CN105158178A - Rapid modeling method for detecting sugar content of navel orange based on spectral peak area in high spectral transmission technology - Google Patents

Abstract

The invention discloses a rapid modeling method for navel orange sugar content detection based on the spectral peak area of hyperspectral transmission technology, and relates to the technical field of non-destructive detection of internal quality of fruits. The method is as follows: ① Obtain the semi-transmission hyperspectral spectrum of the navel orange sample; ② determine the sugar content of the navel orange sample by chemical method; ③ select the average hyperspectral spectrum of the navel orange sample; ④ calculate the peak area of the hyperspectral spectrum; Carry out quality inspection. The present invention is based on hyperspectral technology and collects hyperspectral spectra of navel orange samples in a semi-transmission manner, which can effectively obtain internal quality information of navel oranges and improve the detection level and efficiency of fruit internal quality; the modeling method has high modeling efficiency and high accuracy, and the model operation The speed is fast, and the internal quality index of the sugar content of the fruit can be quickly detected, and the internal quality of the fruit can be evaluated.

Description

Rapid modeling method for sugar content detection of navel orange based on hyperspectral transmission technology spectral peak area

technical field

The invention relates to the technical field of non-destructive detection of fruit quality, in particular to a rapid modeling method for navel orange sugar detection based on the spectral peak area of hyperspectral transmission technology.

Background technique

my country is a big fruit producing country, and its output ranks first in the world all the year round, accounting for about 19% of the world's total fruit output. The fruit industry is developing rapidly, showing an upward trend every year. The fruit industry has developed into a pillar industry for farmers in many parts of our country to generate income, and it has greatly promoted and developed agriculture and the economy. Although my country is a large fruit producing country, the overall export volume is still very small, and it is not a strong country in fruit trade. Citrus fruit is the largest fruit in the world. my country's citrus production ranks third among countries in the world (accounting for 10.8%). The top two are Brazil (accounting for 23.7%) and the United States (accounting for 15.9%); to So far, my country's fruit export volume has not exceeded 4% of the total annual fruit output. At the same time, with the improvement of domestic people's consumption level, the consumption pattern of consumers has evolved from the requirement for product quantity to the requirement for quality. When buying fruits, people not only pay attention to the external quality of fruits, but also to the internal quality of fruits. In particular, the requirements for internal quality such as sugar content, taste and nutritional content have also increased accordingly. In order to promote fruit marketing and increase product value-added, it is necessary to commercialize fruits, that is, to reprocess and reprocess fruits after harvesting, to achieve fruit quality, enhance market competitiveness, expand exports, and improve the economic benefits of fruit planting and operators. It is an indispensable means; among them, how to quickly and non-destructively detect the internal quality of fruits is an important step in a series of fruit commercialization.

Due to the fast, non-destructive and reliable advantages of machine vision technology and spectral technology, they are currently widely used in non-destructive testing of agricultural products. Hyperspectral imaging technology is a fusion technology of image and spectrum, which can simultaneously obtain the spatial and spectral information of the detected object. The combination of image data and spectral data can be combined with external features to obtain the internal physical structure and chemical composition information of objects in a targeted manner, which has a good application prospect. However, for thick-skinned navel oranges, it is difficult to obtain internal information, and the data volume of hyperspectral image information is large, which increases the difficulty and time of data processing.

Contents of the invention

In order to solve the problems of thick navel orange peel, difficulty in obtaining internal information, large amount of hyperspectral image data, and difficulty in accurately and quickly detecting the sugar content in the rapid online detection process based on hyperspectral image technology, the present invention provides a hyperspectral transmission technology based The navel orange sugar detection rapid modeling method based on the spectral peak area establishes a prediction model for the navel orange sugar content, effectively extracts hyperspectral information, and improves the modeling efficiency and detection accuracy.

The purpose of the present invention is achieved like this:

Since the sugar content of navel orange is an important index to measure the quality of navel orange, the transmission method can effectively obtain the internal spectral information related to the sugar content of navel orange without causing internal damage to the navel orange due to excessive light source power; The sugar content of navel oranges can be judged by area, and the complex modeling process with fewer spectral variables has fast calculation speed and high accuracy, which can meet the requirements of rapid and non-destructive detection of sugar content of navel oranges.

Specifically, this method includes the following steps:

①Obtain the semi-transmission hyperspectral spectrum of the navel orange sample

Use a hyperspectral instrument to detect the hyperspectral spectrum of the navel orange sample, and set the acquisition method, exposure time, light source power, wavelength range, resolution and acquisition speed;

② Determination of sugar content of navel orange samples by chemical method

Measure the sugar content of navel orange sample according to the assay method described in national standard GB/T8210;

③Select the average hyperspectral spectrum of navel orange samples

Select the average hyperspectral spectrum of navel orange, according to the characteristics of the hyperspectral spectrum of navel orange, perform spectral preprocessing on the spectral curve in the MATLAB environment, remove non-target information, instrument noise, background interference, and irrelevant variable information such as water peaks, simplify the spectral information, keep important information;

④ Calculate the peak area of the hyperspectral spectrum

In the MATLAB environment, fit the spectral curve, adaptively select the peak and trough values of the spectral region, and calculate the peak area of the hyperspectral spectrum by integral;

⑤Establish a sugar content prediction model for navel orange samples for quality inspection

Using navel orange samples to adaptively select the ratio of the left and right spectrum peak areas in the spectral region as the input variable of the model, the unknown navel orange samples were tested for sugar content quality, and a linear regression quantitative detection model was established for navel orange sugar content. The square root error is used to evaluate the accuracy of the detection model.

The present invention has following advantage and positive effect:

① By collecting the semi-transmission hyperspectral spectrum of the navel orange sample, it can effectively solve the difficulty of obtaining the internal quality of the navel orange peel with thick spectroscopy;

②Using the correlation between the peak area of the hyperspectral spectrum and the sugar content change of the navel orange sample, the influence of the large amount of hyperspectral data on the calculation speed of the model is reduced;

③The rapid modeling method can meet the requirements of high-speed online detection of thick-skinned fruits, and complete the acquisition of effective spectral signals and the establishment of regression models in a short time, which improves the detection efficiency and detection accuracy.

In a word, the present invention collects navel orange sample hyperspectral spectrum through semi-transmission method based on hyperspectral technology, can effectively obtain navel orange internal quality information, and improve the detection level and detection efficiency of fruit internal quality; the modeling method has high modeling efficiency and high accuracy, The calculation speed of the model is fast, and it can quickly detect the internal quality index of the sugar content of the fruit and evaluate the internal quality of the fruit.

Description of drawings

Fig. 1 is the structural representation of this device, among the figure:

0—navel orange sample,

1—computer, 2—box, 3—hyperspectral imager, 4—electric translation stage,

5—manual lifting platform, 6—light source, 7—sample room;

Fig. 2 is the step figure of this law;

Fig. 3 is the semi-transmission mean spectral curve of single navel orange sample;

Fig. 4 is the navel orange spectrum curve after curve fitting;

Fig. 5 is the correlation coefficient figure of navel orange sample modeling set sugar content predicted value and measured value;

Fig. 6 is a graph of the correlation coefficient between the predicted value and the measured value of sugar content in the prediction set of navel orange samples.

Detailed ways

Below in conjunction with accompanying drawing and embodiment describe in detail:

1. Device

As shown in Figure 1, the device includes the working object - navel orange sample 0;

It is equipped with a computer 1, a cabinet 2, a hyperspectral imager 3, an electric translation platform 4, a manual lifting platform 5, a light source 6 and a sample chamber 7;

Its location and connection relationship are:

Bottom in casing 2 is provided with electric displacement stage 4, and the top of electric displacement stage 4 is provided with manual elevating platform 5, is provided with light source 6 in sample room 7, is placed with navel orange sample 0 above manual elevating platform 5 ;

A hyperspectral imager 3 is arranged on the top of the box body 2;

The electric translation stage 4 and the hyperspectral imager 3 are respectively connected to the computer 1 outside the box body 2 .

The above-mentioned functional parts are all common parts.

Its working mechanism is: turn on the computer 1 and the hyperspectral imager 3, connect the light source 6 of the sample chamber 5, preheat for 30 minutes, and set the exposure time, wavelength range and resolution and motor The acquisition speed and the acquisition position of the translation stage 4 adjust the objective lens of the hyperspectral imager 3 to obtain a clear image of the navel orange sample 0; meanwhile, the navel orange sample 0 is placed on the manual lifting platform 5, and the navel orange sample 0 and the navel orange sample 0 in the sample chamber 7 are adjusted. The position of the light source 6 is clicked on the software to start the acquisition button, and the electric translation stage 4 moves to complete the acquisition of the hyperspectral image of the navel orange sample.

2. Method

As shown in Figure 2, this method includes the following steps:

①Obtain the semi-transmission hyperspectral spectrum A of the navel orange sample

A hyperspectral instrument was used to detect the hyperspectral spectrum of the navel orange sample. The acquisition method was semi-transmission mode. The spectral light source was 4 50W halogen lamps. The total power of the light source was 200W. cm/s with a resolution of 32cm ^-1 ; collect the transmission hyperspectral information of the navel orange sample at the equator, and automatically obtain the spectral values within the image range with a pixel size of 120×120pixels in the middle of the image;

② Determination of sugar content B of navel orange samples by chemical methods

Measure the sugar content of navel orange sample according to the assay method described in national standard GB/T8210;

③Choose the average hyperspectral spectrum C of navel orange samples

Select the average hyperspectral spectrum of the navel orange sample, and according to the characteristics of the hyperspectral spectrum of the navel orange, perform S-G smoothing preprocessing on the spectral data in MATLAB software, remove non-target information, instrument noise, background interference, and irrelevant variable information such as water peaks, and simplify the spectrum Information, keep important information;

④ Calculate the peak area D of the hyperspectral spectrum

Fit the spectral curve in MATLAB software, adaptively select the peak and valley values in the spectral region of 390-1055nm, and calculate the peak area of the hyperspectral spectrum by integral;

⑤Establish the sugar content prediction model of navel orange samples, and carry out quality inspection.

Utilize the navel orange sample to adaptively select the ratio of the peak area of the spectral spectrum in the 390-1055nm spectral region as the input variable of the model, carry out sugar content quality detection on the unknown navel orange sample, establish a linear regression quantitative detection model for the sugar content, use the prediction set correlation coefficient and the prediction set The root mean square error was used to evaluate the accuracy of the detection model.

3. Test results

Taking the hyperspectral image spectrum of the navel orange sample as an example, after the integral calculation of the peak area of the hyperspectral spectrum, the sugar content detection model of the navel orange was established, and the detection accuracy of the model was analyzed.

Fig. 2 is the step diagram of this method, and Fig. 3 is the semi-transmission average spectral curve of a single navel orange sample, and the spectral range is 390-1055nm; the hyperspectral spectrum information of the navel orange sample is collected, the collection method is semi-transmission mode, and the spectral light source is 4 50W Halogen lamp, the total power of the light source is 200W, the set exposure time is 100ms, the wavelength range is 300-1100nm, the acquisition speed is 0.2cm/s, and the resolution is 32cm ^-1 ; Automatically acquire the average spectral value within the image range of 120×120pixels in the middle of the image; after collecting the spectrum, measure the sugar content of the navel orange sample according to the measurement method described in the national standard GB/T8210, and the sugar content range is 9.6～12.6 ^O Brix; The samples are divided into modeling set and prediction set according to the ratio of 2:1, and the sugar content value of the samples in the modeling set includes the sugar content value of the samples in the prediction set.

In the described navel orange sugar content rapid modeling method, an image with a pixel size of 120×120pixels is selected from the center of the navel orange hyperspectral image in the ENVI software, and the average spectrum of each image is extracted, and the spectral range is 390-1055nm.

The obtained navel orange sample spectrum was preprocessed by Savitzky-Golay smoothing, and the curve fitting was carried out in Matlab software to obtain the peak and trough positions of the spectrum in the curve, and the peak area of the two peak regions in the navel orange hyperspectral spectrum was calculated.

The ratio of the left and right peak areas in the calculated spectrum is used as the input variable for the sugar content prediction of navel orange, and a linear regression model is established; the linear model RMSEC and RMSEP are respectively 0.312 ^O Brix and 0.306 ^O Brix, which are relatively close. At this time, the modeling set and The correlation coefficients for the prediction sets are 0.832 and 0.846, respectively.

Figures 5 and 6 are the correlation coefficient diagrams of the predicted value and the measured value of the watermelon sugar content of the linear model; it can be seen from the figure that the built linear model has obtained a better prediction result.

Claims (3)

1. a navel orange sugar content detection rapid modeling method based on hyperspectral transmission technology spectral peak area, is characterized in that comprising the following steps: ①Obtain the semi-transmission hyperspectral spectrum of the navel orange sample (A) Use a hyperspectral instrument to detect the hyperspectral spectrum of the navel orange sample, and set the acquisition method, exposure time, light source power, wavelength range, resolution and acquisition speed; ② Determination of sugar content of navel orange samples by chemical method (B) Measure the sugar content of navel orange sample according to the assay method described in national standard GB/T8210; ③ Select the average hyperspectral spectrum of navel orange samples (C) Select the average hyperspectral spectrum of navel orange, according to the characteristics of the hyperspectral spectrum of navel orange, perform spectral preprocessing on the spectral curve in the MATLAB environment, remove non-target information, instrument noise, background interference, and irrelevant variable information such as water peaks, simplify the spectral information, keep important information; ④ Calculate the peak area of the hyperspectral spectrum (D) In the MATLAB environment, fit the spectral curve, adaptively select the peak and trough values of the spectral region, and calculate the peak area of the hyperspectral spectrum by integral; ⑤Establish the sugar content prediction model of navel orange samples for quality inspection (E) Using navel orange samples to adaptively select the ratio of the left and right spectrum peak areas in the spectral region as the input variable of the model, the unknown navel orange samples were tested for sugar content quality, and a linear regression quantitative detection model was established for navel orange sugar content. The square root error is used to evaluate the accuracy of the detection model. 2. by the navel orange sugar content detection fast modeling method described in claim 1, it is characterized in that described step 1. the device that it adopts is: An electric displacement stage (4) is arranged at the bottom of the box body (2), a sample chamber (7) is arranged on the upper part of the electric displacement stage (4), and a light source (6) is arranged in the sample chamber (7). A navel orange sample (0) is placed on the manual lifting platform (5); A hyperspectral imager (3) is arranged on the top of the box (2); The electric displacement stage (4) and the hyperspectral imager (3) are respectively connected with the computer (1) outside the box (2). 3. by the navel orange sugar content detection fast modeling method described in claim 1, it is characterized in that described step 1. the parameter of its setting is: A hyperspectral instrument was used to detect the hyperspectral spectrum of the navel orange sample. The acquisition method was semi-transmission mode. The spectral light source was 4 50W halogen lamps. The total power of the light source was 200W. cm/s with a resolution of 32cm ^-1 ; the transmission hyperspectral information of the navel orange sample at the equator was collected, and the spectral values within the image range with a pixel size of 120×120pixels in the middle of the image were automatically obtained respectively.