CN110516703A - Vehicle identification method, device and storage medium based on artificial intelligence - Google Patents
Vehicle identification method, device and storage medium based on artificial intelligence Download PDFInfo
- Publication number
- CN110516703A CN110516703A CN201910651863.8A CN201910651863A CN110516703A CN 110516703 A CN110516703 A CN 110516703A CN 201910651863 A CN201910651863 A CN 201910651863A CN 110516703 A CN110516703 A CN 110516703A
- Authority
- CN
- China
- Prior art keywords
- vehicle
- image
- artificial intelligence
- vehicle identification
- vehicle 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.)
- Pending
Links
- 238000000034 method Methods 0.000 title claims abstract description 71
- 238000013473 artificial intelligence Methods 0.000 title claims abstract description 61
- 238000003860 storage Methods 0.000 title claims abstract description 30
- 238000012549 training Methods 0.000 claims abstract description 54
- 230000006870 function Effects 0.000 claims description 30
- 238000009826 distribution Methods 0.000 claims description 28
- 238000001914 filtration Methods 0.000 claims description 24
- 230000008569 process Effects 0.000 claims description 19
- 238000012545 processing Methods 0.000 claims description 15
- 238000004422 calculation algorithm Methods 0.000 claims description 13
- 230000002708 enhancing effect Effects 0.000 claims description 11
- 238000011946 reduction process Methods 0.000 claims description 10
- 238000009987 spinning Methods 0.000 claims description 7
- 238000013528 artificial neural network Methods 0.000 description 19
- 238000005516 engineering process Methods 0.000 description 12
- 238000010606 normalization Methods 0.000 description 11
- 238000012360 testing method Methods 0.000 description 10
- 238000010586 diagram Methods 0.000 description 6
- 230000009466 transformation Effects 0.000 description 6
- 230000008901 benefit Effects 0.000 description 5
- 238000004891 communication Methods 0.000 description 4
- 238000010801 machine learning Methods 0.000 description 4
- 230000001537 neural effect Effects 0.000 description 4
- 210000002569 neuron Anatomy 0.000 description 4
- 238000005070 sampling Methods 0.000 description 4
- 235000013399 edible fruits Nutrition 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 230000002779 inactivation Effects 0.000 description 3
- PXFBZOLANLWPMH-UHFFFAOYSA-N 16-Epiaffinine Natural products C1C(C2=CC=CC=C2N2)=C2C(=O)CC2C(=CC)CN(C)C1C2CO PXFBZOLANLWPMH-UHFFFAOYSA-N 0.000 description 2
- 241000239290 Araneae Species 0.000 description 2
- 239000006002 Pepper Substances 0.000 description 2
- 230000004913 activation Effects 0.000 description 2
- 230000006978 adaptation Effects 0.000 description 2
- 230000008859 change Effects 0.000 description 2
- 230000007423 decrease Effects 0.000 description 2
- 238000013135 deep learning Methods 0.000 description 2
- 238000002474 experimental method Methods 0.000 description 2
- 239000004744 fabric Substances 0.000 description 2
- 238000012804 iterative process Methods 0.000 description 2
- 239000004973 liquid crystal related substance Substances 0.000 description 2
- 239000000463 material Substances 0.000 description 2
- 208000019901 Anxiety disease Diseases 0.000 description 1
- 230000036506 anxiety Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 238000004146 energy storage Methods 0.000 description 1
- 210000005036 nerve Anatomy 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
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
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/30—Noise filtering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V2201/00—Indexing scheme relating to image or video recognition or understanding
- G06V2201/08—Detecting or categorising vehicles
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Artificial Intelligence (AREA)
- General Engineering & Computer Science (AREA)
- Evolutionary Computation (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Computational Linguistics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Evolutionary Biology (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Multimedia (AREA)
- Image Analysis (AREA)
Abstract
The present invention relates to field of artificial intelligence, disclose a kind of vehicle identification method based on artificial intelligence, it include: acquisition vehicle image, class label is marked to the vehicle image, the storage vehicle image for having class label, vehicle image data set is established, and pretreatment operation is carried out to the image in the vehicle image data set；Vehicle identification model is constructed, and using the vehicle image data set training vehicle identification model, uses the type of vehicle in vehicle identification model identification image；Vehicle image to be identified is obtained using camera, using the vehicle class identified in the vehicle image to be identified in above-mentioned trained vehicle identification model, and coding output is carried out to the class of vehicle using one-hot coding.The present invention also proposes a kind of vehicle identifier based on artificial intelligence and a kind of computer readable storage medium.The present invention can collect different expenses according to different automobile types to realize with automatic identification class of vehicle.
Description
Technical field
The present invention relates to field of computer technology more particularly to a kind of vehicle identification methods based on artificial intelligence, device
And computer readable storage medium.
Background technique
With the development of the social economy, vehicle guaranteeding organic quantity increases rapidly, the number that resident multiplies car trip increases, perhaps
The problem of more cities " parking difficulty, hardly possible parking ", is increasingly significant, supply and demand anxiety, arbitrary imposition of fees, parking offense, pass in and out parking lot be not easy,
It is generally existing to find phenomena such as parking lot is difficult.More preferable for urban transportation develops and preferably utilizes limited parking stall
Resource, many cities gradually come into effect the charging method that different Parking Fees is collected according to different automobile types.For exhausted big portion
Divide for parking lot, daily vehicle discrepancy amount is huge, judges that confusion easily occurs for type of vehicle only according to manual type.
Summary of the invention
The present invention provides a kind of vehicle identification method based on artificial intelligence, device and computer readable storage medium,
Main purpose is to provide a kind of automatic identification scheme of vehicle class.
To achieve the above object, a kind of vehicle identification method based on artificial intelligence provided by the invention, comprising:
Vehicle image is obtained, class label, the storage vehicle figure for having class label are marked to the vehicle image
Picture establishes vehicle image data set, and carries out pretreatment operation to the image in the vehicle image data set, wherein described
Vehicle image is included the vehicle image obtained from internet using web crawler and uses video camera in predeterminable area
The vehicle image of the discrepancy vehicle of capture；
The vehicle identification model of depth residual error network is constructed, and is known using the vehicle image data set training vehicle
The type of vehicle in other model identification image, and in the vehicle identification model training, enhance institute using random method for deactivating
The generalization ability for stating vehicle identification model, using batch standardized algorithm in the vehicle identification model training process, so that each
The input of layer neural network keeps same distribution；
Vehicle image to be identified is obtained, identifies the vehicle to be identified using above-mentioned trained vehicle identification model
Vehicle class in image, and coding output is carried out to the class of vehicle to be identified using one-hot coding.
Optionally, the image progress pretreatment operation in the vehicle image data set includes:
Noise reduction process is carried out to vehicle image using median filtering algorithm；
Data enhancing processing is carried out to vehicle image using overturning spinning solution.
Optionally, the noise reduction process includes:
Take the filter of N*N size, wherein N is positive integer；
Using presetted pixel point as filter archicenter value, all pixels value in the filter window of N*N size is carried out
Sequence；
Filter archicenter value is replaced with the median after the filter window internal sort of the N*N size.
Optionally, the output layer of the vehicle identification model
The recognition result of the vehicle identification model is exported using regression function, wherein the function is defined as follows institute
Show:
Wherein, ViIt is the output of classifier prime output unit, i indicates classification index, and e is nonterminating and non-recurring decimal, and C is
Total classification number, SiIndicate currentElement index and all elements index and ratio.
Optionally, this method further include:
It calculates using preset expense rule according to the type of the vehicle and exports corresponding charging standard.
In addition, to achieve the above object, the present invention also provides a kind of vehicle identifier based on artificial intelligence, the devices
Including memory and processor, the vehicle based on artificial intelligence that can be run on the processor is stored in the memory
Recognizer, the vehicle identification program based on artificial intelligence realize following steps when being executed by the processor:
Vehicle image is obtained, class label, the storage vehicle figure for having class label are marked to the vehicle image
Picture establishes vehicle image data set, and carries out pretreatment operation to the image in the vehicle image data set, wherein described
Vehicle image is included the vehicle image obtained from internet using web crawler and uses video camera in predeterminable area
The vehicle image of the discrepancy vehicle of capture；
The vehicle identification model of depth residual error network is constructed, and is known using the vehicle image data set training vehicle
The type of vehicle in other model identification image, and in the vehicle identification model training, enhance institute using random method for deactivating
The generalization ability for stating vehicle identification model, using batch standardized algorithm in the vehicle identification model training process, so that each
The input of layer neural network keeps same distribution.
Optionally, the image progress pretreatment operation in the vehicle image data set includes:
Noise reduction process is carried out to vehicle image using median filtering algorithm；
Data enhancing processing is carried out to vehicle image using overturning spinning solution.
Optionally, following steps are realized when the vehicle identification program based on artificial intelligence is executed by the processor:
Vehicle image to be identified is obtained, is identified in the vehicle image to be identified using above-mentioned trained vehicle identification model
Vehicle class, and coding output is carried out to the class of vehicle to be identified using one-hot coding.
Optionally, the output layer of the vehicle identification model
The recognition result of the vehicle identification model is exported using regression function, wherein the function is defined as follows institute
Show:
Wherein, ViIt is the output of classifier prime output unit, i indicates classification index, and e is nonterminating and non-recurring decimal, and C is
Total classification number, SiIndicate currentElement index and all elements index and ratio.
In addition, to achieve the above object, it is described computer-readable the present invention also provides a kind of computer readable storage medium
The vehicle identification program based on artificial intelligence is stored on storage medium, the vehicle identification program based on artificial intelligence can quilt
One or more processor executes, the step of to realize vehicle identification method based on artificial intelligence as described above.
Vehicle identification method based on artificial intelligence, device and computer readable storage medium proposed by the present invention obtain vehicle
Image marks class label to the vehicle image, and the storage vehicle image for having class label establishes vehicle image
Data set, and pretreatment operation is carried out to the image in the vehicle image data set；Vehicle identification model is constructed, and utilizes institute
The vehicle image data set training vehicle identification model is stated, the kind of vehicle in vehicle identification model identification image is used
Class；Obtain vehicle image to be identified using camera, using identified in above-mentioned trained vehicle identification model it is described to
Vehicle class in the vehicle image of identification, and coding output is carried out to the class of vehicle using one-hot coding.Therefore, this hair
It is bright different expenses to be collected according to different automobile types so as to realize with automatic identification class of vehicle.
Detailed description of the invention
Fig. 1 is the flow diagram for the vehicle identification method based on artificial intelligence that one embodiment of the invention provides；
Median filtering algorithm shows in the vehicle identification method based on artificial intelligence that Fig. 2 provides for one embodiment of the invention
It is intended to；
Fig. 3 is the schematic diagram of internal structure for the vehicle identifier based on artificial intelligence that one embodiment of the invention provides；
Vehicle based on artificial intelligence in the vehicle identifier based on artificial intelligence that Fig. 4 provides for one embodiment of the invention
The module diagram of recognizer.
The embodiments will be further described with reference to the accompanying drawings for the realization, the function and the advantages of the object of the present invention.
Specific embodiment
In order to make the objectives, technical solutions, and advantages of the present invention clearer, with reference to the accompanying drawings and embodiments, right
The present invention is further elaborated.It should be appreciated that described herein, specific examples are only used to explain the present invention, not
For limiting the present invention.Based on the embodiments of the present invention, those of ordinary skill in the art are not before making creative work
Every other embodiment obtained is put, shall fall within the protection scope of the present invention.
The description and claims of this application and term " first ", " second ", " third ", " in above-mentioned attached drawing
The (if present)s such as four " are to be used to distinguish similar objects, without being used to describe a particular order or precedence order.It should manage
The data that solution uses in this way are interchangeable under appropriate circumstances, so that the embodiments described herein can be in addition to illustrating herein
Or the sequence other than the content of description is implemented.In addition, the description of " first ", " second " etc. is used for description purposes only, without
It can be interpreted as its relative importance of indication or suggestion or implicitly indicate the quantity of indicated technical characteristic.It defines as a result,
The feature of " first ", " second " can explicitly or implicitly include at least one of the features.
Further, term " includes " and " having " and their any deformation, it is intended that cover non-exclusive packet
Contain, for example, what the process, method, system, product or equipment for containing a series of steps or units were not necessarily limited to be clearly listed
Those step or units, but may include be not clearly listed or it is intrinsic for these process, methods, product or equipment
Other step or units.
It in addition, the technical solution between each embodiment can be combined with each other, but must be with ordinary skill
Based on personnel can be realized, this technical side will be understood that when the combination of technical solution appearance is conflicting or cannot achieve
The combination of case is not present, also not the present invention claims protection scope within.
The present invention provides a kind of vehicle identification method based on artificial intelligence.It is one embodiment of the invention shown in referring to Fig.1
The flow diagram of the vehicle identification method based on artificial intelligence provided.This method can be executed by a device, the device
It can be by software and or hardware realization.
In the present embodiment, the vehicle identification method based on artificial intelligence includes:
S1, vehicle image is obtained, class label, the storage vehicle for having class label is marked to the vehicle image
Image establishes vehicle image data set, and carries out pretreatment operation to the image in the vehicle image data set.
Present pre-ferred embodiments obtain vehicle image using web crawler from internet.
Web crawlers, be otherwise known as webpage spider, network robot, be it is a kind of according to certain rules, automatically grab
The program or script of internet information.
Since training depth residual error network needs a large amount of markd samples, and Image Acquisition is carried out to all types of vehicles
A large amount of manpower and material resources can be expended, thus present pre-ferred embodiments propose it is a kind of using crawlers obtained from internet it is all kinds of
The method of type vehicle image, and class label is marked to every width vehicle image by way of handmarking, generate basic vehicle
Image data set.
Further, present pre-ferred embodiments can also use video camera to obtain predeterminable area, such as parking lot, expressway
Mouthful etc., enter and leave the vehicle image of vehicle.
In machine learning field, only when the sample of training set and test set obeys same distribution, identification model just meeting
Reach preferable classifying quality, therefore, on the basis of basic vehicle image data set, the present invention is for camera, such as stops
The actual vehicle pictures that parking lot camera obtains, the same mark for carrying out class of vehicle to image using the mode manually marked
Note.
The vehicle image data acquisition system in practical application that the present invention obtains basic vehicle image data set and camera
And at a vehicle image data set.The advantage of this method is, first, guarantee that depth residual error network there are enough training sets
It is trained and can achieve convergence.Second, due to joined parking lot in basic vehicle image data set, high speed enters and leaves cause for gossip
Vehicle image in the application scenarios of border, so that the distribution of training set and test set is even more like, this will be to depth residual error network mould
The classifying quality of type generates positive influences.
Since the picture quality in internet is very different, it is therefore desirable to be carried out to the vehicle image of vehicle image data set
Pretreatment operation.
In present pre-ferred embodiments, the pretreatment operation includes:
(1) noise reduction process is carried out to vehicle image using median filtering algorithm.
Median filtering is a kind of nonlinear digital filter technology, for removing the noise in image or other signals.
The main thought of median filtering is to check the sampling in input signal and judge whether it represents signal, uses odd-numbered samples
The observation window of composition realizes this functionality.Numerical value in watch window is ranked up, the intermediate value conduct among observation window
Output, then, abandons earliest value, obtains new sampling, repeats calculating process above.Median filtering is for the spot in image
It is particularly useful for spot noise and salt-pepper noise.
In present pre-ferred embodiments, the specific operation process of the median filtering is as shown in Figure 2.Wherein, the filtering
Device is N*N size, and N is positive integer, using the filter of 3*3 size in Fig. 2.As shown in Fig. 2, median filtering method will be with pre-
It is filter archicenter value, to all pixels value in the window of 3*3 size such as the pixel that pixel value is 1 if pixel
(i.e. shadow region in Fig. 2) is ranked up, and replaces the filtering with the median after the filter window internal sort of the N*N size
Device archicenter value.In inventive embodiments, it is worth and is replaced by 1 pixel value by pixel value 23, such as Fig. 2 of the result after median filtering
Right part shown in.
(2) data enhancing processing is carried out to vehicle image using overturning spinning solution.
The mode that vehicle image to be identified obtains in practical application may be different from vehicle image data set, such as bat
Direction/position/angles taken the photograph etc. are changed, this can cause training set to change with the sample distribution in test set.For
The identification accuracy of depth residual error network model is improved, the sample progress data enhancing processing that data are concentrated in this case is described
The content of data enhancing includes: that the vehicle image is carried out flip horizontal, flip vertical, at rotation transformation and affine transformation
Reason.
The robustness of depth residual error network can be enhanced in data enhancement methods of the present invention.
S2, building vehicle identification model, and using the vehicle image data set training vehicle identification model, it uses
The type of vehicle in the vehicle identification model identification image.
Present pre-ferred embodiments use depth residual error network (ResNet) as vehicle identification model, carry out vehicle class
Classification and Identification.
The overall thought of the ResNet model is that layer is expressed as study residual error function according to input.Experiment shows deep
It spends residual error network to be easier to optimize, and accuracy rate can be improved by increasing comparable depth.The core of ResNet is solution
It has determined increase depth bring side effect, that is, the degenerate problem of deep neural network.It in this way can be by merely increasing
The depth of network improves network performance.According to arbitrary function can be fitted on the neural network theory of multilayer, then
It can use some layers in ResNet and carry out fitting function.
Optionally, present pre-ferred embodiments are when using the vehicle image data set training vehicle identification model
Further include: enhance identification model generalization ability using random inactivation (dropout) technology, and/or uses Batch
Normalization accelerates model training speed and reduces the distributional difference between training set and data set.
The present invention increases the Generalization Capability of depth residual error network using dropout technology.Dropout refers to residual in depth
In the training process of poor network, neural network unit is temporarily abandoned it according to certain probability from network.For random
For gradient decline, due to being random drop, so each mini-batch is substantially in one different network of training.
Dropout technology can be effectively prevented deep neural network and over-fitting occurs, and improve the Generalization Capability of identification model.Each
Taking turns dropout in the iterative process of batch can all force a neural unit and other random select neural units total
With work.This eliminates the simultaneous adaptation between neuron node, so that the Generalization Capability of model is promoted.
Further, the present invention introduces batch standardization (Batch Normalization) skill in depth residual error network
Art.For vehicle identification task, identification model is based on a very important hypothesis, that is, the sample in training set and test set
It originally is independent identically distributed.But in the practical application of vehicle identification, guarantee training set and same point of the stringent obedience of test set
Cloth is relatively difficult.In addition to this, since the time for training depth residual error network to need to expend is more.The present embodiment uses
Batch Normalization technology solves these problems.Batch Normalization is exactly to instruct in depth residual error network
The input of each layer of neural network is made to keep same distribution during practicing.Because depth residual error network is doing nonlinear transformation
Preceding activation input value with network depth deepen perhaps its distribution gradually shifts or changes in the training process, it
So training convergence is slow, usually overall distribution is gradually close toward the bound both ends of the value interval of nonlinear function, so
This causes the gradient of bottom neural network when backpropagation to disappear, this is the slower and slower essence of trained depth residual error network convergence
Reason, and Batch Normalization is exactly by certain standardization means, any neuron of every layer of neural network
It is the standardized normal distribution that 0 variance is 1 that the distribution of input value is tied to mean value by force, substantially about by more and more inclined distribution
Beam returns the distribution of standard of comparison.This make activate input value fall in nonlinear function to input than more sensitive region, it is defeated in this way
The small variation entered will lead to the biggish variation of loss function.
Further, present pre-ferred embodiments use softmax regression function as the output layer of depth residual error network
Export network model recognition result.
Present pre-ferred embodiments use the softmax layers of output layer as depth residual error network to export vehicle identification knot
Fruit.Softmax has a very wide range of applications in machine learning and deep learning.Especially handling more classification problem sides
Face, the last output unit of classifier need softmax function to carry out numerical value processing.
It is defined as follows about softmax function shown:
Wherein, ViIt is the output of classifier prime output unit.I indicates classification index, and total classification number is C, and e is nothing
Limit not circulating decimal.SiIndicate currentElement index and all elements index and ratio.Softmax is by polytypic output
Numerical value is converted into relative probability, it is easier to understand compared with.The output of softmax characterizes relatively general between different classes of
Rate.In vehicle identification task, it is assumed that the corresponding probability value of the first dimension is maximum, then represents vehicle to be identified and belong to first kind vehicle
A possibility that type, is bigger.Softmax converts relative probability for serial number and makes result selection more intuitive.
S3, vehicle image to be identified is obtained using camera, is identified using above-mentioned trained vehicle identification model
The type of vehicle in image, and coding output is carried out to class of vehicle using one-hot coding.
The one-hot coding is also known as an efficient coding, is mainly compiled using N bit status register to N number of state
Code, each state has its independent register-bit, and only have when any one effectively.One-hot coding is classification variable
Expression as binary vector.This method requires tag along sort being mapped to integer value.Then, each integer value is represented as
Binary vector, other than the index position of integer is 1, remaining position is all 0.For vehicle identification task.Assuming that
Let us have ' AA brand A system ' and ' BB brand B system ' both vehicle tags, we can be by the integer value of ' AA brand A system ' point
With being 0, the integer value of ' BB brand B system ' is assigned as 1.The current method of salary distribution is called integer coding.So according to integer coding
The class label of our available vehicles.Being encoded to 0 ' AA brand A system ' will be indicated with binary vector [1,0], wherein the
0 index is marked as 1.Be encoded to 1 ' BB brand B system ' label will use binary vector [0,1] to indicate, wherein first
A index is marked as 1.The classification that one-hot coding carries out data is more accurate, since deep neural network cannot be used directly for counting
According to classification.The classification of data must be converted into number, and the variable that outputs and inputs of classification is the same.Integer coding is only
Suitable between class label there are in the case where comparison, such as ' cold ', ' warm ', ' heat '.But vehicle is known
For other task, due to there is no the presence of comparison between classification, so being only applicable to compile using only heat in this task
Code.
Further, it is calculated simultaneously the invention also includes S4, according to the type of the vehicle using preset expense rule
Export corresponding charging standard.
The present invention also provides a kind of vehicle identifiers based on artificial intelligence.Referring to shown in Fig. 3, implement for the present invention one
The schematic diagram of internal structure for the vehicle identifier based on artificial intelligence that example provides.
In the present embodiment, the vehicle identifier 1 based on artificial intelligence can be PC (Personal Computer, it is a
People's computer), it is also possible to the terminal devices such as smart phone, tablet computer, portable computer.The vehicle based on artificial intelligence is known
Other device 1 includes at least memory 11, processor 12, communication bus 13 and network interface 14.
Wherein, memory 11 include at least a type of readable storage medium storing program for executing, the readable storage medium storing program for executing include flash memory,
Hard disk, multimedia card, card-type memory (for example, SD or DX memory etc.), magnetic storage, disk, CD etc..Memory 11
It can be the internal storage unit of the vehicle identifier 1 based on artificial intelligence in some embodiments, such as should be based on artificial
The hard disk of the vehicle identifier 1 of intelligence.Memory 11 is also possible to the vehicle based on artificial intelligence in further embodiments
The plug-in type hard disk being equipped on the External memory equipment of identification device 1, such as the vehicle identifier 1 based on artificial intelligence, intelligence
Energy storage card (Smart Media Card, SMC), secure digital (Secure Digital, SD) card, flash card (Flash
Card) etc..Further, memory 11 can also both include the storage inside list of the vehicle identifier 1 based on artificial intelligence
Member also includes External memory equipment.Memory 11 can be not only used for storage and be installed on the vehicle identifier based on artificial intelligence
1 application software and Various types of data, such as the code etc. of the vehicle identification program 01 based on artificial intelligence can be also used for temporarily
Ground stores the data that has exported or will export.
Processor 12 can be in some embodiments a central processing unit (Central Processing Unit,
CPU), controller, microcontroller, microprocessor or other data processing chips, the program for being stored in run memory 11
Code or processing data, such as execute the vehicle identification program 01 etc. based on artificial intelligence.
Communication bus 13 is for realizing the connection communication between these components.
Network interface 14 optionally may include standard wireline interface and wireless interface (such as WI-FI interface), be commonly used in
Communication connection is established between the device 1 and other electronic equipments.
Optionally, which can also include user interface, and user interface may include display (Display), input
Unit such as keyboard (Keyboard), optional user interface can also include standard wireline interface and wireless interface.It is optional
Ground, in some embodiments, display can be light-emitting diode display, liquid crystal display, touch-control liquid crystal display and OLED
(Organic Light-Emitting Diode, Organic Light Emitting Diode) touches device etc..Wherein, display can also be appropriate
Referred to as display screen or display unit, for being shown in the information and use that handle in the vehicle identifier 1 based on artificial intelligence
In the visual user interface of display.
Fig. 3 illustrate only the vehicle identification program 01 with component 11-14 and based on artificial intelligence based on artificial intelligence
The vehicle identifier 1 of energy, it will be appreciated by persons skilled in the art that structure shown in fig. 1 is not constituted to based on artificial
The restriction of the vehicle identifier 1 of intelligence, may include than illustrating less perhaps more components or the certain components of combination,
Or different component layout.
In 1 embodiment of device shown in Fig. 3, the vehicle identification program based on artificial intelligence is stored in memory 11
01；Processor 12 realizes following steps when executing the vehicle identification program 01 based on artificial intelligence stored in memory 11:
Step 1: obtaining vehicle image, class label is marked to the vehicle image, has class label described in storage
Vehicle image establishes vehicle image data set, and carries out pretreatment operation to the image in the vehicle image data set.
Present pre-ferred embodiments obtain vehicle image using web crawler from internet.
Web crawlers, be otherwise known as webpage spider, network robot, be it is a kind of according to certain rules, automatically grab
The program or script of internet information.
Since training depth residual error network needs a large amount of markd samples, and Image Acquisition is carried out to all types of vehicles
A large amount of manpower and material resources can be expended, thus present pre-ferred embodiments propose it is a kind of using crawlers obtained from internet it is all kinds of
The method of type vehicle image, and class label is marked to every width vehicle image by way of handmarking, generate basic vehicle
Image data set.
Further, present pre-ferred embodiments can also use video camera to obtain predeterminable area, such as parking lot, expressway
Mouthful etc., enter and leave the vehicle image of vehicle.
In machine learning field, only when the sample of training set and test set obeys same distribution, identification model just meeting
Reach preferable classifying quality, therefore, on the basis of basic vehicle image data set, the present invention is for camera, such as stops
The actual vehicle pictures that parking lot camera obtains, the same mark for carrying out class of vehicle to image using the mode manually marked
Note.
The vehicle image data acquisition system in practical application that the present invention obtains basic vehicle image data set and camera
And at a vehicle image data set.The advantage of this method is, first, guarantee that depth residual error network there are enough training sets
It is trained and can achieve convergence.Second, due to joined parking lot in basic vehicle image data set, high speed enters and leaves cause for gossip
Vehicle image in the application scenarios of border, so that the distribution of training set and test set is even more like, this will be to depth residual error network mould
The classifying quality of type generates positive influences.
Since the picture quality in internet is very different, it is therefore desirable to be carried out to the vehicle image of vehicle image data set
Pretreatment operation.
In present pre-ferred embodiments, the pretreatment operation includes:
(2) noise reduction process is carried out to vehicle image using median filtering algorithm.
Median filtering is a kind of nonlinear digital filter technology, for removing the noise in image or other signals.
The main thought of median filtering is to check the sampling in input signal and judge whether it represents signal, uses odd-numbered samples
The observation window of composition realizes this functionality.Numerical value in watch window is ranked up, the intermediate value conduct among observation window
Output, then, abandons earliest value, obtains new sampling, repeats calculating process above.Median filtering is for the spot in image
It is particularly useful for spot noise and salt-pepper noise.
In present pre-ferred embodiments, the specific operation process of the median filtering is as shown in Figure 2.Wherein, the filtering
Device is N*N size, and N is positive integer, using the filter of 3*3 size in Fig. 2.As shown in Fig. 2, median filtering method will be with pre-
It is filter archicenter value, to all pixels value in the window of 3*3 size such as the pixel that pixel value is 1 if pixel
(i.e. shadow region in Fig. 2) is ranked up, and replaces the filtering with the median after the filter window internal sort of the N*N size
Device archicenter value.In inventive embodiments, it is worth and is replaced by 1 pixel value by pixel value 23, such as Fig. 2 of the result after median filtering
Right part shown in.
(2) data enhancing processing is carried out to vehicle image using overturning spinning solution.
The mode that vehicle image to be identified obtains in practical application may be different from vehicle image data set, such as bat
Direction/position/angles taken the photograph etc. are changed, this can cause training set to change with the sample distribution in test set.For
The identification accuracy of depth residual error network model is improved, the sample progress data enhancing processing that data are concentrated in this case is described
The content of data enhancing includes: that the vehicle image is carried out flip horizontal, flip vertical, at rotation transformation and affine transformation
Reason.
The robustness of depth residual error network can be enhanced in data enhancement methods of the present invention.
Step 2: building vehicle identification model, and the vehicle identification model is trained using the vehicle image data set,
Use the type of vehicle in vehicle identification model identification image.
Present pre-ferred embodiments use depth residual error network (ResNet) as vehicle identification model, carry out vehicle class
Classification and Identification.
The overall thought of the ResNet model is that layer is expressed as study residual error function according to input.Experiment shows deep
It spends residual error network to be easier to optimize, and accuracy rate can be improved by increasing comparable depth.The core of ResNet is solution
It has determined increase depth bring side effect, that is, the degenerate problem of deep neural network.It in this way can be by merely increasing
The depth of network improves network performance.According to arbitrary function can be fitted on the neural network theory of multilayer, then
It can use some layers in ResNet and carry out fitting function.
Optionally, present pre-ferred embodiments are when using the vehicle image data set training vehicle identification model
Further include: enhance identification model generalization ability using random inactivation (dropout) technology, and/or uses Batch
Normalization accelerates model training speed and reduces the distributional difference between training set and data set.
The present invention increases the Generalization Capability of depth residual error network using dropout technology.Dropout refers to residual in depth
In the training process of poor network, neural network unit is temporarily abandoned it according to certain probability from network.For random
For gradient decline.Due to being random drop, so each mini-batch is substantially in one different network of training.
Dropout technology can be effectively prevented deep neural network and over-fitting occurs, and improve the Generalization Capability of identification model.Each
Taking turns dropout in the iterative process of batch can all force a neural unit and other random select neural units total
With work.This eliminates the simultaneous adaptation between neuron node, so that the Generalization Capability of model is promoted.
Further, the present invention introduces batch standardization (Batch Normalization) skill in depth residual error network
Art.For vehicle identification task, identification model is based on a very important hypothesis, that is, the sample in training set and test set
It originally is independent identically distributed.But in the practical application of vehicle identification, guarantee training set and same point of the stringent obedience of test set
Cloth is relatively difficult.In addition to this, since the time for training depth residual error network to need to expend is more.The present embodiment uses
Batch Normalization technology solves these problems.Batch Normalization is exactly to instruct in depth residual error network
The input of each layer of neural network is made to keep same distribution during practicing.Because depth residual error network is doing nonlinear transformation
Preceding activation input value with network depth deepen perhaps its distribution gradually shifts or changes in the training process, it
So training convergence is slow, usually overall distribution is gradually close toward the bound both ends of the value interval of nonlinear function, so
This causes the gradient of bottom neural network when backpropagation to disappear, this is the slower and slower essence of trained depth residual error network convergence
Reason, and Batch Normalization is exactly by certain standardization means, any neuron of every layer of neural network
It is the standardized normal distribution that 0 variance is 1 that the distribution of input value is tied to mean value by force, substantially about by more and more inclined distribution
Beam returns the distribution of standard of comparison.This make activate input value fall in nonlinear function to input than more sensitive region, it is defeated in this way
The small variation entered will lead to the biggish variation of loss function.
Further, present pre-ferred embodiments use softmax regression function as the output layer of depth residual error network
Export network model recognition result.
Present pre-ferred embodiments use the softmax layers of output layer as depth residual error network to export vehicle identification knot
Fruit.Softmax has a very wide range of applications in machine learning and deep learning.Especially handling more classification problem sides
Face, the last output unit of classifier need softmax function to carry out numerical value processing.
It is defined as follows about softmax function shown:
Wherein, ViIt is the output of classifier prime output unit.I indicates classification index, and total classification number is C.SiIt indicates
Be currentElement index and all elements index and ratio.Softmax converts polytypic output numerical value to relatively
Probability, it is easier to understand compared with.The output of softmax characterizes the relative probability between different classes of.Appoint in vehicle identification
In business, it is assumed that the corresponding probability value of the first dimension is maximum, then represents a possibility that vehicle to be identified belongs to first kind type of vehicle
It is bigger.Softmax converts relative probability for serial number and makes result selection more intuitive.
Step 3: obtaining vehicle image to be identified using camera, above-mentioned trained vehicle identification model is used
It identifies the type of vehicle in image, and coding output is carried out to class of vehicle using one-hot coding.
The one-hot coding is also known as an efficient coding, is mainly compiled using N bit status register to N number of state
Code, each state has its independent register-bit, and only have when any one effectively.One-hot coding is classification variable
Expression as binary vector.This method requires tag along sort being mapped to integer value.Then, each integer value is represented as
Binary vector, other than the index position of integer is 1, remaining position is all 0.For vehicle identification task.Assuming that
Let us have ' AA brand A system ' and ' BB brand B system ' both vehicle tags, we can be by the integer value of ' AA brand A system ' point
With being 0, the integer value of ' BB brand B system ' is assigned as 1.The current method of salary distribution is called integer coding.So according to integer coding
The class label of our available vehicles.Being encoded to 0 ' AA brand A system ' will be indicated with binary vector [1,0], wherein the
0 index is marked as 1.Be encoded to 1 ' BB brand B system ' label will use binary vector [0,1] to indicate, wherein first
A index is marked as 1.The classification that one-hot coding carries out data is more accurate, since deep neural network cannot be used directly for counting
According to classification.The classification of data must be converted into number, and the variable that outputs and inputs of classification is the same.Integer coding is only
Suitable between class label there are in the case where comparison, such as ' cold ', ' warm ', ' heat '.But vehicle is known
For other task, due to there is no the presence of comparison between classification, so being only applicable to compile using only heat in this task
Code.
Further, the invention also includes Step 4: utilizing preset expense rule, meter according to the type of the vehicle
It calculates and exports corresponding charging standard.
Optionally, in other embodiments, the vehicle identification program based on artificial intelligence can also be divided into one or
The multiple modules of person, one or more module are stored in memory 11, and (the present embodiment is by one or more processors
Processor 12) it is performed to complete the present invention, the so-called module of the present invention is the series of computation for referring to complete specific function
Machine program instruction section, for describing the vehicle identification program based on artificial intelligence in the vehicle identifier based on artificial intelligence
Implementation procedure.
For example, referring to shown in Fig. 4, for the present invention is based in one embodiment of vehicle identifier of artificial intelligence based on people
The program module schematic diagram of the vehicle identification program of work intelligence, in the embodiment, the vehicle identification program based on artificial intelligence can
Module 10, model training module 20, class of vehicle identification module 30, expense computing module 40 are established to be divided into data set,
Illustratively:
Data set is established module 10 and is used for: obtaining vehicle image, class label is marked to the vehicle image, described in storage
Vehicle image with class label is established vehicle image data set, and is carried out to the image in the vehicle image data set
Pretreatment operation.
Optionally, the image progress pretreatment operation in the vehicle image data set includes:
Noise reduction process is carried out to vehicle image using median filtering algorithm；
Data enhancing processing is carried out to vehicle image using overturning spinning solution.
Optionally, the noise reduction process includes:
Take the filter of N*N size, wherein N is positive integer；
Using presetted pixel point as filter archicenter value, all pixels value in the filter window of N*N size is carried out
Sequence；
Filter archicenter value is replaced with the median after the filter window internal sort of the N*N size.
Model training module 20 is used for: building vehicle identification model, and using described in vehicle image data set training
Vehicle identification model uses the type of vehicle in vehicle identification model identification image.
It is optionally, described to utilize the vehicle image data set training vehicle identification model, further includes:
Enhance the generalization ability of the vehicle identification model using random inactivation technique；
Using Batch Normalization technology in the vehicle identification model training process, so that each layer of nerve
The input of network keeps same distribution；And
Regression function is used to export the identification knot of the vehicle identification model as the output layer of the vehicle identification model
Fruit.
Class of vehicle identification module 30 is used for: being obtained vehicle image to be identified using camera, has been trained using above-mentioned
The vehicle class in the vehicle image to be identified is identified in good vehicle identification model, and using one-hot coding to the vehicle
Classification carries out coding output.
Expense computing module 40 is used for: according to the type of the vehicle, using preset expense rule, being calculated and is exported pair
The charging standard answered.
Above-mentioned data set establishes module 10, model training module 20, class of vehicle identification module 30, expense computing module 40
Etc. program modules be performed realized functions or operations step and be substantially the same with above-described embodiment, details are not described herein.
In addition, the embodiment of the present invention also proposes a kind of computer readable storage medium, the computer readable storage medium
On be stored with the vehicle identification program based on artificial intelligence, the vehicle identification program based on artificial intelligence can be by one or more
A processor executes, to realize following operation:
Vehicle image is obtained, class label, the storage vehicle figure for having class label are marked to the vehicle image
Picture establishes vehicle image data set, and carries out pretreatment operation to the image in the vehicle image data set, wherein described
Vehicle image is included the vehicle image obtained from internet using web crawler and uses video camera in predeterminable area
The vehicle image of the discrepancy vehicle of capture；
The vehicle identification model of depth residual error network is constructed, and is known using the vehicle image data set training vehicle
The type of vehicle in other model identification image, and in the vehicle identification model training, enhance institute using random method for deactivating
The generalization ability for stating vehicle identification model, using batch standardized algorithm in the vehicle identification model training process, so that each
The input of layer neural network keeps same distribution；Vehicle image to be identified is obtained, above-mentioned trained vehicle identification is used
Model identifies the vehicle class in the vehicle image to be identified, and using one-hot coding to the class of vehicle to be identified
Carry out coding output.
Computer readable storage medium specific embodiment of the present invention and the above-mentioned vehicle identifier based on artificial intelligence
It is essentially identical with each embodiment of method, do not make tired state herein.
It should be noted that the serial number of the above embodiments of the invention is only for description, do not represent the advantages or disadvantages of the embodiments.And
The terms "include", "comprise" herein or any other variant thereof is intended to cover non-exclusive inclusion, so that packet
Process, device, article or the method for including a series of elements not only include those elements, but also including being not explicitly listed
Other element, or further include for this process, device, article or the intrinsic element of method.Do not limiting more
In the case where, the element that is limited by sentence "including a ...", it is not excluded that including process, device, the article of the element
Or there is also other identical elements in method.
Through the above description of the embodiments, those skilled in the art can be understood that above-described embodiment side
Method can be realized by means of software and necessary general hardware platform, naturally it is also possible to by hardware, but in many cases
The former is more preferably embodiment.Based on this understanding, technical solution of the present invention substantially in other words does the prior art
The part contributed out can be embodied in the form of software products, which is stored in one as described above
In storage medium (such as ROM/RAM, magnetic disk, CD), including some instructions are used so that terminal device (it can be mobile phone,
Computer, server or network equipment etc.) execute method described in each embodiment of the present invention.
The above is only a preferred embodiment of the present invention, is not intended to limit the scope of the invention, all to utilize this hair
Equivalent structure or equivalent flow shift made by bright specification and accompanying drawing content is applied directly or indirectly in other relevant skills
Art field, is included within the scope of the present invention.
Claims (10)
1. a kind of vehicle identification method based on artificial intelligence, which is characterized in that the described method includes:
Vehicle image is obtained, class label is marked to the vehicle image, the storage vehicle image for having class label is built
Vertical vehicle image data set, and pretreatment operation is carried out to the image in the vehicle image data set, wherein the vehicle figure
As including the vehicle image obtained from internet using web crawler and being captured using video camera in predeterminable area
Enter and leave the vehicle image of vehicle；
The vehicle identification model of depth residual error network is constructed, and utilizes the vehicle image data set training vehicle identification mould
Type identifies the type of vehicle in image, and in the vehicle identification model training, enhances the vehicle using random method for deactivating
The generalization ability of identification model, using batch standardized algorithm in the vehicle identification model training process, so that each layer of mind
Input through network keeps same distribution；
Vehicle image to be identified is obtained, identifies the vehicle figure to be identified using above-mentioned trained vehicle identification model
Vehicle class as in, and coding output is carried out to the class of vehicle to be identified using one-hot coding.
2. as described in claim 1 based on the vehicle identification method of artificial intelligence, which is characterized in that described to the vehicle figure
Include: as the image in data set carries out pretreatment operation
Noise reduction process is carried out to vehicle image using median filtering algorithm；
Data enhancing processing is carried out to vehicle image using overturning spinning solution.
3. as claimed in claim 2 based on the vehicle identification method of artificial intelligence, which is characterized in that the noise reduction process packet
It includes:
Take the filter of N*N size, wherein N is positive integer；
Using presetted pixel point as filter archicenter value, all pixels value in the filter window of N*N size is ranked up；
Filter archicenter value is replaced with the median after the filter window internal sort of the N*N size.
4. as described in claim 1 based on the vehicle identification method of artificial intelligence, which is characterized in that the vehicle identification model
Output layer
The recognition result of the vehicle identification model is exported using regression function, wherein shown in the function is defined as follows:
Wherein, ViIt is the output of classifier prime output unit, i indicates that classification index, e are nonterminating and non-recurring decimal, and C is total
Classification number, SiIndicate currentElement index and all elements index and ratio.
5. the vehicle identification method based on artificial intelligence as described in any one of Claims 1-4, which is characterized in that should
Method further include:
It calculates using preset expense rule according to the type of the vehicle and exports corresponding charging standard.
6. a kind of vehicle identifier based on artificial intelligence, which is characterized in that described device includes memory and processor, institute
The vehicle identification program based on artificial intelligence for being stored with and being run on memory on the processor is stated, it is described based on artificial
The vehicle identification program of intelligence realizes following steps when being executed by the processor:
Vehicle image is obtained, class label is marked to the vehicle image, the storage vehicle image for having class label is built
Vertical vehicle image data set, and pretreatment operation is carried out to the image in the vehicle image data set, wherein the vehicle figure
As including the vehicle image obtained from internet using web crawler and being captured using video camera in predeterminable area
Enter and leave the vehicle image of vehicle；
The vehicle identification model of depth residual error network is constructed, and utilizes the vehicle image data set training vehicle identification mould
Type identifies the type of vehicle in image, and in the vehicle identification model training, enhances the vehicle using random method for deactivating
The generalization ability of identification model, using batch standardized algorithm in the vehicle identification model training process, so that each layer of mind
Input through network keeps same distribution.
7. as claimed in claim 6 based on the vehicle identifier of artificial intelligence, which is characterized in that described to the vehicle figure
Include: as the image in data set carries out pretreatment operation
Noise reduction process is carried out to vehicle image using median filtering algorithm；
Data enhancing processing is carried out to vehicle image using overturning spinning solution.
8. as claimed in claim 6 based on the vehicle identifier of artificial intelligence, which is characterized in that described to be based on artificial intelligence
Vehicle identification program realize following steps when being executed by the processor: obtain vehicle image to be identified, using it is above-mentioned
Trained vehicle identification model identifies the vehicle class in the vehicle image to be identified, and using one-hot coding to described
Class of vehicle to be identified carries out coding output.
9. as claimed in claim 6 based on the vehicle identifier of artificial intelligence, which is characterized in that the vehicle identification model
Output layer
The recognition result of the vehicle identification model is exported using regression function, wherein shown in the function is defined as follows:
Wherein, ViIt is the output of classifier prime output unit, i indicates that classification index, e are nonterminating and non-recurring decimal, and C is total
Classification number, SiIndicate currentElement index and all elements index and ratio.
10. a kind of computer readable storage medium, which is characterized in that be stored on the computer readable storage medium based on people
The vehicle identification program of work intelligence, the vehicle identification program based on artificial intelligence can be held by one or more processor
Row, the step of to realize the vehicle identification method based on artificial intelligence as described in any one of claims 1 to 5.
Priority Applications (2)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910651863.8A CN110516703A (en) | 2019-07-18 | 2019-07-18 | Vehicle identification method, device and storage medium based on artificial intelligence |
| PCT/CN2019/116552 WO2021008018A1 (en) | 2019-07-18 | 2019-11-08 | Vehicle identification method and device employing artificial intelligence, and program and storage medium |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910651863.8A CN110516703A (en) | 2019-07-18 | 2019-07-18 | Vehicle identification method, device and storage medium based on artificial intelligence |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN110516703A true CN110516703A (en) | 2019-11-29 |
Family
ID=68622843
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910651863.8A Pending CN110516703A (en) | 2019-07-18 | 2019-07-18 | Vehicle identification method, device and storage medium based on artificial intelligence |
Country Status (2)
| Country | Link |
|---|---|
| CN (1) | CN110516703A (en) |
| WO (1) | WO2021008018A1 (en) |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109740735A (en) * | 2018-12-29 | 2019-05-10 | 百度在线网络技术（北京）有限公司 | Multiple neural network output method and device, server, computer-readable medium |
| CN111126224A (en) * | 2019-12-17 | 2020-05-08 | 成都通甲优博科技有限责任公司 | Vehicle detection method and classification recognition model training method |
| CN111145365A (en) * | 2019-12-17 | 2020-05-12 | 北京明略软件系统有限公司 | Method, device, computer storage medium and terminal for realizing classification processing |
| CN112566117A (en) * | 2020-11-06 | 2021-03-26 | 厦门大学 | Vehicle node identity recognition method and device based on metric learning |
| CN112598084A (en) * | 2021-03-02 | 2021-04-02 | 深圳金三立视频科技股份有限公司 | Vehicle type identification method and terminal based on image processing |
| CN113343881A (en) * | 2021-06-21 | 2021-09-03 | 浪潮云信息技术股份公司 | Vehicle brand and model fine-grained classification system and method based on deep learning |
| JP2023508639A (en) * | 2020-07-23 | 2023-03-03 | 株式会社アーバンベース | Data Augmented Based Spatial Analysis Model Learning Apparatus and Method |
| JP2023508640A (en) * | 2020-07-23 | 2023-03-03 | 株式会社アーバンベース | Data augmentation-based style analysis model learning device and method |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113822292B (en) * | 2021-02-05 | 2022-05-20 | 深圳信息职业技术学院 | Vehicle characteristic information storage method and device, computer equipment and storage medium |
| CN115601593A (en) * | 2022-10-18 | 2023-01-13 | 苏州魔视智能科技有限公司（Cn） | Image classification method, device, equipment and medium |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108875803A (en) * | 2018-05-30 | 2018-11-23 | 长安大学 | A kind of detection of harmful influence haulage vehicle and recognition methods based on video image |
| CN109063719A (en) * | 2018-04-23 | 2018-12-21 | 湖北工业大学 | A kind of image classification method of co-ordinative construction similitude and category information |
| US20190114511A1 (en) * | 2017-10-16 | 2019-04-18 | Illumina, Inc. | Deep Learning-Based Techniques for Training Deep Convolutional Neural Networks |
Family Cites Families (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US10635927B2 (en) * | 2017-03-06 | 2020-04-28 | Honda Motor Co., Ltd. | Systems for performing semantic segmentation and methods thereof |
| CN108171134A (en) * | 2017-12-20 | 2018-06-15 | 中车工业研究院有限公司 | A kind of operational motion discrimination method and device |
| CN109190444B (en) * | 2018-07-02 | 2021-05-18 | 南京大学 | Method for realizing video-based toll lane vehicle feature recognition system |
-
2019
- 2019-07-18 CN CN201910651863.8A patent/CN110516703A/en active Pending
- 2019-11-08 WO PCT/CN2019/116552 patent/WO2021008018A1/en active Application Filing
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20190114511A1 (en) * | 2017-10-16 | 2019-04-18 | Illumina, Inc. | Deep Learning-Based Techniques for Training Deep Convolutional Neural Networks |
| CN109063719A (en) * | 2018-04-23 | 2018-12-21 | 湖北工业大学 | A kind of image classification method of co-ordinative construction similitude and category information |
| CN108875803A (en) * | 2018-05-30 | 2018-11-23 | 长安大学 | A kind of detection of harmful influence haulage vehicle and recognition methods based on video image |
Non-Patent Citations (1)
| Title |
|---|
| 田萱等: "《基于深度学习的图像语义分割技术》", vol. 978, 31 May 2019, 北京：海洋出版社, pages: 87 - 91 * |
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109740735A (en) * | 2018-12-29 | 2019-05-10 | 百度在线网络技术（北京）有限公司 | Multiple neural network output method and device, server, computer-readable medium |
| US11514296B2 (en) | 2018-12-29 | 2022-11-29 | Baidu Online Network Technology (Beijing) Co., Ltd. | Output method and apparatus for multiple neural network, server and computer readable storage medium |
| CN111126224A (en) * | 2019-12-17 | 2020-05-08 | 成都通甲优博科技有限责任公司 | Vehicle detection method and classification recognition model training method |
| CN111145365A (en) * | 2019-12-17 | 2020-05-12 | 北京明略软件系统有限公司 | Method, device, computer storage medium and terminal for realizing classification processing |
| JP2023508639A (en) * | 2020-07-23 | 2023-03-03 | 株式会社アーバンベース | Data Augmented Based Spatial Analysis Model Learning Apparatus and Method |
| JP2023508640A (en) * | 2020-07-23 | 2023-03-03 | 株式会社アーバンベース | Data augmentation-based style analysis model learning device and method |
| JP7325637B2 (en) | 2020-07-23 | 2023-08-14 | 株式会社アーバンベース | Data augmentation-based style analysis model learning device and method |
| JP7362924B2 (en) | 2020-07-23 | 2023-10-17 | 株式会社アーバンベース | Data augmentation-based spatial analysis model learning device and method |
| CN112566117A (en) * | 2020-11-06 | 2021-03-26 | 厦门大学 | Vehicle node identity recognition method and device based on metric learning |
| CN112566117B (en) * | 2020-11-06 | 2023-12-08 | 厦门大学 | Vehicle node identity recognition method and device based on metric learning |
| CN112598084A (en) * | 2021-03-02 | 2021-04-02 | 深圳金三立视频科技股份有限公司 | Vehicle type identification method and terminal based on image processing |
| CN113343881A (en) * | 2021-06-21 | 2021-09-03 | 浪潮云信息技术股份公司 | Vehicle brand and model fine-grained classification system and method based on deep learning |
Also Published As
| Publication number | Publication date |
|---|---|
| WO2021008018A1 (en) | 2021-01-21 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110516703A (en) | Vehicle identification method, device and storage medium based on artificial intelligence | |
| US7657089B2 (en) | Automatic classification of photographs and graphics | |
| CN107273836A (en) | A kind of pedestrian detection recognition methods, device, model and medium | |
| CN110097051A (en) | Image classification method, device and computer readable storage medium | |
| CN110009556A (en) | Image background weakening method, device, storage medium and electronic equipment | |
| CN111160350B (en) | Portrait segmentation method, model training method, device, medium and electronic equipment | |
| CN110334272A (en) | The intelligent answer method, apparatus and computer storage medium of knowledge based map | |
| CN109492612A (en) | Fall detection method and its falling detection device based on skeleton point | |
| CN110443239A (en) | The recognition methods of character image and its device | |
| CN108596944A (en) | A kind of method, apparatus and terminal device of extraction moving target | |
| CN108229341A (en) | Sorting technique and device, electronic equipment, computer storage media, program | |
| CN109376963A (en) | A kind of criminal case charge law article unified prediction neural network based | |
| Yan et al. | Multiscale convolutional neural networks for hand detection | |
| CN112418360B (en) | Convolutional neural network training method, pedestrian attribute identification method and related equipment | |
| CN106874924A (en) | A kind of recognition methods of picture style and device | |
| Pham et al. | Evaluation of deep models for real-time small object detection | |
| CN110135421A (en) | Licence plate recognition method, device, computer equipment and computer readable storage medium | |
| CN106874913A (en) | A kind of vegetable detection method | |
| CN107392254A (en) | A kind of semantic segmentation method by combining the embedded structural map picture from pixel | |
| CN112308802A (en) | Image analysis method and system based on big data | |
| CN114495060B (en) | Road traffic marking recognition method and device | |
| CN110210479A (en) | A kind of text information extraction method on waste items | |
| Alonso et al. | Enhancing v-slam keyframe selection with an efficient ConvNet for semantic analysis | |
| CN109829484A (en) | A kind of dress ornament classification method, equipment and computer readable storage medium | |
| CN117809236A (en) | Substation illegal behavior adaptation method based on multitask learning |
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 |