CN105300895A - Method for performing early warning against potato germination defects by utilizing feature point tangent included angles - Google Patents

Abstract

The invention discloses a method for performing early warning against potato germination defects by utilizing feature point tangent included angles. Hyperspectral images of multiple sample potatoes are collected under the same condition, early warning time is recorded, regions of interest of the potatoes are selected, spectral data are classified according to different numbers of days of the early warning time, spectral data are intercepted after mean filtering, a spectral fitting function is constructed, a first derivative map is obtained through derivation, included angle cosine values formed by minimum points, maximum points and tangent intersection points thereof serve as characteristic values to undergo discriminant analysis to obtain discriminant coefficient and discriminant constant, the aforementioned steps are repeated on the tested potatoes to obtain corresponding characteristic values, the characteristic values undergo discriminant analysis to obtain early warning results, and the early warning against potato germination is achieved. The potato germination early warning is performed by utilizing two wave bands, the detection efficiency is improved, and the loss caused by potato germination in the circulation process is reduced.

Description

A kind of method utilizing unique point tangent line angle early warning potato sprouting defect

Technical field

The present invention relates to a kind of fruits and vegetables defect inspection method, especially relate to a kind of method utilizing unique point tangent line angle early warning potato sprouting defect.

Background technology

Potato, as one of four generalized grain crops in the world, is extensively planted in worldwide.Potato nutritional is worth high, and have the good reputation of " the second bread " and " underground apple " abroad, the most of nutrition comprised in grain, veterinary antibiotics, potato has substantially.Meanwhile, the arable region of potato is very wide, soil moisture and fertility less demanding, and yield potential is huge, and food and agricultural organization of united state expert regards as when crisis in food appears in future world, can save the cereal crops of the mankind.

In the processes such as results, storage, transport, very easily there is the defects such as various mechanical damage, infection process, germination greening, have a strong impact on Potato Quality, bring economic loss to sweet potato grower and consumer in potato.One of potato GB Testing index is without defects such as frostbite, evil mind, germination, green potatos.Meanwhile, the potato of germination is poisonous (solanine (Solanine)), and unexpected edible meeting threaten to personal safety.

Carrying out the research that potato sprouting detects, is to prevent from preserving or transportation improper and germination commodity that are that cause come into the market, also carry out place mat for the quality restriction of potato and the robotization of classification simultaneously.

At present, many achievements have been obtained for the detection of potato external sort both at home and abroad.At present for potato surface imperfection detection mainly concentrates on mechanical damage, hole, scab, damage in surface, germinate greening etc., wherein in potato sprouting context of detection, Zheng Guannan etc. adopt the gray-scale value differential technique of G passage to detect the potato germinateed, experimentally result, when sprout puts total >10, germination body can be there is in primitive decision image.

For the detection of budded potato, RGB color machines vision system can be utilized to coordinate respective algorithms to complete, and the domestic detection for this direction at present can reach comparatively high-accuracy.But this method relying on the color characteristic on potato surface to carry out germination detection cannot be used for the prediction of potato sprouting time.

Summary of the invention

In order to solve Problems existing in background technology, the object of the present invention is to provide a kind of method utilizing unique point tangent line angle early warning potato sprouting defect, the cosine value of the characteristic angle formed according to the spectral value of area-of-interest 3 wave bands is classified to this region, realizes potato sprouting early warning.

The technical solution adopted for the present invention to solve the technical problems is:

1) high spectrum image of multiple sample potato is gathered under the same conditions:

Described step 1) specifically gather high spectrum image in the following ways: be background with black paperboard, adopt foamed glue to be fixed in paperboard respectively at least 150 potatos, be placed in camera bellows and gather high spectrum image every day, continuous acquisition 5 days.

2) record potato from gather high spectrum image to germinate number of days as pre-warning time, its spectroscopic data, as area-of-interest, is extracted in the germination position choosing potato, classifies to spectroscopic data according to the different number of days of pre-warning time; Concrete enforcement can be the 0th day on the potato sprouting date, and within k days before, backstepping is the front k days that germinates, then within k days, be pre-warning time;

Described step 2) in, the area-of-interest of potato is chosen in the following ways: find and record the eye place-centric (S of budded potato, T), set up centered by eye position (S, T), using the germination area that nine pixels are the length of side as the area-of-interest of data processing.

3) 3 × 3 mean filters are carried out to whole wave band datas of area-of-interest, intercept the discrete spectrum data of area-of-interest place 600-750nm wave band.

Described step 3) in the concrete spectrum simulation function of following formula that adopts matching is carried out to spectroscopic data, and utilize nlinfit function in Matlab to solve:

$f\left(x\right)=\underset{j=1}{\overset{n=5}{\Σ}}{a}_j\×sin(b_j\×x+c_j)$

Wherein, horizontal ordinate x is wavelength value, and ordinate f (x) is spectrum homogenization value, and n represents accumu-late parameter, and j represents the calculating ordinal number of accumu-late parameter.

In concrete enforcement, the root-mean-square error RMSE of Utilization assessment fitting degree determines that the value of n is 5.

4) spectrum simulation function is built, differentiate is carried out to the spectrum simulation function f (χ) of different pre-warning time, obtain the first order derivative figure of area-of-interest, first order derivative figure take wavelength value as horizontal ordinate, with spectrum homogenization value for ordinate, the minimum point A that therefrom selected distance wavelength 680nm is nearest and maximum point C, tangent line is made respectively at minimum point A and maximum point C place, article two, the intersection point of tangent line is intersection points B, calculates the cosine value of the angle ∠ ABC formed by minimum point A, maximum point C and intersection points B as eigenwert cosB;

Described eigenwert cosB specifically calculates in the following ways: the coordinate of minimum point A, maximum point C and intersection points B is labeled as (λ respectively _a, R _a), (λ _b, R _b) and (λ _c, R _c), wherein, λ _a, λ _band λ _cbe respectively the wavelength value that minimum point A, maximum point C and intersection points B are corresponding, R _a, R _band R _cbe respectively the spectrum homogenization value that minimum point A, maximum point C and intersection points B are corresponding, adopt the cosine law to calculate angle ∠ ABC according to the coordinate of three points.

$\begin{array}{c}{\mathrm{\λ}}_B^2+R_B^2+{\mathrm{\λ}}_A\×{\mathrm{\λ}}_C-{\mathrm{\λ}}_A\×{\mathrm{\λ}}_B-{\mathrm{\λ}}_B\×{\mathrm{\λ}}_C+\\ \cos B=\frac{R_A\×R_C-R_A\×R_B-R_B\×R_C}{\sqrt{{({\mathrm{\λ}}_C-{\mathrm{\λ}}_B)}^2+{(R_C-R_B)}^2}\×\sqrt{{({\mathrm{\λ}}_B-{\mathrm{\λ}}_A)}^2+{(R_B-R_A)}^2}}\end{array}.$

5) respectively discriminatory analysis is carried out to the eigenwert cosB of the area-of-interest of different pre-warning time, obtain respective discriminant coefficient p respectively _kwith differentiation constant q _k;

Described step 5) in specifically utilize Fischer discriminant coefficient method of discrimination to represent following formula discriminatory analysis is carried out to eigenwert cosB:

F ₀＝p ₀×X+q ₀

F ₁＝p ₁×X+q ₁

F ₂＝p ₂×X+q ₂

F ₃＝p ₃×X+q ₃

F ₄＝p ₄×X+q ₄

Wherein, F ₀be the score value of the 0th day pre-warning time sample potato, F ₁be the score value of the 1st day pre-warning time sample potato, F ₂be the score value of the 2nd day pre-warning time sample potato, F ₃be the score value of the 3rd day pre-warning time sample potato, F ₄it is the score value of the 4th day pre-warning time sample potato.X represents the set of all pre-warning time lower eigenvalue cosB, p _kfor the discriminant coefficient set that the eigenwert cosB of kth sky pre-warning time sample potato is corresponding, q _kfor differentiating constant;

6) tested potato is repeated above-mentioned steps 1) ~ 4) obtain corresponding eigenwert cosB, and discriminatory analysis acquisition early warning result is carried out to it, realize the early warning to potato sprouting.

Described step 6) in the discriminatory analysis result of eigenwert cosB concrete in the following ways:

Tested potato is extracted all eigenwert cosB obtained and substitutes in following formula the score value F obtained under each pre-warning time _ks, F _ksk in value corresponding to maximal value is pre-warning time:

F _0s＝p ₀×X _s+q ₀

F _1s＝p ₁×X _s+q ₁

F _2s＝p ₂×X _s+q ₂

F _3s＝p ₃×X _s+q ₃

F _4s＝p ₄×X _s+q ₄

Wherein, F _0sbe the score value of the 0th day tested potato of pre-warning time, F _1sbe the score value of the 1st day tested potato of pre-warning time, F _2sbe the score value of the 2nd day tested potato of pre-warning time, F _3sbe the score value of the 3rd day tested potato of pre-warning time, F _4sbe the score value of the 4th day tested potato of pre-warning time, X _srepresent the fitting parameter set of tested potato;

The inventive method utilizes high light spectrum image-forming technology Continuous Observation potato surface eye, and record eye situation is to germination.The note potato sprouting date is the 0th day, and k days before backsteppings are pre-warning time for k days, k before germinateing, and adds up and sorts out the pre-warning time of each eye.Extract the spectroscopic data that each potato bud eye position length of side is the square area of 9 pixels, after mean filter, after mean normalization, Function Fitting are carried out to the spectrum of the 600-750nm wave band in this region, draw the spectral curve at different pre-warning time Potato eye position.Utilize curve first order derivative for the band value λ under extreme value _a, λ _c, obtain point of intersection of tangents B (λ with its derivative value _b, R _b), calculate each eye of potato in corresponding wave band incision wire clamp cosine of an angle value, build discriminant function in this, as variable, realize the early warning of potato sprouting.

The invention has the beneficial effects as follows:

The present invention can obtain the early warning information of the germination of potato, carries out early warning, achieve potato sprouting early warning to the germination of potato, improves detection efficiency, reduces the loss that process of circulation potato sprouting causes.

Accompanying drawing explanation

Fig. 1 is the realization flow figure of the inventive method.

Fig. 2 is potato fixing situation pictorial diagram.

Fig. 3 is the extraction situation of embodiment potato bud eye position area-of-interest.

The curve of spectrum fitted figure of Fig. 4 to be embodiment 600-750nm wave band potato bud eye position pre-warning time be 1 (day1).

The curve of spectrum first order derivative figure of Fig. 5 to be embodiment 600-750nm wave band potato bud eye position pre-warning time be 1 (day1).

The characteristic angle schematic diagram ∠ B of Fig. 6 to be embodiment potato bud eye position pre-warning time be 1 (day1).

Fig. 7 is the spectral curve germinateed the same day (day0) in embodiment 600-750nm wave band potato bud eye position.

Fig. 8 is the curve of spectrum first order derivative figure germinateed the same day (day0) in embodiment 600-750nm wave band potato bud eye position.

Fig. 9 is the characteristic angle schematic diagram ∠ B germinateed the same day (day0) in embodiment potato bud eye position.

The characteristic angle schematic diagram ∠ B of Figure 10 to be embodiment potato bud eye position pre-warning time be 2 (day2).

Embodiment

Below in conjunction with drawings and Examples, the invention will be further described.

Embodiments of the invention are as follows:

As shown in Figure 1, first, be fixed on by potato (shown in Fig. 2) in black paperboard, under being placed in room temperature, shading stores, and gathers its high spectrum image every day.Debugging Hyperspectral imager, mates the parameters such as its object distance, intensity of illumination, camera exposure time, scan area, sweep velocity, is as the criterion can gather out clear, indeformable image, scanning potato high spectrum image.Record eye situation is to germination.Record potato from gather high spectrum image to germinate number of days as pre-warning time, its spectroscopic data, as area-of-interest, is extracted in the germination position choosing potato, classifies to spectroscopic data according to the different number of days of pre-warning time.Extract the spectroscopic data that each potato bud eye position length of side is the square area of 9 pixels, after mean filter, after mean normalization, Function Fitting are carried out to the spectrum of the 600-750nm wave band in this region, draw the spectral curve at different pre-warning time Potato eye position.Utilize curve first order derivative for the band value λ under extreme value _a, λ _c, obtain the band value λ of tangent line included angle B _b, calculate the cosine value of each eye of potato characteristic angle under corresponding wave band, build discriminatory analysis in this, as variable, realize the early warning of potato sprouting.

As shown in Figure 3, main operating process is as follows for potato bud eye position area-of-interest: the eye position (S, T) of record budded potato, with (S, T), centered by, 9 pixels are the area-of-interest of region as data processing of the length of side, are defined as germination position.Extract the spectroscopic data at germination position, sort out spectroscopic data according to germination.Under different pre-warning time, the number at the eye position added up is as shown in table 1.Day0 representative was germinateed the same day, and it is 1-4 days that day1-day4 represents pre-warning time.

The different germination of table 1 lower add up the number at eye position

Fig. 4 to be 600-750nm wave band potato pre-warning time be 1 spectroscopic data matched curve figure.In the present invention, the fitting function form of employing is as follows:

$f\left(x\right)=\underset{j=1}{\overset{5}{\Σ}}{\mathrm{\α}}_j\×sin(b_j\×x+c_j)---\left(1\right)$

Sample size is 117 eye positions, and spectrum discrete value is obtained by the mean normalization of 117 sample light modal data.

Figure 5 shows that 600-750nm wave band potato bud eye position pre-warning time is the curve of spectrum first order derivative figure of 1 (day1).The minimum point A (667.92,0.038489) that selected distance wavelength 680nm is nearest, maximum point C (688.16,0.380951), as the unique point in the present invention.

Figure 6 shows that potato bud eye position pre-warning time is the characteristic angle schematic diagram ∠ B of 1 (day1).In the present invention, unique point A (667.92,0.038489), unique point C (688.16,0.380951), utilize first derivative values to obtain angular position B (678.03,0.048108).Determine λ _a, λ _b, λ _cafter equiwavelength's value, utilize i-th potato sample (667.92, R _ai), B (678.03, R _bi), C (688.16, R _ci) calculate its eigenwert cosB.(i＝1,2,3…117)

Figure 7 shows that the spectral curve germinateed the same day (day0) in 600-750nm wave band potato bud eye position.Sample size is 118 eye positions, and curve is obtained by the mean normalization of 118 sample light modal data.

Figure 8 shows that the curve of spectrum first order derivative figure germinateed the same day (day0) in 600-750nm wave band potato bud eye position.Two extreme point A (667.92,0.038489) that selected distance wavelength 680nm is nearest, C (688.16,0.380951), as the unique point in the present invention.

Figure 9 shows that the characteristic angle schematic diagram ∠ B germinateed the same day (day0) in potato bud eye position.In the present invention, A (667.92,0.099426), B (678.03,0.107984), C (688.16,0.419124).Determine λ _a, λ _b, λ _cafter equiwavelength's value, utilize i-th potato sample (667.92, R _ai), B (678.03, R _bi), C (688.16, R _ci) calculate its eigenwert cosB.(i＝1,2,3…118)

Figure 10 shows that potato bud eye position pre-warning time is the characteristic angle schematic diagram ∠ B of 2 (day2).In the present invention, A (667.92,0.103862), B (678.03,0.111303), C (688.16,0.420053).Determine λ _a, λ _b, λ _cafter equiwavelength's value, utilize i-th potato sample (667.92, R _ai), B (678.03, R _bi), C (688.16, R _ci) calculate its eigenwert cosB.(i＝1,2,3…110)

Discriminatory analysis: the eigenwert cosB of Fischer discriminant coefficient method of discrimination to the area-of-interest of different pre-warning time utilizing SPSS to provide classifies.By the Fischer discriminant function coefficient table in result, as shown in table 2, set up discriminant function.

Table 2 Fischer discriminant function coefficient table

Computing formula is as follows thus:

F ₀＝-66.898X-23.540

F ₁＝-94.664X-45.523

F ₂＝-92.748X-43.764

F ₃＝-95.004X-45.839

F ₄＝-90.524X-41.766

Potato sprouting situation early warning: utilize the discriminant function obtained to carry out early warning.The cosine value of the unique point tangent line angle of new samples is substituted into discriminant function as variable, calculates all kinds of score F _ks, obtaining point maximum kind is early warning class.Through inspection, get germinate the same day 118 samples, pre-warning time is 117 samples of 1, and early warning result is as shown in table 3, to wherein 86.0% sample carried out correct classification.

Table 3 potato early warning result

Visible, improvement that the classifying quality of the single features value that present invention is directed at is perfect, in conjunction with other eigenwerts, can obtain better result for early warning potato sprouting.

Claims (7)

1. utilize a method for unique point tangent line angle early warning potato sprouting defect, it is characterized in that the step of the method is as follows:

1) high spectrum image of multiple sample potato is gathered under the same conditions:

2) record potato from gather high spectrum image to germinate number of days as pre-warning time, its spectroscopic data, as area-of-interest, is extracted in the germination position choosing potato, classifies to spectroscopic data according to the different number of days of pre-warning time;

3) 3 × 3 mean filters are carried out to whole wave band datas of area-of-interest, intercept the discrete spectrum data of area-of-interest place 600-750nm wave band;

4) spectrum simulation function is built, differentiate is carried out to spectrum simulation function f (x) of different pre-warning time, obtain the first order derivative figure of area-of-interest, first order derivative figure take wavelength value as horizontal ordinate, with spectrum homogenization value for ordinate, the minimum point A that therefrom selected distance wavelength 680nm is nearest and maximum point C, tangent line is made respectively at minimum point A and maximum point C place, article two, the intersection point of tangent line is intersection points B, calculates the cosine value of the angle ∠ ABC formed by minimum point A, maximum point C and intersection points B as eigenwert cosB;

5) respectively discriminatory analysis is carried out to the eigenwert cosB of the area-of-interest of different pre-warning time, obtain respective discriminant coefficient p respectively _kwith differentiation constant q _k;

6) tested potato is repeated above-mentioned steps 1) ~ 4) obtain all eigenwert cosB, and discriminatory analysis acquisition early warning result is carried out to it, realize the early warning to potato sprouting.

2. a kind of method utilizing unique point tangent line angle early warning potato sprouting defect according to claim 1, it is characterized in that: described step 1) specifically gather high spectrum image in the following ways: take black paperboard as background, at least 150 potatos are separately fixed in paperboard, be placed in camera bellows and gather high spectrum image every day, continuous acquisition 5 days.

3. a kind of method utilizing unique point tangent line angle early warning potato sprouting defect according to claim 1, it is characterized in that: described step 2) in, the area-of-interest of potato is chosen in the following ways: find and record the eye position (S of budded potato, T), set up centered by eye position (S, T), using the germination area that nine pixels are the length of side as area-of-interest.

4. a kind of method utilizing unique point tangent line angle early warning potato sprouting defect according to claim 1, is characterized in that: described step 3) in the concrete spectrum simulation function of following formula that adopts matching is carried out to spectroscopic data:

$f\left(x\right)=\underset{j=1}{\overset{n=5}{\Σ}}{\mathrm{\α}}_j\×sin(b_j\×x+c_j)$

Wherein, horizontal ordinate x is wavelength value, and ordinate f (x) is spectrum homogenization value, and n represents accumu-late parameter, and j represents the calculating ordinal number of accumu-late parameter.

5. a kind of method utilizing unique point tangent line angle early warning potato sprouting defect according to claim 1, is characterized in that: described step 5) in specifically utilize Fischer discriminant coefficient method of discrimination to represent following formula discriminatory analysis is carried out to eigenwert cosB:

F ₀＝p ₀×X+q ₀

F ₁＝p ₁×X+q ₁

F ₂＝p ₂×X+q ₂

F ₃＝p ₃×X+q ₃

F ₄＝p ₄×X+q ₄

Wherein, F ₀be the score value of the 0th day pre-warning time sample potato, F ₁be the score value of the 1st day pre-warning time sample potato, F ₂be the score value of the 2nd day pre-warning time sample potato, F ₃be the score value of the 3rd day pre-warning time sample potato, F ₄be the score value of the 4th day pre-warning time sample potato, X represents the set of all pre-warning time lower eigenvalue cosB, p _kfor the discriminant coefficient set that the eigenwert cosB of kth sky pre-warning time sample potato is corresponding, q _kfor differentiating constant.

6. a kind of method utilizing unique point tangent line angle early warning potato sprouting defect according to claim 1, is characterized in that: described step 6) in the discriminatory analysis result of all eigenwert cosB concrete in the following ways: tested potato is extracted all eigenwert cosB obtained and substitutes in following formula the score value F obtained under each pre-warning time _ks, F _ksk in value corresponding to maximal value is pre-warning time:

F _0s＝p ₀×X _s+q ₀

F _1s＝p ₁×X _s+q ₁

F _2s＝p ₂×X _s+q ₂

F _3s＝p ₃×X _s+q ₃

F _4s＝p ₄×X _s+q ₄

Wherein, F _0sbe the score value of the 0th day tested potato of pre-warning time, F _1sbe the score value of the 1st day tested potato of pre-warning time, F _2sbe the score value of the 2nd day tested potato of pre-warning time, F _3sbe the score value of the 3rd day tested potato of pre-warning time, F _4sbe the score value of the 4th day tested potato of pre-warning time, X _srepresent the fitting parameter set of tested potato.

7. a kind of method utilizing unique point tangent line angle early warning potato sprouting defect according to claim 1, is characterized in that: described step 4) in eigenwert cosB specifically calculate in the following ways: the coordinate of minimum point A, maximum point C and intersection points B is labeled as (λ respectively _a, R _a), (λ _b, R _b) and (λ _c, R _c), wherein, λ _a, λ _band λ _cbe respectively the wavelength value that minimum point A, maximum point C and intersection points B are corresponding, R _a, R _band R _cbe respectively the spectrum homogenization value that minimum point A, maximum point C and intersection points B are corresponding, calculate angle ∠ ABC according to the cosine law that the coordinate of three points adopts following formula to represent:

$\cos B=\frac{\begin{array}{c}{\mathrm{\λ}}_B^2+R_B^2+{\mathrm{\λ}}_A\×{\mathrm{\λ}}_C-{\mathrm{\λ}}_A\×{\mathrm{\λ}}_B-{\mathrm{\λ}}_B\×{\mathrm{\λ}}_C+\\ R_A\×R_C-R_A\×R_B-R_B\×R_C\end{array}}{\sqrt{{({\mathrm{\λ}}_C-{\mathrm{\λ}}_B)}^2+{(R_C-R_B)}^2}\×\sqrt{{({\mathrm{\λ}}_B-{\mathrm{\λ}}_A)}^2+{(R_B-R_A)}^2}}.$