CN109379557A - Mango insect pest intelligent monitor system based on image recognition - Google Patents
Mango insect pest intelligent monitor system based on image recognition Download PDFInfo
- Publication number
- CN109379557A CN109379557A CN201811163513.9A CN201811163513A CN109379557A CN 109379557 A CN109379557 A CN 109379557A CN 201811163513 A CN201811163513 A CN 201811163513A CN 109379557 A CN109379557 A CN 109379557A
- Authority
- CN
- China
- Prior art keywords
- mango
- pest
- module
- camera
- insect pest
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N7/00—Television systems
- H04N7/18—Closed-circuit television [CCTV] systems, i.e. systems in which the video signal is not broadcast
- H04N7/181—Closed-circuit television [CCTV] systems, i.e. systems in which the video signal is not broadcast for receiving images from a plurality of remote sources
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2413—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on distances to training or reference patterns
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/0002—Inspection of images, e.g. flaw detection
- G06T7/0012—Biomedical image inspection
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30004—Biomedical image processing
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Medical Informatics (AREA)
- Multimedia (AREA)
- Quality & Reliability (AREA)
- Nuclear Medicine, Radiotherapy & Molecular Imaging (AREA)
- General Health & Medical Sciences (AREA)
- Health & Medical Sciences (AREA)
- Signal Processing (AREA)
- Radiology & Medical Imaging (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Evolutionary Computation (AREA)
- General Engineering & Computer Science (AREA)
- Catching Or Destruction (AREA)
Abstract
The invention discloses a kind of mango insect pest intelligent monitor system based on image recognition.The present invention obtains mango realtime graphic by the camera of setting, the screening of automatic batch is carried out to realtime graphic by optical sieving module, and by effective picture be stored as training sample be sent to handled on image analysis processor after be sent to insect pest prewarning unit, insect pest prewarning unit establishes neural network Pest model based on artificial neural network, and is learnt and trained, until deconditioning after Forecasting Pests convergence, then generate artificial neural network prediction of pest attack model.When in use, it is analysed and compared by effective image of the prediction model to input, the prediction result of pest and disease damage can be exported in real time.The present invention integrates the function of autocoding, automatic screening, automatic Prediction result, has forecasting accuracy height, the high beneficial features of self energy degree, and can be applicable in the early warning and prevention and treatment of various mango pest and disease damages.
Description
Technical field
The present invention relates to a kind of wisdom Agricultural Monitoring field, especially a kind of mango insect pest based on image recognition is intelligently supervised
Control system.
Background technique
The popular name (Classification system: Mangifera indicaL.) of mango Shi Mango (Chinese Plants will), mango is one
Kind originates in the evergreen megaphanerophyte of Anacardiaceae of India.Mango Fruit contains sugar, protein, crude fibre etc., and the dimension contained by mango
The precursor carrotene ingredient of raw element A is especially high, is rare in all fruit.Mature Mango Fruit is sweet and delicious, mouthfeel
It is good, and the products such as dried mango, Preserved Mango, mango preserved fruit can also be processed into, it is well received by consumers.
Mango happiness is warm, can not resist cold frost, equal on the Guangdong in China, Guangxi, Hainan, south Fujian and south of Yunnan and other places
There is mango plantation.And Baise Tiandong County in Guangxi is the main product area of mango, once obtains " township of mango " title that country issues,
The mango of output is big, taste color, smell and taste are all good, especially this kind of awns seven, is exported to both at home and abroad.Therefore, the plantation of mango
Biggish economic well-being of workers and staff is brought to orchard workers, people is led to shake off poverty and set out on the road to prosperity.If wanting that the mango in mango plantation is allowed to bear greatly
It is a, produce it is more, without insect pest, completely filled fruit, the strongly fragrant Mango Fruit of taste perfume, in addition to mango orchard planting experience abundant, pungent
Except diligent field supervision is answered, it is also necessary to carry out Added Management by modern advanced intellectualized technology.Such as application No. is
201610865666.2 orchard monitoring system, including video monitoring module, Soil K+adsorption module, power supply, control module, electricity
Brain, desinsection module and pour module, control module respectively with video monitoring module, Soil K+adsorption module, power supply, computer, desinsection
Module is connected with module is poured；Video monitoring module starts the progress of desinsection module for analyzing whether fruit tree has pest if having
Deinsectization；Soil K+adsorption module starts pouring module if having and is poured to fruit tree for detecting whether soil lacks moisture content.It should
Control system can carry out desinsection and pouring to fruit tree automatically, and the real-time condition and monitoring number of fruit tree can be observed by computer
According to solving the problems, such as that existing orchard planting needs to waste a large amount of manpowers and goes to manage and maintain.But its each Testing index acquired
Can only carry out rough induction and control, especially for Mango pest insect this part, can only when pest is broken out sensor ability
It can detect, the response and processing at this point for insect pest are late.Therefore, it is necessary to the characteristic researchs for mango plantation
Out a kind of sensitive monitoring, can effectively prevent the intelligence control system of mango insect pest.
Summary of the invention
Goal of the invention of the invention is, in view of the above-mentioned problems, providing a kind of mango insect pest intelligence prison based on image recognition
Control system, it can also remotely monitor mango orchard planting tree by remote monitoring end, and provide and mango is recognized accurately
Tree whether there is the model of pest and disease damage, have the function that Precision management and prevention and treatment.
In order to achieve the above objectives, the technical scheme adopted by the invention is that:
Mango insect pest intelligent monitor system based on image recognition, including the image as orchard on-site parameters acquisition terminal
Acquisition module and orchard monitor supervision platform as Remote data processing and analysing terminal,
The real time picture in mango plantation and for obtaining mango, described image is arranged in described image acquisition module
Acquisition module mainly by local controller, several cameras in mango plantation and above mango woodlot are set, often
A camera is correspondingly connected with the motor as driving device and is mounted in mango plantation by auxiliary installation device, described
The output end of camera is connect with local controller, and the mango real time picture collection for obtaining shooting is sent to local control
Device；The output end of the local controller is divided into two branches: wherein a branch shoots the camera real-time
Pictures are successively according to code storage, the corresponding mango of a coding, while by the mango real time picture of code storage
Module is sent on the monitor supervision platform of orchard collection by wireless communication；Another branch controls the institute being connected with each camera respectively
Motor is stated, to drive the shooting angle of corresponding camera to obtain multi-angle picture；
The orchard monitor supervision platform, including optical sieving module, image analysis processor, insect pest prewarning unit and information are aobvious
Display screen, module is connect described image screening module with described image acquisition module by wireless communication, to by real time picture collection
It carries out intelligent batch to screen, to identify that pictures are effective pictures or invalid effectively pictures and by effective pictures
Issue image analysis processor；Described image analysis processor receives effective pictures that optical sieving module is sent, and to it
Training sample data are classified as after carrying out picture quality processing；The prediction of pest attack unit is based on artificial nerve network model building disease
Prediction of pest attack model, the training sample data that the artificial neural network is sent using image analysis processor are to artificial neural network
Network is trained, until deconditioning after Forecasting Pests convergence, then generate artificial neural network prediction of pest attack mould
Type analyses and compares to the effective image inputted in real time by artificial neural network prediction of pest attack model, exports disease pest in real time
Harmful prediction result is simultaneously shown, simultaneously according to the corresponding generation control instruction of the prediction result, the control on information display screen
Module is sent on onsite alarming device and/or is sent to mobile device end by GSM network module by wireless communication for system instruction
On end.
Preferably, described image screening module is using in artificial intelligence deep learning technology
Tensorflow system is screened to carry out intelligent batch, specifically by the Tensorflow system and use after study and training
The Opencv java standard library of pictures before storage mango causes pest and disease damage is attached, and the Tensorflow system is automatic
The step of sieving effective pictures is as follows:
1) judge whether the mango real time picture collection of input includes effective initiation mango by Tensorflow system
The pictures of pest and disease damage simultaneously return to judging result；2) decide whether to save as the pictures effectively according to the judging result of return
Initiation mango pest and disease damage picture.Mango real time picture is inputted, after Tensorflow system automatically processes, is identified
Whether the picture is effective image, is to save as the effective picture for causing mango pest and disease damage, while having what screening obtained
The pictures of the initiation mango pest and disease damage of effect are as training sample；If it is invalid picture that Tensorflow system, which exports result,
When, then the picture is deleted or is carried out specified storage.
Preferably, the artificial neural network structure includes n convolutional layer, m pond layer and k full chain
Connect layer, the convolutional layer intersects with pond layer to be set gradually, and pond layer use the pond method based on maximum value, wherein n, m,
K is >=1 integer.
Preferably, the loss function of the artificial neural network uses average cross entropy method, to table
Show the error between the matching result and legitimate reading of full articulamentum output, calculation formula is such as shown in (1):
Wherein, N indicates that the effective picture number of mango trained each time, M indicate to cause the type of mango pest and disease damage
Number, y_The concrete class of the effective picture of N number of mango is respectively indicated with y and by softmax value after convolutional network.
In above scheme, further, the calculation formula of Softmax value such as (2) is shown,
Wherein xiIndicate the value of i-th of element in the last one full articulamentum output result.
Preferably, the auxiliary installation device may include first support bar, second support bar, third support
Bar and bracket for installing camera, the first support bar include the support of inner hollow and be vertically arranged in support and
The motion bar that can be moved up and down along the vertical direction, the lower end of the motion bar are connected with L shape bar, the upper end of motion bar and the
The connection of two support rods, one end of the L shape bar is connect with motion bar, the other end of L shape bar is connected with first driving device, described
The bottom end of third support rod is connect with second support bar, the top of third support rod is equipped with spill spin block, the spill spin block
Side surface integrally formed with horizontally disposed swing arm, the bracket be mounted on it is in swing arm and opposite with spill spin block, for driving
The motor for stating camera is fixed on bracket and is connect by cable with local controller, is additionally provided in bracket for controlling
The driving assembly of swing angle when camera images；The first driving device and motor are connect with local controller, institute
It states local controller and is wirelessly connected with remote manipulator.
It is further preferred that the driving component includes the second motor and the swing piece that is mounted in bracket, the swing
The both ends of piece are socketed in the shaft of the camera shooting both sides of head, and the output shaft and swing piece of the second motor are connected and fixed to described
On bracket, second motor is electrically connected with local controller.
It is further preferred that being equipped with the first horse in the top of the third support rod to make camera convert more perspective
It reaches, the input terminal of first motor is connect with the output end of local controller electrical connection, the first motor with spill spin block, to control
The rotation of swing arm processed；It is equipped with sealing plate in the lower section of the first motor, the sealing plate integrated molding is fixed in third support rod,
It is integrally formed in the bottom end of third support rod and is equipped with push plate, the push plate is contacted with the inner wall of the second support bar.
Preferably, it can be equipped with to be in for Auxiliary support L shape bar below the L shape bar and promote shape
The positioning component of state, the positioning component include at least 2 adjustable springs, the adjustable spring one end connection on the support,
The other end of adjustable spring be connected on L shape bar and adjustable spring in the raw when with L shape bar in the counterclockwise direction at 3~
10 ° of acute angles.
Further, the adjustable spring is memory alloy spring, and the stiffness factor of the memory alloy spring is 500
Between~1000N/m.
Due to the adoption of the above technical scheme, the invention has the following advantages:
1. the present invention obtains mango realtime graphic by the camera of setting, by optical sieving module to real-time figure
Screening as carrying out automatic batch filters out whether the pictures are the effective picture for introducing mango pest and disease damage, and will have
Effect picture be stored as training sample be sent to handled on image analysis processor after be sent to insect pest prewarning unit, the worm
Evil prewarning unit establishes neural network Pest model based on artificial neural network, by instructing to the effective picture sent
Practice, until deconditioning after Forecasting Pests convergence, then generate artificial neural network prediction of pest attack model.It is using
When, it is analysed and compared, can be exported in real time to the effective image inputted in real time by artificial neural network prediction of pest attack model
The prediction result of pest and disease damage is simultaneously shown, simultaneously according to the corresponding generation control instruction of the prediction result, institute on information display screen
Stating control instruction, module is sent on onsite alarming device and/or is sent to mobile set by GSM network module by wireless communication
In standby terminal.The present invention integrates the function of autocoding, automatic screening, automatic Prediction result, high with forecasting accuracy,
The high beneficial features of self energy degree, and the early warning and prevention and treatment of various mango pest and disease damages can be applicable in.
2) present invention is equipped with camera and is mounted in mango orchard by auxiliary installation device, and the camera passes through motor
That comes drives to change the viewing angle up and down of camera, observes convenient for different location of the camera to same plant；It is logical
The second motor for crossing the spill spin block connection, realizes 360 ° of rotation of spill spin block in operational process, and then plays change camera shooting
Head camera position purpose so that the collected picture of camera it is relatively sharp with it is comprehensive.
3) by the first support bar of setting, second support bar, third support rod, first support bar is in first driving device
Drive under carry out up and down displacement, the change of upper and lower position is able to achieve, so as to raise the height of camera in the vertical direction
Degree, so that camera obtains the transformation and camera shooting of the camera angle in the better visual field and camera for different shooting demands
The adjustment of height can be operated by the remote manipulator of setting, be not necessarily to manual hand manipulation, control sensitive, practicability
By force.
Detailed description of the invention
Fig. 1 is composition system block diagram of the invention.
Fig. 2 is the composition block diagram of described image acquisition module.
Fig. 3 is the structural schematic diagram of the auxiliary installation device.
Fig. 4 is the third support rod schematic diagram of internal structure in Fig. 3.
Fig. 5 is A-A in Fig. 1 to structural schematic diagram.
In attached drawing, 1, first driving device；2, L shape bar；3, first support bar；4, Temperature and Humidity device；5, cable；6, it holds in the palm
Frame；7, band；8, second support bar；9, third support rod；10, spill spin block；11, swing arm；12, bracket；13, camera；14, electric
Machine；15, the first motor；16, sealing plate；17, the second motor；18, piece, 19, adjustable spring are swung.
Specific embodiment
It is further illustrated below in conjunction with specific implementation of the attached drawing to invention.
As shown in Figure 1, the mango insect pest intelligent monitor system based on image recognition, including acquired as orchard on-site parameters
The image capture module of terminal and orchard monitor supervision platform as Remote data processing and analysing terminal.
As shown in Fig. 2, the real-time figure in mango plantation and for obtaining mango is arranged in described image acquisition module
Piece.Described image acquisition module mainly by local controller, be arranged in it is several in mango plantation and above mango woodlot
A camera 13.Each camera 13 is correspondingly connected with the motor 14 as driving device and is mounted on by auxiliary installation device
In mango plantation.The output end of the camera 13 is connect with local controller, real-time for that will shoot obtained mango
Pictures are sent to local controller.The output end of the local controller is divided into two branches: wherein a branch is to described
Camera 13 shoots obtained real time picture collection successively according to code storage, the corresponding mango of a coding, while will compile
Module is sent on the monitor supervision platform of orchard the mango real time picture collection of code storage by wireless communication；Another branch is controlled respectively
The motor 14 being connected with each camera 13 is made, to drive the shooting angle of corresponding camera 13 to obtain multi-angle figure
Piece.
The orchard monitor supervision platform, including optical sieving module, image analysis processor, insect pest prewarning unit and information are aobvious
Display screen.
Module is connect described image screening module with described image acquisition module by wireless communication, to by real time picture
Collection carries out intelligent batch and screens, to identify that pictures are effective pictures or invalid effectively pictures and by effective picture
Collection issues image analysis processor.The present embodiment, specifically, described image screening module use artificial intelligence deep learning technology
In Tensorflow system screened to carry out intelligent batch, specifically by the Tensorflow system after study and training
It is attached with the Opencv java standard library of pictures before pest and disease damage is caused for storing mango.In the Opencv java standard library
It is stored with and causes all kinds of pictures that pest and disease damage occurs for mango, such as cause anthracnose, mango white powder by obtaining history mango
The early periods such as disease, scale insect, mid-term or all kinds of pictures when can cause the insect vector of mango pest and disease damage, communication media,
All kinds of pictures can be clearly living fuzzy, and carry out study by Tensorflow system and train just to obtain Opencv mark
Quasi- library.Period can be updated Opencv java standard library.
The step of Tensorflow system automatic sieving effective pictures, is as follows:
1) judge whether the mango real time picture collection of input includes effective initiation mango by Tensorflow system
The pictures of pest and disease damage simultaneously return to judging result；2) decide whether to save as the pictures effectively according to the judging result of return
Initiation mango pest and disease damage picture.Mango real time picture is inputted, after Tensorflow system automatically processes, is identified
Whether the picture is effective image, is to save as the effective picture for causing mango pest and disease damage, while having what screening obtained
The pictures of the initiation mango pest and disease damage of effect are as training sample；If it is invalid picture that Tensorflow system, which exports result,
When, then the picture is deleted or is carried out specified storage.
Described image analysis processor receives effective pictures that optical sieving module is sent, and carries out picture quality to it
Training sample data are classified as after processing.The prediction of pest attack unit is based on artificial nerve network model building pest and disease damage and predicts mould
Type, the training sample data that the artificial neural network is sent using image analysis processor instruct artificial neural network
Practice, until deconditioning after Forecasting Pests convergence, then generate artificial neural network prediction of pest attack model.It is using
When, it is analysed and compared by artificial neural network prediction of pest attack model to the effective image inputted in real time, exports disease pest in real time
Harmful prediction result is simultaneously shown, simultaneously according to the corresponding generation control instruction of the prediction result, the control on information display screen
Module is sent on onsite alarming device and/or is sent to mobile device end by GSM network module by wireless communication for system instruction
On end.The analysis comparison is known for same standard, there are the Opencv java standard libraries in insect pest prewarning unit as extracting
It is clipped to whether characteristics of image is the judgement benchmark of pest and disease damage generation, and final prediction result is obtained based on the image feature information.
The artificial neural network structure includes n convolutional layer, m pond layer and k full linking layers, the convolutional layer with
Layer intersection in pond is set gradually, and pond layer uses the pond method based on maximum value, and wherein n, m, k are >=1 integer.This
In embodiment, n, m, k are identical numerical value, and the convolutional layer in artificial neural network structure intersects with pond layer and successively sets
It sets, and pond layer uses the pond method based on maximum value.Simultaneously using relu function as between convolutional layer and pond layer
Activation primitive.
The loss function of the artificial neural network uses average cross entropy method, to indicate of full articulamentum output
With the error between result and legitimate reading, calculation formula is such as shown in (1):
Wherein, N indicates that the effective picture number of mango trained each time, M indicate to cause the type of mango pest and disease damage
Number, y_The concrete class of the effective picture of N number of mango is respectively indicated with y and by softmax value after convolutional network.
The calculation formula of Softmax value such as (2) is shown,
Wherein xiIndicate the value of i-th of element in the last one full articulamentum output result.
The training step of the artificial neural network is as follows: initialization network first, including the network number of plies, number of nodes, power
Value and transforming function transformation function；K=1 is set, the effectively figure image set of first group of initiation mango pest and disease damage of input is obtained after neural network
Output；Compare the threshold value of output and the setting of Opencv java standard library, the threshold value is at least to contain a kind of initiation mango pest and disease damage
Characteristic information, if neural network exports result every time and is in the error range of setting threshold values, training terminates；If judgement knot
Fruit in the error range of setting threshold values, does not then update neural network parameter, and input next group of data, continues to train；Constantly repeatedly
In generation, until the output of network and the goodness of fit of given threshold are good, training terminates.
As shown in figure 3, figure 4 and figure 5, the auxiliary installation device may include first support bar 3, second support bar 8, third
Support rod 9 and bracket 12 for installing camera 13, the first support bar 3 include the support of inner hollow and are vertically arranged
In support and the motion bar that can move up and down along the vertical direction.The lower end of the motion bar is connected with L shape bar 2, activity
The upper end of bar is connect with second support bar 8.One end of the L shape bar 2 is connect with motion bar, the other end of L shape bar 2 is connected with
One driving device 1.The bottom end of the third support rod 9 is connect with second support bar 8, the top of third support rod 9 is equipped with
Spill spin block 10.The side surface of the spill spin block 10 is mounted on swing arm integrally formed with horizontally disposed swing arm 11, the bracket 12
It is on 11 and opposite with spill spin block 10, for drive the motor 14 of the camera 13 to be fixed on bracket 12 and by cable 5 with
Local controller connection.The driving assembly for controlling swing angle when camera 13 images is additionally provided in bracket 12.
The first driving device 1 and motor 14 are connect with local controller, and the local controller is wirelessly connected with remote control
Device.
Notch is offered on the support, the height of the notch in the vertical direction is equal to L shape bar 2 and moves up most
The high upper limit and the sum of the minimum low limit level moved down.
It motor 17 and the swing piece 18 that is mounted in bracket 12 that the driving component, which includes second, described to swing the two of piece 18
End is socketed in the shaft of 13 two sides of camera, and the output shaft and swing piece 18 of the second motor 17 are connected and fixed to described
On bracket 12, second motor 17 is electrically connected with local controller.
To make camera 13 convert more perspective, equipped with the first motor 15 in the top of the third support rod 9, described the
The input terminal of one motor 15 is electrically connected with local controller, the output end of the first motor 15 is connect with spill spin block 10, to control
The rotation of swing arm 11；It is equipped with sealing plate 16 in the lower section of the first motor 15, the integrated molding of sealing plate 16 is fixed on third branch
In strut 9, it is integrally formed in the bottom end of third support rod 9 and is equipped with push plate, the push plate and the inner wall of the second support bar 8 connect
Touching.
The positioning component for Auxiliary support L shape bar 2 can be installed in the lower section of the L shape bar 2.The positioning component packet
Include at least 2 adjustable springs 19, in the present embodiment, the adjustable spring 19 is 3.One end of the adjustable spring 19 is connected to
On support, the other end of adjustable spring 19 be connected on L shape bar 2 and adjustable spring 19 in the raw when and L shape bar 2 along inverse
Clockwise is at 3~10 ° of acute angles.
In order to guarantee to also ensure the elastic force of flexible paralysis 24, institute while supporting role of the adjustable spring 19 to L shape bar 2
Stating adjustable spring 19 is memory alloy spring, and the stiffness factor of the memory alloy spring is between 500~1000N/m.?
In use, adjustable spring 19 is stretched by L shape bar 2 when L shape bar 2 is jacked up upwards by the output shaft of first driving device 1,3 are stretched
Contracting spring 19 is supported L shape bar 2 by the effect of the support force of support and itself rigid power, to prevent from adjusting camera
The case where when 13 height when appearance shake, trembling or first driving device 1 out of control, the steady of whole device can be significantly improved
Property.
The auxiliary installation device is when in use: camera 13 and orchard monitor supervision platform being carried out signal connection, work people
Member can adjust the camera angle and camera shooting height of camera 13 by operation remote manipulator.
Concrete operations are as follows: the first motor 15 of control takes the photograph 10 circular-rotation of spill spin block to control the rotation of swing arm 11
Rotation is synchronized with swing arm 11 as first 13；The vertical height of first support bar 3 is adjusted by control first driving device 1,
To adjust the vertical camera shooting height of camera 13, observed convenient for plant of the camera 13 to different direction；It was observing
Staff can adjust the upper downwards angle of visibility of camera 13 by control motor 14 in journey, convenient for the different height to same plant
It is observed at position.
Above description is the detailed description for the present invention preferably possible embodiments, but embodiment is not limited to this hair
Bright patent claim, it is all the present invention suggested by technical spirit under completed same changes or modifications change, should all belong to
In the covered the scope of the patents of the present invention.
Claims (10)
1. the mango insect pest intelligent monitor system based on image recognition, including being adopted as the image of orchard on-site parameters acquisition terminal
Collect module and the orchard monitor supervision platform as Remote data processing and analysing terminal, it is characterised in that:
The real time picture in mango plantation and for obtaining mango, described image acquisition is arranged in described image acquisition module
Module mainly by local controller, several cameras in mango plantation and above mango woodlot are set, each take the photograph
The motor as driving device is correspondingly connected with as head and is mounted in mango plantation by auxiliary installation device, the camera shooting
The output end of head is connect with local controller, and the mango real time picture collection for obtaining shooting is sent to local controller；
The output end of the local controller is divided into two branches: the real time picture that wherein a branch shoots the camera
Collection leads to the mango real time picture collection of code storage successively according to code storage, the corresponding mango of a coding
Wireless communication module is crossed to be sent on the monitor supervision platform of orchard；Another branch controls the electricity being connected with each camera respectively
Machine, to drive the shooting angle of corresponding camera to obtain multi-angle picture；
The orchard monitor supervision platform, including optical sieving module, image analysis processor, insect pest prewarning unit and information are shown
Screen, module is connect described image screening module with described image acquisition module by wireless communication, to by real time picture collection into
The intelligent batch of row screens, to identify that pictures are effective pictures or invalid effectively pictures and send out effective pictures
To image analysis processor；Described image analysis processor receives effective pictures for sending of optical sieving module, and to its into
Training sample data are classified as after the processing of row picture quality；The prediction of pest attack unit is based on artificial nerve network model and constructs disease pest
Evil prediction model, the training sample data that the artificial neural network is sent using image analysis processor are to artificial neural network
It is trained, until deconditioning after Forecasting Pests convergence, then generate artificial neural network prediction of pest attack model,
It is analysed and compared by artificial neural network prediction of pest attack model to the effective image inputted in real time, exports pest and disease damage in real time
Prediction result simultaneously shows, simultaneously according to the prediction result corresponding generation control instruction that the control refers on information display screen
Module is sent on onsite alarming device and/or is sent in mobile device terminal by GSM network module by wireless communication for order.
2. the mango insect pest intelligent monitor system according to claim 1 based on image recognition, it is characterised in that: the figure
It is screened as screening module carries out intelligent batch using the Tensorflow system in artificial intelligence deep learning technology, specifically
It is by the Tensorflow system after study and training and to be used to store the pictures before mango causes pest and disease damage
The step of Opencv java standard library is attached, the Tensorflow system automatic sieving effective pictures is as follows:
1) judge whether the mango real time picture collection of input includes effective initiation mango disease pest by Tensorflow system
Harmful pictures simultaneously return to judging result；2) decide whether to save as the pictures according to the judging result of return and effectively draw
Send out mango pest and disease damage picture.
3. the mango insect pest intelligent monitor system according to claim 1 based on image recognition, it is characterised in that: the people
Artificial neural networks structure includes n convolutional layer, m pond layer and k full linking layer, and the convolutional layer intersects successively with pond layer
Setting, and pond layer uses the pond method based on maximum value, wherein n, m, k are >=1 integer.
4. the mango insect pest intelligent monitor system according to claim 3 based on image recognition, it is characterised in that: the people
The loss function of artificial neural networks uses average cross entropy method, to indicate the matching result and true knot of full articulamentum output
Error between fruit, calculation formula is such as shown in (1):
Wherein, N indicates that the effective picture number of mango trained each time, M indicate to cause the type number of mango pest and disease damage,
y_The concrete class of the effective picture of N number of mango is respectively indicated with y and by softmax value after convolutional network.
5. the mango insect pest intelligent monitor system according to claim 4 based on image recognition, it is characterised in that:
The calculation formula of Softmax value such as (2) is shown,
Wherein xiIndicate the value of i-th of element in the last one full articulamentum output result.
6. the mango insect pest intelligent monitor system according to claim 1 based on image recognition, it is characterised in that: it is described auxiliary
Helping mounting device includes first support bar, second support bar, third support rod and the bracket for installing camera, and described first
Support rod includes the support of inner hollow and is vertically arranged in the motion bar that can be moved up and down in support and along the vertical direction,
The lower end of the motion bar is connected with L shape bar, and the upper end of motion bar is connect with second support bar, one end of the L shape bar and work
Lever connects, the other end of L shape bar is connected with first driving device, and the bottom end of the third support rod and second support bar connect
Connect, the top of third support rod is equipped with spill spin block, the side surface of the spill spin block integrally formed with horizontally disposed swing arm,
The bracket be mounted on it is in swing arm and opposite with spill spin block, for driving the motor of the camera to be fixed on bracket and passing through
Cable is connect with local controller, and the driving group for controlling swing angle when camera images is additionally provided in bracket
Part；The first driving device and motor are connect with local controller, and the local controller is wirelessly connected with remote control
Device.
7. the mango insect pest intelligent monitor system according to claim 6 based on image recognition, it is characterised in that: the drive
Dynamic component includes the second motor and the swing piece that is mounted in bracket, and the both ends for swinging piece are socketed in the camera shooting both sides of head
Shaft on, the output shaft of the second motor and swing piece and be connected and fixed on the bracket, second motor and local control
Device electrical connection processed.
8. auxiliary observation device between the garden of mango orchard according to claim 6, it is characterised in that: the third support rod
The first motor is equipped in top, the input terminal of first motor is electrically connected with local controller, the output end of the first motor and
Spill spin block connection, to control the rotation of swing arm；It is equipped with sealing plate in the lower section of the first motor, the sealing plate is integrally formed solid
It is scheduled in third support rod, is integrally formed in the bottom end of third support rod and is equipped with push plate, the push plate and the second support bar
Inner wall contact.
9. the mango insect pest intelligent monitor system according to claim 6 based on image recognition, it is characterised in that: described
The positioning component that promotion state is in for Auxiliary support L shape bar is installed, the positioning component includes at least below L shape bar
2 adjustable springs, one end connection of the adjustable spring is on the support, the other end of adjustable spring is connected on L shape bar and stretches
Contracting spring in the raw when with L shape bar in the counterclockwise direction at 3~100 acute angles.
10. the mango insect pest intelligent monitor system according to claim 9 based on image recognition, it is characterised in that: it is described
Adjustable spring is memory alloy spring, and the stiffness factor of the memory alloy spring is between 500~1000N/m.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811163513.9A CN109379557A (en) | 2018-09-30 | 2018-09-30 | Mango insect pest intelligent monitor system based on image recognition |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811163513.9A CN109379557A (en) | 2018-09-30 | 2018-09-30 | Mango insect pest intelligent monitor system based on image recognition |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109379557A true CN109379557A (en) | 2019-02-22 |
Family
ID=65402936
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811163513.9A Pending CN109379557A (en) | 2018-09-30 | 2018-09-30 | Mango insect pest intelligent monitor system based on image recognition |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109379557A (en) |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109978029A (en) * | 2019-03-13 | 2019-07-05 | 北京邮电大学 | A kind of invalid image pattern screening technique based on convolutional neural networks |
| CN110262603A (en) * | 2019-07-17 | 2019-09-20 | 马明 | A kind of intelligence orchard managing and control system |
| CN110915470A (en) * | 2019-12-11 | 2020-03-27 | 常德鑫芙蓉环保有限公司 | Tobacco leaf pest control system |
| CN111027436A (en) * | 2019-12-03 | 2020-04-17 | 吉林大学 | Northeast black fungus disease and pest image recognition system based on deep learning |
| CN111445062A (en) * | 2020-03-17 | 2020-07-24 | 深圳数联天下智能科技有限公司 | Insect pest position prediction method and device, computer equipment and storage medium |
| CN111507940A (en) * | 2020-03-18 | 2020-08-07 | 仲恺农业工程学院 | Image recognition-based citrus orchard pest and disease identification and alarm system and method |
| CN112217898A (en) * | 2020-10-15 | 2021-01-12 | 江苏斑点狗网络科技有限公司 | Monitoring system of distributed database cluster |
| CN112381156A (en) * | 2020-11-18 | 2021-02-19 | 重庆电子工程职业学院 | Chinese medicinal material pest and disease damage detection system based on image recognition |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN203708332U (en) * | 2013-12-31 | 2014-07-09 | 浙江博采传媒有限公司 | Nine-shaft automatic photographic device |
| US20170083747A1 (en) * | 2015-09-21 | 2017-03-23 | The Climate Corporation | Ponding water detection on satellite imagery |
| CN106682704A (en) * | 2017-01-20 | 2017-05-17 | 中国科学院合肥物质科学研究院 | Method of disease image identification based on hybrid convolutional neural network fused with context information |
| CN108111614A (en) * | 2017-12-27 | 2018-06-01 | 北京工业大学 | Agriculture and forestry pest and disease monitoring manages system |
| CN108363990A (en) * | 2018-03-14 | 2018-08-03 | 广州影子控股股份有限公司 | One boar face identifying system and method |
| CN108388877A (en) * | 2018-03-14 | 2018-08-10 | 广州影子控股股份有限公司 | The recognition methods of one boar face |
-
2018
- 2018-09-30 CN CN201811163513.9A patent/CN109379557A/en active Pending
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN203708332U (en) * | 2013-12-31 | 2014-07-09 | 浙江博采传媒有限公司 | Nine-shaft automatic photographic device |
| US20170083747A1 (en) * | 2015-09-21 | 2017-03-23 | The Climate Corporation | Ponding water detection on satellite imagery |
| CN106682704A (en) * | 2017-01-20 | 2017-05-17 | 中国科学院合肥物质科学研究院 | Method of disease image identification based on hybrid convolutional neural network fused with context information |
| CN108111614A (en) * | 2017-12-27 | 2018-06-01 | 北京工业大学 | Agriculture and forestry pest and disease monitoring manages system |
| CN108363990A (en) * | 2018-03-14 | 2018-08-03 | 广州影子控股股份有限公司 | One boar face identifying system and method |
| CN108388877A (en) * | 2018-03-14 | 2018-08-10 | 广州影子控股股份有限公司 | The recognition methods of one boar face |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109978029A (en) * | 2019-03-13 | 2019-07-05 | 北京邮电大学 | A kind of invalid image pattern screening technique based on convolutional neural networks |
| CN109978029B (en) * | 2019-03-13 | 2021-02-09 | 北京邮电大学 | Invalid image sample screening method based on convolutional neural network |
| CN110262603A (en) * | 2019-07-17 | 2019-09-20 | 马明 | A kind of intelligence orchard managing and control system |
| CN111027436A (en) * | 2019-12-03 | 2020-04-17 | 吉林大学 | Northeast black fungus disease and pest image recognition system based on deep learning |
| CN110915470A (en) * | 2019-12-11 | 2020-03-27 | 常德鑫芙蓉环保有限公司 | Tobacco leaf pest control system |
| CN111445062A (en) * | 2020-03-17 | 2020-07-24 | 深圳数联天下智能科技有限公司 | Insect pest position prediction method and device, computer equipment and storage medium |
| CN111445062B (en) * | 2020-03-17 | 2023-08-29 | 深圳数联天下智能科技有限公司 | Pest location prediction method, device, computer equipment and storage medium |
| CN111507940A (en) * | 2020-03-18 | 2020-08-07 | 仲恺农业工程学院 | Image recognition-based citrus orchard pest and disease identification and alarm system and method |
| CN112217898A (en) * | 2020-10-15 | 2021-01-12 | 江苏斑点狗网络科技有限公司 | Monitoring system of distributed database cluster |
| CN112381156A (en) * | 2020-11-18 | 2021-02-19 | 重庆电子工程职业学院 | Chinese medicinal material pest and disease damage detection system based on image recognition |
| CN112381156B (en) * | 2020-11-18 | 2023-04-18 | 重庆电子工程职业学院 | Chinese medicinal material pest and disease damage detection system based on image recognition |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109379557A (en) | Mango insect pest intelligent monitor system based on image recognition | |
| CN209517198U (en) | A kind of wisdom agricultural standardization management system | |
| CN109035057A (en) | Mango insect pest intelligent monitor system with sightseeing corridor | |
| CN213092145U (en) | Intelligent management measurement and control system for plant growth | |
| CN104381231A (en) | Image-based field insect pest situation monitoring device | |
| CN105547252A (en) | Crop canopy image acquisition device based on context awareness | |
| CN106614440B (en) | Intelligent crop pest forecasting system based on Internet of things | |
| CN110472557A (en) | A kind of method and device of tomato growth monitoring | |
| CN106971409B (en) | Maize canopy leaf color modeling and method | |
| CN105159366B (en) | A kind of user behavior acquisition method for being suitable to greenhouse surroundings monitoring system | |
| CN113273449A (en) | Digital twin body construction method for precise monitoring of sunlight greenhouse | |
| CN105547360A (en) | Crop canopy image collection method based on context awareness | |
| CN110427063A (en) | A kind of tomato growth monitoring management platform based on Internet of Things | |
| CN104521438A (en) | Shearing type shiny-leaved yellowhorn picking machine | |
| CN106034829A (en) | Intelligent greenhouse | |
| CN112465109A (en) | Green house controlling means based on cloud limit is in coordination | |
| CN115238237A (en) | Insect pest situation monitoring method, device and system | |
| CN112565310A (en) | Intelligent plant protection system based on artificial neural network | |
| CN107844149A (en) | A kind of greenhouse intelligent detecting system based on Internet of Things | |
| CN116627193A (en) | Intelligent management and control platform and method for greenhouse | |
| CN113348888A (en) | Tomato grafting and seedling temperature control device and method | |
| CN205865325U (en) | Intelligent big -arch shelter | |
| CN2585550Y (en) | Computer intelligent control system for greenhouse | |
| CN103518686A (en) | Drop irrigation cotton field cotton aphid quick monitoring and early warning method | |
| CN105867193A (en) | Intelligent family disease prevention and treatment equipment, system and method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190222 |