CN110135517A - For obtaining the method and device of vehicle similarity - Google Patents
For obtaining the method and device of vehicle similarity Download PDFInfo
- Publication number
- CN110135517A CN110135517A CN201910441616.5A CN201910441616A CN110135517A CN 110135517 A CN110135517 A CN 110135517A CN 201910441616 A CN201910441616 A CN 201910441616A CN 110135517 A CN110135517 A CN 110135517A
- Authority
- CN
- China
- Prior art keywords
- vehicle
- sample
- similarity
- information
- image
- 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
Classifications
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/22—Matching criteria, e.g. proximity measures
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/74—Image or video pattern matching; Proximity measures in feature spaces
- G06V10/75—Organisation of the matching processes, e.g. simultaneous or sequential comparisons of image or video features; Coarse-fine approaches, e.g. multi-scale approaches; using context analysis; Selection of dictionaries
- G06V10/758—Involving statistics of pixels or of feature values, e.g. histogram matching
Abstract
Embodiment of the disclosure discloses the method and device for obtaining vehicle similarity.One specific embodiment of this method includes: to choose first area and second area from the first image to be processed and the second image to be processed;The first vehicle characteristic information is obtained from first area and obtains the second vehicle characteristic information from second area;According to the similarity of the first vehicle characteristic information of first area and the second vehicle characteristic information of corresponding second area, the first vehicle similarity information is obtained;The third vehicle characteristic information and the 4th vehicle characteristic information of the first image to be processed and the second image to be processed are obtained respectively;Third vehicle characteristic information and the 4th vehicle characteristic information are imported into vehicle identification model, obtain the second vehicle similarity information;According to the first vehicle similarity information and the second vehicle similarity information, the similarity between the corresponding vehicle of the first vehicle image and the corresponding vehicle of the second vehicle image is calculated.This embodiment improves the accuracys to vehicle identification.
Description
Technical field
Embodiment of the disclosure is related to technical field of image processing, and in particular to for obtain vehicle similarity method and
Device.
Background technique
With the rapid development of modern social economy, effective urban traffic management is living in the economy of people, society
Importance in dynamic is increasingly significant.Therefore, further investigation, which solves urban transport problems, just particularly important realistic meaning.Its
In, effective identification to vehicle is the core of modern intelligent transportation research.
Summary of the invention
Embodiment of the disclosure proposes the method and device for obtaining vehicle similarity.
In a first aspect, embodiment of the disclosure provides a kind of method for obtaining vehicle similarity, this method comprises:
Multiple first areas and multiple second areas are chosen from the first image to be processed and the second image to be processed respectively, wherein described
First image to be processed includes the first vehicle image, and the second image to be processed includes the second vehicle image, the multiple firstth area
Vehicle location in vehicle location and the multiple second area in domain in each first area in each second area is corresponding;
From each of the multiple first area first area obtain the first vehicle characteristic information respectively and from the multiple second
Each of region second area obtains the second vehicle characteristic information;It is special according to the first vehicle of each first area
Reference ceases and the similarity of the second vehicle characteristic information of corresponding each second area, the first vehicle similarity of acquisition
Information;Third vehicle characteristic information and the 4th vehicle for obtaining the described first image to be processed and the second image to be processed respectively are special
Reference breath;The third vehicle characteristic information and the 4th vehicle characteristic information are imported into vehicle identification model trained in advance, obtained
To the second vehicle similarity information;According to the first vehicle similarity information and the second vehicle similarity information, calculate
Similarity between the corresponding vehicle of first vehicle image and the corresponding vehicle of the second vehicle image.
In some embodiments, the first vehicle characteristic information of each first area according to and corresponding institute
The similarity of the second vehicle characteristic information of each second area is stated, obtains the first vehicle similarity information, comprising: for more
A first vehicle characteristic information and the first vehicle characteristic information and the second vehicle characteristics letter in multiple second vehicle characteristic informations
Breath, the son calculated between the first vehicle characteristic information and the second vehicle characteristic information of corresponding first vehicle characteristic information are similar
Spend information;Using the sum of multiple sub- similarity informations as the first vehicle similarity information.
In some embodiments, above-mentioned vehicle identification model is obtained by following steps training: obtaining multiple sample vehicles
Image group first sample vehicle characteristics corresponding with sample vehicle image group each in corresponding above-mentioned multiple sample vehicle image groups
Information, the second sample vehicle characteristic information and Sample Similarity information, wherein sample vehicle image group includes comprising vehicle image
First sample vehicle subgraph and the second sample vehicle subgraph comprising vehicle image, first sample vehicle subgraph and
One sample vehicle characteristic information is corresponding, the second sample vehicle subgraph and the second sample vehicle characteristic information, Sample Similarity letter
Breath is for characterizing the similarity between first sample vehicle subgraph and the second sample vehicle subgraph, sample vehicle characteristic information
Include at least one of the following: vehicle wheel profile, vehicle region line;By each sample vehicle of above-mentioned multiple sample vehicle image groups
The first sample vehicle characteristic information of image group and the second sample vehicle characteristic information are as input, by above-mentioned multiple sample vehicles
Above-mentioned Sample Similarity information corresponding to each sample vehicle image group in image group obtains above-mentioned vehicle as output, training
Identification model.
In some embodiments, the first of above-mentioned each sample vehicle image group by above-mentioned multiple sample vehicle image groups
Sample vehicle characteristic information and the second sample vehicle characteristic information, will be every in above-mentioned multiple sample vehicle image groups as input
Above-mentioned Sample Similarity information corresponding to a sample vehicle image group obtains above-mentioned vehicle identification model as output, training,
It include: to execute following training step: by the first sample of each sample vehicle image group in above-mentioned multiple sample vehicle image groups
This vehicle characteristic information and the second sample vehicle characteristic information are sequentially input to initial vehicle identification model, obtain above-mentioned multiple samples
Prediction similarity information corresponding to each sample vehicle image group in this vehicle image group, by above-mentioned multiple sample vehicle figures
Sample corresponding to prediction similarity information and the sample vehicle image group as corresponding to each sample vehicle image group in group
This similarity information is compared, and is obtained the predictablity rate of above-mentioned initial vehicle identification model, is determined above-mentioned predictablity rate
Whether be greater than default accuracy rate threshold value, if more than above-mentioned default accuracy rate threshold value, then using above-mentioned initial vehicle identification model as
The vehicle identification model that training is completed.
In some embodiments, the first of above-mentioned each sample vehicle image group by above-mentioned multiple sample vehicle image groups
Sample vehicle characteristic information and the second sample vehicle characteristic information, will be every in above-mentioned multiple sample vehicle image groups as input
Above-mentioned Sample Similarity information corresponding to a sample vehicle image group obtains above-mentioned vehicle identification model as output, training,
It include: in response to adjusting the parameter of above-mentioned initial vehicle identification model, and continue to execute no more than above-mentioned default accuracy rate threshold value
Above-mentioned training step.
In some embodiments, above-mentioned Sample Similarity information is obtained by following steps: according to sample vehicle image group
The first sample vehicle subgraph and the corresponding first sample vehicle characteristic information of the second sample vehicle subgraph that include and
Second sample vehicle characteristic information determines the second vehicle region of the first vehicle region of sample and sample;Above-mentioned sample is obtained respectively
The first area feature and second area feature of the first vehicle region of sample of the second vehicle region of one vehicle region and sample,
In, provincial characteristics includes at least one of the following: lines assemblage characteristic, region assemblage characteristic;Respectively to above-mentioned first area feature
Fusion Features are carried out with second area feature, obtain first area characteristic value and second area characteristic value;According to above-mentioned firstth area
Difference between characteristic of field value and second area characteristic value calculates Sample Similarity information.
In some embodiments, described to be believed according to the first vehicle similarity information and the second vehicle similarity
Breath calculates the similarity between the corresponding vehicle of first vehicle image and the corresponding vehicle of the second vehicle image, comprising: point
It Wei not the first vehicle similarity information and the second vehicle similarity information the first weight of setting and the second weight;By institute
The first product for stating the first weight and the first vehicle similarity information, second with the second weight and the second vehicle similarity information
Sum between product is as the similarity between vehicle.
Second aspect, embodiment of the disclosure provide a kind of for obtaining the device of vehicle similarity, which includes:
Image-region selecting unit is configured to choose multiple first areas from the first image to be processed and the second image to be processed respectively
With multiple second areas, wherein first image to be processed includes the first vehicle image, and the second image to be processed includes second
Vehicle image, each in the vehicle location and the multiple second area in the multiple first area in each first area
Vehicle location in two regions is corresponding;Fisrt feature information acquisition unit is configured to respectively from the multiple first area
Each first area obtain and the first vehicle characteristic information and obtained from each of the multiple second area second area
Take the second vehicle characteristic information;First vehicle similarity information acquiring unit is configured to according to each described first area
The first vehicle characteristic information and the similarity of the second vehicle characteristic information of corresponding each second area, obtain the
One vehicle similarity information;Second feature information acquisition unit is configured to obtain the described first image to be processed and respectively
The third vehicle characteristic information and the 4th vehicle characteristic information of two images to be processed;Second vehicle similarity information acquiring unit,
It is configured to the third vehicle characteristic information and the 4th vehicle characteristic information importing vehicle identification model trained in advance, obtain
To the second vehicle similarity information;Similarity calculated is configured to according to the first vehicle similarity information and described
Second vehicle similarity information calculates between the corresponding vehicle of first vehicle image and the corresponding vehicle of the second vehicle image
Similarity.
In some embodiments, the first vehicle similarity information acquiring unit includes: that sub- similarity information calculates son
Unit, for the first vehicle characteristic information and second in multiple first vehicle characteristic informations and multiple second vehicle characteristic informations
Vehicle characteristic information, the second vehicle for being configured to calculate the first vehicle characteristic information and corresponding first vehicle characteristic information are special
Sub- similarity information between reference breath;First vehicle similarity information obtains subelement, is configured to multiple sub- similarities
The sum of information is as the first vehicle similarity information.
In some embodiments, above-mentioned apparatus further includes vehicle identification model training unit, is configured to that vehicle is trained to know
Other model, above-mentioned vehicle identification model training unit include: that sample information obtains subelement, are configured to obtain multiple sample vehicles
Image group first sample vehicle corresponding with sample vehicle image group each in corresponding above-mentioned multiple sample vehicle image groups is special
Reference breath, the second sample vehicle characteristic information and Sample Similarity information, wherein sample vehicle image group includes comprising vehicle figure
The first sample vehicle subgraph of picture and the second sample vehicle subgraph comprising vehicle image, first sample vehicle subgraph with
First sample vehicle characteristic information is corresponding, the second sample vehicle subgraph and the second sample vehicle characteristic information, Sample Similarity
Information is used to characterize the similarity between first sample vehicle subgraph and the second sample vehicle subgraph, sample vehicle characteristics letter
Breath includes at least one of the following: vehicle wheel profile, vehicle region line;Vehicle identification model training subelement, being configured to will be upper
State the first sample vehicle characteristic information and the second sample vehicle of each sample vehicle image group of multiple sample vehicle image groups
Characteristic information is as input, by above-mentioned sample corresponding to each sample vehicle image group in above-mentioned multiple sample vehicle image groups
This similarity information obtains above-mentioned vehicle identification model as output, training.
In some embodiments, above-mentioned vehicle identification model training subelement includes: vehicle identification model training module, quilt
Be configured to the first sample vehicle characteristic information of each sample vehicle image group in above-mentioned multiple sample vehicle image groups and
Second sample vehicle characteristic information is sequentially input to initial vehicle identification model, is obtained in above-mentioned multiple sample vehicle image groups
Prediction similarity information corresponding to each sample vehicle image group, by each sample in above-mentioned multiple sample vehicle image groups
Sample Similarity information corresponding to prediction similarity information and the sample vehicle image group corresponding to vehicle image group carries out
Compare, obtain the predictablity rate of above-mentioned initial vehicle identification model, it is default accurate to determine whether above-mentioned predictablity rate is greater than
Rate threshold value then knows the vehicle that above-mentioned initial vehicle identification model is completed as training if more than above-mentioned default accuracy rate threshold value
Other model.
In some embodiments, above-mentioned vehicle identification model training subelement includes: parameter adjustment module, in response to little
It in above-mentioned default accuracy rate threshold value, is configured to adjust the parameter of above-mentioned initial vehicle identification model, and returns to vehicle identification mould
Type training module.
In some embodiments, above-mentioned apparatus further includes Sample Similarity information acquisition unit, is configured to obtain sample
Similarity information, above-mentioned Sample Similarity information acquisition unit include: that sample vehicle region obtains subelement, are configured to basis
The first sample vehicle subgraph and the corresponding first sample of the second sample vehicle subgraph that sample vehicle image group includes
Vehicle characteristic information and the second sample vehicle characteristic information determine the second vehicle region of the first vehicle region of sample and sample;Region
Feature obtains subelement, is configured to obtain the sample the of above-mentioned the first vehicle region of sample and the second vehicle region of sample respectively
The first area feature and second area feature of one vehicle region, wherein provincial characteristics includes at least one of the following: that lines combine
Feature, region assemblage characteristic;Regional characteristic value obtains subelement, is configured to respectively to above-mentioned first area feature and the secondth area
Characteristic of field carries out Fusion Features, obtains first area characteristic value and second area characteristic value;It is single that Sample Similarity information calculates son
Member is configured to calculate Sample Similarity letter according to the difference between above-mentioned first area characteristic value and second area characteristic value
Breath.
In some embodiments, the similarity calculated includes: weight setting subelement, is configured to respectively institute
It states the first vehicle similarity information and the first weight and the second weight is arranged in the second vehicle similarity information;Similarity calculation
Subelement is configured to the first product of first weight and the first vehicle similarity information, with the second weight and second
Sum between second product of vehicle similarity information is as the similarity between vehicle.
The third aspect, embodiment of the disclosure provide a kind of electronic equipment, comprising: one or more processors;Storage
Device is stored thereon with one or more programs, when said one or multiple programs are executed by said one or multiple processors,
So that said one or multiple processors execute the method for obtaining vehicle similarity of above-mentioned first aspect.
Fourth aspect, embodiment of the disclosure provide a kind of computer-readable medium, are stored thereon with computer program,
It is characterized in that, the program realizes the method for obtaining vehicle similarity of above-mentioned first aspect when being executed by processor.
Embodiment of the disclosure provide for obtaining the method and device of vehicle similarity, firstly, respectively from first to
It handles image and the second image to be processed chooses multiple first areas and multiple second areas, respectively from the multiple first area
Each of first area obtain the first vehicle characteristic information and from each of the multiple second area second area
Obtain the second vehicle characteristic information, according to the first vehicle characteristic information of each first area with it is corresponding described each
The similarity of second vehicle characteristic information of a second area obtains the first vehicle similarity information;Then, described in obtaining respectively
The third vehicle characteristic information and the 4th vehicle characteristic information of first image to be processed and the second image to be processed, by the third
Vehicle characteristic information and the 4th vehicle characteristic information import vehicle identification model trained in advance, obtain the second vehicle similarity letter
Breath;Finally, calculating the first vehicle figure according to the first vehicle similarity information and the second vehicle similarity information
As the similarity between corresponding vehicle and the corresponding vehicle of the second vehicle image.The technical solution of the application is improved to vehicle
The accuracy of identification.
Detailed description of the invention
By reading a detailed description of non-restrictive embodiments in the light of the attached drawings below, the disclosure is other
Feature, objects and advantages will become more apparent upon:
Fig. 1 is that one embodiment of the disclosure can be applied to exemplary system architecture figure therein;
Fig. 2 is the flow chart according to one embodiment of the method for obtaining vehicle similarity of the disclosure;
Fig. 3 a is the first image to be processed of an application scenarios of the disclosure;
Fig. 3 b is the second image to be processed of the image to be processed of corresponding diagram 3a first;
Fig. 4 is the flow chart according to one embodiment of the vehicle identification model training method of the disclosure;
Fig. 5 is the structural schematic diagram according to one embodiment of the device for obtaining vehicle similarity of the disclosure;
Fig. 6 is adapted for the electronic devices structure schematic diagram for realizing embodiment of the disclosure.
Specific embodiment
The disclosure is described in further detail with reference to the accompanying drawings and examples.It is understood that this place is retouched
The specific embodiment stated is used only for explaining related invention, rather than the restriction to the invention.It also should be noted that in order to
Convenient for description, part relevant to related invention is illustrated only in attached drawing.
It should be noted that in the absence of conflict, the feature in embodiment and embodiment in the disclosure can phase
Mutually combination.The disclosure is described in detail below with reference to the accompanying drawings and in conjunction with the embodiments.
Fig. 1 is shown can be using embodiment of the disclosure for obtaining the method for vehicle similarity or for obtaining vehicle
The exemplary system architecture 100 of the device of similarity.
As shown in Figure 1, system architecture 100 may include image collecting device 101, network 102 and server 103.Network
102 between vehicle 101 and server 103 to provide the medium of communication link.Network 102 may include various connection classes
Type, such as wired, wireless communication link or fiber optic cables etc..
Image collecting device 101 is interacted by network 102 with server 103, to receive or send message etc..Image Acquisition
Various image processing applications can be installed, such as Image Acquisition application, vehicle identification are applied, vehicle match is answered on device 101
With, marking of cars application etc..
Image collecting device 101 can be hardware, be also possible to software.It, can be with when image collecting device 101 is hardware
The various electronic equipments that there is camera and support image procossing, including but not limited to intelligent video camera head, vehicle-mounted camera,
Monitoring camera head etc..When image collecting device 101 is software, may be mounted in above-mentioned cited electronic equipment.It can
To be implemented as multiple softwares or software module (such as providing Distributed Services), single software or software also may be implemented into
Module is not specifically limited herein.
Server 103 can be to provide the server of various services, such as send to image collecting device 101 to be processed
The server of image progress image procossing.Server can carry out the processing such as analyzing to data such as the images to be processed received,
To obtain corresponding processing result (such as similarity between vehicle).
It should be noted that can be by image for obtaining the method for vehicle similarity provided by embodiment of the disclosure
Acquisition device 101 is individually performed, or can also be executed jointly by image collecting device 101 and server 103.Correspondingly, it uses
It can be set in image collecting device 101, also can be set in server 103 in the device for obtaining vehicle similarity.
It should be noted that server can be hardware, it is also possible to software.When server is hardware, may be implemented
At the distributed server cluster that multiple servers form, individual server also may be implemented into.It, can when server is software
To be implemented as multiple softwares or software module (such as providing Distributed Services), single software or software also may be implemented into
Module is not specifically limited herein.
It should be understood that the number of image collecting device, network and server in Fig. 1 is only schematical.According to reality
It now needs, can have any number of image collecting device, network and server.
With continued reference to Fig. 2, one embodiment of the method for obtaining vehicle similarity according to the disclosure is shown
Process 200.This be used for obtain vehicle similarity method the following steps are included:
Step 201, multiple first areas and multiple the are chosen from the first image to be processed and the second image to be processed respectively
Two regions.
In the present embodiment, for obtaining executing subject (such as the Image Acquisition shown in FIG. 1 of the method for vehicle similarity
Device 101 and/or server 103) the first image to be processed can be obtained by wired connection mode or radio connection
With the second image to be processed.Wherein, the described first image to be processed includes the first vehicle image, and the second image to be processed includes the
Two vehicle images, it is each in the vehicle location and the multiple second area in the multiple first area in each first area
Vehicle location in second area is corresponding.It should be pointed out that above-mentioned radio connection can include but is not limited to 3G/4G company
It connects, WiFi connection, bluetooth connection, WiMAX connection, Zigbee connection, UWB (ultra wideband) connection and other are existing
In known or exploitation in the future radio connection.
In order to compare multiple first vehicle images that the first image to be processed includes and that the second image to be processed includes is more
Similarity between a second vehicle image determines that the corresponding real vehicles of the first vehicle image and the second vehicle image are corresponding
Whether real vehicles are same vehicle (or same type of vehicles), and executing subject can be respectively from the first image to be processed and
Two images to be processed choose multiple first areas and multiple second areas.Wherein, the vehicle location in first area and the secondth area
Vehicle location in domain is corresponding.For example, being the first vehicle image corresponding first tailstock position in first area, in second area
For the second vehicle image, with the first tailstock position corresponding second tailstock position.
Step 202, the first vehicle characteristic information is obtained from each of the multiple first area first area respectively
The second vehicle characteristic information is obtained with from each of the multiple second area second area.
After choosing first area and second area, executing subject can obtain first from first area and second area respectively
Vehicle characteristic information and the second vehicle characteristic information.Wherein, the first vehicle characteristic information (or second vehicle characteristic information) can be with
It is the information such as color characteristic information, texture feature information, vehicle structure characteristic information.It should be noted that the first vehicle characteristics
Information and the second vehicle characteristic information be it is corresponding, i.e. when the first vehicle characteristic information is color characteristic information, the second vehicle is special
Reference breath is also color characteristic information, carries out phase will pass through same characteristic information to the first vehicle image and the second vehicle image
Judge like degree.First vehicle characteristic information and the second vehicle characteristic information can characterize the first image to be processed and second to be processed
The local feature of image.
Step 203, according to the first vehicle characteristic information of each first area with it is corresponding it is described each
The similarity of second vehicle characteristic information in two regions obtains the first vehicle similarity information.
In general, the first vehicle similarity information is identified by way of decimal.For example, the first vehicle similarity information can be with
It is 0.568 etc..First vehicle similarity information may is that high similarity, middle similarity, low similarity;It may also is that 100%
Similarity, 30% similarity etc.;It may also is that vehicle similar with color, vehicle Similar color dissmilarity, vehicle and color not
It is similar etc..According to the actual situation, similarity information can also be other kinds of describing mode.
Executing subject can calculate in several ways first area the first vehicle characteristic information and corresponding secondth area
Similarity between second vehicle characteristic information in domain obtains the first vehicle similarity information.
In some optional implementations of the present embodiment, the first vehicle of each first area according to
It is similar to obtain the first vehicle for the similarity of characteristic information and the second vehicle characteristic information of corresponding each second area
Information is spent, may comprise steps of:
The first step, for the first vehicle characteristics in multiple first vehicle characteristic informations and multiple second vehicle characteristic informations
Information and the second vehicle characteristic information calculate the second vehicle of the first vehicle characteristic information and corresponding first vehicle characteristic information
Sub- similarity information between characteristic information.
First vehicle characteristic information and the second vehicle characteristic information can be converted into corresponding fisrt feature by executing subject
Vector sum second feature vector.Then the COS distance between first eigenvector and second feature vector is calculated, sub- phase is obtained
Like degree information.
Second step, using the sum of multiple sub- similarity informations as the first vehicle similarity information.
Later, executing subject can calculate multiple first vehicle characteristic informations and multiple second vehicle characteristic informations are corresponding
The sum of multiple sub- similarity informations, and the result that sub- similarity information is summed is as the first vehicle similarity information.
Step 204, the third vehicle characteristic information of the described first image to be processed and the second image to be processed is obtained respectively
With the 4th vehicle characteristic information.
Executing subject can carry out the spy of (or globality) of overall importance to the first image to be processed and the second image to be processed
Sign is extracted, and third vehicle characteristic information and the 4th vehicle characteristic information are obtained.
Step 205, the third vehicle characteristic information and the 4th vehicle characteristic information vehicle trained in advance is imported to know
Other model obtains the second vehicle similarity information.
After obtaining third vehicle characteristic information and the 4th vehicle characteristic information, executing subject can believe third vehicle characteristics
Breath and the 4th vehicle characteristic information import vehicle identification model trained in advance, obtain the second vehicle similarity information.For example, holding
Row main body can constitute global characteristics vector by third vehicle characteristic information and the 4th vehicle characteristic information, then utilize joint
The models such as bayesian algorithm model obtain the second vehicle similarity information (i.e. global similarity).Wherein, above-mentioned vehicle identification mould
Type can be used for calculating by third vehicle characteristic information and the 4th vehicle characteristic information the corresponding vehicle of the first vehicle image and
Similarity information between the corresponding vehicle of second vehicle image.Second vehicle similarity information and the first vehicle similarity information
It is identical.
In some optional implementations of the present embodiment, above-mentioned vehicle identification model is trained by following steps
It arrives:
The first step obtains each sample vehicle in multiple sample vehicle image groups and corresponding above-mentioned multiple sample vehicle image groups
The corresponding first sample vehicle characteristic information of image group, the second sample vehicle characteristic information and Sample Similarity information.
In order to train vehicle identification model, executing subject can obtain multiple sample vehicle image groups first and correspondence is above-mentioned
The corresponding first sample vehicle characteristic information of each sample vehicle image group, the second sample vehicle in multiple sample vehicle image groups
Characteristic information and Sample Similarity information.Wherein, sample vehicle image group may include first sample vehicle subgraph and second
Sample vehicle subgraph.First sample vehicle sub-picture pack contains vehicle image, and the second sample vehicle subgraph also includes vehicle figure
Picture.First sample vehicle subgraph is corresponding with first sample vehicle characteristic information, the second sample vehicle subgraph and the second sample
Vehicle characteristic information.That is, first sample vehicle characteristic information is to carry out feature extraction to first sample vehicle subgraph to obtain;
Second sample vehicle characteristic information is to carry out feature extraction to the second sample vehicle subgraph to obtain.Sample Similarity information can
For characterizing the similarity between first sample vehicle subgraph and the second sample vehicle subgraph, sample vehicle characteristic information
It may include at least one of following: vehicle wheel profile, vehicle region line.
Second step, by the first sample vehicle characteristics of each sample vehicle image group of above-mentioned multiple sample vehicle image groups
Information and the second sample vehicle characteristic information are as input, by each sample vehicle figure in above-mentioned multiple sample vehicle image groups
It is exported as the corresponding above-mentioned Sample Similarity information of group is used as, training obtains above-mentioned vehicle identification model.
After getting sample vehicle image group, executing subject can be by each sample of above-mentioned multiple sample vehicle image groups
The first sample vehicle characteristic information of vehicle image group and the second sample vehicle characteristic information are as input, by above-mentioned multiple samples
Above-mentioned Sample Similarity information corresponding to each sample vehicle image group in vehicle image group is obtained as output, training
State vehicle identification model.
In some optional implementations of the present embodiment, above-mentioned Sample Similarity information is obtained by following steps:
The first step, the first sample vehicle subgraph for including according to sample vehicle image group and the second sample vehicle subgraph
Corresponding first sample vehicle characteristic information and the second sample vehicle characteristic information determine the first vehicle region of sample and sample
This second vehicle region.
Seen from the above description, the first sample vehicle subgraph and the second sample vehicle that each sample vehicle image group includes
Subgraph.Executing subject can be respectively to first sample vehicle subgraph and corresponding first sample of the second sample vehicle subgraph
This vehicle characteristic information and the second sample vehicle characteristic information carry out data processing, determine the first vehicle region of sample and sample the
Two vehicle regions.For example, executing subject can be to the vehicle wheel profile and/or vehicle area that first sample vehicle characteristic information includes
Domain line is extended, and then carries out vehicle region division to vehicle, determines the first vehicle region of sample.Wherein, vehicle region can
To include at least one following region: the regions such as headstock region, tailstock region, car door region, car roof area.Vehicle region may be used also
Be vehicle multiple regions juncture area, such as can be the juncture area etc. of tailstock region and car roof area.Execute master
Body can carry out identical processing to the second sample vehicle characteristic information, obtain the second vehicle region of sample.
Second step obtains the first vehicle of sample area of above-mentioned the first vehicle region of sample and the second vehicle region of sample respectively
The first area feature and second area feature in domain.
Later, executing subject can respectively propose the first vehicle region of sample and the second vehicle region of sample progress feature
It takes, the second area of the first area feature and corresponding the second vehicle region of sample that obtain corresponding the first vehicle region of sample is special
Sign.Wherein, provincial characteristics includes at least one of the following: the provincial characteristics such as lines assemblage characteristic, region assemblage characteristic.
Third step carries out Fusion Features to above-mentioned first area feature and second area feature respectively, obtains first area
Characteristic value and second area characteristic value.
Executing subject can carry out Fusion Features to above-mentioned first area feature and second area feature respectively, obtain first
Regional characteristic value and second area characteristic value.For example, executing subject can be according to the corresponding vehicle area of the first vehicle region of sample
Correlation (such as can be positional relationship, color-match relationship etc.) between vehicle region between domain is come to first area
Feature carries out Fusion Features, obtains first area characteristic value.Wherein, first area characteristic value may include at least one type
The characteristic value of (such as can be position, color etc.).Similar, second area characteristic value can also be obtained.
4th step calculates Sample Similarity according to the difference between above-mentioned first area characteristic value and second area characteristic value
Information.
Executing subject can compare first area characteristic value and second area characteristic value obtains difference.Then, executing subject
Sample Similarity information can be calculated according to difference.For example, it is similar that sample can be set when difference is greater than setting first threshold
It is low for spending information;When difference, which is greater than second threshold, is less than first threshold, can be set during Sample Similarity information is;Work as difference
When less than second threshold, Sample Similarity information can be set as height.Wherein, first threshold can be greater than second threshold.
Step 206, according to the first vehicle similarity information and the second vehicle similarity information, described the is calculated
Similarity between the corresponding vehicle of one vehicle image and the corresponding vehicle of the second vehicle image.
After obtaining the first vehicle similarity information and the second vehicle similarity information, executing subject can be according to the first vehicle
Similarity information and the second vehicle similarity information are corresponding to calculate the corresponding vehicle of the first vehicle image and the second vehicle image
Vehicle between similarity.
It is described according to the first vehicle similarity information and described in some optional implementations of the present embodiment
Second vehicle similarity information calculates between the corresponding vehicle of first vehicle image and the corresponding vehicle of the second vehicle image
Similarity, may comprise steps of:
The first step, the first power of the respectively described first vehicle similarity information and the second vehicle similarity information setting
Weight and the second weight.
The local feature and global characteristics of image in order to balance, executing subject can be respectively the first vehicle similarity
The first weight and the second weight is arranged in information and the second vehicle similarity information.First weight and the specific of the second weight take
Value can be depending on actual needs.
Second step, by the first product of first weight and the first vehicle similarity information, with the second weight and second
Sum between second product of vehicle similarity information is as the similarity between vehicle.
Executing subject can calculate separately the first product of the first weight and the first vehicle similarity information, the second weight and
Second product of the second vehicle similarity information.Then, executing subject can be using the sum of first the second product of sum of products as vehicle
Similarity between.
With continued reference to Fig. 3 a, Fig. 3 a is the application scenarios according to the method for obtaining vehicle similarity of the present embodiment
One schematic diagram.Fig. 3 a is the first image to be processed, and Fig. 3 b is the second image to be processed of corresponding diagram 3a.Executing subject (such as
Image collecting device 101 shown in FIG. 1) first area and black of a white box can be chosen in the first image to be processed
The first area of frame;The second area an of white box and the secondth area of black box can be chosen in the second image to be processed
Domain.Wherein, the second area of the first area of white box and white box is by car roof area, rear window region and vehicle rear window region together
Juncture area.The first area of black box and the second area of black box are all Rear lamp for vehicle region.Then, executing subject is distinguished
It is obtained from the second area of the first area of white box and white box, the first area of black box and the second area of black box
First vehicle characteristic information and the second vehicle characteristic information.Later, according to the first vehicle characteristic information of each first area
With the similarity of the second vehicle characteristic information of each corresponding second area, the first vehicle similarity information is obtained;It executes
Main body obtains the third vehicle characteristic information and the 4th vehicle characteristics letter of the first image to be processed and the second image to be processed respectively
Breath;Third vehicle characteristic information and the 4th vehicle characteristic information are imported into vehicle identification model, obtain the second vehicle similarity letter
Breath;Finally, calculating the corresponding vehicle of the first vehicle image according to the first vehicle similarity information and the second vehicle similarity information
And the second similarity between the corresponding vehicle of vehicle image may is that 0.9.
The method provided by the above embodiment of the disclosure, firstly, respectively from the first image to be processed and the second figure to be processed
As choosing multiple first areas and multiple second areas, obtained respectively from each of the multiple first area first area
First vehicle characteristic information and from each of the multiple second area second area obtain the second vehicle characteristic information, root
It is special according to the first vehicle characteristic information and the second vehicle of corresponding each second area of each first area
The similarity of reference breath, obtains the first vehicle similarity information;Then, obtain respectively the described first image to be processed and second to
The third vehicle characteristic information and the 4th vehicle characteristic information for handling image, by the third vehicle characteristic information and the 4th vehicle
Characteristic information imports vehicle identification model trained in advance, obtains the second vehicle similarity information;Finally, according to first vehicle
Similarity information and the second vehicle similarity information calculate the corresponding vehicle of first vehicle image and the second vehicle
Similarity between the corresponding vehicle of image.The technical solution of the application improves the accuracy to vehicle identification.
With further reference to Fig. 4, it illustrates the processes 400 of another embodiment of vehicle identification model training method.It should
The process 400 of vehicle identification model training method, comprising the following steps:
Step 401, each sample in multiple sample vehicle image groups and corresponding above-mentioned multiple sample vehicle image groups is obtained
The corresponding first sample vehicle characteristic information of vehicle image group, the second sample vehicle characteristic information and Sample Similarity information.
In the present embodiment, executing subject (such as the image collecting device shown in FIG. 1 of vehicle identification model training method
101 and/or server 103) it is every in available multiple sample vehicle image groups and corresponding above-mentioned multiple sample vehicle image groups
The corresponding first sample vehicle characteristic information of a sample vehicle image group, the second sample vehicle characteristic information and Sample Similarity letter
Breath.
Step 402, by the first sample vehicle of each sample vehicle image group in above-mentioned multiple sample vehicle image groups
Characteristic information and the second sample vehicle characteristic information are sequentially input to initial vehicle identification model, obtain above-mentioned multiple sample vehicles
Prediction similarity information corresponding to each sample vehicle image group in image group.
Executing subject can be by the first sample vehicle of each sample vehicle image group in multiple sample vehicle image groups
Characteristic information and the second sample vehicle characteristic information are sequentially input to initial vehicle identification model, to obtain multiple sample vehicles
Prediction similarity information corresponding to each sample vehicle image group in image group.Here, executing subject can be by each sample
The first sample vehicle characteristic information of this vehicle image group and the second sample vehicle characteristic information are from initial vehicle identification model
Input side input, successively by the processing of the parameter of each layer in initial vehicle identification model, and from initial vehicle identification model
Outlet side output, outlet side output information be prediction similarity information corresponding to the sample vehicle image group.Wherein,
The depth that initial vehicle identification model can be unbred deep learning model (convolutional neural networks etc.) or training is not completed
Learning model is spent, each layer is provided with initiation parameter, and initiation parameter can be by the training process of deep learning model
It continuously adjusts.
Step 403, by prediction phase corresponding to each sample vehicle image group in above-mentioned multiple sample vehicle image groups
It is compared like degree information with Sample Similarity information corresponding to the sample vehicle image group, obtains above-mentioned initial vehicle identification
The predictablity rate of model.
In the present embodiment, based on each sample vehicle figure in the obtained multiple sample vehicle image groups of step 402
As the corresponding prediction similarity information of group, executing subject can be by each sample vehicle figure in multiple sample vehicle image groups
The prediction similarity information as corresponding to group is compared with Sample Similarity information corresponding to the sample vehicle image group, from
And obtain the predictablity rate of initial vehicle identification model.
Step 404, determine whether above-mentioned predictablity rate is greater than default accuracy rate threshold value.
In the present embodiment, the predictablity rate based on the obtained initial vehicle identification model of step 403, executing subject
The predictablity rate of initial vehicle identification model can be compared with default accuracy rate threshold value.If more than default accuracy rate threshold
Value, thens follow the steps 405;If thening follow the steps 406 no more than default accuracy rate threshold value.
Step 405, the vehicle identification model above-mentioned initial vehicle identification model completed as training.
In the present embodiment, the case where the predictablity rate of initial vehicle identification model is greater than default accuracy rate threshold value
Under, illustrate that the vehicle identification model training is completed, at this point, executing subject can be completed initial vehicle identification model as training
Vehicle identification model.
Step 406, the parameter of above-mentioned initial vehicle identification model is adjusted.
In the present embodiment, the case where the predictablity rate of initial vehicle identification model is not more than default accuracy rate threshold value
Under, the parameter of the adjustable initial vehicle identification model of executing subject, and return to step 402.
With further reference to Fig. 5, as the realization to method shown in above-mentioned each figure, present disclose provides one kind for obtaining vehicle
One embodiment of the device of similarity, the Installation practice is corresponding with embodiment of the method shown in Fig. 2, and the device is specific
It can be applied in various electronic equipments.
As shown in figure 5, the device 500 for obtaining vehicle similarity of the present embodiment may include: image-region selection
Unit 501, fisrt feature information acquisition unit 502, the first vehicle similarity information acquiring unit 503, second feature information obtain
Take unit 504, the second vehicle similarity information acquiring unit 505 and similarity calculated 506.Wherein, image-region selects
Unit 501 is configured to choose multiple first areas and multiple the from the first image to be processed and the second image to be processed respectively
Two regions, wherein first image to be processed includes the first vehicle image, and the second image to be processed includes the second vehicle figure
Picture, each second area in the vehicle location and the multiple second area in the multiple first area in each first area
Interior vehicle location is corresponding;Fisrt feature information acquisition unit 502 is configured to respectively from every in the multiple first area
One first area obtains the first vehicle characteristic information and obtains the from each of the multiple second area second area
Two vehicle characteristic informations;First vehicle similarity information acquiring unit 503 is configured to according to each first area
The similarity of first vehicle characteristic information and the second vehicle characteristic information of corresponding each second area obtains first
Vehicle similarity information;Second feature information acquisition unit 504 is configured to obtain the described first image to be processed and respectively
The third vehicle characteristic information and the 4th vehicle characteristic information of two images to be processed;Second vehicle similarity information acquiring unit
505, it is configured to the third vehicle characteristic information and the 4th vehicle characteristic information importing vehicle identification mould trained in advance
Type obtains the second vehicle similarity information;Similarity calculated 506 is configured to be believed according to the first vehicle similarity
Breath and the second vehicle similarity information, calculate the corresponding vehicle of first vehicle image and the second vehicle image is corresponding
Similarity between vehicle.
In some optional implementations of the present embodiment, the first vehicle similarity information acquiring unit 503 can
To include: that sub- similarity information computation subunit (not shown) and the first vehicle similarity information obtain subelement (in figure
It is not shown).Wherein, sub- similarity information computation subunit, it is special for multiple first vehicle characteristic informations and multiple second vehicles
The first vehicle characteristic information and the second vehicle characteristic information in reference breath, are configured to calculate the first vehicle characteristic information and right
It should sub- similarity information between the second vehicle characteristic information of the first vehicle characteristic information;First vehicle similarity information obtains
Subelement is taken to be configured to using the sum of multiple sub- similarity informations as the first vehicle similarity information.
In some optional implementations of the present embodiment, the above-mentioned device 500 for obtaining vehicle similarity may be used also
To include vehicle identification model training unit (not shown), it is configured to train vehicle identification model, above-mentioned vehicle identification
Model training unit may include: that sample information obtains subelement (not shown) and vehicle identification model training subelement
(not shown).Wherein, sample information acquisition subelement is configured to obtain multiple sample vehicle image groups and correspondence is above-mentioned
The corresponding first sample vehicle characteristic information of each sample vehicle image group, the second sample vehicle in multiple sample vehicle image groups
Characteristic information and Sample Similarity information, wherein sample vehicle image group includes first sample vehicle comprising vehicle image
Image and the second sample vehicle subgraph comprising vehicle image, first sample vehicle subgraph and first sample vehicle characteristics are believed
Breath corresponds to, and the second sample vehicle subgraph and the second sample vehicle characteristic information, Sample Similarity information is for characterizing the first sample
Similarity between this vehicle subgraph and the second sample vehicle subgraph, sample vehicle characteristic information include following at least one
: vehicle wheel profile, vehicle region line;Vehicle identification model training subelement is configured to above-mentioned multiple sample vehicle images
The first sample vehicle characteristic information and the second sample vehicle characteristic information of each sample vehicle image group of group, will as input
Above-mentioned Sample Similarity information corresponding to each sample vehicle image group in above-mentioned multiple sample vehicle image groups is as defeated
Out, training obtains above-mentioned vehicle identification model.
In some optional implementations of the present embodiment, above-mentioned vehicle identification model training subelement may include:
Vehicle identification model training module (not shown) is configured to each sample in above-mentioned multiple sample vehicle image groups
The first sample vehicle characteristic information of vehicle image group and the second sample vehicle characteristic information sequentially input to initial vehicle and identify
Model obtains prediction similarity information corresponding to each sample vehicle image group in above-mentioned multiple sample vehicle image groups,
By prediction similarity information corresponding to each sample vehicle image group in above-mentioned multiple sample vehicle image groups and the sample
Sample Similarity information corresponding to vehicle image group is compared, and the prediction for obtaining above-mentioned initial vehicle identification model is accurate
Rate, determines whether above-mentioned predictablity rate is greater than default accuracy rate threshold value, then will be above-mentioned if more than above-mentioned default accuracy rate threshold value
The vehicle identification model that initial vehicle identification model is completed as training.
In some optional implementations of the present embodiment, above-mentioned vehicle identification model training subelement may include:
Parameter adjustment module (not shown) is configured to adjust above-mentioned initial in response to being not more than above-mentioned default accuracy rate threshold value
The parameter of vehicle identification model, and return to vehicle identification model training module.
In some optional implementations of the present embodiment, the above-mentioned device 500 for obtaining vehicle similarity may be used also
To include Sample Similarity information acquisition unit (not shown), it is configured to obtain Sample Similarity information, above-mentioned sample
Similarity information acquiring unit may include: that sample vehicle region obtains subelement (not shown), provincial characteristics obtains son
Unit (not shown), regional characteristic value obtain subelement (not shown) and Sample Similarity information computation subunit
(not shown).Wherein, sample vehicle region obtains subelement and is configured to include according to sample vehicle image group first
Sample vehicle subgraph and the corresponding first sample vehicle characteristic information of the second sample vehicle subgraph and the second sample vehicle
Characteristic information determines the second vehicle region of the first vehicle region of sample and sample;Provincial characteristics obtains subelement and is configured ingredient
The first area feature of the first vehicle region of sample of above-mentioned the first vehicle region of sample and the second vehicle region of sample is not obtained
With second area feature, wherein provincial characteristics includes at least one of the following: lines assemblage characteristic, region assemblage characteristic;Region is special
Value indicative obtains subelement and is configured to carry out Fusion Features to above-mentioned first area feature and second area feature respectively, obtains the
One regional characteristic value and second area characteristic value;Sample Similarity information computation subunit is configured to according to above-mentioned first area
Difference between characteristic value and second area characteristic value calculates Sample Similarity information.
In some optional implementations of the present embodiment, the similarity calculated 506 may include: that weight is set
Set subelement (not shown) and similarity calculation subelement (not shown).Wherein, weight setting subelement is configured
At the respectively described first vehicle similarity information and the second vehicle similarity information, the first weight and the second weight are set;
Similarity calculation subelement is configured to the first product of first weight and the first vehicle similarity information, with the second power
Sum between weight and the second product of the second vehicle similarity information is as the similarity between vehicle.
The present embodiment additionally provides a kind of electronic equipment, comprising: one or more processors;Memory is stored thereon with
One or more programs, when said one or multiple programs are executed by said one or multiple processors, so that said one
Or multiple processors execute the above-mentioned method for obtaining vehicle similarity.
The present embodiment additionally provides a kind of computer-readable medium, is stored thereon with computer program, and the program is processed
Device realizes the above-mentioned method for obtaining vehicle similarity when executing.
Below with reference to Fig. 6, it illustrates the electronic equipments for being suitable for being used to realize embodiment of the disclosure (for example, in Fig. 1
Server 103) computer system 600 structural schematic diagram.Electronic equipment shown in Fig. 6 is only an example, should not be right
The function and use scope of embodiment of the disclosure bring any restrictions.
As shown in fig. 6, electronic equipment 600 may include processing unit (such as central processing unit, graphics processor etc.)
601, random access can be loaded into according to the program being stored in read-only memory (ROM) 602 or from storage device 608
Program in memory (RAM) 603 and execute various movements appropriate and processing.In RAM 603, it is also stored with electronic equipment
Various programs and data needed for 600 operations.Processing unit 601, ROM 602 and RAM 603 pass through the phase each other of bus 604
Even.Input/output (I/O) interface 605 is also connected to bus 604.
In general, following device can connect to I/O interface 605: including such as touch screen, touch tablet, keyboard, mouse, taking the photograph
As the input unit 606 of head, microphone, accelerometer, gyroscope etc.;Including such as liquid crystal display (LCD), loudspeaker, vibration
The output device 607 of dynamic device etc.;Storage device 608 including such as tape, hard disk etc.;And communication device 609.Communication device
609, which can permit electronic equipment 600, is wirelessly or non-wirelessly communicated with other equipment to exchange data.Although Fig. 6 shows tool
There is the electronic equipment 600 of various devices, it should be understood that being not required for implementing or having all devices shown.It can be with
Alternatively implement or have more or fewer devices.Each box shown in Fig. 6 can represent a device, can also root
According to needing to represent multiple devices.
Particularly, in accordance with an embodiment of the present disclosure, it may be implemented as computer above with reference to the process of flow chart description
Software program.For example, embodiment of the disclosure includes a kind of computer program product comprising be carried on computer-readable medium
On computer program, which includes the program code for method shown in execution flow chart.In such reality
It applies in example, which can be downloaded and installed from network by communication device 609, or from storage device 608
It is mounted, or is mounted from ROM 602.When the computer program is executed by processing unit 601, the implementation of the disclosure is executed
The above-mentioned function of being limited in the method for example.
It is situated between it should be noted that the above-mentioned computer-readable medium of embodiment of the disclosure can be computer-readable signal
Matter or computer readable storage medium either the two any combination.Computer readable storage medium for example can be with
System, device or the device of --- but being not limited to --- electricity, magnetic, optical, electromagnetic, infrared ray or semiconductor, or it is any more than
Combination.The more specific example of computer readable storage medium can include but is not limited to: have one or more conducting wires
Electrical connection, portable computer diskette, hard disk, random access storage device (RAM), read-only memory (ROM), erasable type are programmable
Read-only memory (EPROM or flash memory), optical fiber, portable compact disc read-only memory (CD-ROM), light storage device, magnetic are deposited
Memory device or above-mentioned any appropriate combination.In embodiment of the disclosure, computer readable storage medium, which can be, appoints
What include or the tangible medium of storage program that the program can be commanded execution system, device or device use or and its
It is used in combination.And in embodiment of the disclosure, computer-readable signal media may include in a base band or as carrier wave
The data-signal that a part is propagated, wherein carrying computer-readable program code.The data-signal of this propagation can be adopted
With diversified forms, including but not limited to electromagnetic signal, optical signal or above-mentioned any appropriate combination.Computer-readable signal is situated between
Matter can also be any computer-readable medium other than computer readable storage medium, which can be with
It sends, propagate or transmits for by the use of instruction execution system, device or device or program in connection.Meter
The program code for including on calculation machine readable medium can transmit with any suitable medium, including but not limited to: electric wire, optical cable,
RF (radio frequency) etc. or above-mentioned any appropriate combination.
Above-mentioned computer-readable medium can be included in above-mentioned electronic equipment;It is also possible to individualism, and not
It is fitted into the electronic equipment.Above-mentioned computer-readable medium carries one or more program, when said one or more
When a program is executed by the electronic equipment, so that the electronic equipment: respectively from the first image to be processed and the second image to be processed
Choose multiple first areas and multiple second areas, wherein first image to be processed include the first vehicle image, second to
Handling image includes the second vehicle image, vehicle location in the multiple first area in each first area and the multiple
Vehicle location in second area in each second area is corresponding;Respectively from the firstth area of each of the multiple first area
Domain obtains the first vehicle characteristic information and obtains the second vehicle characteristics from each of the multiple second area second area
Information;According to the first vehicle characteristic information of each first area and the second of corresponding each second area
The similarity of vehicle characteristic information obtains the first vehicle similarity information;The described first image to be processed and second are obtained respectively
The third vehicle characteristic information and the 4th vehicle characteristic information of image to be processed;By the third vehicle characteristic information and the 4th vehicle
Characteristic information imports vehicle identification model trained in advance, obtains the second vehicle similarity information;According to first vehicle
Similarity information and the second vehicle similarity information calculate the corresponding vehicle of first vehicle image and the second vehicle figure
As the similarity between corresponding vehicle.
The behaviour for executing embodiment of the disclosure can be write with one or more programming languages or combinations thereof
The computer program code of work, above procedure design language include object oriented program language-such as Java,
Smalltalk, C++ further include conventional procedural programming language-such as " C " language or similar program design language
Speech.Program code can be executed fully on the user computer, partly be executed on the user computer, as an independence
Software package execute, part on the user computer part execute on the remote computer or completely in remote computer or
It is executed on server.In situations involving remote computers, remote computer can pass through the network of any kind --- packet
It includes local area network (LAN) or wide area network (WAN)-is connected to subscriber computer, or, it may be connected to outer computer (such as benefit
It is connected with ISP by internet).
Flow chart and block diagram in attached drawing illustrate system, method and the computer of the various embodiments according to the disclosure
The architecture, function and operation in the cards of program product.In this regard, each box in flowchart or block diagram can be with
A part of a module, program segment or code is represented, a part of the module, program segment or code includes one or more
Executable instruction for implementing the specified logical function.It should also be noted that in some implementations as replacements, institute in box
The function of mark can also occur in a different order than that indicated in the drawings.For example, two boxes succeedingly indicated are practical
On can be basically executed in parallel, they can also be executed in the opposite order sometimes, and this depends on the function involved.Also it wants
It is noted that the combination of each box in block diagram and or flow chart and the box in block diagram and or flow chart, Ke Yiyong
The dedicated hardware based system of defined functions or operations is executed to realize, or can be referred to specialized hardware and computer
The combination of order is realized.
Being described in unit involved in embodiment of the disclosure can be realized by way of software, can also be passed through
The mode of hardware is realized.Described unit also can be set in the processor, for example, can be described as: a kind of processor
Including image-region selecting unit, fisrt feature information acquisition unit, the first vehicle similarity information acquiring unit, second feature
Information acquisition unit, the second vehicle similarity information acquiring unit and similarity calculated.Wherein, the title of these units exists
The restriction to the unit itself is not constituted in the case of certain, such as similarity calculated is also described as " for leading to
It crosses the first vehicle similarity information and the second vehicle similarity information calculates the unit of similarity between two vehicle images ".
Above description is only the preferred embodiment of the disclosure and the explanation to institute's application technology principle.Those skilled in the art
Member is it should be appreciated that invention scope involved in the disclosure, however it is not limited to technology made of the specific combination of above-mentioned technical characteristic
Scheme, while should also cover in the case where not departing from foregoing invention design, it is carried out by above-mentioned technical characteristic or its equivalent feature
Any combination and the other technical solutions formed.Such as features described above has similar function with (but being not limited to) disclosed in the disclosure
Can technical characteristic replaced mutually and the technical solution that is formed.
Claims (16)
1. a kind of method for obtaining vehicle similarity, comprising:
Multiple first areas and multiple second areas are chosen from the first image to be processed and the second image to be processed respectively, wherein
First image to be processed includes the first vehicle image, and the second image to be processed includes the second vehicle image, and the multiple the
Vehicle location in vehicle location and the multiple second area in one region in each first area in each second area
It is corresponding;
From each of the multiple first area first area obtain the first vehicle characteristic information respectively and from the multiple
Each of second area second area obtains the second vehicle characteristic information;
According to the first vehicle characteristic information of each first area and the second of corresponding each second area
The similarity of vehicle characteristic information obtains the first vehicle similarity information;
Third vehicle characteristic information and the 4th vehicle for obtaining the described first image to be processed and the second image to be processed respectively are special
Reference breath;
The third vehicle characteristic information and the 4th vehicle characteristic information are imported to vehicle identification model trained in advance, obtain the
Two vehicle similarity informations;
According to the first vehicle similarity information and the second vehicle similarity information, first vehicle image pair is calculated
Similarity between the vehicle answered and the corresponding vehicle of the second vehicle image.
2. according to the method described in claim 1, wherein, the first vehicle characteristics of each first area according to are believed
Breath is believed with the similarity of the second vehicle characteristic information of corresponding each second area, the first vehicle similarity of acquisition
Breath, comprising:
For the first vehicle characteristic information and second in multiple first vehicle characteristic informations and multiple second vehicle characteristic informations
Vehicle characteristic information, calculate the first vehicle characteristic information and corresponding first vehicle characteristic information the second vehicle characteristic information it
Between sub- similarity information;
Using the sum of multiple sub- similarity informations as the first vehicle similarity information.
3. according to the method described in claim 1, wherein, the vehicle identification model is obtained by following steps training:
Obtain each sample vehicle image group pair in multiple sample vehicle image groups and corresponding the multiple sample vehicle image group
First sample vehicle characteristic information, the second sample vehicle characteristic information and the Sample Similarity information answered, wherein sample vehicle figure
As group includes the first sample vehicle subgraph comprising vehicle image and the second sample vehicle subgraph comprising vehicle image, the
One sample vehicle subgraph is corresponding with first sample vehicle characteristic information, and the second sample vehicle subgraph and the second sample vehicle are special
Reference breath, Sample Similarity information are similar between first sample vehicle subgraph and the second sample vehicle subgraph for characterizing
Degree, sample vehicle characteristic information include at least one of the following: vehicle wheel profile, vehicle region line;
By the first sample vehicle characteristic information and second of each sample vehicle image group of the multiple sample vehicle image group
Sample vehicle characteristic information, will be corresponding to each sample vehicle image group in the multiple sample vehicle image group as input
The Sample Similarity information as output, training obtain the vehicle identification model.
4. according to the method described in claim 3, wherein, each sample vehicle by the multiple sample vehicle image group
The first sample vehicle characteristic information of image group and the second sample vehicle characteristic information are as input, by the multiple sample vehicle
The Sample Similarity information corresponding to each sample vehicle image group in image group obtains the vehicle as output, training
Identification model, comprising:
Execute following training step: by the first sample of each sample vehicle image group in the multiple sample vehicle image group
Vehicle characteristic information and the second sample vehicle characteristic information are sequentially input to initial vehicle identification model, obtain the multiple sample
Prediction similarity information corresponding to each sample vehicle image group in vehicle image group, by the multiple sample vehicle image
Sample corresponding to prediction similarity information and the sample vehicle image group corresponding to each sample vehicle image group in group
Similarity information is compared, and obtains the predictablity rate of the initial vehicle identification model, determines that the predictablity rate is
It is no to be greater than default accuracy rate threshold value, if more than the default accuracy rate threshold value, then using the initial vehicle identification model as instruction
Practice the vehicle identification model completed.
5. according to the method described in claim 4, wherein, each sample vehicle by the multiple sample vehicle image group
The first sample vehicle characteristic information of image group and the second sample vehicle characteristic information are as input, by the multiple sample vehicle
The Sample Similarity information corresponding to each sample vehicle image group in image group obtains the vehicle as output, training
Identification model, comprising:
In response to being not more than the default accuracy rate threshold value, the parameter of the initial vehicle identification model is adjusted, and continue to execute
The training step.
6. according to the method described in claim 3, wherein, the Sample Similarity information is obtained by following steps:
The first sample vehicle subgraph and the second sample vehicle subgraph for including according to sample vehicle image group are corresponding
First sample vehicle characteristic information and the second sample vehicle characteristic information determine the second vehicle of the first vehicle region of sample and sample
Region;
The firstth area of the first vehicle region of sample of first vehicle region of sample and the second vehicle region of sample is obtained respectively
Characteristic of field and second area feature, wherein provincial characteristics includes at least one of the following: lines assemblage characteristic, region assemblage characteristic;
Fusion Features are carried out to the first area feature and second area feature respectively, obtain first area characteristic value and second
Regional characteristic value;
Sample Similarity information is calculated according to the difference between the first area characteristic value and second area characteristic value.
7. according to claim 1 to method described in 6 any one, wherein described according to the first vehicle similarity information
With the second vehicle similarity information, the corresponding vehicle of first vehicle image and the corresponding vehicle of the second vehicle image are calculated
Similarity between, comprising:
The first weight and the second power is arranged in the respectively described first vehicle similarity information and the second vehicle similarity information
Weight;
By the first product of first weight and the first vehicle similarity information, believe with the second weight and the second vehicle similarity
Sum between second product of breath is as the similarity between vehicle.
8. a kind of for obtaining the device of vehicle similarity, comprising:
Image-region selecting unit is configured to choose multiple first from the first image to be processed and the second image to be processed respectively
Region and multiple second areas, wherein first image to be processed includes the first vehicle image, and the second image to be processed includes
Second vehicle image, it is every in the vehicle location and the multiple second area in the multiple first area in each first area
Vehicle location in a second area is corresponding;
Fisrt feature information acquisition unit is configured to obtain from each of the multiple first area first area respectively
First vehicle characteristic information and from each of the multiple second area second area obtain the second vehicle characteristic information;
First vehicle similarity information acquiring unit is configured to the first vehicle characteristics letter according to each first area
Breath is believed with the similarity of the second vehicle characteristic information of corresponding each second area, the first vehicle similarity of acquisition
Breath;
Second feature information acquisition unit is configured to obtain the described first image to be processed and the second image to be processed respectively
Third vehicle characteristic information and the 4th vehicle characteristic information;
Second vehicle similarity information acquiring unit is configured to believe the third vehicle characteristic information and the 4th vehicle characteristics
Breath imports vehicle identification model trained in advance, obtains the second vehicle similarity information;
Similarity calculated is configured to be believed according to the first vehicle similarity information and the second vehicle similarity
Breath calculates the similarity between the corresponding vehicle of first vehicle image and the corresponding vehicle of the second vehicle image.
9. device according to claim 8, wherein the first vehicle similarity information acquiring unit includes:
Sub- similarity information computation subunit, in multiple first vehicle characteristic informations and multiple second vehicle characteristic informations
First vehicle characteristic information and the second vehicle characteristic information are configured to calculate the first vehicle characteristic information and corresponding first vehicle
Sub- similarity information between second vehicle characteristic information of characteristic information;
First vehicle similarity information obtains subelement, is configured to using the sum of multiple sub- similarity informations as the first vehicle phase
Like degree information.
10. device according to claim 8, wherein described device further includes vehicle identification model training unit, is configured
At training vehicle identification model, the vehicle identification model training unit includes:
Sample information obtains subelement, is configured to obtain multiple sample vehicle image groups and corresponding the multiple sample vehicle figure
As the corresponding first sample vehicle characteristic information of sample vehicle image group each in group, the second sample vehicle characteristic information and sample
Similarity information, wherein sample vehicle image group include comprising vehicle image first sample vehicle subgraph and comprising vehicle
Second sample vehicle subgraph of image, first sample vehicle subgraph is corresponding with first sample vehicle characteristic information, the second sample
This vehicle subgraph and the second sample vehicle characteristic information, Sample Similarity information for characterize first sample vehicle subgraph and
Similarity between second sample vehicle subgraph, sample vehicle characteristic information include at least one of the following: vehicle wheel profile, vehicle
Region line;
Vehicle identification model training subelement is configured to each sample vehicle image of the multiple sample vehicle image group
The first sample vehicle characteristic information and the second sample vehicle characteristic information of group are as input, by the multiple sample vehicle image
As output, training obtains the vehicle and knows the Sample Similarity information corresponding to each sample vehicle image group in group
Other model.
11. device according to claim 10, wherein the vehicle identification model training subelement includes:
Vehicle identification model training module is configured to each sample vehicle image in the multiple sample vehicle image group
The first sample vehicle characteristic information and the second sample vehicle characteristic information of group are sequentially input to initial vehicle identification model, are obtained
Prediction similarity information corresponding to each sample vehicle image group in the multiple sample vehicle image group, will be the multiple
Prediction similarity information corresponding to each sample vehicle image group in sample vehicle image group and the sample vehicle image group
Corresponding Sample Similarity information is compared, and obtains the predictablity rate of the initial vehicle identification model, determine described in
Whether predictablity rate is greater than default accuracy rate threshold value, if more than the default accuracy rate threshold value, then knows the initial vehicle
The vehicle identification model that other model is completed as training.
12. device according to claim 11, wherein the vehicle identification model training subelement includes:
Parameter adjustment module is configured to adjust the initial vehicle identification in response to being not more than the default accuracy rate threshold value
The parameter of model, and return to vehicle identification model training module.
13. device according to claim 10, wherein described device further includes Sample Similarity information acquisition unit, quilt
It is configured to obtain Sample Similarity information, the Sample Similarity information acquisition unit includes:
Sample vehicle region obtains subelement, is configured to the first sample vehicle subgraph for including according to sample vehicle image group
First sample vehicle characteristic information corresponding with the second sample vehicle subgraph and the second sample vehicle characteristic information determine
The second vehicle region of the first vehicle region of sample and sample;
Provincial characteristics obtains subelement, is configured to obtain first vehicle region of sample and the second vehicle region of sample respectively
The first vehicle region of sample first area feature and second area feature, wherein provincial characteristics includes at least one of the following:
Lines assemblage characteristic, region assemblage characteristic;
Regional characteristic value obtains subelement, is configured to carry out feature to the first area feature and second area feature respectively
Fusion, obtains first area characteristic value and second area characteristic value;
Sample Similarity information computation subunit, be configured to according to the first area characteristic value and second area characteristic value it
Between difference calculate Sample Similarity information.
14. according to device described in claim 8 to 13 any one, wherein the similarity calculated includes:
Subelement is arranged in weight, is configured to the respectively described first vehicle similarity information and the second vehicle similarity letter
Breath the first weight of setting and the second weight;
Similarity calculation subelement, is configured to the first product of first weight and the first vehicle similarity information, with
Sum between second weight and the second product of the second vehicle similarity information is as the similarity between vehicle.
15. a kind of electronic equipment, comprising:
One or more processors;
Memory is stored thereon with one or more programs,
When one or more of programs are executed by one or more of processors, so that one or more of processors
Perform claim requires any method in 1 to 7.
16. a kind of computer-readable medium, is stored thereon with computer program, which is characterized in that the program is executed by processor
Method of the Shi Shixian as described in any in claim 1 to 7.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910441616.5A CN110135517B (en) | 2019-05-24 | 2019-05-24 | Method and device for obtaining vehicle similarity |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201910441616.5A CN110135517B (en) | 2019-05-24 | 2019-05-24 | Method and device for obtaining vehicle similarity |
Publications (2)
Publication Number | Publication Date |
---|---|
CN110135517A true CN110135517A (en) | 2019-08-16 |
CN110135517B CN110135517B (en) | 2023-04-07 |
Family
ID=67581863
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201910441616.5A Active CN110135517B (en) | 2019-05-24 | 2019-05-24 | Method and device for obtaining vehicle similarity |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110135517B (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111738228A (en) * | 2020-08-04 | 2020-10-02 | 杭州智诚惠通科技有限公司 | Multi-view vehicle feature matching method for hypermetrological evidence chain verification |
Citations (15)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20090164059A1 (en) * | 2007-12-20 | 2009-06-25 | Denso Corporation | Vehicle verification apparatus and vehicle control system using the same |
CN103246876A (en) * | 2013-05-10 | 2013-08-14 | 苏州祥益网络科技有限公司 | Image feature comparison based counterfeit vehicle registration plate identification method |
CN105279971A (en) * | 2015-10-09 | 2016-01-27 | 北京易华录信息技术股份有限公司 | Vehicle information matching method and system, and monitoring device |
CN106228179A (en) * | 2016-07-13 | 2016-12-14 | 乐视控股(北京)有限公司 | The method and system of vehicle comparison |
CN106446754A (en) * | 2015-08-11 | 2017-02-22 | 阿里巴巴集团控股有限公司 | Image identification method, metric learning method, image source identification method and devices |
CN106548145A (en) * | 2016-10-31 | 2017-03-29 | 北京小米移动软件有限公司 | Image-recognizing method and device |
CN107016061A (en) * | 2017-03-15 | 2017-08-04 | 海尔优家智能科技(北京)有限公司 | Video monitoring document handling method and device |
US20170255902A1 (en) * | 2016-03-02 | 2017-09-07 | International Business Machines Corporation | Vehicle identification and interception |
CN107688819A (en) * | 2017-02-16 | 2018-02-13 | 平安科技(深圳)有限公司 | The recognition methods of vehicle and device |
WO2018036277A1 (en) * | 2016-08-22 | 2018-03-01 | 平安科技(深圳)有限公司 | Method, device, server, and storage medium for vehicle detection |
CN108388888A (en) * | 2018-03-23 | 2018-08-10 | 腾讯科技(深圳)有限公司 | A kind of vehicle identification method, device and storage medium |
CN108596277A (en) * | 2018-05-10 | 2018-09-28 | 腾讯科技(深圳)有限公司 | A kind of testing vehicle register identification method, apparatus and storage medium |
CN109166120A (en) * | 2018-09-11 | 2019-01-08 | 百度在线网络技术(北京)有限公司 | For obtaining the method and device of information |
CN109446905A (en) * | 2018-09-26 | 2019-03-08 | 深圳壹账通智能科技有限公司 | Sign electronically checking method, device, computer equipment and storage medium |
CN109784415A (en) * | 2019-01-25 | 2019-05-21 | 北京地平线机器人技术研发有限公司 | The method and device of image-recognizing method and device, training convolutional neural networks |
-
2019
- 2019-05-24 CN CN201910441616.5A patent/CN110135517B/en active Active
Patent Citations (15)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
US20090164059A1 (en) * | 2007-12-20 | 2009-06-25 | Denso Corporation | Vehicle verification apparatus and vehicle control system using the same |
CN103246876A (en) * | 2013-05-10 | 2013-08-14 | 苏州祥益网络科技有限公司 | Image feature comparison based counterfeit vehicle registration plate identification method |
CN106446754A (en) * | 2015-08-11 | 2017-02-22 | 阿里巴巴集团控股有限公司 | Image identification method, metric learning method, image source identification method and devices |
CN105279971A (en) * | 2015-10-09 | 2016-01-27 | 北京易华录信息技术股份有限公司 | Vehicle information matching method and system, and monitoring device |
US20170255902A1 (en) * | 2016-03-02 | 2017-09-07 | International Business Machines Corporation | Vehicle identification and interception |
CN106228179A (en) * | 2016-07-13 | 2016-12-14 | 乐视控股(北京)有限公司 | The method and system of vehicle comparison |
WO2018036277A1 (en) * | 2016-08-22 | 2018-03-01 | 平安科技(深圳)有限公司 | Method, device, server, and storage medium for vehicle detection |
CN106548145A (en) * | 2016-10-31 | 2017-03-29 | 北京小米移动软件有限公司 | Image-recognizing method and device |
CN107688819A (en) * | 2017-02-16 | 2018-02-13 | 平安科技(深圳)有限公司 | The recognition methods of vehicle and device |
CN107016061A (en) * | 2017-03-15 | 2017-08-04 | 海尔优家智能科技(北京)有限公司 | Video monitoring document handling method and device |
CN108388888A (en) * | 2018-03-23 | 2018-08-10 | 腾讯科技(深圳)有限公司 | A kind of vehicle identification method, device and storage medium |
CN108596277A (en) * | 2018-05-10 | 2018-09-28 | 腾讯科技(深圳)有限公司 | A kind of testing vehicle register identification method, apparatus and storage medium |
CN109166120A (en) * | 2018-09-11 | 2019-01-08 | 百度在线网络技术(北京)有限公司 | For obtaining the method and device of information |
CN109446905A (en) * | 2018-09-26 | 2019-03-08 | 深圳壹账通智能科技有限公司 | Sign electronically checking method, device, computer equipment and storage medium |
CN109784415A (en) * | 2019-01-25 | 2019-05-21 | 北京地平线机器人技术研发有限公司 | The method and device of image-recognizing method and device, training convolutional neural networks |
Non-Patent Citations (2)
Title |
---|
LI-HUNG WANG ETAL.: "A real-time architecture of multiple features extraction for vehicle verification", 《2014 IEEE ASIA PACIFIC CONFERENCE ON CIRCUITS AND SYSTEMS (APCCAS)》 * |
蔡晓东等: "基于多分支卷积神经网络的车辆图像比对方法", 《电视技术》, vol. 40, no. 11 * |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111738228A (en) * | 2020-08-04 | 2020-10-02 | 杭州智诚惠通科技有限公司 | Multi-view vehicle feature matching method for hypermetrological evidence chain verification |
Also Published As
Publication number | Publication date |
---|---|
CN110135517B (en) | 2023-04-07 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110163153A (en) | The method and device on traffic mark board boundary for identification | |
CN110110811A (en) | Method and apparatus for training pattern, the method and apparatus for predictive information | |
CN109902186A (en) | Method and apparatus for generating neural network | |
CN109858445A (en) | Method and apparatus for generating model | |
CN109214584A (en) | Method and apparatus for passenger flow forecast amount | |
CN108898185A (en) | Method and apparatus for generating image recognition model | |
CN108985259A (en) | Human motion recognition method and device | |
CN107609502A (en) | Method and apparatus for controlling automatic driving vehicle | |
CN109191514A (en) | Method and apparatus for generating depth detection model | |
CN109410253B (en) | For generating method, apparatus, electronic equipment and the computer-readable medium of information | |
CN107481292A (en) | The attitude error method of estimation and device of vehicle-mounted camera | |
CN109767130A (en) | Method for controlling a vehicle and device | |
CN110009059A (en) | Method and apparatus for generating model | |
CN110119725A (en) | For detecting the method and device of signal lamp | |
CN109858444A (en) | The training method and device of human body critical point detection model | |
CN107918963A (en) | Information generating method and device for vehicle | |
CN108491823A (en) | Method and apparatus for generating eye recognition model | |
CN110059623A (en) | Method and apparatus for generating information | |
CN107515607A (en) | Control method and device for unmanned vehicle | |
CN109472264A (en) | Method and apparatus for generating object detection model | |
CN110348463A (en) | The method and apparatus of vehicle for identification | |
CN108509921A (en) | Method and apparatus for generating information | |
CN108133197A (en) | For generating the method and apparatus of information | |
CN107765691A (en) | Method and apparatus for controlling automatic driving vehicle | |
CN109902624A (en) | The method and apparatus of information for rendering |
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 |