CN105891327A - Plant water shortage detection device based on vibration information and method thereof - Google Patents

Abstract

The invention discloses a plant water shortage detection device based on vibration information and a method thereof, which relate to the technical field of plant water shortage detection. The device comprises a distance measurement module, a singlechip microcomputer module, a relay, an electromagnetic valve, an air pump, an air inlet hose, an air outlet hose, a fixing bracket, a fixing rod, a computer and a plant detection software system. According to the device, the singlechip microcomputer module drives the relay to be turned on by changing high and low levels, the electromagnetic valve is controlled to be switched on, compressed air in the air pump blows plant leaves through the air inlet hose and the air outlet hose and enables the plant leaves to vibrate, the distance measurement module acquires a vibration signal of the plant leaves, and the vibration signal is subjected to AD conversion through the singlechip microcomputer module, transmitted to the computer and subjected to power spectrum processing and mode identification to realize the water shortage detection of plants; the device can be used for high-precision nondestructive detection of the plants in a water shortage detection process.

Description

A kind of plant hydropenia based on vibration information detection devices and methods therefor

Technical field

The present invention relates to a kind of Agricultural Moisture detection Apparatus and method for, be based especially on vibration information plant hydropenia detection device and Its method, belongs to plant hydropenia detection technique field.

Background technology

Shortage of water resources is the key factor affecting China's agricultural development.In process of Industrial Agriculture production, for realize high yield, High-quality, it is necessary to improve the diagnosis to damage caused by a drought and control technique.Whether moisture is most important for plant growing, and lack for plant Water passes through the more difficult observation of naked eyes.Existing plant hydropenia diagnostic method has form detection method, soil moisture content detection method, the signal of telecommunication to examine Survey method etc..

Wherein, Morphology observation method aspect, patent " vegetation growth state based on computer vision and Internet of Things monitoring system " (CN102564593A) a kind of vegetation growth state based on computer vision and Internet of Things monitoring system, P.Foucher are disclosed Deng (2004) at paper " Morphological image analysis for the detection of water stress in Potted forsythia. " in have studied the whole Form Development of plant under Water Stress Conditions is described under artificial vision identifies. Form monitoring method involved by both the above method or system is strictly limited to the condition such as floristics etc. fixed.

Wherein, electrical signal detection method aspect, Bao Yidan etc. (2005) at paper " based on vane electrical characteristics and the plant of leaf water potential Hydropenia degree study " in have studied the Changing Pattern between plant hydropenia information and plant leaf blade electrical characteristics and leaf water potential.But the method The magnitude of voltage measured easily is affected by the water of irrigating plant or culture medium intermediate ion amount and inaccurate.

Wherein, soil moisture content detection method aspect, patent " a kind of plant water-depletion detection water applicator " (CN201174917Y) is open A kind of plant water-depletion detection water applicator of testing circuit based on resistance wire change in resistance.But this detecting system is more difficult in real time Detect plant hydropenia situation.

So, the above-mentioned plant hydropenia detection method utilizing Morphology observation, electrical signal detection, soil moisture content to detect is carrying out plant During hydropenia detection, there is the deficiency that floristics limits, accuracy rate is relatively low.

Summary of the invention

It is desirable to provide a kind of plant hydropenia based on vibration information detection devices and methods therefor, impact leaves of plants by gas The mode of sheet carries out plant hydropenia detection, overcome in prior art to plant hydropenia detect in floristics limit, accuracy rate relatively Low problem.

For achieving the above object, the present invention is by the following technical solutions: a range finder module, an one-chip computer module, a relay, One electromagnetic valve, an air pump, an air induction hose, an air-out hose, a fixed support, one fix bar, a computer and plant inspection Surveying software system, described plnat monitoring software system includes signal acquisition module, signal processing module, signal characteristic abstraction mould Block, pattern recognition module and display module；Wherein, fixing bar is fixed on fixed support, and range finder module is fixed on fixing bar On, range finder module is positioned at above plant leaf blade, and range finder module is connected with one-chip computer module by electric wire, and one-chip computer module is by number Being connected with computer USB mouth according to line, one-chip computer module is connected with relay incoming end by electric wire, and relay output end passes through electric wire Being connected with electromagnetic valve, air induction hose one end is connected to electromagnetic valve inlet end, and the other end is connected on air pump, and air-out hose one end is solid Being scheduled on fixed support, the other end is connected to electromagnetic valve outlet side, and computer USB mouth provides power supply and monolithic for one-chip computer module Machine module carries out data transmission, and one-chip computer module drives relay to open by changing low and high level, controls solenoid valve conduction, Making the compressed air in air pump be blown by air induction hose and air-out hose hit plant leaf blade and make it vibrate, range finder module gathers this Plant leaf blade vibration signal is input to one-chip computer module, and one-chip computer module is transferred to computer USB mouth by after this signal AD conversion, letter Number acquisition module gathers the plant leaf blade vibration signal of computer USB mouth, and by the signal processing module of plnat monitoring software system, Signal characteristic abstraction module, pattern recognition module process, and pattern recognition module is delivered to display module after processing and shown.

A kind of plant hydropenia detection method based on vibration information: be positioned over by plant leaf blade below range finder module, electricity connected by air pump Source works, and one-chip computer module drives relay to open, and controls solenoid valve conduction 0.5 second, and air pump compressed air was at this 0.5 second In be input to electromagnetic valve by air induction hose and exported by air-out hose and blow and hit plant leaf blade surface, range finder module gathers this and plants simultaneously Thing blade vibration signal is input to one-chip computer module, and one-chip computer module is transferred to computer USB mouth, signal after this signal AD conversion Acquisition module gathers the plant leaf blade vibration signal of computer USB mouth, and is transferred to signal processing module, signal characteristic abstraction according to this Module, pattern recognition module process, and the result of pattern recognition module is sent to display module and shows.

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

Step is 1.: be filtered, by the wavedec wavelet function in matlab software, the plant leaf blade vibration signal collected After save as array S；

Step is 2.: carry out segmentation overlay counting by the pwelch function in matlab software is 50, and it is 160 that FFT counts Welch power spectrum processes, the plant leaf blade vibration signal data after being processed；

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

Step 1: plant leaf blade when coercing different water carries out repeatedly vibration signals collecting, sets numerical value n as global vibration signal Times of collection；

Step 2: plant leaf blade vibration signal data save as array Si after signal processing module processes；

Step 3: found out front 2 main constituent frequencies of array Si by the princomp function in matlab software；

Step 4: according to these 2 main constituent frequencies, extracts three eigenvalues X1, X2, X3 from array Si, wherein, described X1, X2, X3 are respectively maximum main peak peak value, the maximum frequency corresponding to main peak peak value, the meansigma methods of three maximum main peak peak values, Computing formula is as follows:

X1=Max (Si)

X2=f_max(Si)

$X3=\frac{\underset{n=1}{\overset{3}{\mathrm{\Σ}}}{\mathrm{Si}}_{\max }^n}{3};$

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

N group X1, X2, X3 being input to " plant hydropenia detection model ", wherein how defeated " plant hydropenia detection model " use Entering radial basis function neural network (RBFNN) disaggregated model of single output, the dynamic state of parameters of RBFNN net adjusts process according to following Step calculates:

Set numerical value k=0, set three-dimensional array A and three-dimensional array B；

Step a: wait new X1, X2, X3 to be entered；

Step b: if k≤20, then A [k]=X1, X2, X3, k=k+1, proceed to step a；K ＞ 20 then proceeds to else if Step c；

Step c:B [20]=X1, X2, X3, sets numerical value m=1, B [20-m]=A [k-m]；

Step d: if m ＜ 21, then proceed to step c, otherwise proceed to step e；

Step e: by array B as input different moisture content plant characteristics sequence, the parameter of training RBFNN net；

Step f: calculate the output data that different moisture content plant is corresponding, i.e. plant whether hydropenia predictive value by RBFNN net；

Step g:k=k+1, is transferred to step a；

Wherein, the RBFNN network parameters training in step e only need to train smoothing factor, and training method is according to following steps:

I () makes smoothing factor with increment incremental variations in certain scope；

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

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

(iiii) repeat step (ii), (iii), until all of training sample is the most once used for testing, try to achieve forecast error Meansigma methods and as the object function E of optimizing；

The work process of display module of the present invention is: be shown as plant hydropenia state when pattern recognition module is output as 1, When pattern recognition module is output as 0, it is shown as plant moisture abundance state.

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

Existing plant hydropenia detection method, as by Morphology observation method, soil moisture content detection method and electrical signal detection method etc., though So plant moisture can be carried out Non-Destructive Testing, but there is the deficiency that floristics limits, accuracy rate is relatively low.

The present invention carries out plant hydropenia detection by the way of gas impact plant leaf blade, drives relay by one-chip computer module Open thus control solenoid valve conduction, by signal processing module, plant leaf blade vibration signal is processed, pass through signal characteristic Extraction module extracts plant leaf blade feature when different water is coerced, and realizes treating measuring plants whether hydropenia by pattern recognition module Prediction, can realize the high accuracy Non-Destructive Testing to plant whether hydropenia.

Accompanying drawing explanation

Fig. 1 is plant hydropenia structure of the detecting device schematic diagram based on vibration information；

Fig. 2 is that plnat monitoring software system internal module connects block diagram；

In Fig. 1,1. range finder module；2. one-chip computer module；3. relay；4. electromagnetic valve；5. air pump；6. air induction hose；7. Air-out hose；8. fixed support；9, fixing bar；10. computer；11. plnat monitoring software systems；12. computer USB mouths；13. Plant leaf blade.

Detailed description of the invention

Below in conjunction with the accompanying drawings and specific embodiment the invention will be further described.

Embodiment 1

As it is shown in figure 1, a kind of plant hydropenia based on vibration information detection device:

It includes that range finder module 1, one-chip computer module 2, relay 3, electromagnetic valve 4, air pump 5, air inlet is soft Pipe 6, one air-out hose 7, fixed support 8, one fixes bar 9, computer 10 and plnat monitoring software system 11, institute The plnat monitoring software system 11 stated includes signal acquisition module, signal processing module, signal characteristic abstraction module, pattern recognition Module and display module；Wherein, fixing bar 9 is fixed on fixed support 8, and range finder module 1 is fixed on fixing bar 9, Range finder module 1 is positioned at above plant leaf blade 13, and range finder module 1 is connected with one-chip computer module 2 by electric wire, one-chip computer module 2 Being connected with computer USB mouth 12 by data wire, one-chip computer module 2 is connected with relay 3 incoming end by electric wire, relay 3 Outfan is connected with electromagnetic valve 4 by electric wire, and air induction hose 6 one end is connected to electromagnetic valve 4 inlet end, and the other end is connected to gas On pump 5, air-out hose 7 one end is fixed on fixed support 8, and the other end is connected to electromagnetic valve 4 outlet side, computer USB mouth 12 provide power supply for one-chip computer module 2 and carry out data transmission with one-chip computer module 2, and one-chip computer module 2 is by changing height electricity Put down and drive relay 3 to open, control electromagnetic valve 4 and turn on, make the compressed air in air pump 5 by air induction hose 6 and give vent to anger Flexible pipe 7 blows and hits plant leaf blade 13 and make it vibrate, and range finder module 1 gathers this plant leaf blade vibration signal and is input to single-chip microcomputer mould Block 2, one-chip computer module 2 will be transferred to computer USB mouth 12 after this signal AD conversion, signal acquisition module gathers computer USB mouth The plant leaf blade vibration signal of 12, and by the signal processing module of plnat monitoring software system 11, signal characteristic abstraction module, Pattern recognition module processes, and pattern recognition module is delivered to display module after processing and shown.

Embodiment 2

As in figure 2 it is shown, plnat monitoring software system in the present invention, it includes signal acquisition module, signal processing module, signal characteristic Extraction module, pattern recognition module and display module.

Embodiment 3

A kind of plant hydropenia detection method based on vibration information: be positioned over by plant leaf blade below range finder module, air pump switches on power work Making, one-chip computer module drives relay to open, and controls solenoid valve conduction 0.5 second, and air pump compressed air is logical in this is 0.5 second Crossing air induction hose be input to electromagnetic valve and blown hit plant leaf blade surface by air-out hose output, range finder module gathers this leaves of plants simultaneously Sheet vibration signal is input to one-chip computer module, and one-chip computer module is transferred to computer USB mouth, signals collecting after this signal AD conversion Module gather computer USB mouth plant leaf blade vibration signal, and be transferred to according to this signal processing module, signal characteristic abstraction module, Pattern recognition module processes, and the result of pattern recognition module is sent to display module and shows.

Embodiment 4

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

Step is 1.: be filtered, by the wavedec wavelet function in matlab software, the plant leaf blade vibration signal collected After save as array S；

Step is 2.: carry out segmentation overlay counting by the pwelch function in matlab software is 50, and it is 160 that FFT counts Welch power spectrum processes, the plant leaf blade vibration signal data after being processed.

Embodiment 5

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

Step 1: plant leaf blade when coercing different water carries out repeatedly vibration signals collecting, sets numerical value n as global vibration signal Times of collection；

Step 2: plant leaf blade vibration signal data save as array Si after signal processing module processes；

Step 3: found out front 2 main constituent frequencies of array Si by the princomp function in matlab software；

Step 4: according to these 2 main constituent frequencies, extracts three eigenvalues X1, X2, X3 from array Si, wherein, described X1, X2, X3 are respectively maximum main peak peak value, the maximum frequency corresponding to main peak peak value, the meansigma methods of three maximum main peak peak values, Computing formula is as follows:

X1=Max (Si)

X2=f_max(Si)

$X3=\frac{\underset{n=1}{\overset{3}{\mathrm{\Σ}}}{\mathrm{Si}}_{\max }^n}{3}.$

Embodiment 6

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

N group X1, X2, X3 being input to " plant hydropenia detection model ", wherein how defeated " plant hydropenia detection model " use Entering radial basis function neural network (RBFNN) disaggregated model of single output, the dynamic state of parameters of RBFNN net adjusts process according to following Step calculates:

Set numerical value k=0, set three-dimensional array A and three-dimensional array B；

Step a: wait new X1, X2, X3 to be entered；

Step b: if k≤20, then A [k]=X1, X2, X3, k=k+1, proceed to step a；K ＞ 20 then proceeds to else if Step c；

Step c:B [20]=X1, X2, X3, sets numerical value m=1, B [20-m]=A [k-m]；

Step d: if m ＜ 21, then proceed to step c, otherwise proceed to step e；

Step e: by array B as input different moisture content plant characteristics sequence, the parameter of training RBFNN net；

Step f: calculate the output data that different moisture content plant is corresponding, i.e. plant whether hydropenia predictive value by RBFNN net；

Step g:k=k+1, is transferred to step a；

Wherein, the RBFNN network parameters training in step e only need to train smoothing factor, and training method is according to following steps:

I () makes smoothing factor with increment incremental variations in certain scope；

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

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

(iiii) repeat step (ii), (iii), until all of training sample is the most once used for testing, try to achieve forecast error Meansigma methods and as the object function E of optimizing.

Embodiment 7

In the present invention, the work process of display module is: is shown as plant hydropenia state when pattern recognition module is output as 1, works as pattern When identification module is output as 0, it is shown as plant moisture abundance state.

Claims (2)

1. plant hydropenia based on a vibration information detection device, it is characterised in that: include range finder module 1, one-chip computer module 2, relay 3, electromagnetic valve 4, air pump 5, air induction hose 6, air-out hose 7, fixed support 8, is solid Fixed pole 9, one computer 10 and plnat monitoring software system 11, described plnat monitoring software system 11 includes signals collecting mould Block, signal processing module, signal characteristic abstraction module, pattern recognition module and display module；Wherein, fixing bar 9 is fixed On fixed support 8, range finder module 1 is fixed on fixing bar 9, and range finder module 1 is positioned at above plant leaf blade 13, mould of finding range Block 1 is connected with one-chip computer module 2 by electric wire, and one-chip computer module 2 is connected with computer USB mouth 12 by data wire, single-chip microcomputer Module 2 is connected with relay 3 incoming end by electric wire, and relay 3 outfan is connected with electromagnetic valve 4 by electric wire, and air inlet is soft Pipe 6 one end is connected to electromagnetic valve 4 inlet end, and the other end is connected on air pump 5, and air-out hose 7 one end is fixed on fixed support On 8, the other end is connected to electromagnetic valve 4 outlet side, computer USB mouth 12 for one-chip computer module 2 provide power supply and with single-chip microcomputer mould Block 2 carries out data transmission.

2. a plant hydropenia detection method based on vibration information, it is characterised in that: plant leaf blade is positioned over below range finder module, Air pump switches on power work, and one-chip computer module drives relay to open, and controls solenoid valve conduction 0.5 second, air pump compressed air In this is 0.5 second, it is input to electromagnetic valve by air induction hose and is blown hit plant leaf blade surface by air-out hose output, mould of simultaneously finding range Block gathers this plant leaf blade vibration signal and is input to one-chip computer module, and one-chip computer module is transferred to computer after this signal AD conversion USB port, signal acquisition module gather computer USB mouth plant leaf blade vibration signal, and be transferred to according to this signal processing module, Signal characteristic abstraction module, pattern recognition module process, and the result of pattern recognition module is sent to display module and carries out Display；

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

Step is 1.: be filtered, by the wavedec wavelet function in matlab software, the plant leaf blade vibration signal collected After save as array S；

Step is 2.: carry out segmentation overlay counting by the pwelch function in matlab software is 50, and it is 160 that FFT counts Welch power spectrum processes, the plant leaf blade vibration signal data after being processed；

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

Step 1: plant leaf blade when coercing different water carries out repeatedly vibration signals collecting, sets numerical value n as global vibration signal Times of collection；

Step 2: plant leaf blade vibration signal data save as array Si after signal processing module processes；

Step 3: found out front 2 main constituent frequencies of array Si by the princomp function in matlab software；

Step 4: according to these 2 main constituent frequencies, extracts three eigenvalues X1, X2, X3 from array Si, wherein, described X1, X2, X3 are respectively maximum main peak peak value, the maximum frequency corresponding to main peak peak value, the meansigma methods of three maximum main peak peak values, Computing formula is as follows:

X1=Max (Si)

X2=f_max(Si)

$X3=\frac{\underset{n=1}{\overset{3}{\mathrm{\Σ}}}{\mathrm{Si}}_{\max }^n}{3};$

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

N group X1, X2, X3 being input to " plant hydropenia detection model ", wherein how defeated " plant hydropenia detection model " use Entering radial basis function neural network (RBFNN) disaggregated model of single output, the dynamic state of parameters of RBFNN net adjusts process according to following Step calculates:

Set numerical value k=0, set three-dimensional array A and three-dimensional array B；

Step a: wait new X1, X2, X3 to be entered；

Step b: if k≤20, then A [k]=X1, X2, X3, k=k+1, proceed to step a；K ＞ 20 then proceeds to else if Step c；

Step c:B [20]=X1, X2, X3, sets numerical value m=1, B [20-m]=A [k-m]；

Step d: if m ＜ 21, then proceed to step c, otherwise proceed to step e；

Step e: by array B as input different moisture content plant characteristics sequence, the parameter of training RBFNN net；

Step f: calculate the output data that different moisture content plant is corresponding, i.e. plant whether hydropenia predictive value by RBFNN net；

Step g:k=k+1, is transferred to step a；

Wherein, the RBFNN network parameters training in step e only need to train smoothing factor, and training method is according to following steps:

I () makes smoothing factor with increment incremental variations in certain scope；

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

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

(iiii) repeat step (ii), (iii), until all of training sample is the most once used for testing, try to achieve forecast error Meansigma methods and as the object function E of optimizing；

2.4 display modules according to claim 2, is characterized by: be shown as plant when pattern recognition module is output as 1 and lack Water state, when pattern recognition module is output as 0, is shown as plant moisture abundance state.