CN107463959A - A kind of fruit fly recognition methods based on BP neural network - Google Patents
A kind of fruit fly recognition methods based on BP neural network Download PDFInfo
- Publication number
- CN107463959A CN107463959A CN201710663088.9A CN201710663088A CN107463959A CN 107463959 A CN107463959 A CN 107463959A CN 201710663088 A CN201710663088 A CN 201710663088A CN 107463959 A CN107463959 A CN 107463959A
- Authority
- CN
- China
- Prior art keywords
- drosophila
- neural network
- fruit fly
- characteristic point
- grader
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
- G06F18/2148—Generating training patterns; Bootstrap methods, e.g. bagging or boosting characterised by the process organisation or structure, e.g. boosting cascade
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/044—Recurrent networks, e.g. Hopfield networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/10—Image acquisition
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Biology (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Computational Linguistics (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Multimedia (AREA)
- Image Processing (AREA)
Abstract
A kind of fruit fly recognition methods based on BP neural network, drosophila map picture is acquired by using terminal device for methods described and feature extraction, according to the characteristics of drosophila, characteristic point coordinate value extraction is carried out with the method for mark drosophila wing characteristic point, forms sample data.Then learning training is carried out to sample data by using the grader that BP neural network algorithm is realized.Identification drosophila finally is identified using grader.Methods described includes:IMAQ is carried out using drosophila image collecting device, chooses image characteristic point, drosophila identification, return qualification result.Technical solution of the present invention has broken traditional drosophila identification method, realizes realtime qualification process, substantially increases operating efficiency.
Description
Technical field
The present invention relates to a kind of fruit fly recognition methods based on BP neural network, belongs to mobile communication technology field.
Background technology
In order to protect the safety of agricultural production, various countries have carried out strict animals and plants inspection and quarantine, have prevented that high risk has
The diffusion and invasion of evil biology, but with the continuous improvement of international free trade degree, the continuous growth of trade quantity, to each
The problem of inspection and quarantine of kind plants and plant product already becomes a sternness.At present, generally existing in inspection and quarantine work
Following drawback:(1) very high is required to inspection and quarantine worker specialized capability, plant quarantine is related to plant pathogenetic bacteria, true
The harmful organism of numerous classifications such as bacterium, nematode, weeds, insect, quarantine person must possess substantial amounts of relevant biological classification side
The professional knowledge in face;(2) in quarantine procedures, the various excessive risk harmful organisms do not known well are difficult to accurately identify;(3) to mirror
Surely, it is necessary to look through a great amount of information, also Xu Song correlative studys department and expert are identified the harmful organism having any problem sometimes, are wasted
A large amount of manpower and materials, and need to spend longer time;(4) biology to be identified need to connect PC computer ends and can just be identified;
Therefore it provides a kind of fruit fly recognition methods based on BP neural network is very necessary.
The content of the invention
The object of the present invention is in order to solve problem present in above-mentioned drosophila identification process, there is provided one kind is based on BP god
Fruit fly recognition methods through network.
Technical scheme is as follows:A kind of fruit fly recognition methods based on BP neural network, methods described utilize
Drosophila map picture is acquired terminal device and feature extraction;According to the characteristics of drosophila, with mark drosophila wing feature point coordinates
The method of value carries out characteristic point coordinate value extraction, forms sample data;Then realized by using BP neural network algorithm
Grader carries out classification based training to sample data;Identification drosophila finally is identified using grader.The inventive method includes
Following steps:
(1)IMAQ is carried out using drosophila image collecting device;
(2)Choose image characteristic point;
(3)Drosophila is identified;
(4)Return to qualification result;
(5)Terminate this identification.
The step is completed in mobile terminal.
The drosophila image collecting device, using existing Samsung flat board Note8.0, additional very energy wide-angle ultra micro away from mobile phone
Camera lens W-67 is assembled into a hand-held drosophila image collecting device.
The drosophila map includes image collecting device part and mobile terminal as device for picking;Image collecting device part passes through number
Mobile terminal is connected according to line;
Described image harvester part is mobile lens, including wide-angle lens, micro-lens and camera len;The wide-angle lens
For 110 degree of wide-angles, the micro-lens is 10 times of micro-lens;The camera lens uses aluminum alloy casing and optical glass lens,
The camera len is connected with mobile terminal;
The mobile terminal is provided with:
Photo module:For taking pictures, connect mobile terminal by mobile lens and complete;
Picture processing module:Realize and picture is scaled and rotated;
Feature point extraction module:For carrying out characteristic point selection to image;
Threshold module is set:Thresholding size is adjusted according to actual conditions, improves identification accuracy rate;
Add sample module:Identified drosophila categorical data is added to characteristic value file;
Train classifier modules:After obtaining trypetid wing characteristic value, grader is trained with characteristic value;
Drosophila identifies module:For carrying out category authentication to drosophila map picture.
Described image carries out feature point extraction, is carried from the method progress characteristic point coordinate value of mark drosophila wing characteristic point
Take, the point on definition and the selection of mark point in geometric shape surveying is met can react the base of the essential characteristic of drosophila
On plinth, 15 groups of relative distance values between 11 drosophila wing characteristic points are have chosen as characteristic of division.
The setting threshold value, refer to set critical value, with increasing for sample data, can dynamically adjust the big of threshold value
It is small, improve the accuracy and reliability of drosophila identification.
It is described to drosophila map as carry out category authentication through the following steps that realize:
(1)It is loaded into known class drosophila map picture;
(2)Image characteristic point extracts;
(3)Form characteristic value file and preserve;
(4)Train grader.
The training grader, refer to and learning training carried out to the grader that BP neural network algorithm is realized,
First, input layer receives drosophila wing sample data, and then, the information transmission received is passed through to hidden layer, drosophila data
After processing, hidden layer so, just completes the forward-propagating of a data again the information transmission after processing to output layer
Journey, when the reality output of output layer is not consistent with expected output result, then into the back-propagation phase of error;Institute
The backpropagation of meaning refers to that be transmitted to hidden layer by output layer is transmitted to input layer again;Pass through the data message forward-propagating to move in circles
With error back propagation process so that error is declined by gradient direction, the weights and threshold value of each layer is adjusted, until reality output is use up
It is possible be consistent with desired output just complete the learning training process of grader.
The invention has the advantages that the invention provides a kind of drosophila recognition methods based on BP neural network, quarantine
Personnel carry out situ appraisal using the mobile application system portable set of the present invention, solve in conventional gatherer process to identification
The shortcomings of personnel specialty competency profiling is higher and process is relatively complicated, the working strength of quarantine functionary is alleviated, in addition, this image
Gatherer process is simple, easy to use, and accuracy is considerable, effectively improves drosophila identification situ appraisal efficiency, has very extensive
Application prospect.
Brief description of the drawings
Fig. 1 is a kind of flow chart of the drosophila recognition methods based on BP neural network of the present invention;
Fig. 2 is a kind of flow chart for being used to realize the classifier design of this method of the present invention;
Fig. 3 is a kind of precedence diagram for drosophila image characteristic point mark of the present invention.
Fig. 4 is a kind of flow chart for being used to realize the BP neural network algorithm of this method of the present invention.
Embodiment
The embodiment of the present invention is as shown in the figure.
Fig. 1 is a kind of flow chart of the drosophila recognition methods based on BP neural network of the present invention, the specific steps of this method
For:
(1)IMAQ is carried out using drosophila image collecting device:Use the additional very energy wide-angle Mobile phone mirrors of Samsung flat board Note8.0
The hand-held drosophila image collecting device that head W-67 is assembled into carries out IMAQ to drosophila to be identified.
(2)Choose image characteristic point:Pretreatment to characteristic point;Then it is corresponding by marking image characteristic point to obtain image
Coordinate value, it should be noted that marker characteristic point has sequencing, as shown in Figure 3 numeral it is ascending be mark it is special
Levy the sequencing of point.
(3)Drosophila is identified:According to the characteristic point parameter of acquisition, combining classification device is identified drosophila and shown in terminal
Show result, complete identification.
(4)Return to qualification result;
(5)Terminate this identification.
When being pre-processed to characteristic point, including following link:
(1)Choosing is that can accurately realize the key factor of drosophila identification to characteristic point, when mark point is chosen, it is necessary to follow
Following two principles:First, the selection of trypetid wing characteristic point has to comply with geometric shape surveying and determined on mark point
Justice;Second, the point of selection will can react the essential characteristic of trypetid wing, and mutually echoed with trypetid traditional taxonomy feature.Abide by
Following two above principle, the present invention have chosen 15 groups of relative distance values between 11 drosophila wing characteristic points as characteristic of division,
Meanwhile also by Fisher discriminant analysis methods demonstrate characteristic parameter be used for identify feasibility.
(2)The trypetid wing picture extracted need to be standardized, to eliminate influence of the non-form parameter to characteristic point,
Specifically include displacement, scaling and rotation.The processing procedure of displacement refers to, in mark point, specified coordinate origin, and other mark points
Coordinate value do corresponding change.The processing of scaling refers to coordinate value divided by can react a numerical value of trypetid wing size, so as to
Eliminate the difference of the coordinate value caused by the difference of size of trypetid wing itself.The process of rotation refers to, wing is surrounded into the origin of coordinates
Rotated, so that the distance between mark point of each same position of trypetid of the same race meets the requirement of least square method.
As shown in Fig. 2 grader through the following steps that realize:(1)It is loaded into known class drosophila map picture;(2)Extraction
Characteristic point;(3)Form characteristic value file and preserve;(4)Grader is trained, grader is come real by BP neural network algorithm
Existing, particular flow sheet is as shown in Figure 4:At the beginning of selection including initialization weights and threshold value, acquisition drosophila sample data, parameters
Beginningization, calculate the output of hidden layer output layer each unit, calculate output layer error E (q), being wanted as output layer error E (q) is not met
Ask, then calculation error, correct weight threshold;Continue to calculate the output of hidden layer output layer each unit and output layer error E (q), directly
Met the requirements to output layer error.
Claims (7)
1. a kind of fruit fly recognition methods based on BP neural network, it is characterised in that methods described is using terminal device to fruit
Fly image is acquired and feature extraction;According to the characteristics of drosophila, characteristic point is carried out with the method for mark drosophila wing characteristic point
Coordinate value extracts, and forms sample data;Then the grader realized by using BP neural network algorithm enters to sample data
Row classification based training;Identification drosophila finally is identified using grader;
It the described method comprises the following steps:
(1)IMAQ is carried out using drosophila image collecting device;
(2)Choose image characteristic point;
(3)Drosophila is identified;
(4)Return to qualification result;
(5)Terminate this identification;
The step is completed in mobile terminal.
A kind of 2. fruit fly recognition methods based on BP neural network according to claim 1, it is characterised in that the fruit
Fly image collecting device, it is assembled into using existing Samsung flat board Note8.0, additional very energy wide-angle ultra micro away from mobile lens W-67
One portable drosophila image collecting device.
A kind of 3. fruit fly recognition methods based on BP neural network according to claim 2, it is characterised in that the fruit
Fly image device for picking includes image collecting device part and mobile terminal;Image collecting device part is connected by data wire and moved
End;
Described image harvester part is mobile lens, including wide-angle lens, micro-lens and camera len;The wide-angle lens
For 110 degree of wide-angles, the micro-lens is 10 times of micro-lens;The camera lens uses aluminum alloy casing and optical glass lens,
The camera len is connected with mobile terminal;
The mobile terminal is provided with:
Photo module:For taking pictures, connect mobile terminal by mobile lens and complete;
Picture processing module:Realize and picture is scaled and rotated;
Feature point extraction module:For carrying out characteristic point selection to image;
Threshold module is set:Thresholding size is adjusted according to actual conditions, improves identification accuracy rate;
Add sample module:Identified drosophila categorical data is added to characteristic value file;
Train classifier modules:After obtaining trypetid wing characteristic value, grader is trained with characteristic value;
Drosophila identifies module:For carrying out category authentication to drosophila map picture.
A kind of 4. fruit fly recognition methods based on BP neural network according to claim 3, it is characterised in that the figure
As carrying out feature point extraction, characteristic point coordinate value extraction is carried out from the method for mark drosophila wing characteristic point, is meeting geometric form
On the basis of point in state surveying on definition and the selection of mark point can react the essential characteristic of drosophila, 11 are have chosen
15 groups of relative distance values between drosophila wing characteristic point are as characteristic of division.
5. a kind of fruit fly recognition methods based on BP neural network according to claim 3, it is characterised in that described to set
Threshold value is put, refers to set critical value, with increasing for sample data, the size of threshold value can be dynamically adjusted, improve drosophila identification
Accuracy and reliability.
6. a kind of fruit fly recognition methods based on BP neural network according to claim 3, it is characterised in that described right
Drosophila map is as carrying out category authentication through the following steps that realizing:
(1)It is loaded into known class drosophila map picture;
(2)Image characteristic point extracts;
(3)Form characteristic value file and preserve;
(4)Train grader.
A kind of 7. fruit fly recognition methods based on BP neural network according to claim 3, it is characterised in that the instruction
Practice grader, refer to and learning training is carried out to the grader that BP neural network algorithm is realized,
First, input layer receives drosophila wing sample data, and then, the information transmission received is passed through to hidden layer, drosophila data
After processing, hidden layer so, just completes the forward-propagating of a data again the information transmission after processing to output layer
Journey, when the reality output of output layer is not consistent with expected output result, then into the back-propagation phase of error;Institute
The backpropagation of meaning refers to that be transmitted to hidden layer by output layer is transmitted to input layer again;Pass through the data message forward-propagating to move in circles
With error back propagation process so that error is declined by gradient direction, the weights and threshold value of each layer is adjusted, until reality output is use up
It is possible be consistent with desired output just complete the learning training process of grader.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710663088.9A CN107463959A (en) | 2017-08-05 | 2017-08-05 | A kind of fruit fly recognition methods based on BP neural network |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201710663088.9A CN107463959A (en) | 2017-08-05 | 2017-08-05 | A kind of fruit fly recognition methods based on BP neural network |
Publications (1)
Publication Number | Publication Date |
---|---|
CN107463959A true CN107463959A (en) | 2017-12-12 |
Family
ID=60548311
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201710663088.9A Pending CN107463959A (en) | 2017-08-05 | 2017-08-05 | A kind of fruit fly recognition methods based on BP neural network |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107463959A (en) |
Cited By (3)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108280483A (en) * | 2018-01-30 | 2018-07-13 | 华南农业大学 | Trypetid adult image-recognizing method based on neural network |
CN109555521A (en) * | 2019-01-29 | 2019-04-02 | 冀中能源峰峰集团有限公司 | A kind of cutting head of roadheader combined positioning method |
CN110942063A (en) * | 2019-11-21 | 2020-03-31 | 望海康信(北京)科技股份公司 | Certificate text information acquisition method and device and electronic equipment |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20160086015A1 (en) * | 2007-01-09 | 2016-03-24 | Si Corporation | Method and system for automated face detection and recognition |
CN105643157A (en) * | 2016-03-02 | 2016-06-08 | 湘潭大学 | Automatic girder welding obstacle predicting method for optimizing GRNN based on correction type fruit fly algorithm |
-
2017
- 2017-08-05 CN CN201710663088.9A patent/CN107463959A/en active Pending
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20160086015A1 (en) * | 2007-01-09 | 2016-03-24 | Si Corporation | Method and system for automated face detection and recognition |
CN105643157A (en) * | 2016-03-02 | 2016-06-08 | 湘潭大学 | Automatic girder welding obstacle predicting method for optimizing GRNN based on correction type fruit fly algorithm |
Non-Patent Citations (1)
Title |
---|
娄丽霞: "基于Android的高风险果实蝇图像识别系统研究与实现", 《中国优秀硕士学位论文全文数据库信息科技辑》 * |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN108280483A (en) * | 2018-01-30 | 2018-07-13 | 华南农业大学 | Trypetid adult image-recognizing method based on neural network |
CN109555521A (en) * | 2019-01-29 | 2019-04-02 | 冀中能源峰峰集团有限公司 | A kind of cutting head of roadheader combined positioning method |
CN110942063A (en) * | 2019-11-21 | 2020-03-31 | 望海康信(北京)科技股份公司 | Certificate text information acquisition method and device and electronic equipment |
CN110942063B (en) * | 2019-11-21 | 2023-04-07 | 望海康信(北京)科技股份公司 | Certificate text information acquisition method and device and electronic equipment |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107463959A (en) | A kind of fruit fly recognition methods based on BP neural network | |
CN110147794A (en) | A kind of unmanned vehicle outdoor scene real time method for segmenting based on deep learning | |
CN108629368B (en) | Multi-modal foundation cloud classification method based on joint depth fusion | |
CN111985274B (en) | Remote sensing image segmentation method based on convolutional neural network | |
CN110108704A (en) | A kind of automatic monitoring and pre-alarming method of cyanobacteria and its automatic monitoring and alarming system | |
CN110516723B (en) | Multi-modal foundation cloud picture identification method based on depth tensor fusion | |
CN104063686A (en) | System and method for performing interactive diagnosis on crop leaf segment disease images | |
CN106874961A (en) | A kind of indoor scene recognition methods using the very fast learning machine based on local receptor field | |
CN115457006B (en) | Unmanned aerial vehicle inspection defect classification method and device based on similarity consistency self-distillation | |
CN112464766A (en) | Farmland automatic identification method and system | |
CN114626476A (en) | Bird fine-grained image recognition method and device based on Transformer and component feature fusion | |
CN110288033A (en) | A kind of identification of sugarcane sugarcane section feature and localization method based on convolutional neural networks | |
CN106991449A (en) | A kind of living scene reconstruct assists in identifying the method for blueberry kind | |
CN110321774A (en) | Crops evaluation methods for disaster condition, device, equipment and computer readable storage medium | |
CN113012111A (en) | Rice lodging degree grading monitoring method based on edge calculation | |
CN116597332A (en) | Unmanned aerial vehicle-based tea tree bud detection method and device | |
CN110287990A (en) | Microalgae image classification method, system, equipment and storage medium | |
CN109063756A (en) | A kind of cloud atlas classification method based on multi-modal production confrontation network | |
Wang et al. | Identification of growing points of cotton main stem based on convolutional neural network | |
CN110728205A (en) | AI cloud computing-based portable intelligent rice seed test system | |
CN104615987B (en) | A kind of the wreckage of an plane intelligent identification Method and system based on error-duration model neutral net | |
CN115458151A (en) | Diagnostic method for cryptocaryon irritans disease of marine fishes based on image recognition technology | |
CN212322280U (en) | Corn plant heart identification system based on machine vision | |
CN110796639B (en) | Pinellia ternata quality grading method based on neural network | |
CN112115885A (en) | Fruit tree bearing branch shearing point positioning method for picking based on deep convolutional neural network |
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 | ||
WD01 | Invention patent application deemed withdrawn after publication | ||
WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20171212 |