CN205762405U - Sorting unit for online Non-Destructive Testing Apple Mould Core equipment - Google Patents

Abstract

nullA kind of sorting unit for online Non-Destructive Testing Apple Mould Core equipment，Including detection conveyer，It is provided with some Fructus Mali pumilae pallets on the conveyer belt of detection conveyer，It is characterized in that，Detection conveyer is provided with photographic head and detection black box，The spectral detection module arranged in detection black box carries out spectral detection to the Fructus Mali pumilae through detection black box，Rear class at detection conveyer is provided with weighing unit，Complete the entrance weighing unit of the Fructus Mali pumilae after detection to weigh，Weighing unit rear class is provided with inclined-plane sortation conveyor，According to testing result with weigh and size carries out fruit sorting，This utility model is based near infrared detection technology，Achieve Apple Mould Core on-line checking，Reduce transmission light path and the impact on spectral detection of the Fructus Mali pumilae density，The reliability of the adjustment model is high，Improve disease discriminant accuracy，Autonomous test speed is only 1 2s，Disclosure satisfy that detection demand on line，Morbidity Fructus Mali pumilae can be effectively identified during Fructus Mali pumilae material storage，Reduce production and processing and storage phase sickness rate，Quality assurance.

Description

Sorting device for on-line nondestructive testing of apple moldy core equipment

Technical Field

The utility model belongs to the technical field of agricultural intelligent equipment, in particular to a sorting unit that is used for online nondestructive test apple moldy core disease equipment.

Background

The apple moldy core disease is a common apple disease, fruits rot from ventricles after the disease occurs, the fruits gradually spread outwards, and the whole fruits become moldy and rot in severe cases. The appearance of fruits at the early stage of the onset of the core rot is not obviously characterized, the onset of the core rot is from inside to outside, the diseases are difficult to identify in the fruit picking and sorting links, the detection of the core rot apple is a great problem in the fruit identification and deep processing processes, if the core rot apple can be effectively identified in a nondestructive mode, the quality of the fruits can be effectively improved, and the core rot apple identification method has important significance for improving the quality and the benefit of the apple industry in China.

In recent years, with the increasing concern of fruit quality and food safety, experts in related fields carry out deep research on the mechanism and detection method of apple moldy core, and the moldy core detection is explored from the aspects of biological impedance characteristics, machine vision, optical characteristics and the like. The spectral analysis technology has incomparable advantages compared with other technologies when processing objects which cannot be contacted and are not damaged, and is applied to fruit nondestructive testing more. The near-infrared analysis has the advantages of easy acquisition of visible region spectral analysis signals and rich infrared region spectral analysis information, and is more effective in qualitative and quantitative analysis of organic substances. The disease detection based on the optical characteristics mainly adopts a near infrared spectrum analysis technology, carries out disease judgment according to a substance characteristic absorption peak, has high precision and good effect, but has complex data analysis and model establishment, mostly adopts a special computer to carry out data analysis and processing, and has high price of a spectrometer, so that research results are difficult to be popularized and applied in actual fruit production and processing.

Disclosure of Invention

In order to overcome the shortcoming of the prior art, the utility model aims to provide a sorting unit for online nondestructive test apple heartburn equipment, can be on line nondestructive test apple heartburn back, with apple automatic separation, have with low costs, easy operation carries out the disease automatically and distinguishes and selects separately the characteristics such as.

In order to realize the purpose, the utility model discloses a technical scheme is:

the utility model provides a sorting unit for online nondestructive test apple mould heart disease equipment, including detecting conveyer 8, be provided with a plurality of apple trays 9 on detecting conveyer 8's the conveyer belt, be provided with camera 1 and detect black case 2 on detecting conveyer 8, the spectral detection module 15 that detects the setting in the black case 2 carries out spectral detection to the apple through detecting black case 2, be provided with weighing platform 4 at the back level that detects conveyer 8, the apple after the completion detected gets into weighing platform 4 and weighs, weighing platform 4 back level is provided with inclined plane sorting conveyor 6, carry out fruit sorting according to testing result and weighing and size.

Do the processing of caving in the apple tray 9, along direction of delivery single-row setting on the conveyer belt, camera 1 sets up before detecting black case 2, and the apple of process is taken a picture and is discerned, carries the result to the treater and judges its size.

The detection black box 2 is internally provided with a cross sliding table 10, two coaxial stepping sliding table motors are arranged in the vertical direction of the cross sliding table 10, a reverse lead screw motor is arranged in the horizontal direction, the coaxial stepping sliding table motors drive the reverse lead screw motors to move in the vertical direction, and the spectrum detection module 15 is fixed at two ends of the reverse lead screw motors through a support.

The spectrum detection module 15 comprises a light source and a receiving device, wherein the light source adopts an LED with the central wavelength of 710nm and the half-wave width of 20nm, the working voltage is 3.4V, and the scattering angle of the light source is 120 degrees; the receiving device adopts an avalanche diode as a sensitive device, adopts a filter amplifying circuit as a signal processing circuit, and converts the transmitted light intensity into an electric signal through a sampling module to obtainTo the transmission intensity M, the discrimination model is:if the result is 0, it is a healthy fruit, and if the result is 1, it is a heartburn fruit, wherein,

is the Largrange coefficient, b^*As a weight matrix, the weight matrix is,x is an input sample, x_iTo support the vector, x_i＝[M_i,P_i,R_i,G_i]^T，＝1,2,…，y_i1, is a geometric spacing, M_iIs the sample transmission intensity, P_iIs the fruit shape index, R, of the sample_iIs the diameter of the sample, G_iIs the sample mass.

The apple weighing machine is characterized in that a front shifting piece 3 and a rear shifting piece 5 are arranged on the weighing platform 4, apples enter the weighing platform 4 from a detection conveyor 8 under the action of the front shifting piece 3, and apples enter the inclined plane sorting conveyor 6 from the weighing platform 4 under the action of the rear shifting piece 5.

The inclined surface sorting conveyor 6 is provided with a plurality of check bars, a first outlet 12, a second outlet 13 and a third outlet 14 are sequentially arranged below the inclined surface along the conveying direction, and each outlet is provided with a movable baffle 7.

The inlet diameter of the second outlet 13 is 70mm, and the inlet diameter of the third outlet 14 is larger than 70 mm.

The camera 1, the weighing platform 4 and the spectrum detection module 15 are all connected with the processor, the control end of the processor is connected with each movable baffle 7, the detection result of the spectrum detection module 15 is input into the processor, if the processor judges that the mould core fruit is the mould core fruit, the movable baffle 7 of the first outlet 12 is opened, the mould core fruit enters the first outlet 12, the identification result of the camera 1 is input into the processor, if the processor judges that the ruler diameter of the processor is smaller than 70mm, the movable baffle 7 of the second outlet 13 is opened, the mould core fruit enters the second outlet 13, and if the processor judges that the ruler diameter of the processor is larger than 70mm, the movable baffle 7 of the third outlet 14 is opened, and the mould core fruit enters the third outlet 14.

Compared with the prior art, the beneficial effects of the utility model are that:

the utility model discloses based on near-infrared detection technique, provided apple mould core disease on-line measuring method, independently design apple mould core disease sorting unit, detection speed only is 1-2s, can satisfy on-line measuring demand, provides new thinking to apple mould core disease on-line nondestructive test detection theory and method, can effectively discern the sick apple at the apple raw materials warehouse entry in-process, reduces production and processing and storage period morbidity, guarantee quality.

The self-designed narrow-band LED light source and the detection system have high detection sensitivity, can replace a spectrometer on the existing disease sorting line, and effectively reduce the equipment cost.

The utility model discloses can be used to the sorting of online nondestructive test apple heartburn equipment, on line after nondestructive test apple heartburn, with apple automatic separation.

Drawings

FIG. 1 is a graph showing the relationship between apple transmission spectrum and the degree of morbidity.

Fig. 2 is a schematic structural diagram of the sorting device of the present invention.

Fig. 3 is a schematic structural diagram of the black box detection device of the present invention.

Fig. 4 is a schematic structural diagram of the high-frequency driving module of the present invention.

Fig. 5 is a schematic diagram of the discriminant model construction of the present invention.

Fig. 6 is a schematic diagram of the training result of the present invention.

Detailed Description

The following describes embodiments of the present invention in detail with reference to the drawings and examples.

The utility model discloses a theoretical foundation:

in the near-infrared band range, specific original groups all have characteristic absorption peaks corresponding to the specific original groups, and the proportion of light absorption meets the Lambert-beer law. After the apple is suffered from the mildew heart disease, the species and the proportion of all substances in the apple are changed, and the absorption, reflection and scattering capabilities of a spectrum are obviously influenced, so that the near-infrared transmission spectrum curves are different. A transmission spectrum data acquisition platform for apple core rot is set up in a test, 1 high-concentration halogen lamp with the power of 50w is used as a light source, a portable spectrometer OFS1100(Ocean Optics company) is used as a spectrometer, the test is completed in a dark box for eliminating ambient light interference, the inner wall of the dark box is sprayed with sub-light, and light interference such as diffuse reflection is reduced by light absorption sponge. The number of the tested apple samples is 304, the weight, the height in the direction of the fruit stem and the diameter in the direction of the equator are measured for each apple, the spectrum measurement is carried out on the 304 apple samples by using a detection platform and the apple samples are cut along the fruit stem, samples with different disease degrees in the tested diseased fruits are taken to analyze the spectrum data as shown in figure 1, the obtained healthy fruits have good permeability near the wavelength of 710nm, the diseased fruits have poor permeability near the wavelength of 710nm, and the obtained apple spectrum data and the core rot disease are subjected to correlation analysis along with the negative correlation between the disease degree and the permeability to obtain the table 1.

TABLE 1 relationship of Corset disease to Transmission Spectrum

Serial number	Wave band	Correlation coefficient	Serial number	Wave band	Correlation coefficient
						1	709.36	-0.720	6	711.14	-0.711
2	709.8	-0.719	7	708.47	-0.711
						3	708.91	-0.717	8	708.02	-0.710
4	710.69	-0.715	9	706.69	-0.709
						5	710.25	-0.713	10	707.58	-0.705

The wave band with the strongest correlation with the apple moldy core in the transmission spectrum is in the area near 706-710nm, wherein the correlation is the largest at 709nm, so that the finally selected characteristic wave band related to the apple moldy core is 709 nm. Meanwhile, according to the lambert beer theorem, the distance between the light source and the spectrum receiving device is caused by the difference of the diameters of the fruits in the detection direction of the apples, namely the equator direction, namely, the change of the optical path, the light transmission capacity is influenced by the difference of the densities of the apples, and the detection is also greatly influenced. Therefore, according to the principle, the apple moldy core disease can be accurately judged by knowing the spectrum transmission absorption intensity, the fruit type index and the weight.

As shown in figure 2, the utility model relates to an apple sorting device for online nondestructive detection of moldy core, which comprises a detection conveyor 8, wherein the material of the conveyor belt is PVC belt, an apple tray 9 is arranged on the belt, the tray is internally sunk, the bottom of the tray is fixed on the conveyor through bolts, the detection conveyor 8 drives apples on the apple tray 9 to move, a camera 1 is fixed on the detection, the apples within the shooting range of the camera are shot and identified, a detection black box 2 detects disease parameters such as spectrum and the like of the apples entering the black box, a cross sliding table 10 is arranged inside the black box 2 and is fixed on the inner wall of the black box through bolts, wherein, the vertical direction is a coaxial motor, a reverse screw rod motor fixed on the sliding table is driven to move in the vertical direction, a spectrum detection module 15 is fixed on the reverse screw rod through a bracket, after the apples are detected on the detection conveyor 8, under the action of the front shifting piece 3, apples enter the weighing table 4 from the detection conveyor 8, the rear shifting piece 5 pushes the apples out of the weighing table 4 after weighing and detection are finished, the apples enter the inclined surface sorting conveyor 6, fruit sorting is carried out according to detection results and different sizes of the apples, when the mildewed fruits pass through a sorting line, a switch on the first outlet 12 is opened, the mildewed fruits roll out of the sorting line and enter the first outlet under the action of gravity, the healthy fruits are divided into fruits with diameters smaller than 70mm and larger than 70mm, the fruits with diameters smaller than 70mm enter the second outlet 13, and the fruits with diameters larger than 70mm enter the third outlet 14.

The utility model discloses in, CAMERA 1 adopts the PC CAMERA of the nine tripod companies in hong Kong, and the mode through USB transmission will sample image information and upload to the treater, adopts OPENCV database technology to establish the analysis program based on C + + + language, obtains the high H and the equator direction fruit footpath R of the apple that awaits measuring through program analysis to obtain fruit shape index P ═ H/R.

Weighing platform 4 is a part of weighing platform 11, weighing platform 11 core adopts sun facing instrument company's split type electronic balance, measuring range 1kg, measurement accuracy 0.01, data accessible serial ports upload to the treater and carry out data storage and processing, the pan of steelyard both sides set up preceding plectrum 3 and back plectrum 5 respectively, the plectrum comprises 42 type step motor, thunder plug drive module DM320C and plectrum body of thunder plug company, dial to weighing platform 4 by preceding plectrum 3, carry out weight measurement to it, obtain apple weight G, dial the apple to the inclined conveyor belt 6 by back plectrum 5 after the weight measurement is accomplished.

The conveyer belt inclination 10 of inclined conveyor belt 6, the inclined plane lower extreme sets up adjustable fender 7 and export 12, No. two exports 13, No. three exports 14, and the host computer carries out disease judgement and classification according to the apple characteristic, lets it fall into different exports according to its different characteristics, and when the apple entered its matching export, adjustable fender 7 in exit lifted, and the apple rolls into this export.

As shown in fig. 3, the detection black box 2 is a shading detection darkroom, and is internally provided with a cross sliding table 10, two stepping sliding table motors in the vertical direction and a reverse screw rod motor in the horizontal direction, and two ends of the reverse screw rod motor are provided with light transmission detection modules 15.

The light transmission detection module 15 is composed of a light source and a receiving device, the light source adopts an LED with the central wavelength of 710nm and the half-wave width of 20nm, the working voltage is 3.4V, the scattering angle of the light source is 120 degrees, an LED high-frequency driving circuit is built based on an NSI45030AZT1 constant current diode, and the circuit is shown in figure 4. The receiving device adopts an avalanche diode as a sensitive device, adopts a filter amplifying circuit as a signal processing circuit, and converts the transmission light intensity into an electric signal through a sampling module to obtain the transmission intensity M.

A discriminant function:

wherein,is the Largrange coefficient, b^*As a weight matrix, the weight matrix is,x is an input sample, x_iTo support the vector, x_i＝[M_i,P_i,R_i,G_i]^T，i＝1,2,…，y_i1, is a geometric spacing, M_iIs the sample transmission intensity, P_iIs the fruit shape index, R, of the sample_iIs the diameter of the sample, G_iIs the sample mass.

The distinguishing method comprises the following steps:

let sample apple be { (X)_i,Y_i),i＝1,2,…,l},X_iNumbering the apples, Y_iFor apple health, Y_i0 is a healthy fruit, Y_i1 is a heartburn fruit.

Presence classification hyperplane

wx_i+b＝0 (1)

Apples can be correctly classified into two categories according to diseases. W and b are one-dimensional parameter vectors, and a sample point X is defined_iTo the classification hyperplane indicated by the formula (1)Partition_iIs composed of

_i＝y_i(wx_i+b)＝|wx_i+b| (2)

Normalizing w and b in (2), and defining the normalized interval as the geometric interval

${\mathrm{\δ}}_i=\frac{{\mathrm{wx}}_i+b}{\left|\right|w\left|\right|}---\left(3\right)$

Optimal classification performance even if the geometric separation of the sample set and the classification hyperplane_iMaximum, i.e. problem conversion to

$\left\{\begin{array}{c}\min \frac{\left|\right|w|{|}^2}{2}\\ s.t.y_i({\mathrm{wx}}_i+b)\≥1,i=1,2,\mathrm{...}l\end{array}\right.---\left(4\right)$

Because the calculation is complex, the direct solution is not generally carried out, the formula (4) is converted into a dual problem according to the Largrange dual theory,

$\left\{\begin{array}{c}\max Q\left(\mathrm{\α}\right)={\Σ}_{i=1}^l{\mathrm{\α}}_i-{\Σ}_{i=1}^l{\Σ}_{j=1}^l{\mathrm{\α}}_i{\mathrm{\α}}_j\left(x_i{x}_j\right)\\ s.t.{\Σ}_{i=1}^l{\mathrm{\α}}_i{y}_i=0,{\mathrm{\α}}_i\≥0\end{array}\right.---\left(5\right)$

when solving the dual problem, the dot product of the sample point vectors needs to be calculated, and a kernel function K (x) meeting the Mercer condition is adopted_i,x_j) Instead of dot product operation, the method can simplify calculation and improve processingSpeed, then equation 5 translates to:

$\left\{\begin{array}{c}\max Q\left(\mathrm{\α}\right)={\Σ}_{i=1}^l{\mathrm{\α}}_i-\frac{1}{2}{\Σ}_{i=1}^l{\Σ}_{j=1}^l{\mathrm{\α}}_i{\mathrm{\α}}_j(x_i,x_j)\\ s.t.\{\begin{array}{c}{\Σ}_{i=1}^l{\mathrm{\α}}_i{y}_i=0\\ 0\≤{\mathrm{\α}}_i\≤C\end{array},\mathrm{\α}\≥0\end{array}\right.---\left(6\right)$

the optimal solution solvable by equation 6 is:

${\mathrm{\α}}^{*}={\[{\mathrm{\α}}_1^{*},{\mathrm{\α}}_2^{*},\mathrm{...}{\mathrm{\α}}_l^{*}\]}^T$

$w^{*}=\underset{i=1}{\overset{l}{\Σ}}{\mathrm{\α}}_i^{*}{y}_i\mathrm{\Φ}\left(x_i\right)$

the final discriminant function is thus:

and designing a discrimination program based on MFC technology in a C + + language environment, and inputting various data measured in the detection link into an SVM algorithm program for discrimination by calling a Matlab software interface.

A Support Vector Machine (Support Vector Machine) is a new learning method proposed according to the statistical learning theory, which is based on the structural risk minimization criterion, the optimal hyperplane is constructed with the maximum classification interval, when the SVM solves the nonlinear classification problem, the conversion from a low-cheg-dimensional space to a high-dimensional space is realized by introducing a kernel function, the operand is small and is independent of the dimension of a sample, meanwhile, the model parameters comprise a punishment parameter C, a radial basis kernel function parameter g, an order p, an off-training error and the like, wherein the penalty factor C is a coefficient specified by a user and represents the penalty of adding to the point with misjudgment during model training, when C is increased in a reasonable range, the point with misjudgment can be obviously reduced, when sample data is unbalanced or needs to be adjusted artificially, the prediction precision can be effectively improved by optimizing the model parameter C. The radial basis kernel function parameter g is a characteristic space for converting nonlinear separable samples into linear separable samples, and different kernel function selections can lead classification hyperplanes generated by the SVM model to be different, so that larger difference is generated, and the performance of the SVM model is directly influenced.

The utility model discloses well apple moldy core classification problem belongs to the non-linear classification, because moldy core fruit morbidity process is by inside to outside, is difficult to distinguish with normal fruit in the outward appearance, leads to healthy fruit and sick fruit proportion serious imbalance in the sample, in the characteristic wave band extraction test in earlier stage, healthy fruit 250 in 304 samples, sick fruit only 54, good fruit bad fruit proportion reaches 5:1, and the sample proportion is serious unbalance, has proposed very high requirement to the selection of modeling approach. Aiming at the actual situation, an SVM algorithm is selected for model construction, a penalty factor C is set, and the influence of sample number unbalance on the discrimination model is eliminated.

Through a large amount of parameter optimization, a good fruit punishment factor in the discrimination model is set to be 1, a bad fruit punishment factor is set to be 1.53, the influence of a small number of bad fruit samples on model construction can be reduced, and a model construction flow chart is shown in fig. 5. Test data are extracted according to the early-stage features, a support vector machine is adopted for training, and finally a mould core disease distinguishing model is obtained, the distinguishing accuracy of the training set is 100%, the distinguishing accuracy of the training set is 92.3%, the training result is shown in figure 6, only 1 sample is wrongly judged, the training error is small, the distinguishing performance is good, and the model can be used as a mould core disease sorting line distinguishing model.

Claims (8)

1. The utility model provides a sorting unit for online nondestructive test apple mould heart disease equipment, including detecting conveyer (8), be provided with a plurality of apple trays (9) on the conveyer belt that detects conveyer (8), be provided with camera (1) and detect black case (2) on detecting conveyer (8), spectrum detection module (15) that set up in detecting black case (2) carry out spectral detection to the apple through detecting black case (2), a serial communication port, be provided with weighing platform (4) at the back level that detects conveyer (8), the apple after accomplishing the detection gets into weighing platform (4) and weighs, weighing platform (4) the back level is provided with inclined plane sorting conveyor (6), carry out the fruit according to testing result and weighing and size and select separately.

2. The sorting device for the on-line nondestructive testing of apple moldy core of the claim 1 is characterized in that the apple trays (9) are internally sunken and arranged in a single row along the conveying direction on the conveying belt, the camera (1) is arranged before the detection black box (2) to photograph and identify the passing apples, and the result is conveyed to the processor to judge the size of the passing apples.

3. The sorting device for the on-line nondestructive testing of apple moldy core equipment according to claim 1, characterized in that a cross sliding table (10) is arranged in the detection black box (2), two coaxial stepping sliding table motors are arranged in the vertical direction of the cross sliding table (10), one reverse lead screw motor is arranged in the horizontal direction, the coaxial stepping sliding table motors drive the reverse lead screw motors to move in the vertical direction, and the spectrum detection module (15) is fixed at two ends of the reverse lead screw motors through a support.

4. The sorting device for the on-line nondestructive testing equipment for the apple moldy core as claimed in claim 3, wherein the spectrum detection module (15) comprises a light source and a receiving device, the light source adopts an LED with the central wavelength of 710nm and the half-wave width of 20nm, the working voltage is 3.4V, and the scattering angle of the light source is 120 degrees; the receiving device adopts an avalanche diode as a sensitive device, adopts a filter amplifying circuit as a signal processing circuit, and converts the transmission light intensity into an electric signal through a sampling module to obtain the transmission intensity.

5. The sorting device for the on-line nondestructive testing equipment for the apple moldy core as claimed in claim 1, wherein a front shifting piece (3) and a rear shifting piece (5) are arranged on the weighing platform (4), the apples enter the weighing platform (4) from the detection conveyor (8) under the action of the front shifting piece (3), and enter the inclined sorting conveyor (6) from the weighing platform (4) under the action of the rear shifting piece (5).

6. The sorting device for the on-line nondestructive testing apple moldy core equipment of claim 1, wherein a plurality of grids are arranged on the inclined sorting conveyor (6), a first outlet (12), a second outlet (13) and a third outlet (14) are sequentially arranged below the inclined plane along the conveying direction, and each outlet is provided with a movable baffle (7).

7. The sorting device for the on-line nondestructive testing apple moldy core device of claim 6, wherein the inlet diameter of the second outlet (13) is 70mm, and the inlet diameter of the third outlet (14) is more than 70 mm.

8. The sorting device for the on-line nondestructive testing apple moldy core equipment according to claim 6, wherein the camera (1), the weighing platform (4) and the spectrum detection module (15) are all connected with the processor, the control end of the processor is connected with each movable baffle (7), the detection result of the spectrum detection module (15) is input into the processor, if the apple moldy core is detected, the processor outputs a control signal to open the movable baffle (7) of the first outlet (12) so that the apple moldy core enters the first outlet (12), the identification result of the camera (1) is input into the processor, if the ruler diameter of the camera is smaller than 70mm, the processor outputs a control signal to open the movable baffle (7) of the second outlet (13) so that the apple moldy core enters the second outlet (13), and if the ruler diameter of the apple moldy core is larger than 70mm, the processor outputs a control signal to open the movable baffle (7) of the third outlet (14), so that the moldy core fruits enter the third outlet (14).