CN111445469A - Hyperspectrum-based apple damage parameter lossless prediction method after impact - Google Patents
Hyperspectrum-based apple damage parameter lossless prediction method after impact Download PDFInfo
- Publication number
- CN111445469A CN111445469A CN202010293066.XA CN202010293066A CN111445469A CN 111445469 A CN111445469 A CN 111445469A CN 202010293066 A CN202010293066 A CN 202010293066A CN 111445469 A CN111445469 A CN 111445469A
- Authority
- CN
- China
- Prior art keywords
- hyperspectral
- damage
- original
- image
- sample
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 41
- 238000001228 spectrum Methods 0.000 claims abstract description 36
- 238000004364 calculation method Methods 0.000 claims abstract description 22
- 238000012937 correction Methods 0.000 claims abstract description 13
- 230000003595 spectral effect Effects 0.000 claims description 18
- 238000002310 reflectometry Methods 0.000 claims description 15
- 238000009499 grossing Methods 0.000 claims description 11
- 230000008569 process Effects 0.000 claims description 5
- 238000005259 measurement Methods 0.000 claims description 3
- 230000001066 destructive effect Effects 0.000 claims description 2
- 238000009530 blood pressure measurement Methods 0.000 claims 1
- 235000013399 edible fruits Nutrition 0.000 abstract description 35
- 238000011161 development Methods 0.000 abstract description 9
- 238000001514 detection method Methods 0.000 abstract description 7
- 230000001953 sensory effect Effects 0.000 abstract description 4
- 238000011156 evaluation Methods 0.000 abstract description 3
- 241000220225 Malus Species 0.000 description 29
- 238000005516 engineering process Methods 0.000 description 7
- 238000012360 testing method Methods 0.000 description 6
- 235000021016 apples Nutrition 0.000 description 4
- 238000002474 experimental method Methods 0.000 description 3
- 238000011002 quantification Methods 0.000 description 3
- 229910000831 Steel Inorganic materials 0.000 description 2
- 238000012512 characterization method Methods 0.000 description 2
- 238000002790 cross-validation Methods 0.000 description 2
- 239000000284 extract Substances 0.000 description 2
- 238000000605 extraction Methods 0.000 description 2
- 239000011159 matrix material Substances 0.000 description 2
- 229920001343 polytetrafluoroethylene Polymers 0.000 description 2
- 239000004810 polytetrafluoroethylene Substances 0.000 description 2
- 238000007781 pre-processing Methods 0.000 description 2
- 239000010959 steel Substances 0.000 description 2
- 208000034656 Contusions Diseases 0.000 description 1
- -1 Polytetrafluoroethylene Polymers 0.000 description 1
- 235000010724 Wisteria floribunda Nutrition 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000015556 catabolic process Effects 0.000 description 1
- 230000008859 change Effects 0.000 description 1
- 238000004040 coloring Methods 0.000 description 1
- 230000001186 cumulative effect Effects 0.000 description 1
- 238000000354 decomposition reaction Methods 0.000 description 1
- 238000006731 degradation reaction Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000000556 factor analysis Methods 0.000 description 1
- 238000003306 harvesting Methods 0.000 description 1
- 238000005286 illumination Methods 0.000 description 1
- 238000012417 linear regression Methods 0.000 description 1
- 239000003094 microcapsule Substances 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 235000016709 nutrition Nutrition 0.000 description 1
- 230000035764 nutrition Effects 0.000 description 1
- 238000004806 packaging method and process Methods 0.000 description 1
- 238000002203 pretreatment Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000000611 regression analysis Methods 0.000 description 1
- 238000013097 stability assessment Methods 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01D—MEASURING NOT SPECIALLY ADAPTED FOR A SPECIFIC VARIABLE; ARRANGEMENTS FOR MEASURING TWO OR MORE VARIABLES NOT COVERED IN A SINGLE OTHER SUBCLASS; TARIFF METERING APPARATUS; MEASURING OR TESTING NOT OTHERWISE PROVIDED FOR
- G01D21/00—Measuring or testing not otherwise provided for
- G01D21/02—Measuring two or more variables by means not covered by a single other subclass
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q10/00—Administration; Management
- G06Q10/04—Forecasting or optimisation specially adapted for administrative or management purposes, e.g. linear programming or "cutting stock problem"
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/90—Dynamic range modification of images or parts thereof
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10032—Satellite or aerial image; Remote sensing
- G06T2207/10036—Multispectral image; Hyperspectral image
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30168—Image quality inspection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30181—Earth observation
- G06T2207/30188—Vegetation; Agriculture
Landscapes
- Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Business, Economics & Management (AREA)
- Human Resources & Organizations (AREA)
- Quality & Reliability (AREA)
- Strategic Management (AREA)
- Economics (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Game Theory and Decision Science (AREA)
- Development Economics (AREA)
- Entrepreneurship & Innovation (AREA)
- Marketing (AREA)
- Operations Research (AREA)
- Tourism & Hospitality (AREA)
- General Business, Economics & Management (AREA)
- Investigating Or Analysing Materials By Optical Means (AREA)
Abstract
The invention provides a hyperspectral-based apple post-impact damage parameter nondestructive prediction method which comprises the following steps of obtaining the average pressure and contact force of a sample falling damage; acquiring a hyperspectral image of a sample; performing black-and-white correction on the hyperspectral image, and extracting an average spectrum of a damage area of the hyperspectral image after correction; acquiring the characteristic wavelength of the average spectrum; and establishing a prediction model of hyperspectral data and damage parameters. The method has the advantages that the mechanical parameters of the fruits are predicted, the damage degree of the fruits is quantized, important basis can be provided for evaluating the mechanical damage of the fruits, compared with the traditional manual sensory detection and related parameter calculation method, the time can be saved, the efficiency is improved, the rapid and nondestructive evaluation and prediction can be realized, and the method has important significance for the industrial development of the fruit market.
Description
Technical Field
The invention belongs to the technical field of lossless prediction of fruits, and particularly relates to a hyperspectral-based lossless prediction method for damage parameters of apples after impact.
Background
The apple is popular with consumers because of rich nutrition and crisp taste, and is one of the most popular fruits in the market. According to the data source of the national statistical bureau: in 2018, 3923.4 million tons of apples are produced in China, and the total export value reaches $ 13 hundred million. Apple yield is high, but is inevitably affected by external forces during harvesting, packaging, shipping and storage. The economic loss of fruit due to mechanical damage accounts for about 30% of the total. Mechanical damage is investigated as a major cause of apple quality and value degradation, with impact damage being most common. The physiological change of the fruit is intensified in the storage process after the fruit is impacted and damaged, so that the rotting of the fruit is accelerated, and the shelf life is shortened. Due to the lack of objective quantitative assessment of apple damage degree, economic losses cannot be accurately estimated. The manual sensory detection of the damage degree of the apples wastes time and labor, and is difficult to adapt to the trend of industrial large-scale apple quality grading.
Disclosure of Invention
In view of the above problems, the present invention provides a hyperspectral apple damage parameter lossless prediction method after impact, so as to solve the above or other former problems in the prior art.
In order to solve the technical problems, the invention adopts the technical scheme that: a hyperspectral-based apple damage parameter lossless prediction method after impact comprises the following steps,
obtaining the average pressure and contact force of the falling damage of the sample;
acquiring an original hyperspectral image of a sample;
performing black-and-white correction on the original hyperspectral image, and extracting an average spectrum of a damage area of the corrected hyperspectral image;
acquiring the characteristic wavelength of the average spectrum;
and establishing a prediction model of hyperspectral data and damage parameters.
Further, in the step of obtaining the average pressure and the contact force of the falling damage of the sample, a pressure measuring device is adopted to measure the average pressure and the contact force of the falling damage of the sample, and the parameter values of the average pressure and the contact force are obtained through a pressure scanning system.
Furthermore, the pressure measuring device is a pressure sensing film, the pressure sensing film is an ultra-low pressure LLL W type film, and the measuring range is 0.2-0.6 MPa.
Further, in the step of obtaining the hyperspectral image of the original sample, the original hyperspectral data of the undamaged sample and the hyperspectral data of the damaged sample are respectively obtained.
Further, the step of performing black-and-white correction on the original hyperspectral image and extracting the average spectrum of the damage area of the corrected hyperspectral image comprises the following steps:
correcting the reflectivity of the original hyperspectral image: under the same conditions and parameter settings as those of the original hyperspectral image, acquiring a white board calibration image by acquiring a white calibration board, and acquiring a full-black calibration image with the reflectivity of 0% by covering a camera lens by a camera cover;
and selecting an interested area of the corrected hyperspectral image, and extracting a spectral mean value of the interested area.
Further, in the process of correcting the reflectivity of the original hyperspectral image, calculation is performed through the following formula:
wherein I is a corrected hyperspectral image, IrFor the original hyperspectral image, IdDemarcating the image for total black, IwAnd calibrating the white board with images.
Further, in the step of obtaining the characteristic wavelength of the average spectrum, SG2 is usedndAnd computing the original hyperspectral image by the Der, and selecting characteristic wavelength.
Further, SG2 is adoptedndWhen the Der calculates the original hyperspectral image, the calculation steps are as follows:
smoothing the average spectrum, and calculating by adopting least square convolution, wherein the calculation equation is as follows:
where Y is the original spectrum, Y is the smoothed spectrum, CiIs a smoothing window, N is the number of convolution integers, j is the running index of the original coordinate data, the smoothing array consists of 2m +1 points, where m is half the width of the smoothing window;
and sequentially performing first derivative calculation and second derivative calculation on the smoothed spectral curve to obtain the characteristic wavelength, wherein the first derivative calculation formula is as follows:
the second derivative is calculated by the formula:
wherein Δ λ is an interval between adjacent bands, and Δ λ ═ λk-λj＝λj-λi,λk>λj>λi；
And acquiring characteristic wavelengths, wherein the characteristic wavelengths are wave crests and wave troughs.
Further, the characteristic wavelengths are 967nm, 1001nm, 1100nm, 1154nm, 1190nm, 1407nm and 1443nm, respectively.
Further, in the step of establishing a prediction model of the hyperspectral data and the damage parameters, a P L S regression model is adopted to respectively establish a prediction model among the hyperspectral data, the contact force and the average pressure according to the acquired characteristic wavelength.
Due to the adoption of the technical scheme, the near-infrared hyperspectral technology is utilized and Savitzky-Golay second derivative (SG 2) is usedndDer) selecting characteristic wavelength, measuring contact force and average pressure parameter by using pressure-sensitive film based on apple drop test, and performing partial least squares (P L S) regression modelThe quantitative relation between the hyperspectral data and the measurement parameters is established, the optimal prediction model is searched, the hyperspectral data value of the fruit sample to be detected is input into the model, the predicted value corresponding to the damage parameter can be obtained, the quantitative characterization of the apple damage degree can be quickly and effectively realized, the quantification of the fruit damage degree is promoted, the economic loss of agricultural products is reduced, and the development of a hyperspectral agricultural product detection technology is facilitated;
the hyperspectral apple post-impact damage parameter nondestructive prediction method is a quick and reliable method for realizing prediction of mechanical parameters of fruits, quantifies the damage degree of the fruits, can provide important basis for evaluating the mechanical damage of the fruits, can save time and improve efficiency compared with the traditional manual sensory detection and related parameter calculation method, can realize quick and nondestructive evaluation and prediction, and has important significance for industrial development of the fruit market;
by using SG2 before modellingndThe method for extracting the characteristic wavelength greatly improves the operation efficiency of the model and improves the accuracy of the prediction model to a certain extent.
Drawings
FIG. 1 is a flow chart of an embodiment of the present invention;
FIG. 2 is a graph of raw spectra of a reflectance-corrected hyperspectral image of an apple sample according to an embodiment of the invention;
FIG. 3 shows an embodiment of the present invention using SG2ndExtracting a result graph of the characteristic spectrum by the Der;
FIG. 4 is a schematic diagram of contact force scatter obtained based on the P L S regression model according to an embodiment of the present invention.
Detailed Description
The invention is further described with reference to the following figures and specific embodiments.
Fig. 1 shows a flow chart of a method according to an embodiment of the invention, which relates to a hyperspectral-based nondestructive prediction method for damage parameters of an apple after impact, and is used for quantifying the damage degree of a fruit, and quantitatively predicting the damage degree of the fruit by using a near-infrared hyperspectral technology, so that the quantitative characterization of the damage degree of the fruit can be quickly and effectively realized, the economic loss of agricultural products can be reduced, and the development of a hyperspectral agricultural product detection technology can be facilitated.
A hyperspectral-based apple damage parameter lossless prediction method after impact is used for achieving quantification of fruit damage degree, and SG2 is adopted through a near-infrared hyperspectral technology in combination with the use of a pressure-sensitive filmndThe method comprises the steps of extracting characteristic wavelengths by a Der, establishing a model for predicting damage parameters, namely measuring the average pressure and contact force of sample damage by combining a pressure-sensitive film through a drop impact damage experiment of fruits, collecting a hyperspectral image of a sample damage area by using a hyperspectral collection system, collecting a black-and-white calibration image for black-and-white correction, extracting the average spectrum of the damage area in the center of the corrected hyperspectral image, and using SG2 based on the extracted spectrumndAnd extracting characteristic wavelengths by using a Der method, and establishing a P L S regression model based on the characteristic wavelengths and the damage parameters respectively.
Specifically, the hyperspectral non-destructive apple damage parameter prediction method is described by taking an apple sample as an example, and as shown in fig. 1, comprises the following steps,
obtaining the average pressure and contact force of the falling damage of the sample, and collecting the damage parameters of the sample through a falling experiment:
in the step of obtaining the average pressure and the contact force of the falling damage of the sample, the pressure measuring device is adopted to measure the average pressure and the contact force of the falling damage of the sample, and the parameter values of the average pressure and the contact force are obtained through the pressure scanning system, so that the pressure and the contact force of the falling damage of the sample are quantized, and specific parameter values can be measured, specifically:
carrying out a drop test on the apple sample, and determining the damage parameters of the sample: the apple samples with regular shapes are selected for average grouping, one group is used as a control group, other groups are used as experimental groups, the number of the experimental groups is selected according to actual requirements, and no specific requirement is made here.
The method comprises the steps of placing samples of an experimental group on a tray of a drop test machine, enabling the samples to freely drop from different heights to generate impact damage, placing a pressure measuring device on an impact point steel plate of the drop test machine to measure damage parameters of the samples, and reading the damage parameters by using a pressure scanning system to obtain parameter values, wherein the pressure measuring device is a pressure sensitive film, the pressure sensitive film is preferably an ultra-low pressure LLL W type film, the measuring range of the film is 0.2-0.6MPa, the measuring precision is less than or equal to +/-10%, the pressure sensitive film bears impact force to cause microcapsules of a color development layer to break, the color development layer is red, the coloring area and the color development depth of the film represent different values, the pressure scanning system collects the parameters of the pressure sensitive film to obtain contact force and average pressure, and the pressure scanning system is preferably a FPD-8010 special scanner.
The quantity of the fruits in each group is the same, the quantity of the fruits is selected according to actual requirements, no specific requirement is made here, the fruits in the experimental group freely fall from different heights respectively when falling tests are carried out, the height of the fruits is selected according to the actual requirements, and no specific requirement is made here. For example: in this embodiment, 240 red fuji apples without any bruises and with regular shapes are selected according to the following selection criteria: the height of the sample was about 7.5cm, the equatorial diameter was about 8cm and the mass was 185 g. + -.18 g. The samples were divided on average into 6 groups, each group containing 40 fruits. One group was used as a control group to obtain raw spectral data, and the other 5 groups were used as experimental groups. The samples are placed on a tray of a drop test machine, and fall freely from different heights to generate impact damage, the falling heights of each group are different, and the falling heights of each group are respectively 0.3m, 0.6m, 0.9m, 1.2m and 1.5 m.
The pressure-sensitive film was placed on a steel plate of a drop tester to measure the contact force and the average pressure, and the parameters were read using a scanner dedicated to FPD-8010 and a calibration plate to obtain parameters of the contact force and the average pressure, and the like.
The method comprises the following steps of obtaining an original hyperspectral image of a sample, and respectively obtaining original hyperspectral data of an undamaged sample and original hyperspectral data of a damaged sample, wherein the method specifically comprises the following steps:
obtaining raw hyperspectral data of samples of a control group: collecting hyperspectral images of a contrast group by using a spectral camera to obtain original hyperspectral data;
obtaining raw hyperspectral data for samples of an experimental group: and (4) collecting the hyperspectral images of the experimental group by adopting a spectral camera, and collecting the whole information.
In the process of collecting the hyperspectral data of the experimental group samples, the samples are placed on a moving platform, a damaged area is right opposite to a spectrum camera, overall information is collected through the movement of the platform, and hyperspectral images of the experimental group samples are collected after relevant parameters of a hyperspectral collection system are adjusted.
The parameters are set as follows: the distance between the sample and the lens of the spectrum camera is 28-36cm, the exposure time is 18-22ms, the advancing speed of the moving platform is 0.72-0.80cm/s, the size of the image of each wave band in one dimension is a fixed value, and the size of the image is 320x174 pixels-320 x214 pixels.
Because of the nonuniformity of dark current and illumination in the CCD camera, firstly, the reflectivity of all the acquired original hyperspectral images is corrected, the hyperspectral images are subjected to black-and-white correction, and the average spectrum of the damage areas of the corrected hyperspectral images is extracted, which comprises the following steps:
correcting the reflectivity of the original hyperspectral image: under the same conditions and parameter settings as those for obtaining the original hyperspectral image, a white calibration plate is collected to obtain a white board calibration image, the white calibration plate is standard Polytetrafluoroethylene (PTFE), the reflectivity is close to 100%, a camera cover covers a camera lens to obtain a full-black calibration image with the reflectivity of 0%, the camera cover is completely opaque, and the reflectivity is 0%.
In the process of correcting the reflectivity of the original hyperspectral image, calculation is carried out through the following formula:
wherein I is a corrected hyperspectral image, IrFor the original hyperspectral image, IdDemarcating the image for total black, IwAnd calibrating the white board with images.
Fig. 2 shows an original spectrum curve graph of a hyperspectral image of an apple sample subjected to reflectivity correction in the embodiment, a region of interest of the hyperspectral image after the reflectivity correction is selected, a spectrum mean value of the region of interest (ROI) is extracted, a middle part of a damaged area of a fruit in the image after the reflectivity correction is selected as the ROI, the spectrum mean values of all pixel points in the region are extracted through ENVI5.1 software, and an average spectrum of the ROI is extracted.
Analyzing the sample spectrum data to obtain the characteristic wavelength of an average spectrum:
preprocessing the spectral data, wherein the preprocessing method comprises the following steps: standard normal transformation (SNV), Multivariate Scatter Correction (MSC), Norris first derivative (1)stDer) and Norris second derivative (2)ndDer). The modeling results of the data after different pretreatments show that the results are optimal based on the original spectral data, so that the following spectral data processing is based on the original spectral data, and SG2 is adopted to improve the analysis efficiency of the model and keep higher stabilityndThe Der calculates the original hyperspectral data, selects characteristic wavelengths, extracts the characteristic wavelengths through a second derivative, is beneficial to eliminating redundant information, and extracts characteristic wave bands quickly, so that the model is more suitable for online monitoring.
Using SG2ndWhen the Der calculates the original hyperspectral image, the basic principle and the calculation steps are as follows:
savitzky and Golay smooth and compute the derivatives of a set of continuous values using a simplified least squares fit convolution, and therefore, the average spectrum is smoothed and computed using a least squares convolution whose equation is:
where Y is the original spectrum, Y is the smoothed spectrum, CiIs a smoothing window, i.e. the coefficient of the ith spectral value of the filter, N is the number of convolution integers, j is the running index of the original coordinate data, the smoothing array consists of 2m +1 points, where m is half the width of the smoothing window;
the spectral derivative is very sensitive to the spectral signal-to-noise ratio, the first derivative calculation and the second derivative calculation are sequentially performed on the smoothed spectral curve to obtain the characteristic wavelength, and the first derivative calculation formula is as follows:
the second derivative is calculated by the formula:
wherein Δ λ is an interval between adjacent bands, and Δ λ ═ λk-λj＝λj-λi,λk>λj>λi；
And after the operation of the algorithm is finished, a second-order derivative spectrum curve graph can be obtained, characteristic wavelengths are obtained, the characteristic wavelengths are wave crests and wave troughs, and the wavelength selection of the characteristic wavelengths and the quantity selection of the characteristic wavelengths are carried out according to the actual second-order derivative spectrum curve graph.
FIG. 3 shows SG2 used in the present embodimentndAnd (3) extracting a result graph of the characteristic spectrum by the Der, selecting a characteristic peak value as a characteristic wavelength establishing model, and selecting 7 variables, wherein the characteristic wavelengths are 967nm, 1001nm, 1100nm, 1154nm, 1190nm, 1407nm and 1443nm respectively.
And establishing a prediction model of the hyperspectral data and the damage parameters, and respectively establishing a prediction model among the hyperspectral data, the contact force and the average pressure by adopting a P L S regression model according to the acquired characteristic wavelength.
Selecting 3/4 from samples as a modeling set, using the rest 1/4 as prediction sets, respectively establishing prediction models among spectral data, contact force and average pressure by adopting a P L S regression model according to the selected characteristic wave bands, combining factor analysis and regression analysis, simultaneously carrying out principal component decomposition on a spectral matrix X and a parameter matrix Y, obtaining latent variables, calculating the square sum of prediction residuals by adopting leave-one-out cross validation, searching the optimal number of the latent variables according to the cumulative contribution rate of the latent variables and the square sum of the prediction residuals, associating the X with the Y, establishing a linear regression model of the X and the Y, and establishing the correlation coefficients (R) of the prediction setsP) And prediction set root mean square error (RMSEP) as a basis for stability assessment of the model. RPCloser to 1, the smaller the RMSEP, the more stable the model is. Similarly, for the samples in the calibration set and the cross-validation set, the corresponding correlation coefficient and root mean square error are RCAnd RMSEC, RCVAnd RMSECV.
Table 1 shows that the embodiment is based on SG2ndPrediction results of P L S regression model for Der' S apple parameters from the data in the Table, SG2 is basedndBetter results are obtained by a prediction model established by Der extraction of characteristic wavelengths, and table 1 shows that the results of a prediction set are close to those of a correction set and a verification set. For contact force, RCAnd RCVThe values were all 0.91. R of contact forcePAnd RMSEP prediction results were 0.91 and 65.91N, respectively. Further shows that based on SG2ndThe characteristic wavelength extracted by the Der method basically covers the characteristic information of the apple, and the P L S regression can accurately predict the damage parameters.
TABLE 1 based on SG2ndPerformance of P L S regression model for Der extraction of characteristic wavelengths
Fig. 4 shows a scatter plot of predicted values and measured values for the modeling and prediction set of the present embodiment, where the x-axis represents actual measured values of parameters, the y-axis represents predicted values, and the sample points are distributed near the regression line, which is relatively close to the regression line, indicating that there is a significant linear relationship. The scatter plot results show that based on SG2ndThe characteristic wavelength extracted by the Der algorithm basically covers the characteristic information of apple fruits, and damage parameters can be accurately predicted by combining P L S regression.
The prediction of the mean pressure is slightly worse, RP0.73, robustness is to be improved. The undesirable results are predicted because the average pressure is calculated by dividing the contact force acquired by the pressure sensitive film by the area of damage, and because of the limitations of the type of film selected, an out-of-range contact force does not show accurate resultsThe data, resulting in a range of errors in the calculations, may still provide a degree of reference for the prediction of the average pressure parameter.
Nondestructive rapid determination of sample damage parameters: the hyperspectral data of the damaged fruits to be detected are collected, and the characteristic spectrum data of the sample are input into a prediction model to obtain the average pressure and the contact force quickly, so that the hyperspectral data can be used as the judgment basis of the damage degree.
Due to the adoption of the technical scheme, the near-infrared hyperspectral technology is utilized and SG2 is usedndThe method comprises the following steps that Der selects characteristic wavelengths, pressure-sensitive films are used for measuring contact force and average pressure parameters based on an apple drop experiment, a quantitative relation between hyperspectral data and measurement parameters is established based on a P L S regression model, an optimal prediction model is searched, hyperspectral data values of fruit samples to be detected are input into the model, and then prediction values corresponding to damage parameters can be obtained, quantitative representation of apple damage degree can be quickly and effectively achieved, quantification of the fruit damage degree is promoted, reduction of economic loss of agricultural products is facilitated, and development of hyperspectral agricultural product detection technology is facilitated;
the hyperspectral apple post-impact damage parameter nondestructive prediction method is a quick and reliable method for realizing prediction of mechanical parameters of fruits, quantifies the damage degree of the fruits, can provide important basis for evaluating the mechanical damage of the fruits, can save time and improve efficiency compared with the traditional manual sensory detection and related parameter calculation method, can realize quick and nondestructive evaluation and prediction, and has important significance for industrial development of the fruit market;
by using SG2 before modellingndThe method for extracting the characteristic wavelength greatly improves the operation efficiency of the model and improves the accuracy of the prediction model to a certain extent.
The embodiments of the present invention have been described in detail, but the description is only for the preferred embodiments of the present invention and should not be construed as limiting the scope of the present invention. All equivalent changes and modifications made within the scope of the present invention shall fall within the scope of the present invention.
Claims (10)
1. A hyperspectral-based apple post-impact damage parameter lossless prediction method is characterized by comprising the following steps: comprises the following steps of (a) carrying out,
obtaining the average pressure and contact force of the falling damage of the sample;
acquiring an original hyperspectral image of the sample;
performing black-and-white correction on the original hyperspectral image, and extracting an average spectrum of a damage area of the corrected hyperspectral image;
acquiring the characteristic wavelength of the average spectrum;
and establishing a prediction model of hyperspectral data and damage parameters.
2. The hyperspectral-based apple post-impact damage parameter lossless prediction method according to claim 1, wherein: in the step of obtaining the average pressure and the contact force of the falling damage of the sample, a pressure measuring device is adopted to measure the average pressure and the contact force of the falling damage of the sample, and parameter values of the average pressure and the contact force are obtained through a pressure scanning system.
3. The hyperspectral non-destructive prediction method for the damage parameters after the apple impact according to the claim 2 is characterized in that the pressure measurement device is a pressure sensitive film, the pressure sensitive film is an ultra-low pressure LLL W type film, and the measurement range is 0.2-0.6 MPa.
4. The hyperspectral-based apple post-impact damage parameter lossless prediction method according to any of claims 1-3, characterized in that: in the step of obtaining the original hyperspectral image of the sample, original hyperspectral data of an undamaged sample and original hyperspectral data of a damaged sample are respectively obtained.
5. The hyperspectral based apple post-impact damage parameter lossless prediction method according to any of claims 1-4, characterized in that: the step of performing black-and-white correction on the original hyperspectral image and extracting the average spectrum of the damage area of the corrected hyperspectral image comprises the following steps:
correcting the reflectivity of the original hyperspectral image: under the same conditions and parameter settings as those for obtaining the original hyperspectral image, acquiring a white board calibration image by collecting a white calibration board, and acquiring a full-black calibration image with the reflectivity of 0% by covering a camera lens by a camera cover;
and selecting an interested area of the corrected hyperspectral image, and extracting a spectral mean value of the interested area.
6. The hyperspectral-based apple post-impact damage parameter lossless prediction method according to claim 5, wherein: in the process of correcting the reflectivity of the original hyperspectral image, calculation is carried out through the following formula:
wherein I is a corrected hyperspectral image, IrFor the original hyperspectral image, IdDemarcating the image for total black, IwAnd calibrating the white board with images.
7. The hyperspectral-based apple post-impact damage parameter lossless prediction method according to any of claims 1-6, characterized in that: in the step of obtaining the characteristic wavelength of the average spectrum, SG2 is adoptedndAnd the Der calculates the original hyperspectral image and selects the characteristic wavelength.
8. The hyperspectral-based apple post-impact damage parameter lossless prediction method according to claim 7, wherein: said use of SG2ndWhen the Der calculates the original hyperspectral image, the calculation steps are as follows:
smoothing the average spectrum, and calculating by adopting least square convolution, wherein the calculation equation is as follows:
where Y is the original spectrum, Y is the smoothed spectrum, CiIs a smoothing window, N is the number of convolution integers, j is the running index of the original coordinate data, the smoothing array consists of 2m +1 points, where m is half the width of the smoothing window;
and sequentially performing first derivative calculation and second derivative calculation on the smoothed spectral curve to obtain the characteristic wavelength, wherein the first derivative calculation formula is as follows:
the calculation formula of the second derivative is as follows:
wherein Δ λ is an interval between adjacent bands, and Δ λ ═ λk-λj＝λj-λi,λk>λj>λi；
And acquiring the characteristic wavelength, wherein the characteristic wavelength is a wave crest and a wave trough.
9. The hyperspectral-based apple post-impact damage parameter lossless prediction method according to claim 8, wherein: the characteristic wavelengths are 967nm, 1001nm, 1100nm, 1154nm, 1190nm, 1407nm and 1443nm respectively.
10. The hyperspectral-based apple damage parameter lossless prediction method according to claim 1, wherein in the step of establishing a hyperspectral data and damage parameter prediction model, a P L S regression model is adopted to respectively establish a prediction model among spectral data, contact force and average pressure according to the acquired characteristic wavelength.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010293066.XA CN111445469A (en) | 2020-04-15 | 2020-04-15 | Hyperspectrum-based apple damage parameter lossless prediction method after impact |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010293066.XA CN111445469A (en) | 2020-04-15 | 2020-04-15 | Hyperspectrum-based apple damage parameter lossless prediction method after impact |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111445469A true CN111445469A (en) | 2020-07-24 |
Family
ID=71656018
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010293066.XA Pending CN111445469A (en) | 2020-04-15 | 2020-04-15 | Hyperspectrum-based apple damage parameter lossless prediction method after impact |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111445469A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112329296A (en) * | 2020-10-28 | 2021-02-05 | 天河超级计算淮海分中心 | Fruit mechanical damage calculation method and device, computer equipment and storage medium |
| CN113008815A (en) * | 2021-02-24 | 2021-06-22 | 浙江工业大学 | Hyperspectral image information-based method for nondestructive detection of total flavonoids in spina date seeds |
| CN113643388A (en) * | 2021-10-14 | 2021-11-12 | 深圳市海谱纳米光学科技有限公司 | Black frame calibration and correction method and system for hyperspectral image |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPH11148865A (en) * | 1997-11-17 | 1999-06-02 | Kurabo Ind Ltd | Method for processing and correcting two-dimensional spectrum data |
| WO2015147403A1 (en) * | 2014-03-28 | 2015-10-01 | 강원대학교산학협력단 | Method for simultaneously analyzing, by using near-infrared spectroscopy, quantities of nutritional content contained in various kinds of food having different raw materials and shapes |
| CN105021617A (en) * | 2015-07-14 | 2015-11-04 | 华中农业大学 | Hyperspectral imaging-based measuring equipment of chlorophyll content of whole rice plant and method thereof |
| CN107271375A (en) * | 2017-07-21 | 2017-10-20 | 石河子大学 | A kind of high spectral image detecting method of quality of mutton index |
| CN109100323A (en) * | 2018-08-20 | 2018-12-28 | 江苏大学 | A kind of transmitted spectrum harmless quantitative evaluation method of apple water core |
| CN110320165A (en) * | 2019-08-08 | 2019-10-11 | 华南农业大学 | The Vis/NIR lossless detection method of banana soluble solid content |
| CN110596117A (en) * | 2019-08-15 | 2019-12-20 | 山东科技大学 | Hyperspectral imaging-based rapid nondestructive detection method for apple surface damage |
-
2020
- 2020-04-15 CN CN202010293066.XA patent/CN111445469A/en active Pending
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JPH11148865A (en) * | 1997-11-17 | 1999-06-02 | Kurabo Ind Ltd | Method for processing and correcting two-dimensional spectrum data |
| WO2015147403A1 (en) * | 2014-03-28 | 2015-10-01 | 강원대학교산학협력단 | Method for simultaneously analyzing, by using near-infrared spectroscopy, quantities of nutritional content contained in various kinds of food having different raw materials and shapes |
| CN105021617A (en) * | 2015-07-14 | 2015-11-04 | 华中农业大学 | Hyperspectral imaging-based measuring equipment of chlorophyll content of whole rice plant and method thereof |
| CN107271375A (en) * | 2017-07-21 | 2017-10-20 | 石河子大学 | A kind of high spectral image detecting method of quality of mutton index |
| CN109100323A (en) * | 2018-08-20 | 2018-12-28 | 江苏大学 | A kind of transmitted spectrum harmless quantitative evaluation method of apple water core |
| CN110320165A (en) * | 2019-08-08 | 2019-10-11 | 华南农业大学 | The Vis/NIR lossless detection method of banana soluble solid content |
| CN110596117A (en) * | 2019-08-15 | 2019-12-20 | 山东科技大学 | Hyperspectral imaging-based rapid nondestructive detection method for apple surface damage |
Non-Patent Citations (4)
| Title |
|---|
| DOUGLAS BARBIN等: "Near-infrared hyperspectral imaging for grading and classification of pork", 《MEAT SCIENCE》 * |
| 吴瑾光: "《近代傅里叶变换红外光谱技术及应用 下》", 31 December 1994 * |
| 徐夺花: "基于高光谱成像技术的典型水果冲击损伤的研究", 《硕士电子期刊工程科技Ⅰ辑》 * |
| 王志慧, 黄河水利出版社 * |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN112329296A (en) * | 2020-10-28 | 2021-02-05 | 天河超级计算淮海分中心 | Fruit mechanical damage calculation method and device, computer equipment and storage medium |
| CN113008815A (en) * | 2021-02-24 | 2021-06-22 | 浙江工业大学 | Hyperspectral image information-based method for nondestructive detection of total flavonoids in spina date seeds |
| CN113643388A (en) * | 2021-10-14 | 2021-11-12 | 深圳市海谱纳米光学科技有限公司 | Black frame calibration and correction method and system for hyperspectral image |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Zhang et al. | Nondestructive measurement of soluble solids content in apple using near infrared hyperspectral imaging coupled with wavelength selection algorithm | |
| CN111968080B (en) | Method for detecting quality of inside and outside of Feicheng peaches based on hyperspectral and deep learning | |
| Mendoza et al. | Integrated spectral and image analysis of hyperspectral scattering data for prediction of apple fruit firmness and soluble solids content | |
| Talens et al. | Prediction of water and protein contents and quality classification of Spanish cooked ham using NIR hyperspectral imaging | |
| Zhang et al. | Fast prediction of sugar content in Dangshan pear (Pyrus spp.) using hyperspectral imagery data | |
| Hu et al. | Estimating blueberry mechanical properties based on random frog selected hyperspectral data | |
| ElMasry et al. | Chemical-free assessment and mapping of major constituents in beef using hyperspectral imaging | |
| CN108663339B (en) | On-line detection method for mildewed corn based on spectrum and image information fusion | |
| CN110411957B (en) | Nondestructive rapid prediction method and device for shelf life and freshness of fruits | |
| CN111445469A (en) | Hyperspectrum-based apple damage parameter lossless prediction method after impact | |
| CN109100323B (en) | Nondestructive quantitative evaluation method for transmission spectrum of apple water core disease | |
| Huang et al. | Rapid and nondestructive prediction of amylose and amylopectin contents in sorghum based on hyperspectral imaging | |
| Tang et al. | Multispectral imaging for predicting sugar content of ‘Fuji’apples | |
| CN108956545B (en) | Fruit internal quality nondestructive testing model establishing method and system | |
| CN108169165B (en) | Maltose mixture quantitative analysis method based on terahertz spectrum and image information fusion | |
| Xing et al. | Bruise detection on Jonagold apples by visible and near-infrared spectroscopy | |
| CN111044483A (en) | Method, system and medium for determining pigment in cream based on near infrared spectrum | |
| Xu et al. | Factors influencing near infrared spectroscopy analysis of agro-products: a review | |
| Yu et al. | Rapid and visual measurement of fat content in peanuts by using the hyperspectral imaging technique with chemometrics | |
| Zhu et al. | Study on the quantitative measurement of firmness distribution maps at the pixel level inside peach pulp | |
| Liu et al. | Digital image analysis method for rapid measurement of rice degree of milling | |
| Lin et al. | Prediction of protein content in rice using a near-infrared imaging system as diagnostic technique | |
| Zhao et al. | Determination of apple firmness using hyperspectral imaging technique and multivariate calibrations | |
| CN110609011A (en) | Near-infrared hyperspectral detection method and system for starch content of single-kernel corn seeds | |
| CN116008225A (en) | Method for detecting total flavone content of ginkgo leaf for hyperspectral imaging leaves |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20200724 |
|
| RJ01 | Rejection of invention patent application after publication |