Long Xue

Hyperspectral imaging technology for determination of dichlorvos residue on the surface of navel orange

A hyperspectral imaging system is developed to detect dichlorvos residue on the surface of navel orange. After acquiring hyperspectral images of 400 navel oranges, the actual content of dichlorvos residue is measured by gas chromatography. Optimal wavelengths are extracted using the regression coefficients of partial least squares (PLS), and a PLS model with 12 factors is established. In the prediction set of 0.2282-11.652-mg/kg pesticide residue, the correlation coefficient and the root mean standard error are 0.8320 and 1.3416, respectively. The hyperspectral imaging technology can meet the requirement of online fast nondestructive detection.

Vitamin C content estimation of chilies using Vis/NIR spectroscopy

Vitamin C has value in treatment or prevention of scurvy and it can be obtained from vegetable and fruit. Vitamin C is usually determined by traditional chemical methods which are destructive, time-consuming. This paper was conducted to study the vitamin C (VC) content estimation in chilies using quantitative analysis technique based on visible/near infrared (Vis/NIR) diffuse reflectance spectroscopy. Total 141 fresh chilies were purchased from market. After the samples have been washed and air-dried, Vis/NIR spectral data were collected using a QualitySpec® Pro Vis/NIR spectrometer (ASD Inc.). Then the vitamin C contents in samples were determined by 2, 6-dichloro-indophenol titration method. Spectral preprocessing techniques, including standard normal variate (SNV), multiplicative scatter corrections (MSC), first derivative (FD), second derivative (SD) and smoothing methods were applied to the spectral data and examined for their effectiveness at reducing or eliminating scatter effects. Partial least squares (PLS) regression was applied to examine the impact of the preprocessing transforms on assessing the content of vitamin C in chilies. The result shows that the best calibration model can be obtained by the first derivative preprocessing method in the spectral range of 450–1000nm. The prediction results are 0.803 and 0.509 for correlation coefficient (r) and root mean square errors of prediction (RMSEP), respectively. The study shows that vitamin C (VC) content in chilies can be effectively estimated using Vis/NIR spectroscopy technology.

Detecting Navel Orange Canker with Hyperspectral Imaging

Detection of navel orange canker by hyper spectral imaging technique was proposed in this work. Navel orange was adopted as the experimental object. The hyper spectral images of navel oranges were collected between 400 nm and 1000 nm wavelength. Principal component analysis (PCA) was performed to determine optical wavelength (672nm). The feature images under 672nm were selected to establish region of interest and build the mask. After analyzing the feature image by PCA, we found the second principal component (PC2) image to be most suitable for identifying the presence of canker. Finally, the feature of canker was extracted through determination of optional threshold and morphological image processing. The result shows that navel orange canker can be detected with an accuracy of 92.5% by hyper spectral imaging technique.

Study of Pesticide Contaminated Navel Orange Recognition Using near Infrared Spectroscopy

Based on Support Vector Machine (SVM) and genetic algorithm (GA), this paper intends to search for the characteristic spectral ranges and wavelengths of near infrared spectroscopy of navel oranges contaminated by different pesticides, and set up recognition models. The pesticides in the experiment were Lannate®L insecticide, fenvalerate and omethoate, and three different concentrations were given to each pesticide. Preparing ten groups of navel oranges, each group was sprayed with a different pesticide and the 10th group without pesticide spraying was used for comparison. Searching the whole spectral range through GA, 5 best spectral ranges (165 wavelengths) were obtained and the recognition rate reached 98.86%. Then based on the chosen spectral ranges, 85 feature wavelengths were extracted with continual GA-SVM optimization, and the recognition rate was 99.14%. Experiment results showed that the application of SVM combining with GA can not only improve recognition accuracy, but also simplify the model effectively

Determination of dichlorvos contamination on navel orange surface based on least squares support vector machines

Spectral technique can provide a rapid, nondestructive means to assess quality and safety of agricultural commodities for human consumption. A procedure for determination of dichlorvos contamination has been developed with Vis-NIR spectroscopy. Four spectral preprocessing methods, including multiplicative scatter corrections (MSC), standard normal variate (SNV), first derivative (FD) and second derivative (SD) were used to reduce or eliminate scatter effects. Based on the spectral preprocessing methods, least squares support vector machines (LS-SVM) was performed. A total of 160 navel oranges were separated into two sets, one was calibration set (including 110 samples), and the other was prediction set (including 50 samples). It was found that LS-SVM model with the application of the preprocessing method of FD could predict dichlorvos residue. In the prediction set, the root mean squared error of prediction samples (RMSEP) was 6.2598 and the correlation coefficient Rpre was 0.8174.

The Detection of Early-Maturing Pear’s Effective Acidity Based on Hyperspectral Imaging Technology

The hyperspectral imaging technology is used to detect early-maturing pear’s effective acidity nondestructively, and effective prediction model is established. 145 pears’ hyperspectral images are obtained in the wavelength range of 400nm-1000nm. Total 145 pears are separated into the calibration set (77 samples) and prediction set (68 samples). Early-maturing pear’s effective acidity partial least squares (PLS) prediction model is built in different range of spectrum band. By comparison, the range 498 nm - 971 nm was selected in using partial least squares (PLS) to build early-maturing pear’s effective acidity prediction model. The experimental results show that, PLS prediction model of early-maturing pear’s effective acidity has the best effect in this range of wavelength. The correlation coefficient R between early-maturing pear’s actual effective acidity and predicted effective acidity is 0.9944 and 0.9233 for calibration set and prediction set respectively, the root mean squared error of prediction samples (RMSEP) is 0.022 and 0.072 for calibration set and prediction set respectively.

Nondestructive Measurement of Sugar Content in Navel Orange Based on Vis-NIR Spectroscopy

The Vis-NIR spectroscopy technology was studied on nondestructive measurement of sugar content in navel orange. Three different spectral ranges of 450-1000nm,1000-1800nm and 450-1800nm were selected, respectively and five different spectral pre-processing methods of Standard Normal Variate (SNV), first derivative (FD), second derivative (SD), multiplicative scatter correction (MSC) and smoothing were used for establishing the partial least squares(PLS)models, which was determined the sugar content in navel orange. The results showed that the model developed from 450-1800nm spectroscopy after SNV pre-processing achieved the optimal performance. The correlation coefficient (r2) of the calibration set and validation set were 0.9349 and 0.8514 respectively, and the root mean squared error of calibration and validation sets were 0.8017 and 1.1649, respectively. The research indicated that the Vis-NIR spectroscopy technique was feasible to nondestructive measurement of sugar content in navel orange.

Application of laser to nondestructive detection of fruit quality

In this study, a hyperspectral imaging system using a laser source was developed and two experiments were carried out. The first experiment was detection of pesticide residue on navel orange surface. We calculated the mean intensity of regions of interest to plot the curves between 629nm to 638nm. The analysis of the mean intensity curves showed that the mean intensity can be described by a characteristic Gaussian curve equation. The coefficients a in characteristic equations of 0%, 0.1% and 0.5% fenvalerate residue images were more than 2400, 1570-2400 and less than 1570, respectively. So we suggest using equation coefficient a to detect pesticide residue on navel orange surface. The second experiment was predicting firmness, sugar content and vitamin C content of kiwi fruit. The optimal wavelength range of the kiwi fruit firmness, sugar content, vitamin C content line regressing prediction model were 680-711nm, 674-708nm, 669-701nm. The correlation coefficients (R) of prediction models for firmness, sugar content and vitamin C content were 0.898, 0.932 and 0.918. The mean errors of validation results were 0.35×105Pa, 0.32%Brix and 7mg/100g. The experimental results indicate that a hyperspectral imaging system based on a laser source can detect fruit quality effectively.

Visible and Near infrared reflectance spectroscopy for determining soluble solids content of navel orange

Soluble solids content (SSC) is one of the most important quality attributes of navel oranges, either for fresh or for processing market. Since, SSC can be measured only destructively, the results are representative only if carried out on large samples and do not allow classifying marketable fruit one by one according to their specific SSC. As a nondestructive method, Vis/NIR measurement of SSC of fruits is widely studied all over the world. The objective of this research is to determine how Vis/NIR measurements of SSC of navel oranges were affected by spectra pretreatment and various spectroscopic ranges (Vis:400–799 nm; NIR1:800–999 nm; NIR2:1000–1800 nm) using partial least squares (PLS) regression. Spectra correction algorithms, such as multiplication scatter correction (MSC), first derivative (FD) and second derivative (SD) were used and evaluated in this work. The FD correction and PLS produced best noise removing capability and obtained optimal calibration models. This model could predict SSC of navel oranges of prediction set with correlation coefficient (Rpre) of 0.8553 and standard errors of prediction (RMSEP) of 1.6330.

Determination of Moisture Content in Ginger Using PSO Combined with Vis/NIR

Vis/NIR spectroscopy was used to measure the moisture content of ginger. 330 samples were separated into two groups, as training and validation. Vis/NIR reflection spectral data from 350 to 1800 nm were collected using ginger within the training and validation sets. PSO was used to establish the PLS model. In comparison to the full spectrum model (contained 1451 variables), the prediction capability was improved after using PSO for PLS models. The number of selected variables and LVs were 300 and 6, respectively. The correlation of determination in validation set (), root mean square error of prediction (RMSEP), and bias by PSO-PLS were 0.9881, 4.7827, and 0.1751.