A. Moghimi; A. Sazgarnia; M. H. Aghkhani
Abstract
IntroductionPistachio production has been adversely affected by Psylla, which is a devastating insect. The primary goal of this study was to select sensitive spectral bands to distinguish pistachio leaves infected by Psylla from healthy leaves. Diagnosis of psylla disease before the onset of visual cues ...
Read More
IntroductionPistachio production has been adversely affected by Psylla, which is a devastating insect. The primary goal of this study was to select sensitive spectral bands to distinguish pistachio leaves infected by Psylla from healthy leaves. Diagnosis of psylla disease before the onset of visual cues is crucial for making decisions about topical garden management. Since it is not possible to diagnose psylla disease even after the onset of symptoms with the help of color images by drones, hyperspectral and multispectral sensors are needed. The main purpose of this study was to extract spectral bands suitable for distinguishing healthy leaves from psylla leaves. For this purpose, in this paper, a new method for selecting sensitive spectral properties from hyperspectral data with the high spectral resolution is presented. The intelligent selection of sensitive bands is a convenient way to build multispectral sensors for a specific application (in this article, the diagnosis of psylla leaves). Knowledge of disease-sensitive wavelengths can also help researchers analyze multispectral and hyperspectral aerial images captured by satellites or drones.Materials and MethodsA total number of 160 healthy and diseased leaves were scanned in 64 spectral bands between 400-1100 nm with 10 nm spectral resolution. A random forest algorithm was used to identify the importance of features in classifying the dataset into diseased and healthy leaves. After computing the importance of the features, a clustering algorithm was developed to cluster the most important features into six clusters such that the center of clusters was 50 nm apart. To transfer the hyperspectral dataset into a multispectral dataset, the reflectance was averaged in spectral bands within ±15 nm of each cluster center and achieved six broad multispectral bands. Afterwards a support vector machine algorithm was utilized to classify the diseased and healthy leaves using both hyperspectral and multispectral datasets.Results and DiscussionThe center of clusters were 468 nm, 598 nm, 710 nm, 791 nm, 858 nm, and 1023 nm, which were calculated by taking the average of all the members assigned to the individual clusters. These are the most informative spectral bands to distinguish the pistachio leaves infected by Psylla from the healthy leaves. The F1-score was 90.91 when the hyperspectral dataset (all bands) was used, while the F1-score was 88.69 for the multispectral dataset. The subtle difference between the F1-scores indicates that the proposed pipeline in this study was able to select appropriately the sensitive bands while retaining all relevant information.ConclusionThe importance of spectral bands in the visible and near-infrared region (between 400 and 1100 nm) was obtained to identify pistachio tree leaves infected with psylla disease. Based on the importance of spectral properties and using a clustering algorithm, six wavelengths were obtained as the best wavelengths for classifying healthy and diseased pistachio leaves. Then, by averaging the wavelengths at a distance of 15 nm from these six centers, the hyperspectral data (64 bands) became multispectral (6 bands). Since the correlation between the wavelengths in the near-infrared region was very high (more than 95%), out of the three selected wavelengths in the near-infrared region (710, 791, and 1023), only the 710-nm wavelength, which was closer to the visible region, was selected. The results of classification of infected and diseased leaves using hyperspectral and multispectral data showed that the degree of classification accuracy decreases by about 2% and if only 4 bands are used, the degree of accuracy decreases by about 3%.The results of this study revealed that the proposed framework could be used for selecting the most informative spectral bands and accordingly develop custom-designed multispectral sensors for disease detection in pistachio. In addition, we could reduce the dimensionality of the hyperspectral datasets and avoid the issues related to the curse of dimensionalitylity.
T. Mohammadi Moghaddam; S. M. A. Razavi; M. Taghizadeh; A. Sazgarnia; B. Pardhan
Abstract
Introduction: Pistachio nut is one of the most delicious and nutritious nuts in the world and it is being used as a salted and roasted product or as an ingredient in snacks, ice cream, desserts, etc. (Maghsudi, 2010; Kashaninejad et al. 2006). Roasting is one of the most important food processes which ...
Read More
Introduction: Pistachio nut is one of the most delicious and nutritious nuts in the world and it is being used as a salted and roasted product or as an ingredient in snacks, ice cream, desserts, etc. (Maghsudi, 2010; Kashaninejad et al. 2006). Roasting is one of the most important food processes which provides useful attributes to the product. One of the objectives of nut roasting is to alter and significantly enhance the flavor, texture, color and appearance of the product (Ozdemir, 2001). In recent years, spectral imaging techniques (i.e. hyperspectral and multispectral imaging) have emerged as powerful tools for safequality inspection of various agricultural commodities (Gowen et al., 2007). The objectives of this study were to apply reflectance hyperspectral imaging for non-destructive determination of moisture content and hardness of pistachio kernels roasted in different conditions.
Materials and methods: Dried O’hadi pistachio nuts were supplied from a local market in Mashhad. Pistachio nuts were soaked in 5L of 20% salt solution for 20min (Goktas Seyhan, 2003). For roasting process, three temperatures (90, 120 and 150°C), three times (20, 35 and 50 min) and three air velocities (0.5, 1.5 and 2.5 m s-1) were applied. The moisture content of pistachio kernels was measured in triplicate using oven drying (3 gr samples at 105 °C for 12 hours). Uniaxial compression test by a 35mm diameter plastic cylinder, was made on the pistachio kernels, which were mounted on a platform. Samples were compressed at a depth of 2mm and speed of 30 mm min-1. A hyperspectral imaging system in the Vis-NIR range (400-1000 nm) was employed. The spectral pre-processing techniques: first derivative and second derivative, median filter, Savitzkye-Golay, wavelet, multiplicative scatter correction (MSC) and standard normal variate transformation (SNV) were used. To make models at PLSR and ANN methods, ParLeS software and Matlab R2009a were used, respectively. The coefficient of determination (R2), the root mean square error of prediction (RMSEP) and the ratio of the standard deviation of the response variable to RMSEP (known as relative performance determinant (RPD)) were calculated.
Results and discussion:
Interpretation of hyperspectral data: The results showed that the spectra of the shell, the whole kernel and the internal part of the kernel have different patterns. The internal part of thekernel had 2 peaks at 630 nm and 690 nm, while the shell and the whole kernel had 1 peak at 670 nm and 720 nm, respectively and the peak of the whole kernel was sharper than that of the shell. The highest and lowest intensities were for the internal part of the kernel and the whole kernel, respectively. The spectral slope of the internal part is higher than that of the shell and the whole kernel at 500-700 nm.
The effect of different pre-processing techniques and analysis on prediction of pistachio kernels properties: In the absence of pre-processing techniques, low correlation coefficients were observed for prediction of moisture content and hardness. However, with the use of pre-processing techniques, in some models, correlation coefficient and RPD increased and the RMSEP decreased. The results revealed that ANN models would predict moisture content and textural characteristics of roasted pistachio kernels better than PLSR models.
Moisture content: ANN models can predict moisture content of roasted pistachio kernels better than PLSR models. In total, PLSR models showed low RPD and R2. For all samples, RPD was lower than 1.5, indicating that the developed models do not give an accurate prediction for moisture content. The best results with ANN method were achieved using a combination of SNV, wavelet and D1 for predicting moisture content with R2 =0.907 and RMSEP=0.179.
Hardness: The results indicated that ANN models can predict the hardness better than PLSR models. The best results with PLSR models were achieved using a combination of SNV, wavelet and D1 with R2= 0.643, RMSEP=10.78, RDP= 1.48 and 2 PLSR factors. However, due to high RMSEP and low R2 and RPD, it can be mentioned that prediction of hardness values with ANN model was not sufficiently desirable. However it was better than the PLSR models. The best results with ANN models were achieved using a combination of SNV, wavelet and D2 with R2=0.876 and RMSEP=5.216.
Conclusions: The results of this study showed that employing pre-processing methods causesa decrease in prediction error and improves the quality of the models. ANN models could predict moisture content and hardness of roasted pistachio kernels better than the PLSR models.