is known as an important source of sugar and vitamins in the human diet. Due to the increase in expectations for food products with high quality and safety standards, accurate, fast and targeted determination of the characteristics of food products is necessary. In the nectarine product, quality evaluation after the harvesting stage is necessary to provide a reliable and uniform product to the market. The purpose of this study is to identify and classify nectarines by extracting characteristics from the response patterns of the spectrometer and using chemometrics methods. A near-infrared spectrometer can detect the spectrum of reflected light by estimating its concentration or determining some of its inherent properties. The quality assessment of agricultural products includes two main methods, quality grading systems based on the external characteristics of agricultural products and quality grading systems based on internal quality assessment, which has gained outstanding points in recent years. In the meantime, several methods have been invented for the qualitative grading of agricultural products based on the assessment of their internal properties in a non-destructive manner, and only some of them have been able to meet the above conditions and have been justified in terms of technical and industrial aspects. To be meanwhile, spectrometry can be highly efficient in determining the quality of cultivars. Spectroscopy is a type of system that has a different structure and approach from other methods (image processing, neural network, etc.) and can perform classification and determination of digit quality. With increasing expectations for food products with high quality and safety standards, the need for accurate, fast and targeted determination of the characteristics of food products is now essential. Because manual methods do not have automatic control, they are very tiring, difficult and expensive, and they are easily affected by environmental factors. Today, spectroscopic systems are non-destructive and cost-effective and are ideally used for routine inspections and quality assurance in the food industry and related products. This technology allows inspection works to be carried out using wavelength data analysis techniques and is a non-destructive method for measuring quality parameters. In this research, using spectrometry and chemometrics methods, the variety of nectarine fruit was identified.

For this study, 5 different nectarine cultivars were prepared from the gardens of Moghan city (Ardebil province) and were tested and data collected. A spectroradiometer model PS-100 (Apogee Instruments, INC., Logan, UT, USA) was used to acquire the spectrum of the samples. This spectroradiometer is very small, light, and portable, has a single-wavelength sputtering type with a resolution of 1 nm and a linear silicon CCD array detector with 2048 pixels that covers the spectral range of 250-1150 nm (Vis/NIR) well. Also, there is the ability to connect the optical fibre to the PS-100 spectroradiometer and transfer the data to the computer with the purpose of displaying and storing the acquired spectra in the Spectra Wiz software through the USB port. With the aim of creating optimal light in contrast mode measurements, an OPTC (Halogen Light Source) model halogen-tungsten light source, which can be connected to an optical fibre, was used. This light source has three output powers of 10, 20, and 30 watts, which were used in this research. Also, a two-branch optical fibre probe model (Apogee Instruments, INC., Logan, Utah, USA), which includes 7 parallel optical fibres with a diameter of 400 micrometres, was used in counter-mode measurements. After providing the necessary equipment, the optimal spectroscopic arrangement was designed and implemented in order to facilitate the experiments and minimize the effect of environmental factors during the spectroscopic process. The data obtained from spectral imaging may be affected by the scattering of light by the detector with sample change, sample size change, surface roughness in the sample, the noise created due to the increase in temperature of the device and many other factors, and unwanted information affect the accuracy of calibration models. Therefore, to achieve stable, accurate and reliable calibration models, data pre-processing is needed (Rossel, 2008). In this research, Savitzky-Golay smoothing, first and second derivatives, baseline, standard normal distribution, and incremental scatter correction were applied to the data. The use of non-destructive methods based on spectroscopy in the full range of wavelengths requires spending time and very high costs, which makes the practical application of this method almost impossible; therefore, one should look for a way to find the optimal wavelengths and limit the wavelengths to the minimum possible value. Chemometrics uses multivariate statistics to extract useful information from complex analytical data. The chemometrics used in this study started with principal component analysis (PCA) to explore the output response of the sensors and reduce the dimensionality of the data. In the next step, linear diagnostic analysis (LDA) was also used to classify 5 varieties of Shail. (PCA) is one of the most common statistical data reduction methods. This method is an unsupervised technique used to explore and reduce the dimensionality of a dataset. The analysis itself involves the determination of variable components, which are linear combinations of many investigated characteristics. In this research, in order to construct the LDA model, the data were randomly divided into two parts: 70% of the samples were used for training and cross-validation, and the rest of the data were used for independent validation.

Based on the results of the PCA analysis presented in Figure 2, the first principal component (PC-1) describes 72% and the second principal component (PC-2) 13% of the variance of the tested samples. As a result, the first two principal components together express 85% of the data. Considering that it is possible that the degree of correlation between the properties of different samples during the tests, due to various reasons such as technical problems of the equipment, data collection, incorrect sampling, etc., in some samples, inappropriate or socalled outliers The LDA method is a supervised method that is used to find the most distinct eigenvectors and maximizes the between-class and intra-class variance ratios and is capable of classifying two or more groups of samples. The LDA method was used to identify the nectarine cultivars based on the output response of the spectrometer. Unlike the PCA method, the LDA method can extract the resulting information to optimize the resolution between classes. Therefore, this method was used to detect 5 nectarine cultivars based on the output response of the spectrometer. The results of the identification of figures equal to 100% were obtained.

Potato is considered one of the most important food sources in the world (4th rank) and studying its various aspects is very important to ensure that the produced product has the necessary qualifications and can satisfy the customer. In the food industry, this product is transformed into various products such as baked potatoes, fried potatoes, potato chips, potato starch, dry fried potatoes, etc.The complexity of food odours makes it difficult to analyze them with conventional analytical techniques such as gas chromatography. However, expert sensory analysis is costly and requires trained people who can only work for a relatively short period. Problems such as the human subjectivity of the response to smell and the variation between people should also be considered. Hence, there is a need for a tool such as an electronic nose with high sensitivity and correlation with human sensory panel data for specific applications in food control. Due to its easy construction, cheapness and the need for little time for analysis, the electronic nose is becoming an automatic non-destructive method to describe the smell of food.An olfactory machine can recognize the fragrance composition by estimating its concentration or determining some of its intrinsic properties, which the human nose is hardly able to do. In general, the human olfactory system is a five-step process including smelling, receiving the scent, evaluating, detecting and erasing the effect of the scent. The olfactory phenomenon begins with inhaling the intended smell and ends with breathing fresh air to remove the effect of the scent. The human olfactory system, with all its unique capabilities, also has disadvantages that limit its use in quality control processes, including subjectivity, low reproducibility (for example, results depending on time, people's health, analysis before the presence of odour and fatigue is variable), time-consuming, high labour cost, adaptation of people (less sensitivity when exposed to odour for a long time). In addition, it cannot be used to evaluate dangerous odours.Meanwhile, the electronic nose can detect the volatile compounds of potatoes. The electronic nose has been used in extensive research to identify and classify food and agricultural products.
The purpose of this research was to evaluate the ability of the electronic nose using one of the chemometrics methods to detect 5 different potato cultivars.

First, 5 varieties of potato were prepared from the agricultural research centre of Ardabil city. These 5 varieties included Colombo, Milwa, Agria, Esprit and Sante.After preparing the cultivars, first, the samples were placed in a closed container (sample compartment) for 1 day to saturate the space of the container with the aroma and smell of potatoes, and then the sample compartments were used for data collection with the electronic nose.In this research, the electronic nose made in the Biosystems Engineering Department of Mohaghegh Ardabili University was used. In this device, 9 metal oxide semiconductor (MOS) sensors with low power consumption are used, which are listed in Table 1.The sample chamber was connected to the electronic nose device and data collection was done. This data collection was done in such a way that first, clean air was passed through the sensor chamber for 150 seconds to clean the sensors from the presence of odours and other gases. Then, the smell of the sample was sucked from the sample chamber by the pump for 150 seconds and directed to the sensors, and finally, clean air was injected into the sensor chamber for 150 seconds to prepare the device for repetition and subsequent tests. 15 repetitions were considered for each sample.Through the mentioned steps, the output voltage of the sensors was changed due to exposure to gases emitted from the sample (potato smell) and their olfactory responses were collected and recorded by data collection cards, the sensor signals were recorded and stored at 1-second intervals. . A fractional method was used to correct the baseline, in which noise or possible deviations were removed and the responses of the sensors were normalized and dimensionless.
By chemometrics method in this research, it started with principal component analysis (PCA) to discover the output response of the sensors and reduce the dimension of the data.Principal component analysis (PCA) is one of the simplest multivariatemethods and is known as an unsupervised technique for clustering data according to groups. It is usually used to reduce the dimensionality of the data and the best results are obtained when the data are positively or negatively correlated. Another advantage of PCA is that this technique reduces the volume of multidimensional data while removing redundant data without losing important information.

The scores chart (Figure 1) showed that the variance of the total data is equal to PC-1 (94%) and PC-2 (3%), respectively, and the first two principal components account for 97% of the variance of the total normalized data. When the total variance is higher than 90%, it means that the first two PCs are sufficient to explain the total variance of the data set. So it can be concluded that the electronic nose has a good response to the smell of potatoes and its cultivars can be distinguished, which shows the high accuracy of the electronic nose in identifying the smell of different products.With the correlation loadings plot, the relationships between all variables can be shown. The loading diagram (Figure 2) shows the relative role of sensors for each main component. The inner oval represents 50% and the outer oval represents 100% of the total variance of the data. The higher the loading coefficient of a sensor is, the greater the role of that sensor in identification and classification. Therefore, the sensors that are located on the outer circle have a greater role in data classification. According to the figure, it is clear that all the sensors have an important role in identifying the rice variety, including the role of sensors number 1 and 9, which are respectively the same sensors as MQ9 (to detect carbon dioxide and combustible gases) and MQ3 (to detect alcohol, methane, natural gases), it was less than the rest of the sensors, and by removing these two sensors, the cost of making an olfactory device (to distinguish genuine and fake rice) can be reduced and costs can be saved. In this research, an electronic nose with 9 metal oxide sensors was used to identify and distinguish potato cultivars. The Chemometrics PCA method was used for qualitative and quantitative analysis of complex data from the electronic sensor array. PCA was used to reduce the data and with two main components PC1 and PC2, it described 97% of the variance of the data set and provided an initial classification. The electronic nose has the ability to be used as a fast and non-destructive method to identify potato varieties. Using this method will be very useful for consumers, especially restaurants and processing units, in order to choose high-quality cultivars.