Large thick-peel melon and fruit diffuse transmission spectrum nondestructive testing method and portable device

Abstract

The invention discloses a diffuse transmission spectrum nondestructive testing method and a portable device for large thick-peel melons and fruits, belongs to the field of nondestructive testing of fruit quality, and provides a real-time and portable internal quality nondestructive testing method and device suitable for the large thick-peel melons and fruits. The system comprises a camera obscura, a shading sucker, an annular shading sponge, a light source module, a spectrum detection module, a collimating lens, an optical fiber, a Raspberry Pi controller and a touch screen. During working, the halogen light source emits a focused light beam to penetrate through melons and fruits, the focused light beam is converted into an electric signal by the spectrum detection module after being scattered and transmitted, and the Raspberry Pi analyzes in real time to obtain and display indexes such as hardness, pH, soluble solids and the like. A 220V industrial power supply is adopted to drive a high-power light source, a cooling fan and a shading sucker structure are combined to reduce environmental interference, and an annular shading sponge ensures that a light path is closed. An integrated portable device of a high-power light source is realized, and the device can be used for on-site rapid nondestructive detection of key quality indexes of large thick-peel melons and fruits such as Hami melons and the like.

Description

Technical Field

[0001] The invention relates to the field of nondestructive testing of fruit quality, and in particular to a diffuse transmission spectrum nondestructive testing method for large thick-skinned melons and fruits and a portable device. Background Art

[0002] With the continuous advancement of agricultural technology, precision agriculture has become an important means to improve agricultural production efficiency and ensure the quality of agricultural products. This requires us to be able to detect the quality of large thick-skinned fruits and melons with high efficiency and low loss. The quality inspection of large thick-skinned fruits and melons is not only related to their external qualities such as color, smell, skin defects, shape and size, but also depends on key internal quality indicators such as hardness, soluble solids, pH, etc. At present, the quality inspection of large thick-skinned fruits and melons mainly relies on manual experience and chemometrics. Manual experience is easily affected by subjective factors, and the judgment results are random and have large errors. Chemometrics mostly uses destructive testing, which is complicated to operate and is only suitable for small batch sampling. Therefore, it is necessary to provide a portable diffuse transmission spectrum detection device for the internal quality of large thick-skinned fruits and melons to achieve rapid, accurate and non-destructive detection of key internal quality indicators of large thick-skinned fruits and melons.

[0003] The Chinese patent number CN107831133A was published on March 23, 2018. The invention name is "A near-infrared spectroscopy online detection system and implementation method for the internal quality of large fruits and melons". The application discloses "A near-infrared spectroscopy online detection system and implementation method for the internal quality of large fruits and melons. The method uses a spectrometer to collect diffuse transmission spectra of large fruits and melons, connects the spectrometer to a computer via a USB cable, and cooperates with spectrum collection software to collect and calculate the original static spectrum and dynamic spectrum of the sample, and displays the sugar content value of the sample on the computer detection interface". The "near-infrared spectroscopy online detection system and implementation method for the internal quality of large fruits and melons" is characterized by designing an adjustable angle light source fixing frame, processing and analyzing the collected spectrum through the software system, realizing non-destructive detection of the internal quality of fruits and melons, and combining with a spiral conveying device to realize online detection of the internal quality of large fruits and melons. The problem with this method is that the detection device is large in size, and a computer is used to collect and calculate spectral data, which is not convenient to carry.

[0004] China's public patent number CN114509387A, published on May 17, 2022, is titled "Portable kiwifruit internal quality nondestructive detection method and device". The application discloses "a portable kiwifruit internal quality nondestructive detection method and device", which is characterized by: "including a model building step and a detection device design step, the model building step includes collecting the sample black and white correction spectrum and the sample original spectrum, and obtaining the sample corrected spectrum by calculation, thereby establishing a kiwifruit internal quality prediction model to obtain the specific values ​​of the predicted sample's internal quality, namely, sugar content, hardness, and moisture content; the detection device design step is characterized by the selection of detection device components; the detection device component selection includes a light shield, a detection probe, a bracket, a control processing module, a wireless communication module, a button fixing shaft, a measurement button, a lithium battery, a function button, a display screen, a casing, and a correction whiteboard; the detection probe is composed of a photoelectric sensor, a light-emitting diode, and an optical lens". The light source used in this invention is an LED light source, and the collected spectrum is a diffuse reflection spectrum, which realizes nondestructive detection of the internal quality of small, thin-skinned fruits such as kiwifruit. The problems with this method are: the light source intensity of the detection device is low, and only small thin-skinned fruits can be detected; and the diffuse reflection mode contains more fruit skin information and surrounding stray light information, which will affect the detection accuracy of the model.

[0005] In summary, the current visible-near infrared spectroscopy technology has achieved certain results in the non-destructive testing of small thin-skinned fruits, but there is little research on the non-destructive testing of the internal quality of large thick-skinned fruits. At present, the non-destructive testing of the internal quality of large thick-skinned fruits still mainly relies on imported equipment, and the spectral detection mode is mostly diffuse reflection mode. This type of technology has low detection accuracy, high cost, and is not portable. Summary of the invention

[0006] In view of the problems existing in the prior art, the present invention provides a diffuse transmission spectrum nondestructive testing method and a portable device for large thick-skinned melons and fruits.

[0007] The present invention is achieved through the following technical solutions: A portable diffuse transmission spectrum detection device for the internal quality of large thick-skinned melons and fruits, comprising an annular shading sponge, a dark box, a cooling fan, a 12V DC voltage to 5V DC voltage transformer, a touch display screen, a control button, a 220V AC voltage to 12V DC voltage DC power supply, a shading suction cup, a light source module, a 74UV collimating lens, a heat dissipation hole, a circuit board heat dissipation fan, a Raspberry Pi controller, an optical fiber, a power cord, a spectrum detection module, a 220V ship-shaped power switch, a support foot, a light source frame fixing screws, and a spectrum collection hole; the light source module comprises a halogen light source, a light source bracket, and a fixing hole; the halogen light source comprises four 12V, 20W halogen lamps with a wavelength range of 350-2500nm, each halogen lamp is spaced 90° apart, and is distributed in a surrounding manner with the sample detection site; the light source frame is a four-sided light source bracket, which is fixed by fixing holes.

[0008] The spectrum collection hole is opened at the center of the top of the dark box, and the diameter of the circular hole is 12 cm, which is smaller than the average horizontal diameter of the sample; the annular shading sponge is installed and fixed on the spectrum collection hole on the dark box; the cooling fan is symmetrically installed on the left and right sides of the dark box, and is flush with the height of the light source module; the heat dissipation hole is fixed to the cooling fan installation position; the 12V DC voltage to 5V DC voltage transformer is installed and fixed on the back of the dark box; the 220V AC voltage to 12V DC voltage DC power supply is fixed inside the dark box shell; the touch display screen is fixedly installed on the left side of the front of the dark box; the control button is fixedly installed on the right side of the front of the dark box; the shading suction cup is fixed on the top of the 74UV collimating lens, and The same height as the annular shading sponge; the light source module is fixed on the top of the dark box; the bottom of the 74UV collimating lens is fixed to the center hole of the light source module; the Raspberry Pi controller is fixed to the bottom of the dark box with a rubber column; the circuit board cooling fan is fixed to the upper part of the Raspberry Pi controller with a rubber column; the spectrum detection module is fixed inside the dark box shell; one end of the optical fiber is connected to the SMA905 interface of the 74UV collimating lens, and the other end is connected to the SMA905 interface of the spectrum detection module; the 220V boat-shaped power switch is installed and fixed on the lower right side of the left part of the dark box; the left end of the power cord jack is inserted into the 220V boat-shaped power switch plug; the support feet are symmetrically installed at the four corners of the bottom of the dark box; In the light source module, the light source bracket is a four-sided light source bracket with a horizontal angle of 150°, which is fixed to the top of the dark box shell through fixing holes and light source bracket fixing screws. Four circular holes with the same diameter as the halogen lamp are opened on the four sides of the light source bracket for fixing the four halogen lamps in the halogen light source, ensuring that the emitted light of the halogen light source forms an angle of 30° with the surface of the 74UV collimating lens; the bottom of the 74UV collimating lens is fixed to the center hole of the light source bracket of the light source module; the top optical lens of the 74UV collimating lens is fixed to the center of the light-shielding suction cup.

[0009] The power cord jack is inserted into the 220V boat-shaped power switch three-pin socket; the 220V boat-shaped power switch and the 220V AC voltage to 12V DC voltage DC power supply are connected by a wire to output a 12V power supply; one power supply end of the 220V AC voltage to 12V DC voltage DC power supply is connected to the halogen light source, the cooling fan and the control button by a wire to form a light source cooling circuit; the other power supply end of the 220V AC voltage to 12V DC voltage DC power supply is connected to the 12V DC voltage to 5V DC voltage transformer by a wire to output a 5V power supply; the 12V DC voltage to 5V DC voltage transformer is connected to the Raspberry Pi controller, the circuit board cooling fan, the touch display screen, the spectrum detection module and the control button by a wire to form a control circuit; the circuit board cooling fan, the touch display screen and the spectrum detection module are all controlled by the Raspberry Pi controller; The three-pin plug of the power cord is inserted into a 220V industrial power supply, and the main circuit of the device is powered under the control of the 220V boat-shaped power switch; the halogen light source is lit under the control of the control button, and the cooling fan, Raspberry Pi controller, circuit board cooling fan and touch display screen are powered; the spectrum detection module receives the sample spectrum signal collected by the 74UV collimating lens, converts it into an electrical signal and transmits it to the Raspberry Pi controller for processing and calculation, and displays the result on the touch display screen.

[0010] The dark box and light source bracket are both made of aluminum alloy, and the inner and outer walls are sprayed with black matte material to prevent reflection caused by the halogen light source irradiating the inner wall; the annular shading sponge is made of black sponge material and has a shading function, which can be used to prevent the influence of external stray light; the shading suction cup is made of black rubber material, which is soft and can block light, so as to prevent the sample spectrum from carrying surface texture feature information and realize diffuse transmission spectrum collection.

[0011] The length, width and height of the portable large-scale thick-skinned melon and fruit internal quality diffuse transmission spectrum detection device are 30 cm, 24 cm and 15 cm respectively.

[0012] A diffuse transmission spectrum nondestructive testing method for large thick-skinned melons and fruits, characterized in that visible-near infrared light emitted by a halogen light source passes through a large thick-skinned melon and fruit sample, is received by a collimating lens with a light-shielding sucker below after multiple scattering and transmission, is transmitted to a spectrum detection module through an optical fiber, converted into an electrical signal, and then transmitted to a Raspberry Pi controller for processing, and the calculation result is displayed on a touch display screen, comprising the following steps: a. Turn on the main power: insert the power cord plug into the 220V industrial power supply, and insert the jack into the 220V rocker-shaped power switch, press the rocker-shaped switch on the 220V rocker-shaped power switch to complete the power supply operation of the device system; b. Preheating: Use the control button to turn on the light source module for preheating for about 30 minutes, and use the control button to turn on the heat dissipation module at the same time; c. Collect reference spectrum T w and T d :Use the control button to turn on the light source module, cover the detection hole of the shading sponge with a calibration white plate and close to the shading suction cup above the 74UV collimating lens, set and collect the white reference spectrum T w ; Use the control button to turn off the light source module, cover the shading sponge detection hole with a black shading cloth and close to the shading suction cup above the 74UV collimating lens, set and collect the black reference spectrum T d ; d. Collect the sample diffuse transmission spectrum T s :Put the sample horizontally on the annular light-shielding sponge so that its curved surface fits the annular light-shielding sponge to form a closed light-shielding environment. At the same time, make its equatorial detection part close to the light-shielding suction cup. Then, through the collimating lens and optical fiber connected to the light-shielding suction cup, the spectral information is transmitted to the spectral detection module to complete the sample diffuse transmission spectrum T s collection; e. Calculate the sample diffuse transmission correction spectrum T c :According to the collected black and white reference spectrum T d 、T w and the original spectrum T s , using the formula Calculate the sample diffuse transmission correction spectrum T c ; The diffuse transmission spectrum T c Wavelength range is 500-1050nm; f. Establishment of the internal quality prediction model of large thick-skinned melons and fruits: The sample diffuse transmission correction spectrum obtained by the detection device is used as the input, and the physical and chemical values ​​of the key quality indicators of the samples obtained by the traditional destructive detection methods of the sugar refractometer, pH meter and texture analyzer are used as the output; after detrending DT, SG smoothing, standard normal transformation SNV, multivariate scattering correction MSC, first-order derivative FD, second-order derivative 2D and other preprocessing methods; competitive adaptive reweighted sampling algorithm CARS, uninformative variable elimination algorithm UVE, continuous projection algorithm SPA, collaborative interval partial least squares iPLS, random frog leaping algorithm RF and other characteristic band screening methods, a partial least squares regression PLSR quantitative prediction model is established, and the optimal large thick-skinned melon and fruit internal quality prediction model is obtained by comparing the correction set correlation coefficient, correction set root mean square error RMSEC, prediction set correlation coefficient, prediction set root mean square error RMSEP and residual prediction deviation RPD and other evaluation indicators; g. Model deployment: Deploy the optimal large thick-skinned melon internal quality prediction model on the Raspberry Pi controller; h. Predict key internal quality indicators of large thick-skinned melons and fruits: Use the control button to turn on the light source module and the Raspberry Pi controller to collect the spectrum of the sample to be tested. After correction by the formula, it is brought into the optimal internal quality prediction model to obtain the specific values ​​of the key internal quality indicators of large thick-skinned melons and fruits, namely soluble solids content, pH and hardness, and the results are displayed on the touch screen.

[0013] The detection principle of the present invention is: Large thick-skinned fruits with different quality characteristics have specific wavelengths of double frequency and combined frequency absorption in the visible-near infrared spectrum region, and the absorption spectra are also different. Based on this difference, the internal quality of large thick-skinned fruits can be quantitatively analyzed, among which the main substance that determines the soluble solid content and pH value of the sample is carbohydrates; in the organizational structure of fruit cells, the binding force has a certain relationship with the pectin content, among which the cellulose content is more related to the hardness value. The cellulose monomer is glucose, and it is composed of CH groups and OH groups like carbohydrates. The visible-near infrared light emitted by the device of the present invention is transmitted into the sample, and its CH groups and OH groups are absorbed by double frequency and combined frequency of different bands, and corresponding absorption peaks are generated at specific bands, generating diffuse transmission spectra of the internal substances of the sample. According to the mathematical relationship between the diffuse transmission spectrum and the sample material composition, a quantitative prediction model is established to predict the soluble solid content, pH value and hardness value of large thick-skinned fruits.

[0014] Compared with the prior art, the advantages of the present invention are: a. All components of the portable detection device are integrated inside the device, which is convenient to carry and easy to operate; b. Choose diffuse transmission mode for detection and device design. Compared with traditional diffuse reflection and full transmission methods, it can better penetrate the peel of large and thick-skinned fruits and obtain the internal pulp spectral information, and reduce the interference of external stray light and skin texture information on the detection results; c. Use the power cord and 220V boat-type switch socket to connect the Raspberry Pi controller, high-power halogen light source, cooling fan and 220V industrial power supply, so as to keep the circuit running stably without disassembling and replacing the battery; d. An optimal internal quality detection model suitable for large thick-skinned melons and fruits is proposed, which can realize accurate detection of key quality of samples, and the added characteristic band screening algorithm can improve the model detection speed.

[0015] Reference numerals 1. Annular shading sponge; 2. Dark box; 3. Cooling fan; 4. 12V DC to 5V DC voltage transformer; 5. Touch screen; 6. Control button; 7. 220V AC to 12V DC power supply; 8. Shading suction cup; 9. Light source module; 10. 74UV collimating lens; 11. Heat dissipation hole; 12. Circuit board cooling fan; 13. Raspberry Pi controller; 14. Optical fiber; 15. Power cord; 16. Spectral detection module; 17. 220V boat-shaped power switch; 18. Support foot; 19. Light source frame fixing screw; 20. Spectral collection hole; 9-1. Halogen light source; 9-2. Light source bracket; 9-3. Fixing hole. BRIEF DESCRIPTION OF THE DRAWINGS

[0016] Figure 1 This is the overall structure diagram of the portable diffuse transmission spectrum detection device for the internal quality of large thick-skinned melons and fruits of the present invention; Figure 2 It is a top view of the portable diffuse transmission spectrum detection device for the internal quality of large thick-skinned melons and fruits of the present invention; Figure 3 It is a schematic diagram of the structure of the light source module in the portable diffuse transmission spectrum detection device for the internal quality of large thick-skinned melons and fruits of the present invention; Figure 4 This is a physical picture of the portable diffuse transmission spectrum detection device for the internal quality of large thick-skinned melons and fruits of the present invention; Figure 5 The original visible-near infrared diffuse transmission spectrum of the sample in Example 1 of the present invention and the spectrum after SNV processing; Figure 6 is a schematic diagram of feature band extraction after SNV processing in Example 1 of the present invention; Figure 7 It is a scatter plot of the predicted value and the measured value of the soluble solid content in Example 1 of the present invention. DETAILED DESCRIPTION

[0017] The present invention is further described below in conjunction with an embodiment and the accompanying drawings. The described embodiment is only a part of the embodiments of the present invention, rather than all the embodiments. like Figure 1 , Figure 2 , Figure 3As shown, the portable large-scale thick-skinned melon and fruit internal quality diffuse transmission spectrum detection device comprises a ring-shaped light-shielding sponge 1, a dark box 2, a cooling fan 3, a 12V DC voltage to 5V DC voltage transformer 4, a touch screen 5, a control button 6, a 220V AC voltage to 12V DC voltage DC power supply 7, a light-shielding sucker 8, a light source module 9, a 74UV collimating lens 10, a heat dissipation hole 11, a circuit board cooling fan 12, a Raspberry Pi controller 13, an optical fiber 14, a power cord 15, and a spectrum detection module 1. 6. 220V boat-shaped power switch 17, supporting feet 18, light source frame fixing screws 19, and spectrum collection holes 20; the light source module 9 includes a halogen light source 9-1, a light source bracket 9-2, and a fixing hole 9-3; the halogen light source 9-1 includes four 12V, 20W halogen lamps with a wavelength range of 350-2500nm, each halogen lamp is 90° apart, and is distributed in a surrounding manner with the sample detection site; the light source bracket 9-2 is a four-sided light source bracket, which is fixed by fixing holes 9-3. The spectrum collection hole 20 is opened at the center of the top of the dark box 2, and the diameter of the circular hole is 12 cm, which is smaller than the average horizontal diameter of the sample; the annular shading sponge 1 is installed and fixed to the spectrum collection hole 20 on the dark box 2; the cooling fan 3 is symmetrically installed on the left and right sides of the dark box 2, and is flush with the light source module 9; the heat dissipation hole 11 is fixed to the installation position of the cooling fan 3; the 12V DC voltage to 5V DC voltage transformer 4 is installed and fixed on the back of the dark box 2; the 220V AC voltage to 12V DC voltage DC power supply 7 is fixed inside the shell of the dark box 2; the touch display screen 5 is fixedly installed on the left side of the front of the dark box 2; the control button 6 is fixedly installed on the right side of the front of the dark box 2; the shading suction cup 8 is fixed on the top of the 74UV collimating lens 10, and is aligned with the annular shading The light sponge 1 is at the same height; the light source module 9 is fixed on the top of the dark box 2; the bottom of the 74UV collimating lens 10 is fixed to the central hole of the light source module 9; the Raspberry Pi controller 13 is installed and fixed to the bottom of the dark box 2 with a rubber column; the circuit board cooling fan 12 is installed and fixed to the upper part of the Raspberry Pi controller 13 with a rubber column; the spectrum detection module 16 is fixed inside the shell of the dark box 2; one end of the optical fiber 14 is connected to the SMA905 interface of the 74UV collimating lens 10, and the other end is connected to the SMA905 interface of the spectrum detection module 16; the 220V boat-shaped power switch 17 is installed and fixed on the lower right side of the left part of the dark box 2; the left end jack of the power cord 15 is inserted into the 220V boat-shaped power switch 17 plug; the supporting feet 18 are symmetrically installed at the four corners of the bottom of the dark box 2.In the light source module 9, the light source bracket 9-2 is a four-sided light source bracket with a horizontal angle of 150°, which is fixed to the top of the outer shell of the dark box 2 through the fixing holes 9-3 and the light source bracket fixing screws 19. The four sides of the light source bracket are provided with four circular holes with the same diameter as the halogen lamp for fixing the four halogen lamps in the halogen light source 9-1, ensuring that the emitted light of the halogen light source 9-1 forms an angle of 30° with the surface of the 74UV collimating lens 10; the bottom of the 74UV collimating lens 10 is fixed to the center hole of the light source bracket 9-2 of the light source module 9; the top optical lens of the 74UV collimating lens 10 is fixed to the center of the light-shielding suction cup 8. The power cord 15 jack is inserted into the 220V rocker-shaped power switch 17 three-pin socket; the 220V rocker-shaped power switch 17 and the 220V AC voltage to 12V DC voltage DC power supply 7 are connected by a wire to output a 12V power supply; one power supply end of the 220V AC voltage to 12V DC voltage DC power supply 7 is connected to the halogen light source 9-1, the cooling fan 3 and the control button 6 by a wire to form a light source cooling circuit; the other power supply end of the 220V AC voltage to 12V DC voltage DC power supply 7 is connected to the 12V DC voltage to 5V DC voltage transformer 4 by a wire to output a 5V power supply; the 12V DC voltage to 5V DC voltage transformer 4 is connected to the Raspberry Pi controller 13, the circuit board cooling fan 12, the touch display screen 5, the spectrum detection module 16 and the control button 6 by a wire to form a control circuit; the circuit board cooling fan 12, the touch display screen 5 and the spectrum detection module 16 are all controlled by the Raspberry Pi controller 13. The three-pin plug of the power cord 15 is inserted into a 220V industrial power supply, and the main circuit of the device is powered under the control of the 220V boat-shaped power switch 17; under the control of the control button 6, the halogen light source 9-1 is lit, and the cooling fan 3, the Raspberry Pi controller 13, the circuit board cooling fan 12 and the touch display screen 5 are powered; the spectrum detection module 16 receives the sample spectrum signal collected by the 74UV collimating lens 10, and converts it into an electrical signal and transmits it to the Raspberry Pi controller 13 for processing and calculation, and displays the result on the touch display screen 5. The dark box 2 and the light source bracket 9-2 are both made of aluminum alloy, and the inner and outer walls are sprayed with black matte material to prevent reflection caused by the halogen light source 9-1 irradiating the inner wall; the annular shading sponge 1 is made of black sponge material, has a shading function, and can be used to prevent the influence of external stray light; the shading suction cup 8 is made of black rubber material, which is soft and can shield light, so as to prevent the sample spectrum from carrying surface texture feature information and realize diffuse transmission spectrum collection. The length, width and height of the portable large-scale thick-skinned melon and fruit internal quality diffuse transmission spectrum detection device are 30 cm, 24 cm and 15 cm respectively.

[0018] like Figure 4As shown, there are 4 control buttons on the darkbox display panel. The "horizontal light" and "vertical light" control buttons are used to control the horizontal halogen light source power supply and the vertical halogen light source power supply respectively; the "heat dissipation" button is used to power the cooling fan; the "detection" button is used to power the Raspberry Pi controller and the touch display.

[0019] The diffuse transmission spectroscopy nondestructive testing method for large thick-skinned melons and fruits in the present invention is further described in detail using an embodiment. Example 1

[0020] Sample preparation: In this embodiment, 260 Hami melon samples with uniform shape and size, no surface damage, and no pests and diseases were selected, numbered, and placed at room temperature of 25° C. and relative humidity of 50-55% for 24 hours.

[0021] Visible-NIR spectroscopy acquisition includes the following steps: a. Turn on the main power supply: insert the power cord 15 plug into the 220V industrial power supply, and insert the jack into the 220V rocker-shaped power switch 17, press the rocker-shaped switch on the 220V rocker-shaped power switch 17 to complete the power supply operation of the device system; b. Preheating: Use the control button 6 to turn on the light source module 9 to preheat for about 30 minutes, and at the same time use the control button 6 to turn on the heat dissipation module 3; c. Collect reference spectrum T w and T d : Use the control button 6 to turn on the light source module 9, use the calibration white board to cover the detection hole of the shading sponge 1 and close to the shading suction cup 8 above the 74UV collimating lens 10, set and collect the white reference spectrum T w Use the control button 6 to turn off the light source module 9, cover the shading sponge 1 with a black shading cloth to detect the hole and close to the shading suction cup 8 above the 74UV collimating lens 10, set and collect the black reference spectrum T d ; d. Collect the sample diffuse transmission spectrum T s : Place the sample horizontally on the annular light-shielding sponge 1, so that its curved surface fits the annular light-shielding sponge 1 to form a closed light-shielding environment, and at the same time, make its equatorial detection part close to the light-shielding suction cup 8, and then transmit the spectral information to the spectral detection module 16 through the collimating lens 10 and the optical fiber 14 connected to the light-shielding suction cup 8 to complete the sample diffuse transmission spectrum T s collection; e. Calculate the sample diffuse transmission correction spectrum T c :According to the collected black and white reference spectrum T d 、T w and the original spectrum T s , using the formula Calculate the sample diffuse transmission correction spectrum T c ; The diffuse transmission spectrum T c The wavelength range is 500-1050nm.

[0022] Determination of physical and chemical indicators: The physical and chemical values ​​of key quality indicators of samples, such as soluble solids, pH and hardness, are obtained through traditional destructive testing methods using a sugar refractometer, pH meter and texture analyzer.

[0023] Construction of internal quality prediction model for large thick-skinned melons and fruits: the sample diffuse transmission corrected spectrum obtained by the detection device is taken as input, and the physical and chemical values ​​of key quality indicators of the samples obtained by the physical and chemical index determination method are taken as output; after preprocessing methods such as detrending DT, SG smoothing, standard normal transformation SNV, multivariate scattering correction MSC, first-order derivative FD, second-order derivative 2D, etc.; characteristic band screening methods such as competitive adaptive reweighted sampling algorithm CARS, uninformation variable elimination algorithm UVE, continuous projection algorithm SPA, collaborative interval partial least squares iPLS, random frog jumping algorithm RF, etc., a partial least squares regression PLSR quantitative prediction model is established, and the optimal internal quality prediction model for large thick-skinned melons and fruits is obtained by comparing evaluation indicators such as the correction set correlation coefficient Rc, the correction set root mean square error RMSEC, the prediction set correlation coefficient Rp, the prediction set root mean square error RMSEP and the residual prediction deviation RPD.

[0024] Model deployment: The optimal large thick-skinned melon and fruit internal quality prediction model is deployed on the Raspberry Pi controller 13.

[0025] Predicting key internal quality indicators of large thick-skinned melons and fruits: Use the control button 6 to turn on the light source module 9 and the Raspberry Pi controller 13 to collect the spectrum of the sample to be tested. After correction by the formula, the optimal internal quality prediction model is used to obtain the specific values ​​of the key internal quality indicators of large thick-skinned melons and fruits, namely, the soluble solids content, pH and hardness.

[0026] Display analysis results: By touching the display screen 5 to detect the pre-processing operation on the interface, the collected diffuse transmission original spectrum curve is pre-processed by SG smoothing, detrending DT, standard normal transformation SNV, multivariate scattering correction MSC, first-order derivative FD, second-order derivative 2D, etc., to reduce external noise interference. Figure 5 The diffuse transmission original spectrum and the diffuse transmission spectrum after SNV preprocessing collected by the portable large-scale thick-skinned melon and fruit internal quality diffuse transmission spectrum detection device described in Example 1; CARS, UVE, SPA, RF and iPLS characteristic bands are extracted from the spectrum curve after SNV preprocessing, Figure 6 The schematic diagram of CARS characteristic band extraction of the spectral curve after SNV processing is shown in Figure 1. 161 characteristic bands are extracted from the original 1044 raw bands to improve the model detection speed; Figure 7It is a scatter plot of the predicted value and the measured value of the soluble solids in Example 1 obtained through the optimal prediction model. The correlation coefficients between the correction set and the prediction set are 93.1% and 90.8% respectively, the root mean square errors between the correction set and the prediction set are 0.583 and 0.663 respectively, and the residual prediction deviation RPD is 2.211; the correlation coefficients between the correction set and the prediction set obtained through the optimal prediction model for hardness are 94.5% and 86.8% respectively, the root mean square errors between the correction set and the prediction set are 0.363 and 0.604 respectively, and the residual prediction deviation RPD is 1.994; the correlation coefficients between the correction set and the prediction set obtained through the optimal prediction model for pH are 92.2% and 91.9% respectively, the root mean square errors between the correction set and the prediction set are 0.026 and 0.473 respectively, and the residual prediction deviation RPD is 2.158.

[0027] The above are only preferred embodiments of the present invention and do not limit the protection scope of the present invention. Modifications and partial technical replacements based on the scheme described in the present invention are still within the protection scope of the patent application.

Claims

1. A nondestructive testing method for diffuse transmission spectroscopy of large thick-skinned melons and fruits, characterized in that: The visible-near infrared light emitted by the halogen light source passes through a large thick-skinned fruit sample, and after multiple scattering and transmission, is received by the collimating lens with a light-shielding suction cup below. It is transmitted to the spectrum detection module through the optical fiber, converted into an electrical signal, and then transmitted to the Raspberry Pi controller for processing. The calculation results are displayed on the touch screen, including the following steps: a. Turn on the main power: Plug the power cord (15) into the 220V industrial power supply, and insert the jack into the 220V rocker-shaped power switch (17), press the rocker switch on the 220V rocker-shaped power switch (17) to complete the power supply operation of the device system; b. Preheating: Use the control button (6) to turn on the light source module (9) and preheat for about 30 minutes, and at the same time use the control button (6) to turn on the cooling module (3); c. Collect reference spectrum T w and T d :Use the control button (6) to turn on the light source module (9), use the calibration white board to cover the detection hole of the shading sponge (1) and close to the shading suction cup (8) above the 74UV collimating lens (10), set and collect the white reference spectrum T w Use the control button (6) to turn off the light source module (9), cover the shading sponge (1) with a black shading cloth to detect the hole and close to the shading suction cup (8) above the 74UV collimating lens (10), set and collect the black reference spectrum T d ; d. Collect the sample diffuse transmission spectrum T s :The sample is placed horizontally on the annular light-shielding sponge (1), so that its arc surface fits the annular light-shielding sponge (1) to form a closed light-shielding environment, and at the same time, its equatorial detection part is closely attached to the light-shielding sucker (8), and then the spectral information is transmitted to the spectral detection module (16) through the collimating lens (10) and the optical fiber (14) connected to the light-shielding sucker (8), so as to complete the sample diffuse transmission spectrum T s collection; e. Calculate the sample diffuse transmission correction spectrum T c :According to the collected black and white reference spectrum T d 、T w and the original spectrum T s , using the formula Calculate the sample diffuse transmission correction spectrum T c ; The diffuse transmission spectrum T c Wavelength range is 500-1050nm; f. Establishment of the internal quality prediction model of large thick-skinned melons and fruits: The sample diffuse transmission correction spectrum obtained by the detection device is used as the input, and the physical and chemical values ​​of the key quality indicators of the samples obtained by the traditional destructive detection methods of the sugar refractometer, pH meter and texture analyzer are used as the output; after detrending DT, SG smoothing, standard normal transformation SNV, multivariate scattering correction MSC, first-order derivative FD, second-order derivative 2D and other preprocessing methods; competitive adaptive reweighted sampling algorithm CARS, uninformative variable elimination algorithm UVE, continuous projection algorithm SPA, collaborative interval partial least squares iPLS, random frog leaping algorithm RF and other characteristic band screening methods, a partial least squares regression PLSR quantitative prediction model is established, and the optimal large thick-skinned melon and fruit internal quality prediction model is obtained by comparing the correction set correlation coefficient, correction set root mean square error RMSEC, prediction set correlation coefficient, prediction set root mean square error RMSEP and residual prediction deviation RPD and other evaluation indicators; g. Model deployment: deploy the optimal large thick-skinned melon internal quality prediction model on the Raspberry Pi controller (13); h. Predicting key internal quality indicators of large thick-skinned melons and fruits: Use the control button (6) to turn on the light source module (9) and the Raspberry Pi controller (13) to collect the spectrum of the sample to be tested, and after correction by the formula, bring it into the optimal internal quality prediction model to obtain the specific values ​​of the key internal quality indicators of large thick-skinned melons and fruits, namely, the soluble solid content, pH and hardness, and display the results on the touch screen (5).

2. A portable diffuse transmission spectrum detection device for the internal quality of large thick-skinned fruits and vegetables, characterized in that: The device comprises a ring-shaped light-shielding sponge (1), a dark box (2), a cooling fan (3), a 12V DC voltage to 5V DC voltage transformer (4), a touch screen (5), a control button (6), a 220V AC voltage to 12V DC voltage DC power supply (7), a light-shielding sucker (8), a light source module (9), a 74UV collimating lens (10), a heat dissipation hole (11), a circuit board cooling fan (12), a Raspberry Pi controller (13), an optical fiber (14), a power cord (15), a spectrum detection module (16), and a 220V ship-type power supply. A switch (17), a supporting foot (18), a light source frame fixing screw (19), and a spectrum collection hole (20); the light source module (9) comprises a halogen light source (9-1), a light source frame (9-2), and a fixing hole (9-3); the halogen light source (9-1) comprises four 12V, 20W halogen lamps with a wavelength range of 350-2500nm, each halogen lamp is spaced 90° apart and is distributed in a surrounding manner with respect to the sample detection part; the light source frame (9-2) is a four-sided light source frame, which is fixed by the fixing hole (9-3).

3. The portable diffuse transmission spectrum detection device for the internal quality of large thick-skinned melons and fruits according to claim 2 is characterized by: The spectrum collection hole (20) is opened at the center of the top of the dark box (2), and the diameter of the circular hole is 12 cm, which is smaller than the average transverse diameter of the sample; the annular light-shielding sponge (1) is installed and fixed at the spectrum collection hole (20) on the dark box (2); the cooling fan (3) is symmetrically installed on the left and right sides of the dark box (2), and is flush with the light source module (9); the heat dissipation hole (11) is fixed to the installation position of the cooling fan (3); the 12V DC voltage to 5V DC voltage transformer (4) is installed and fixed on the back of the dark box (2); the 220V AC voltage to 12V DC voltage DC power supply (7) is fixed inside the shell of the dark box (2); the touch display screen (5) is fixedly installed on the left side of the front of the dark box (2); the control button (6) is fixedly installed on the right side of the front of the dark box (2); the light-shielding suction cup (8) is fixed on the top of the 74UV collimating lens (10), and is aligned with the annular light-shielding sponge (11). The light sponge (1) is at the same height; the light source module (9) is fixed on the top of the dark box (2); the bottom of the 74UV collimating lens (10) is fixed to the central hole of the light source module (9); the Raspberry Pi controller (13) is fixed to the bottom of the dark box (2) by means of a rubber column; the circuit board cooling fan (12) is fixed to the upper part of the Raspberry Pi controller (13) by means of a rubber column; the spectrum detection module (16) is fixed inside the shell of the dark box (2); one end of the optical fiber (14) is connected to the SMA905 interface of the 74UV collimating lens (10), and the other end is connected to the SMA905 interface of the spectrum detection module (16); the 220V ship-shaped power switch (17) is fixed to the lower right side of the left part of the dark box (2); the left end jack of the power cord (15) is inserted into the plug of the 220V ship-shaped power switch (17); the support feet (18) are symmetrically installed at the four corners of the bottom of the dark box (2); In the light source module (9), the light source bracket (9-2) is a four-sided light source bracket with a lateral angle of 150°, which is fixed to the top of the outer shell of the dark box (2) through fixing holes (9-3) and light source bracket fixing screws (19); four circular holes with the same diameter as the halogen lamps are opened on the four sides of the light source bracket (9-2) for fixing the four halogen lamps in the halogen light source (9-1), so as to ensure that the emitted light of the halogen light source (9-1) forms an angle of 30° with the surface of the 74UV collimating lens (10); the bottom of the 74UV collimating lens (10) is fixed to the central hole position of the light source bracket (9-2) of the light source module (9); and the top optical lens of the 74UV collimating lens (10) is fixed to the central position of the light shielding sucker (8).

4. The portable diffuse transmission spectrum detection device for the internal quality of large thick-skinned melons and fruits according to claim 3 is characterized by: The power cord (15) jack is inserted into the three-pin socket of the 220V rocker-shaped power switch (17); the 220V rocker-shaped power switch (17) and the 220V AC voltage to 12V DC voltage DC power supply (7) are connected by a wire to output a 12V power supply; a power supply end of the 220V AC voltage to 12V DC voltage DC power supply (7) is connected by a wire to the halogen light source (9-1), the cooling fan (3) and the control button (6) to form a light source cooling circuit; the 220V AC voltage to 12V DC voltage DC power supply (7) is connected by a wire to form a light source cooling circuit. The other power supply end of the DC power source (7) is connected to the 12V DC voltage to 5V DC voltage transformer (4) by a wire to output a 5V power supply; the 12V DC voltage to 5V DC voltage transformer (4) is connected to the Raspberry Pi controller (13), the circuit board cooling fan (12), the touch screen (5), the spectrum detection module (16) and the control button (6) by a wire to form a control circuit; the circuit board cooling fan (12), the touch screen (5) and the spectrum detection module (16) are all controlled by the Raspberry Pi controller (13); The three-pin plug of the power cord (15) is inserted into a 220V industrial power supply, and the main circuit of the device is powered under the control of the 220V boat-shaped power switch (17); under the control of the control button (6), the halogen light source (9-1) is lit, and the cooling fan (3), the Raspberry Pi controller (13), the circuit board cooling fan (12) and the touch display screen (5) are powered; the spectrum detection module (16) receives the sample spectrum signal collected by the 74UV collimating lens (10), converts it into an electrical signal, transmits it to the Raspberry Pi controller (13) for processing and calculation, and displays the result on the touch display screen (5).

5. The portable diffuse transmission spectrum detection device for the internal quality of large thick-skinned melons and fruits according to claim 3 is characterized by: The dark box (2) and the light source bracket (9-2) are both made of aluminum alloy, and the inner and outer walls are sprayed with black matte material to prevent reflection caused by the halogen light source (9-1) irradiating the inner wall; the annular light-shielding sponge (1) is made of black sponge material and has a light-shielding function, which can be used to prevent the influence of external stray light; the light-shielding suction cup (8) is made of black rubber material, which is soft and can shield light, so as to prevent the sample spectrum from carrying surface texture feature information, thereby realizing diffuse transmission spectrum collection.

6. The portable diffuse transmission spectrum detection device for the internal quality of large thick-skinned melons and fruits according to claim 3 is characterized by: The length, width and height of the portable large-scale thick-skinned melon and fruit internal quality diffuse transmission spectrum detection device are 30 cm, 24 cm and 15 cm respectively.

Images