CN105424795A - Egg quality nondestructive detection device and method based on magnetostriction technology - Google Patents

Abstract

The invention discloses an egg quality nondestructive detection device and method based on a magnetostriction technology, and relates to the technical field of egg detection. The device comprises a hollow shell, a magnetostrictive actuator, an audio power amplification circuit, a rubber pad, a microphone, a computer and an egg detection software system. According to the device, a computer software programming control sound card generates swept frequency excitation on the magnetostrictive actuator so that eggs placed on the magnetostrictive actuator can generate vibration audio signals, the microphone receives the vibration audio signals of the eggs and transmits the vibration audio signals into the computer in real time, and computer software performs egg quality analysis according to the vibration audio signals. According to the method, egg quality information in the audio signals is enhanced through a swept frequency vibration method, and in combination with a signal processing and mode recognition method, nondestructive detection of egg quality is achieved. The egg quality nondestructive detection device and method based on the magnetostriction technology can be applied to nondestructive detection of egg quality and particularly can achieve high-precision detection of eggs with fine cracks.

Description

A kind of quality of poultry eggs the cannot-harm-detection device based on magnetostrictive technology and method thereof

Technical field

The present invention relates to a kind of agriculture Quality Detection Apparatus and method for, especially based on magnetostrictive quality of poultry eggs the cannot-harm-detection device and method thereof, belong to quality of poultry eggs detection technique field.

Background technology

Birds, beasts and eggs are a kind of nutritious food easily digested again, and its sales revenue occupies certain status in national economy.It is very important that eggshell crack in birds, beasts and eggs detects in sorting and packaging link.Existing eggshell lossless detection method has digital image understanding method and sound identification method, and these two kinds of methods all instead of the method that traditional artificial naked eyes are selected.

Wherein, digital image understanding method aspect, patent " quality of poultry eggs automatic grouping system Apparatus for () and method therefor " (CN102179374A) discloses a kind of detection realizing birds, beasts and eggs inside and external sort by a series of dynamic image pro cess and pattern-recognition, patent " egg quality the cannot-harm-detection device and detection method based on DSP machine vision " (CN102445456A) discloses a kind of Image Acquisition based on DSP and process detects egg cracks, shape, the apparatus and method of freshness, patent " incubated egg quality online automatic detection screening installation and method thereof " (CN102217559A) discloses the Non-Destructive Testing realizing whole dish incubated egg quality by a series of images process and pattern-recognition.Although above patent can realize the Non-Destructive Testing of birds, beasts and eggs, be difficult to identify the fresh crack on bird egg and micro-cracks.

Wherein, sound identification method aspect, Mei Jinghua (2011) describe in its Master's thesis " dynamic birds, beasts and eggs knock the research of sound-producing device and eggshell crackle Acoustic detection automatically " a kind ofly the bird egg surface on supply line to be knocked automatically, the analysis of audible signal synchronous acquisition and crack identification the system rejected, Sun Li (2013) in its PhD dissertation " quality of poultry eggs on-line intelligenceization detects key technology research " to knocking Fast nondestructive evaluation crackle birds, beasts and eggs by single.But this quality of poultry eggs Dynamic Non-Destruction Measurement involved by two sections of articles and method are for the more difficult detection of birds, beasts and eggs micro-cracks.

So, the above-mentioned quality of poultry eggs detection method utilizing digital image understanding method and sound identification method, all more difficult detection birds, beasts and eggs micro-cracks.

Summary of the invention

The present invention aims to provide a kind of quality of poultry eggs the cannot-harm-detection device based on magnetostrictive technology and method thereof, by magnetostriction principle, information enhancement is carried out to birds, beasts and eggs vibration signal, overcome the problem of prior art to the more difficult detection of micro-cracks in quality of poultry eggs detection.

For achieving the above object, the present invention is by the following technical solutions: a kind of quality of poultry eggs the cannot-harm-detection device based on magnetostrictive technology, it comprises a hollow housing, a magnetic deformation actuator, an audio power amplifier circuit, a rubber blanket, a microphone, a computer and birds, beasts and eggs inspection software system, and described birds, beasts and eggs inspection software system comprises swept-frequency signal generation module, signal acquisition module, signal processing module, signal characteristic abstraction module, pattern recognition module and display module, wherein, a magnetic deformation actuator is provided with in hollow housing inner upper, a rubber blanket is provided with above magnetic deformation actuator, birds, beasts and eggs to be detected are placed above rubber blanket, below the magnetic deformation actuator of hollow housing inside, be provided with audio frequency power amplifier module, be provided with a microphone on hollow housing side, microphones head is placed in above hollow housing, be provided with a computer on microphone side, birds, beasts and eggs inspection software system is installed in computer, magnetic deformation actuator is connected with audio power amplifier circuit by electric wire, audio power amplifier circuit is connected with the audio output port of computer sound card by electric wire, audio power amplifier circuit is connected with computer USB mouth by electric wire, and microphone is connected with the audio input port of computer sound card by electric wire, computer USB mouth provides power supply for audio power amplifier circuit, swept-frequency signal generation module in the birds, beasts and eggs inspection software system of installing in computer produces sound signal and is input to audio power amplifier circuit by the audio output port of computer sound card, audio power amplifier circuit drives magnetic deformation actuator vibration, thus the vibration of excitation egg produces sound signal, this sound signal is input to the audio input port of computer sound card after being detected by microphone, the signal acquisition module of birds, beasts and eggs inspection software system gathers the audio input signal of computer sound card audio input port, and pass through the signal processing module of birds, beasts and eggs inspection software system, signal characteristic abstraction module, deliver to display module after pattern recognition module processes to show.

A kind of quality of poultry eggs lossless detection method based on magnetostrictive technology: birds, beasts and eggs are positioned on rubber blanket, swept-frequency signal generation module passes through audio output port and the vibration of audio power amplifier circuit driving magnetic deformation actuator of computer sound card, the vibration of excitation birds, beasts and eggs produces sound signal, the audio input port that this sound signal is input to sound card is gathered by microphone, signal acquisition module gathers the audio input signal of computer sound card audio input port, and be transferred to signal processing module successively, signal characteristic abstraction module, pattern recognition module processes, the result of pattern recognition module is sent to display module and shows.

The course of work of swept-frequency signal generation module of the present invention is: by software control computer sound card in the frequency range of 10Hz to 10000Hz, is gradient, produces sinusoidal wave swept-frequency signal with 10Hz, and the signal time length of each frequency is 0.5 second;

The course of work of signal processing module of the present invention is:

Step is 1.: by wavedec in matlab software (x, 3, ' db10 ') wavelet function saves as array S after carrying out filtering to the birds, beasts and eggs vibration sound signal collected;

Step is 2.: choose 4 rank AR models by the pburg function in matlab software and carry out the process of Burg power spectrum, obtains the vibration audio signal data after processing and saves as array Si;

The course of work of signal characteristic abstraction module of the present invention is:

Setting numerical value i is different quality birds, beasts and eggs testing time to be measured;

Step a: find out this i front 4 the major component frequencies of different quality birds, beasts and eggs Si data by princomp (Si) function in matlab software;

Step b: according to these 4 major component frequencies, five eigenwert X1, X2, X3, X4, X5 is extracted from i array Si, wherein, described X1, X2, X3, X4, X5 are respectively the ratio of maximum main peak peak value, the frequency corresponding to maximum main peak peak value, half bandwidth length of maximum main peak peak atenuation 10db, the average amplitude amplitude of 5000-8000Hz, maximum main peak peak value and second largest main peak peak value, and computing formula is as follows:

X1＝Max(Si)

X2＝f _max(Si)

X3＝f _max(Si)-f _0.1*max(Si)

$X4=\frac{{\mathrm{\Σ}}_{5000}^{8000}\left(\mathrm{Si}\right)}{3000}$

$X5=\frac{{\mathrm{Si}}_{f_{\max }}}{{\mathrm{Si}}_{f_{\sec \mathrm{ond}}}};$

The course of work of pattern recognition module of the present invention is:

I group X1, X2, X3, X4, X5 are input to " quality of poultry eggs forecast model ", wherein " quality of poultry eggs forecast model " adopts the support vector machine (SupportVectorMachine of multiple input single output, SVM) disaggregated model, the dynamic state of parameters adjustment process of SVM is as follows:

Step I: wait new X1, X2, X3, X4, X5 to be entered;

Step II: adopt the mode of cross validation to train the parameter of SVM;

Step II I: the output data being calculated this group X1, X2, X3, X4, X5 by SVM, namely this organizes birds, beasts and eggs to be measured prediction quality;

Wherein, the SVM parameter training in Step II only need train penalty factor c and kernel functional parameter g, and training method is according to following steps:

I () setting c=0.1, g=0.1 are initial value;

(ii) in learning sample, remove one, two sample, with remaining sample training SVM, with this, two samples test;

(iii) Error Absolute Value of test sample book is calculated with the network model built, i.e. predicated error;

(iiii) repeat step (ii), (iii), until all training samples are all once for test, try to achieve the mean value of predicated error and it can be used as the objective function E of optimizing.

The course of work of display module of the present invention is: be shown as damaged birds, beasts and eggs when pattern recognition module output is 1, when pattern recognition module output is 0, is shown as harmless birds, beasts and eggs.

Compared with prior art, beneficial effect of the present invention is as follows:

Existing quality of poultry eggs detection technique, as by digital image understanding method and sound identification method etc., although can carry out Non-Destructive Testing to quality of poultry eggs, for all more difficult detection of birds, beasts and eggs micro-cracks.

The present invention carries out information enhancement by magnetostriction principle to birds, beasts and eggs vibration signal, by signal processing module, birds, beasts and eggs vibration sound signal is processed, different quality birds, beasts and eggs feature is extracted by signal characteristic abstraction module, the prediction to quality of poultry eggs to be measured is realized by pattern recognition module, the high precision Non-Destructive Testing to quality of poultry eggs can be realized, and effectively can detect the birds, beasts and eggs of micro-cracks.

Accompanying drawing explanation

Fig. 1 is the quality of poultry eggs the cannot-harm-detection device structural representation based on magnetostrictive technology;

Fig. 2 is birds, beasts and eggs inspection software internal system model calling block diagram;

In Fig. 1,1. hollow housing; 2. needle expansion actuator; 3. audio power amplifier circuit; 4. rubber blanket; 5. microphone; 6. computer; 7. the audio output port of computer sound card; 8. the audio input port of computer sound card; 9. computer USB mouth; 10. birds, beasts and eggs inspection software system; 11. birds, beasts and eggs to be measured.

Embodiment

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

Embodiment 1

As shown in Figure 1, a kind of quality of poultry eggs the cannot-harm-detection device based on magnetostrictive technology:

It comprises hollow housing 1, magnetic deformation actuator 2, audio power amplifier circuit 3, rubber blanket 4, microphone 5, computer 6 and a birds, beasts and eggs inspection software system 10, and described birds, beasts and eggs inspection software system 10 comprises swept-frequency signal generation module, signal acquisition module, signal processing module, signal characteristic abstraction module, pattern recognition module and display module, wherein, a magnetic deformation actuator 2 is provided with in hollow housing 1 inner upper, a rubber blanket 4 is provided with above magnetic deformation actuator 2, birds, beasts and eggs 11 to be detected are placed above rubber blanket 4, below the magnetic deformation actuator 2 of hollow housing 1 inside, be provided with audio power amplifier circuit 3, be provided with a microphone 5 on hollow housing 1 side, microphone 5 Receiver is placed in above hollow housing 1, be provided with a computer 6 on microphone 5 side, birds, beasts and eggs inspection software system 10 is installed in computer 6, magnetic deformation actuator 2 is connected with audio power amplifier circuit 3 by electric wire, audio power amplifier circuit 3 is connected with the audio output port 7 of computer sound card by electric wire, audio power amplifier circuit 3 is connected with computer USB mouth 9 by electric wire, and microphone is connected with the audio input port 8 of computer sound card by electric wire, computer USB mouth 9 provides power supply for audio power amplifier circuit 3, swept-frequency signal generation module in the birds, beasts and eggs inspection software system 10 of installing in computer 6 produces sound signal and is input to audio power amplifier circuit 3 by the audio output port 8 of computer sound card, audio power amplifier circuit 3 drives magnetic deformation actuator 2 to vibrate, thus excitation birds, beasts and eggs 11 vibrate generation sound signal, this sound signal is input to the audio input port 8 of computer sound card after being detected by microphone 5, the signal acquisition module of birds, beasts and eggs inspection software system 10 gathers the audio input signal of computer sound card audio input port 8, and pass through the signal processing module of birds, beasts and eggs inspection software system 10, signal characteristic abstraction module, deliver to display module after pattern recognition module processes to show.

Embodiment 2

As shown in Figure 2, birds, beasts and eggs inspection software system in the present invention, it comprises swept-frequency signal generation module, signal acquisition module, signal processing module, signal characteristic abstraction module, pattern recognition module and display module.

Embodiment 3

A kind of quality of poultry eggs lossless detection method based on magnetostrictive technology: birds, beasts and eggs are positioned on rubber blanket, swept-frequency signal generation module passes through audio output port and the vibration of audio power amplifier circuit driving magnetic deformation actuator of computer sound card, the vibration of excitation birds, beasts and eggs produces sound signal, the audio input port that this sound signal is input to sound card is gathered by microphone, signal acquisition module gathers the audio input signal of computer sound card audio input port, and be transferred to signal processing module successively, signal characteristic abstraction module, pattern recognition module processes, the result of pattern recognition module is sent to display module and shows.

Embodiment 4

In the present invention, the course of work of swept-frequency signal generation module is: by software control computer sound card in the frequency range of 10Hz to 10000Hz, is gradient, produces sinusoidal wave swept-frequency signal with 10Hz, and the signal time length of each frequency is 0.5 second;

Embodiment 5

In the present invention, the course of work of signal processing module is:

Step is 1.: by wavedec in matlab software (x, 3, ' db10 ') wavelet function saves as array S after carrying out filtering to the birds, beasts and eggs vibration sound signal collected;

Step is 2.: choose 4 rank AR models by the pburg function in matlab software and carry out the process of Burg power spectrum, obtains the vibration audio signal data after processing and saves as array Si;

Embodiment 6

In the present invention, the course of work of signal characteristic abstraction module is:

Setting numerical value i is different quality birds, beasts and eggs testing time to be measured;

Step a: find out this i front 4 the major component frequencies of different quality birds, beasts and eggs Si data by princomp (Si) function in matlab software;

Step b: according to these 4 major component frequencies, five eigenwert X1, X2, X3, X4, X5 is extracted from i array Si, wherein, described X1, X2, X3, X4, X5 are respectively the ratio of maximum main peak peak value, the frequency corresponding to maximum main peak peak value, half bandwidth length of maximum main peak peak atenuation 10db, the average amplitude amplitude of 5000-8000Hz, maximum main peak peak value and second largest main peak peak value, and computing formula is as follows:

X1＝Max(Si)

X2＝f _max(Si)

X3＝f _max(Si)-f _0.1*max(Si)

$X4=\frac{{\mathrm{\Σ}}_{5000}^{8000}\left(\mathrm{Si}\right)}{3000}$

$X5=\frac{{\mathrm{Si}}_{f_{\max }}}{{\mathrm{Si}}_{f_{\sec \mathrm{ond}}}}.$

Embodiment 7

In the present invention, the course of work of pattern recognition module is:

I group X1, X2, X3, X4, X5 are input to " quality of poultry eggs forecast model ", wherein " quality of poultry eggs forecast model " adopts the support vector machine (SupportVectorMachine of multiple input single output, SVM) disaggregated model, the dynamic state of parameters adjustment process of SVM is as follows:

Step I: wait new X1, X2, X3, X4, X5 to be entered;

Step II: adopt the mode of cross validation to train the parameter of SVM;

Step II I: the output data being calculated this group X1, X2, X3, X4, X5 by SVM, namely this organizes birds, beasts and eggs to be measured prediction quality;

Wherein, the SVM parameter training in Step II only need train penalty factor c and kernel functional parameter g, and training method is according to following steps:

I () setting c=0.1, g=0.1 are initial value;

(ii) in learning sample, remove one, two sample, with remaining sample training SVM, with this, two samples test;

(iii) Error Absolute Value of test sample book is calculated with the network model built, i.e. predicated error;

(iiii) repeat step (ii), (iii), until all training samples are all once for test, try to achieve the mean value of predicated error and it can be used as the objective function E of optimizing.

Embodiment 8

In the present invention, the course of work of display module is: be shown as damaged birds, beasts and eggs when pattern recognition module output is 1, when pattern recognition module output is 0, is shown as harmless birds, beasts and eggs.

Claims (2)

1. quality of poultry eggs the cannot-harm-detection device based on magnetostrictive technology, it is characterized in that: comprise hollow housing 1, magnetic deformation actuator 2, audio power amplifier circuit 3, rubber blanket 4, microphone 5, computer 6 and a birds, beasts and eggs inspection software system 10, described birds, beasts and eggs inspection software system 10 comprises swept-frequency signal generation module, signal acquisition module, signal processing module, signal characteristic abstraction module, pattern recognition module and display module; Wherein, a magnetic deformation actuator 2 is provided with in hollow housing 1 inner upper, a rubber blanket 4 is provided with above magnetic deformation actuator 2, birds, beasts and eggs 11 to be detected are placed above rubber blanket 4, below the magnetic deformation actuator 2 of hollow housing 1 inside, be provided with audio power amplifier circuit 3, be provided with a microphone 5 on hollow housing 1 side, microphone 5 Receiver is placed in above hollow housing 1, be provided with a computer 6 on microphone 5 side, birds, beasts and eggs inspection software system 10 is installed in computer 6; Magnetic deformation actuator 2 is connected with audio power amplifier circuit 3 by electric wire, audio power amplifier circuit 3 is connected with the audio output port 7 of computer sound card by electric wire, audio power amplifier circuit 3 is connected with computer USB mouth 9 by electric wire, microphone is connected with the audio input port 8 of computer sound card by electric wire, and computer USB mouth 9 provides power supply for audio power amplifier circuit 3.

2. the quality of poultry eggs lossless detection method based on magnetostrictive technology, it is characterized in that: birds, beasts and eggs are positioned on rubber blanket, swept-frequency signal generation module sequentially passes through audio output port and the vibration of audio power amplifier circuit driving magnetic deformation actuator of computer sound card, the vibration of excitation birds, beasts and eggs produces sound signal, microphone gathers the audio input port that this sound signal is input to sound card, signal acquisition module gathers the audio input signal of computer sound card audio input port, and be transferred to signal processing module successively, signal characteristic abstraction module, pattern recognition module processes, the result of pattern recognition module is sent to display module and shows,

2.1 swept-frequency signal generation modules according to claim 2, it is characterized by: by software control computer sound card in the frequency range of 10Hz to 10000Hz, be gradient with 10Hz, produce sinusoidal wave swept-frequency signal, and the signal time length of each frequency be 0.5 second;

2.2 signal processing modules according to claim 2, is characterized by according to following steps work:

Step is 1.: by wavedec in matlab software (x, 3, ' db10 ') wavelet function saves as array S after carrying out filtering to the birds, beasts and eggs vibration sound signal collected;

Step is 2.: choose 4 rank AR models by the pburg function in matlab software and carry out the process of Burg power spectrum, obtains the vibration audio signal data after processing and saves as array Si;

2.3 signal characteristic abstraction modules according to claim 2, is characterized by according to following steps work:

Setting numerical value i is different quality birds, beasts and eggs testing time to be measured;

Step a: find out this i front 4 the major component frequencies of different quality birds, beasts and eggs Si data by princomp (Si) function in matlab software;

Step b: according to these 4 major component frequencies, five eigenwert X1, X2, X3, X4, X5 is extracted from i array Si, wherein, described X1, X2, X3, X4, X5 are respectively the ratio of maximum main peak peak value, the frequency corresponding to maximum main peak peak value, half bandwidth length of maximum main peak peak atenuation 10db, the average amplitude amplitude of 5000-8000Hz, maximum main peak peak value and second largest main peak peak value, and computing formula is as follows:

X1＝Max(Si)

X2＝f _max(Si)

X3＝f _max(Si)-f _0.1*max(Si)

$X4=\frac{{\mathrm{\Σ}}_{5000}^{8000}\left(\mathrm{Si}\right)}{3000}$

$X5=\frac{{\mathrm{Si}}_{f_{\max }}}{{\mathrm{Si}}_{f_{\sec \mathrm{ond}}}}$

2.4 pattern recognition modules according to claim 2, is characterized by according to following steps work:

I group X1, X2, X3, X4, X5 are input to " quality of poultry eggs forecast model ", wherein " quality of poultry eggs forecast model " adopts the support vector machine (SupportVectorMachine of multiple input single output, SVM) disaggregated model, the dynamic state of parameters adjustment process of SVM is as follows:

Step I: wait new X1, X2, X3, X4, X5 to be entered;

Step II: adopt the mode of cross validation to train the parameter of SVM;

Step II I: the output data being calculated this group X1, X2, X3, X4, X5 by SVM, namely this organizes birds, beasts and eggs to be measured prediction quality;

Wherein, the SVM parameter training in Step II only need train penalty factor c and kernel functional parameter g, and training method is according to following steps:

I () setting c=0.1, g=0.1 are initial value;

(ii) in learning sample, remove 1 sample, with remaining sample training SVM, test with this 1 sample;

(iii) Error Absolute Value of test sample book is calculated with the network model built, i.e. predicated error;

(iiii) repeat step (ii), (iii), until all training samples are all once for test, try to achieve the mean value of predicated error and it can be used as the objective function E of optimizing.

2.5 display modules according to claim 2, is characterized by: be shown as damaged birds, beasts and eggs when pattern recognition module output is 1, when pattern recognition module output is 0, are shown as harmless birds, beasts and eggs.