CN110008912A - A kind of social platform matching process and system based on plants identification - Google Patents
A kind of social platform matching process and system based on plants identification Download PDFInfo
- Publication number
- CN110008912A CN110008912A CN201910286271.0A CN201910286271A CN110008912A CN 110008912 A CN110008912 A CN 110008912A CN 201910286271 A CN201910286271 A CN 201910286271A CN 110008912 A CN110008912 A CN 110008912A
- Authority
- CN
- China
- Prior art keywords
- image
- user
- blade
- information
- matching
- 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.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 41
- 230000008569 process Effects 0.000 title claims abstract description 19
- 210000003462 vein Anatomy 0.000 claims abstract description 33
- 230000011218 segmentation Effects 0.000 claims abstract description 28
- 238000013528 artificial neural network Methods 0.000 claims abstract description 14
- 230000001537 neural effect Effects 0.000 claims abstract description 11
- 230000004913 activation Effects 0.000 claims description 18
- 238000000605 extraction Methods 0.000 claims description 18
- 230000006870 function Effects 0.000 claims description 9
- 230000005484 gravity Effects 0.000 claims description 7
- 238000012545 processing Methods 0.000 claims description 7
- 238000001914 filtration Methods 0.000 claims description 3
- 238000012549 training Methods 0.000 claims description 2
- 241000196324 Embryophyta Species 0.000 description 23
- 238000010586 diagram Methods 0.000 description 7
- 238000012360 testing method Methods 0.000 description 4
- 230000000694 effects Effects 0.000 description 3
- 239000011159 matrix material Substances 0.000 description 3
- 230000009466 transformation Effects 0.000 description 3
- 230000009286 beneficial effect Effects 0.000 description 2
- 238000006243 chemical reaction Methods 0.000 description 2
- 238000002360 preparation method Methods 0.000 description 2
- 210000001367 artery Anatomy 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 210000000988 bone and bone Anatomy 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000005286 illumination Methods 0.000 description 1
- 238000003709 image segmentation Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000000877 morphologic effect Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 238000000638 solvent extraction Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
Classifications
-
- 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
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Information and communication technology [ICT] specially adapted for implementation of business processes of specific business sectors, e.g. utilities or tourism
- G06Q50/01—Social networking
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/26—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/46—Descriptors for shape, contour or point-related descriptors, e.g. scale invariant feature transform [SIFT] or bags of words [BoW]; Salient regional features
- G06V10/462—Salient features, e.g. scale invariant feature transforms [SIFT]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/20—Scenes; Scene-specific elements in augmented reality scenes
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02P—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN THE PRODUCTION OR PROCESSING OF GOODS
- Y02P90/00—Enabling technologies with a potential contribution to greenhouse gas [GHG] emissions mitigation
- Y02P90/30—Computing systems specially adapted for manufacturing
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Computing Systems (AREA)
- Business, Economics & Management (AREA)
- Health & Medical Sciences (AREA)
- General Health & Medical Sciences (AREA)
- Marketing (AREA)
- Biophysics (AREA)
- Strategic Management (AREA)
- General Business, Economics & Management (AREA)
- Primary Health Care (AREA)
- Human Resources & Organizations (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Tourism & Hospitality (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Molecular Biology (AREA)
- General Engineering & Computer Science (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Economics (AREA)
- Image Analysis (AREA)
Abstract
The present invention relates to a kind of social platform matching process and system based on plants identification;The method of the present invention includes: to obtain the photographing information of the leaf image and the image of registered users shooting in social platform;It is handled for leaf image, obtains blade segmented image and vein image;Obtain foundation characteristic, wave character, framework characteristic and information of shooting angles;Foundation characteristic, wave character and framework characteristic are input to Classification Neural trained in advance, obtain blade styles information;At least one user is obtained as matching result using the blade styles information of user to be matched and whole users, user's registration information, information of shooting angles photographing information and matching neural network trained in advance;The method of the present invention segmentation result close to real blade shape, noise is small, segmentation precision is high, and category identification result is accurate, and furthermore social platform user matching, which combines category identification result and user information, has preferable practicability.
Description
Technical field
The invention belongs to social technical field more particularly to a kind of social platform matching process based on plants identification and it is
System.
Background technique
The partitioning algorithm of leaf image carries out figure using super green feature (2G-R-B) under existing research complex background
It is more sensitive for the shade performance in the reflection of blade face caused by illumination and figure as segmentation, it is highly prone to the influence of light, only
In background, single and colour-difference is just ideal away from the extraction effect for carrying out area-of-interest in biggish situation;
It is combined pair firstly, the identification of existing plant leaf blade converts algorithm with bottom cap using super green algorithm (hereinafter referred to as EXG)
The leaf blade photo of complex background carries out image segmentation;As shown in fig. 6, this method includes three steps altogether: (1) to acquisition
Blade original image carries out super green algorithm and bottom cap conversion process respectively and obtains binary picture using OTSU maximum between-cluster variance algorithm
Picture.Wherein the background that G channel components in RGB channel and other two components differ greatly is removed using EXG algorithm, and it is right
It is removed in G channel components and R, the lesser situation of channel B difference using the transformation of bottom cap, meanwhile, bottom cap conversion process corrects light
According to non-uniform situation, relatively clear blade edge and vein is obtained;(2) using XOR operation obtain having obvious vein and
The binaryzation leaf image at edge.During XOR operation, without information such as veins in super green processing, and exist after the transformation of bottom cap,
It more can completely retain the image details such as sawtooth and the vein of prophyll picture;(3) result after XOR operation is carried out
Morphological scale-space and refinement segmentation.The morphology basic operations such as operation, and benefit can be opened and closed after obtaining complete edge
It is split to obtain goal tree impeller exterior feature with watershed algorithm, algorithm of region growing etc., carries out dot product with leaf original image matrix
After final segmentation result can be obtained;Wherein when super green arithmetic result and bottom cap transformation arithmetic result are carried out XOR operation, pole
Easily due to the presence of vein that original blade is separated, the segmentation result of parts of images is accordingly changed into half leaf, causes
The defect of area-of-interest is not particularly suited for most of blades, and not only the degree of automation is not high, also will cause the feelings of over-segmentation
Condition.
Secondly, as shown in fig. 7, being matched at present for strange user in user's matching, generally using the matched side of personality
Method, such as personality test is carried out, user is matched according to lattice test result and restrictive condition;It can be seen that existing strange use
Matching process classification in family is simple, lacks learning functionality or learning ability is poor, and personalized strange user can not be carried out to user
Match;When the matching demand of user changes, can not be matched accordingly in time;Not by blade information and strange user
Match contacts are got up.
Summary of the invention
(1) technical problems to be solved
Prior art blade segmentation effect is poor, category identification accuracy rate is low in order to solve, and in user's matching not according to
Floristics information cooperates the technical issues of object, on the one hand, the present invention provides a kind of social platform based on plants identification
Method of completing the square, on the other hand, the present invention provides a kind of social platform matching system based on plants identification.
(2) technical solution
In order to achieve the above object, the main technical schemes of the method for the present invention include:
S1, the photographing information for obtaining the leaf image and the image of registered users shooting in social platform;
S2, it is handled for the leaf image, obtains blade segmented image and vein image;
S3, feature extraction is carried out for the blade segmented image and the vein image, obtains the blade segmentation figure
The framework characteristic of the foundation characteristic and wave character of picture and the vein image, and according to foundation characteristic acquisition
The information of shooting angles of leaf image;
S4, the foundation characteristic, wave character and framework characteristic are input to Classification Neural trained in advance, obtained
Blade styles information;
S5, by the blade styles information, user's registration information, the shooting angle of user to be matched and whole user
Information and photographing information are input to matching neural network trained in advance after being standardized reason, obtain user to be matched and whole
The matching degree of user, and be that user to be matched recommends at least one user as matching result according to the matching degree;
The photographing information includes shooting time and the shooting location of leaf image, and the user's registration information includes
Personality, gender and the age of user.
Optionally, before further including step S1 further include:
S0, the foundation characteristic using obtaining in advance, wave character and framework characteristic sample data are as point constructed in advance
The input of Connectionist model, using the floristics marked in advance as the defeated of the Classification Neural model constructed in advance
Out, Classification Neural trained in advance is obtained.
Optionally, in step s 2, it obtains blade segmented image and vein image includes:
S21, super green algorithm process and HSV algorithm are utilized respectively for the leaf image, obtain super green image and
HSV image;
S22, growth image is obtained using algorithm of region growing for the super green image;
S23, Threshold segmentation acquisition Threshold segmentation figure is carried out for the super green image, and compares the face of the growth image
The long-pending area with area-of-interest in the Threshold segmentation figure;
If the area of the growth image is less than the half of the pixel number of the growth image or is greater than the threshold value
In segmentation figure at two times of the area of area-of-interest;Then adjust threshold value, region growing step-length and seed point, and return step S22;
Otherwise, image will currently be grown as the blade segmented image;
S24, the blade segmented image and the HSV image are successively carried out to dot product, gray processing and rotation acquisition HSV ash
Image is spent, and carries out edge extracting as the vein image for the HSV gray level image.
Optionally, obtaining wave character in step s3 includes:
S31, original waveform is obtained according to the distance of point to the blade center of gravity of blade edge in the blade segmented image;
S32, one of gaussian filtering, curve matching or wavelet transform process, acquisition are utilized for the original waveform
Shaped wave;
S33, the original waveform is subtracted to the shaped wave, obtains blade edge intelligence wave;
S34, the shaped wave and the blade edge intelligence wave are carried out to 64 equal parts respectively and by 128 numbers of acquisition
According to as the wave character.
Optionally, include: in step also S3
By the skeleton bifurcated number of the vein image, Skeleton pixel points and the subduplicate ratio of minimum circumscribed rectangle area
As the framework characteristic.
Optionally, step S3 also in include:
Various features are extracted as the foundation characteristic for the blade segmented image;
The foundation characteristic and area-of-interest are had to the elliptical long axis and x-axis of identical standard second-order moment around mean
The angle of cut is as the information of shooting angles.
Optionally, in step s 5, include: to the standardization of the input of the matching neural network of training in advance
The age gap section of the user is changed into 0-1;
The gender of the user is indicated with 0.1 and 1, Xiang Tongwei 1 is not all 0.1;
It is 1 by of a sort plant label, is 0.5 by the plant label of same category, the plant label by same section is
0.25, it is 0.125 by same purpose plant label, is 0.0625 by the plant label of same guiding principle, remaining plants mark
It is 0;
Utilize following formula one to three difference acquisition time difference T, the distance difference D and shooting angle difference of shooting location
θ;
Formula one:
Formula two:
Formula three:
Wherein, t is two users' picture shooting time difference, and t unit enables T=0 if T < 0 for hour;D is two users' figure
Piece shooting distance difference, d unit enable D=0 if D < 0 for km;α is to have identical standard second-order central with area-of-interest
The elliptical long axis of square and the angle of cut of x-axis, α unit be degree, section be α ∈ (- 90,90].
Optionally, the matching neural network trained in advance include: input layer, the first hidden layer, the second hidden layer and
Output layer, and the matching neural network trained in advance is trained using incremental learning method;
The input layer input is the data of 0-1;
First hidden layer is convolutional layer, and there are four the filters of 1*4 to carry out the convolution operation that step-length is 1, after
By a nonlinear activation layer, and ReLU function is as activation primitive;
Second hidden layer is convolutional layer, and there are four the filters of 1*4 to carry out the convolution operation that step-length is 1, after
By a nonlinear activation layer, and ReLU function is as activation primitive;
The output layer be convolutional layer, there are four 1*4 filter carry out step-length be 1 convolution operation, after pass through
One nonlinear activation layer and, tanh function is as activation primitive.
The main technical schemes of present system include:
Acquiring unit, storage unit, image processing unit, feature extraction unit and subscriber matching unit;
The acquiring unit is used to obtain the registration information of user and the leaf image and the image of user's shooting
Photographing information;
Image procossing is applied alone in generating blade segmented image and vein image according to the leaf image, and according to
The foundation characteristic obtains the information of shooting angles of the leaf image;
The feature extraction unit is used to generate foundation characteristic and wave character, Yi Jiyong according to the blade segmented image
According to vein image generation framework characteristic;
Subscriber matching unit is used to be that user to be matched recommends at least one user's conduct according to the demand of user to be matched
Matching result;
The storage unit is used to store the registration information of the user, the leaf image that the user shoots and is somebody's turn to do
The photographing information of image, the information of shooting angles, the foundation characteristic, the wave character and the framework characteristic;
Wherein, the user's registration information includes the personality, gender and age of user.
Optionally, the feature extraction unit is also used to according to the point of blade edge in the blade segmented image to blade
The distance of center of gravity generates original waveform, generates shaped wave according to the original waveform, according to the original waveform and the shape
Shape waveform generates blade edge intelligence wave, and it is special to generate the waveform according to the shaped wave and the blade edge intelligence wave
Sign.(3) beneficial effect
The beneficial effect of the method for the present invention is: firstly, can be obtained in reasonable segmentation times it is ideal, more connect
The segmentation result of nearly real blade shape, while final result does not have the appearance of big noise, greatly improves the precision of segmentation,
Be characterized extraction and image classification and etc. done good preparation;Secondly, social platform user matching combines category identification
As a result there is preferable practicability with user information.
The advantageous effects of present system are: improving the precision of blade segmentation and classification, form more accurate
And the higher plant leaf blade segmentation of the degree of automation, feature extraction, classification process and the matched process of user.
Detailed description of the invention
Fig. 1 is a kind of process signal for social platform matching process based on plants identification that one embodiment of the invention provides
Figure;
Fig. 2 be one embodiment of the invention provide utilize leaf image to generate blade segmented image and vein image
Schematic diagram;
Fig. 3 a is the schematic diagram that foundation characteristic is generated using blade segmented image that one embodiment of the invention provides;
Fig. 3 b is the schematic diagram that framework characteristic is generated using vein image that one embodiment of the invention provides;
Fig. 3 c is the schematic diagram that wave character is generated using blade segmented image that one embodiment of the invention provides;
Fig. 4 is the user's matching figure inventing an embodiment and providing;
Fig. 5 is the structural schematic diagram for inventing the matching neural network that an embodiment provides;
Fig. 6 is traditional plant leaf identification method schematic diagram;
Fig. 7 is traditional strange user matching method flow diagram
Specific embodiment
In order to preferably explain the present invention, in order to understand, with reference to the accompanying drawing, by specific embodiment, to this hair
It is bright to be described in detail.
Embodiment one
As shown in Figure 1, present embodiments provide a kind of social platform matching process based on plants identification, specifically include with
Lower step:
S0, the foundation characteristic using obtaining in advance, wave character and framework characteristic sample data are as point constructed in advance
The input of Connectionist model, using the floristics marked in advance as the defeated of the Classification Neural model constructed in advance
Out, Classification Neural trained in advance is obtained, the type of Classification Neural plant for identification;
S1, the photographing information for obtaining the leaf image and the image of registered users shooting in social platform;
S2, it is handled for leaf image a, obtains blade segmented image d and vein image g;As shown in Fig. 2,
Specific steps include:
S21, it is utilized respectively super green algorithm process and HSV algorithm for leaf image a, obtains super green image b and HSV
Image e;For example, HSV image can be substituted with HIS, YUV or YcbCr image in the present embodiment, and the present embodiment only uses HSV
Image illustrates;
S22, growth image c is obtained using algorithm of region growing for super green image b;
S23, Threshold segmentation acquisition Threshold segmentation figure is carried out for super green image b, and compare the area and threshold of growth image c
It is worth the area of area-of-interest in segmentation figure;
If the area of growth image c is less than the half of the pixel number of growth image or greater than feeling emerging in Threshold segmentation figure
At two times of the area in interesting region;Then adjust threshold value, region growing step-length and seed point, and return step S22;
Otherwise, image c will currently be grown as blade segmented image d;
S24, blade segmented image c and HSV image are successively carried out to dot product, gray processing and rotation acquisition HSV gray level image
F, and edge extracting is carried out as vein image g for HSV gray level image f.
Specifically, the feature extraction of vein information is introduced, feature extraction combines HSV image with super green image, energy
It is enough that segmentation result ideal, closer to real blade shape, while final result are obtained in reasonable segmentation times
Do not have the appearance of big noise, greatly improve the precision of segmentation, be characterized extraction and image classification and etc. done it is good
Preparation;
S3, feature extraction is carried out for blade segmented image d and vein image g, the basis for obtaining blade segmented image d is special
It seeks peace wave character and the framework characteristic of vein image g, and obtains the shooting angle of leaf image according to foundation characteristic
Information;
As best shown in figures 3 a and 3b, various features are extracted for blade segmented image d and are used as basic feature, for example root,
21 kinds of features are extracted according to blade segmented image d, are specifically included: total number of pixels in image each region, comprising corresponding region
Minimum rectangle, each region mass center (center of gravity), with region have identical standard second-order moment around mean elliptical long axis length (as
Under plain meaning), with region there is the elliptical minor axis length (under pixel meaning) of identical standard second-order moment around mean, have with region
The elliptical eccentricity (can be used as feature) of identical standard second-order moment around mean has the ellipse of identical standard second-order moment around mean with region
The angle of cut (degree) of round long axis and x-axis, the logic matrix with certain region with same size have same size with certain region
Filling logic matrix, the minimal convex polygon comprising certain region, draws above-mentioned zone at the on number of pixels in filling region image
Minimal convex polygon, the on number of pixels in filling region convex polygon image, a topological novariable in geometry topology
Amount --- Euler's numbers have to local extremum point, with region diameter of a circle of the same area, simultaneously in region and it is minimum from all directions
The rope of pixel ratio in convex polygon while the pixel ratio in region and its minimum boundary rectangle, storage region pixel
The perimeter for drawing subscript, storing the corresponding pixel coordinate of above-mentioned index, image each region forntier region;By the bone of vein image g
Frame bifurcated number, Skeleton pixel points and the subduplicate ratio of minimum circumscribed rectangle area are as framework characteristic.
As shown in Figure 3c, in step s3 obtain wave character the following steps are included:
S31, original waveform is obtained according to the distance of point to the blade center of gravity of blade edge in blade segmented image d;
S32, one of gaussian filtering, curve matching or wavelet transform process, acquisition are utilized for the original waveform
Shaped wave;
S33, original waveform is subtracted to shaped wave, obtains blade edge intelligence wave;
S34, shaped wave and blade edge intelligence wave are carried out to 64 equal parts respectively and using 128 data of acquisition as wave
Shape feature.
By the elliptical long axis with area-of-interest with identical standard second-order moment around mean of foundation characteristic and the angle of cut of x-axis
As information of shooting angles.
S4, foundation characteristic, wave character and framework characteristic are input to Classification Neural trained in advance, obtain blade
Information;For example the present embodiment also can be used SVM method and obtain blade styles information;This method uses completely new feature
The bifurcated situation of leaf vein is described, floristics identification is more acurrate.;
S5, by blade styles information, user's registration information, information of shooting angles and the bat of user to be matched and whole user
It takes the photograph after information is standardized reason and is input to matching neural network trained in advance, obtain of user to be matched and whole users
It is that user to be matched recommends at least one user as matching result with degree, and according to matching degree;
As shown in figure 4, photographing information includes shooting time and the shooting location of leaf image, user's registration information packet
Include the personality, gender and age of user;For example, in user's matching step, social platform user matching combines type knowledge
Other result and user information have preferable practicability;
Specifically for example, include: to the standardization of matching neural network input data in the present embodiment
The age gap section of user is changed into 0-1;
The gender of the user is indicated with 0.1 and 1, Xiang Tongwei 1 is not all 0.1;It is 1 by of a sort plant label,
It is 0.5 by the plant label of same category, is 0.25 by the plant label of same section, is 0.125 by same purpose plant label, it will
The plant label of same guiding principle is 0.0625, and remaining plants, which mark, is;
Utilize following formula 1 to 3 difference acquisition time difference T, the distance difference D and shooting angle difference θ of shooting location;
Formula 1:
Formula 2:
Formula 3:
Wherein, t is two users' picture shooting time difference, and t unit enables T=0 if T < 0 for hour;D is two users' figure
Piece shooting distance difference, d unit enable D=0 if D < 0 for km;α is to have identical standard second-order central with area-of-interest
The elliptical long axis of square and the angle of cut of x-axis, α unit be degree, section be α ∈ (- 90,90];
For example, the personality information in the present embodiment in user's registration information is tested by personality and is obtained, can be in user
There is provided nine type Personality tests when registering as user, if user's personality be it is of the same race if output be 1, if a different or at least user
Personality test is not carried out then exports 0.
As shown in figure 5, each user data is stored by server, but do not matched;For example, work as user
1 wants to get in touch with strange user, needs to amount to the matching degree for calculating user 1 and user 2 or user 3, finally by matching
Degree sorts other users, is put into the recommendation list of user 1;
The matching neural network trained in advance of the present embodiment includes: input layer, the first hidden layer, the second hidden layer and defeated
Layer out, and matching neural network trained in advance is trained using incremental learning method;
Input layer input is the data of 0-1;
First hidden layer be convolutional layer, there are four 1*4 filter carry out step-length be 1 convolution operation, after pass through
One nonlinear activation layer, and ReLU function is as activation primitive;
Second hidden layer be convolutional layer, there are four 1*4 filter carry out step-length be 1 convolution operation, after pass through
One nonlinear activation layer, and ReLU function is as activation primitive;
Output layer is convolutional layer, there are four 1*4 filter carry out step-length be 1 convolution operation, after pass through one
Nonlinear activation layer and, tanh function is as activation primitive.
Embodiment two
A kind of social platform matching system based on plants identification is present embodiments provided, is specifically included:
Acquiring unit, storage unit, image processing unit, feature extraction unit and subscriber matching unit;
Acquiring unit is used to obtain the bat of the registration information of user and the leaf image and the image of user's shooting
Take the photograph information;
Image procossing is applied alone in generating blade segmented image and vein image according to leaf image, and according to basis
The information of shooting angles of feature acquisition leaf image;
Feature extraction unit is used to generate foundation characteristic and wave character according to blade segmented image, and for according to leaf
Arteries and veins image generates framework characteristic;
Subscriber matching unit is used to be that user to be matched recommends at least one user's conduct according to the demand of user to be matched
Matching result;
Storage unit is used to store the shooting letter of the registration information of user, the leaf image of user's shooting and the image
Breath, information of shooting angles, foundation characteristic, wave character and framework characteristic;
Wherein, user's registration information includes the personality, gender and age of user.
Preferably, feature extraction unit be also used to according to the point of blade edge in blade segmented image to blade center of gravity away from
From generating original waveform, generating shaped wave according to original waveform, according to original waveform and shaped wave, blade edge letter is generated
Wave is ceased, wave character is generated according to shaped wave and blade edge intelligence wave.
Finally, it should be noted that above-described embodiments are merely to illustrate the technical scheme, rather than to it
Limitation;Although the present invention is described in detail referring to the foregoing embodiments, those skilled in the art should understand that:
It can still modify to technical solution documented by previous embodiment, or to part of or all technical features into
Row equivalent replacement;And these modifications or substitutions, it does not separate the essence of the corresponding technical solution various embodiments of the present invention technical side
The range of case.
Claims (10)
1. a kind of social platform matching process based on plants identification characterized by comprising
S1, the photographing information for obtaining the leaf image and the image of registered users shooting in social platform;
S2, it is handled for the leaf image, obtains blade segmented image and vein image;
S3, feature extraction is carried out for the blade segmented image and the vein image, obtains the blade segmented image
The framework characteristic of foundation characteristic and wave character and the vein image, and the plant is obtained according to the foundation characteristic
The information of shooting angles of leaf image;
S4, the foundation characteristic, wave character and framework characteristic are input to Classification Neural trained in advance, obtain blade
Information;
S5, by the blade styles information, user's registration information, the information of shooting angles of user to be matched and whole user
It is standardized after reason with photographing information and is input to matching neural network trained in advance, obtain user to be matched and whole users
Matching degree, and according to the matching degree be user to be matched recommend at least one user as matching result;
The photographing information includes shooting time and the shooting location of leaf image, and the user's registration information includes user
Personality, gender and age.
2. the method as described in claim 1, which is characterized in that before further including step S1 further include:
S0, the foundation characteristic using obtaining in advance, wave character and framework characteristic sample data are as the classification mind constructed in advance
Input through network model is obtained using the floristics marked in advance as the output of the Classification Neural model constructed in advance
Take Classification Neural trained in advance.
3. method according to claim 2, which is characterized in that in step s 2, obtain blade segmented image and vein image
Include:
S21, it is utilized respectively super green algorithm process and HSV algorithm for the leaf image, obtains super green image and HSV figure
Picture;
S22, growth image is obtained using algorithm of region growing for the super green image;
S23, carry out Threshold segmentation for the super green image and obtain Threshold segmentation figure, and compare the area of the growth image with
The area of area-of-interest in the Threshold segmentation figure;
If the area of the growth image is less than the half of the pixel number of the growth image or is greater than the Threshold segmentation
In figure at two times of the area of area-of-interest;Then adjust threshold value, region growing step-length and seed point, and return step S22;
Otherwise, image will currently be grown as the blade segmented image;
S24, the blade segmented image and the HSV image are successively carried out to dot product, gray processing and rotation acquisition HSV grayscale image
Picture, and edge extracting is carried out as the vein image for the HSV gray level image.
4. method as claimed in claim 3, which is characterized in that obtaining wave character in step s3 includes:
S31, original waveform is obtained according to the distance of point to the blade center of gravity of blade edge in the blade segmented image;
S32, one of gaussian filtering, curve matching or wavelet transform process, acquisition shape are utilized for the original waveform
Waveform;
S33, the original waveform is subtracted to the shaped wave, obtains blade edge intelligence wave;
S34, the shaped wave and the blade edge intelligence wave are carried out to 64 equal parts respectively and make 128 data of acquisition
For the wave character.
5. method as claimed in claim 4, which is characterized in that include: in step also S3
Using the skeleton bifurcated number of the vein image, Skeleton pixel points and the subduplicate ratio of minimum circumscribed rectangle area as
The framework characteristic.
6. method as claimed in claim 5, which is characterized in that step S3 also in include:
Various features are extracted as the foundation characteristic for the blade segmented image;
By the elliptical long axis with area-of-interest with identical standard second-order moment around mean of the foundation characteristic and the angle of cut of x-axis
As the information of shooting angles.
7. method as claimed in claim 6, which is characterized in that in step s 5, to the matching neural network of training in advance
The standardization of input includes:
The age gap section of the user is changed into 0-1;
The gender of the user is indicated with 0.1 and 1, Xiang Tongwei 1 is not all 0.1;
It is 1 by of a sort plant label, is 0.5 by the plant label of same category, is 0.25 by the plant label of same section, it will
Same purpose plant label is 0.125, is 0.0625 by the plant label of same guiding principle, remaining plants, which mark, is;
Utilize following formula one to three difference acquisition time difference T, the distance difference D and shooting angle difference θ of shooting location;
Formula one:
Formula two:
Formula three:
Wherein, t is two users' picture shooting time difference, and t unit enables T=0 if T < 0 for hour;D is the bat of two users' picture
Photographic range difference, d unit enable D=0 if D < 0 for km;α is to have identical standard second-order moment around mean with area-of-interest
The angle of cut of elliptical long axis and x-axis, α unit be degree, section be α ∈ (- 90,90].
8. the method according to claim 1 to 7, which is characterized in that the matching neural network packet trained in advance
It includes: input layer, the first hidden layer, the second hidden layer and output layer, and the matching neural network trained in advance uses increment
Learning method is trained;
The input layer input is the data of 0-1;
First hidden layer be convolutional layer, there are four 1*4 filter carry out step-length be 1 convolution operation, after pass through
One nonlinear activation layer, and ReLU function is as activation primitive;
Second hidden layer be convolutional layer, there are four 1*4 filter carry out step-length be 1 convolution operation, after pass through
One nonlinear activation layer, and ReLU function is as activation primitive;
The output layer be convolutional layer, there are four 1*4 filter carry out step-length be 1 convolution operation, after pass through one
Nonlinear activation layer and, tanh function is as activation primitive.
9. a kind of social platform matching system based on plants identification characterized by comprising
Acquiring unit, storage unit, image processing unit, feature extraction unit and subscriber matching unit;
The acquiring unit is used to obtain the bat of the registration information of user and the leaf image and the image of user's shooting
Take the photograph information;
Image procossing is applied alone in generating blade segmented image and vein image according to the leaf image, and according to described
Foundation characteristic obtains the information of shooting angles of the leaf image;
The feature extraction unit is used to generate foundation characteristic and wave character according to the blade segmented image, and is used for root
Framework characteristic is generated according to the vein image;
Subscriber matching unit is used to be that user to be matched recommends at least one user as matching according to the demand of user to be matched
As a result;
The storage unit is used to store the leaf image and the image of the registration information of the user, user shooting
Photographing information, the information of shooting angles, the foundation characteristic, the wave character and the framework characteristic;
Wherein, the user's registration information includes the personality, gender and age of user.
10. system as claimed in claim 9, which is characterized in that the feature extraction unit is also used to according to the blade point
The distance for cutting point to the blade center of gravity of blade edge in image generates original waveform, generates shape wave according to the original waveform
Shape, according to the original waveform and the shaped wave, blade edge intelligence wave is generated, according to the shaped wave and the leaf
Piece marginal information wave generates the wave character.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910286271.0A CN110008912B (en) | 2019-04-10 | 2019-04-10 | Social platform matching method and system based on plant identification |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910286271.0A CN110008912B (en) | 2019-04-10 | 2019-04-10 | Social platform matching method and system based on plant identification |
Publications (2)
Publication Number | Publication Date |
---|---|
CN110008912A true CN110008912A (en) | 2019-07-12 |
CN110008912B CN110008912B (en) | 2023-04-18 |
Family
ID=67170933
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910286271.0A Active CN110008912B (en) | 2019-04-10 | 2019-04-10 | Social platform matching method and system based on plant identification |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110008912B (en) |
Cited By (4)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110533605A (en) * | 2019-07-26 | 2019-12-03 | 遵义师范学院 | A kind of accurate noise scaling method |
CN111833367A (en) * | 2020-06-24 | 2020-10-27 | 中国第一汽车股份有限公司 | Image processing method and device, vehicle and storage medium |
CN111860330A (en) * | 2020-07-21 | 2020-10-30 | 陕西工业职业技术学院 | Apple leaf disease identification method based on multi-feature fusion and convolutional neural network |
CN112542163A (en) * | 2019-09-04 | 2021-03-23 | 百度在线网络技术(北京)有限公司 | Intelligent voice interaction method, equipment and storage medium |
Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103984775A (en) * | 2014-06-05 | 2014-08-13 | 网易(杭州)网络有限公司 | Friend recommending method and equipment |
CN104281650A (en) * | 2014-09-15 | 2015-01-14 | 南京锐角信息科技有限公司 | Friend search recommendation method and friend search recommendation system based on interest analysis |
CN106599925A (en) * | 2016-12-19 | 2017-04-26 | 广东技术师范学院 | Plant leaf identification system and method based on deep learning |
CN108182228A (en) * | 2017-12-27 | 2018-06-19 | 北京奇虎科技有限公司 | User social contact method, device and the computing device realized using augmented reality |
US20180246899A1 (en) * | 2017-02-28 | 2018-08-30 | Laserlike Inc. | Generate an index for enhanced search based on user interests |
CN109241454A (en) * | 2018-07-18 | 2019-01-18 | 广东工业大学 | A kind of point of interest recommended method merging social networks and picture material |
US20190034976A1 (en) * | 2017-07-26 | 2019-01-31 | Jehan Hamedi | Systems and Methods for Automating Content Design Transformations Based on User Preference and Activity Data |
CN109325529A (en) * | 2018-09-06 | 2019-02-12 | 安徽大学 | A kind of application of sketch recognition method and this method in commodity retrieval |
CN109359675A (en) * | 2018-09-28 | 2019-02-19 | 腾讯科技(武汉)有限公司 | Image processing method and equipment |
-
2019
- 2019-04-10 CN CN201910286271.0A patent/CN110008912B/en active Active
Patent Citations (9)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN103984775A (en) * | 2014-06-05 | 2014-08-13 | 网易(杭州)网络有限公司 | Friend recommending method and equipment |
CN104281650A (en) * | 2014-09-15 | 2015-01-14 | 南京锐角信息科技有限公司 | Friend search recommendation method and friend search recommendation system based on interest analysis |
CN106599925A (en) * | 2016-12-19 | 2017-04-26 | 广东技术师范学院 | Plant leaf identification system and method based on deep learning |
US20180246899A1 (en) * | 2017-02-28 | 2018-08-30 | Laserlike Inc. | Generate an index for enhanced search based on user interests |
US20190034976A1 (en) * | 2017-07-26 | 2019-01-31 | Jehan Hamedi | Systems and Methods for Automating Content Design Transformations Based on User Preference and Activity Data |
CN108182228A (en) * | 2017-12-27 | 2018-06-19 | 北京奇虎科技有限公司 | User social contact method, device and the computing device realized using augmented reality |
CN109241454A (en) * | 2018-07-18 | 2019-01-18 | 广东工业大学 | A kind of point of interest recommended method merging social networks and picture material |
CN109325529A (en) * | 2018-09-06 | 2019-02-12 | 安徽大学 | A kind of application of sketch recognition method and this method in commodity retrieval |
CN109359675A (en) * | 2018-09-28 | 2019-02-19 | 腾讯科技(武汉)有限公司 | Image processing method and equipment |
Non-Patent Citations (2)
Title |
---|
B.SATHYA BAMA.ET AL: ""CONTENT BASED LEAF IMAGE RETRIEVAL (CBLIR) USING SHAPE, COLOR AND TEXTURE FEATURES"" * |
翟传敏等: ""基于形状上下文特征的植物叶片图像匹配方法"" * |
Cited By (7)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN110533605A (en) * | 2019-07-26 | 2019-12-03 | 遵义师范学院 | A kind of accurate noise scaling method |
CN110533605B (en) * | 2019-07-26 | 2023-06-02 | 遵义师范学院 | Accurate noise point calibration method |
CN112542163A (en) * | 2019-09-04 | 2021-03-23 | 百度在线网络技术(北京)有限公司 | Intelligent voice interaction method, equipment and storage medium |
CN112542163B (en) * | 2019-09-04 | 2023-10-27 | 百度在线网络技术(北京)有限公司 | Intelligent voice interaction method, device and storage medium |
CN111833367A (en) * | 2020-06-24 | 2020-10-27 | 中国第一汽车股份有限公司 | Image processing method and device, vehicle and storage medium |
CN111860330A (en) * | 2020-07-21 | 2020-10-30 | 陕西工业职业技术学院 | Apple leaf disease identification method based on multi-feature fusion and convolutional neural network |
CN111860330B (en) * | 2020-07-21 | 2023-08-11 | 陕西工业职业技术学院 | Apple leaf disease identification method based on multi-feature fusion and convolutional neural network |
Also Published As
Publication number | Publication date |
---|---|
CN110008912B (en) | 2023-04-18 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110008912A (en) | A kind of social platform matching process and system based on plants identification | |
CN104050471B (en) | Natural scene character detection method and system | |
CN112950661B (en) | Attention-based generation method for generating network face cartoon | |
CN108108699A (en) | Merge deep neural network model and the human motion recognition method of binary system Hash | |
CN108416353B (en) | Method for quickly segmenting rice ears in field based on deep full convolution neural network | |
CN108090906B (en) | Cervical image processing method and device based on region nomination | |
CN109815826A (en) | The generation method and device of face character model | |
CN113591766B (en) | Multi-source remote sensing tree species identification method for unmanned aerial vehicle | |
CN104143079A (en) | Method and system for face attribute recognition | |
CN108717524A (en) | It is a kind of based on double gesture recognition systems and method for taking the photograph mobile phone and artificial intelligence system | |
CN110287882A (en) | A kind of big chrysanthemum kind image-recognizing method based on deep learning | |
CN108038504A (en) | A kind of method for parsing property ownership certificate photo content | |
CN110599463A (en) | Tongue image detection and positioning algorithm based on lightweight cascade neural network | |
CN109034012A (en) | First person gesture identification method based on dynamic image and video sequence | |
Sun et al. | An improved CNN-based apple appearance quality classification method with small samples | |
CN109977899A (en) | A kind of training, reasoning and the method and system for increasing New raxa of article identification | |
Zeppelzauer et al. | Interactive segmentation of rock-art in high-resolution 3D reconstructions | |
CN109190571A (en) | A kind of detection recognition method and its device of grazing sheep feeding typical plant type | |
CN114758132A (en) | Fruit tree pest and disease identification method and system based on convolutional neural network | |
CN107909073A (en) | Multidimensional local binary patterns and the hand-written music score spectral line delet method of machine learning | |
CN113239932A (en) | Tesseract-OCR-based identification method for air velocity scale in PFD (flight display device) | |
CN117437691A (en) | Real-time multi-person abnormal behavior identification method and system based on lightweight network | |
CN116957610A (en) | Agricultural product whole-flow information traceability input management method, system and medium | |
CN109165551B (en) | Expression recognition method for adaptively weighting and fusing significance structure tensor and LBP characteristics | |
CN106570459A (en) | Face image processing 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 | ||
GR01 | Patent grant | ||
GR01 | Patent grant |