CN110458136A - A kind of traffic sign recognition method, device and equipment - Google Patents

A kind of traffic sign recognition method, device and equipment Download PDF

Info

Publication number
CN110458136A
CN110458136A CN201910764738.8A CN201910764738A CN110458136A CN 110458136 A CN110458136 A CN 110458136A CN 201910764738 A CN201910764738 A CN 201910764738A CN 110458136 A CN110458136 A CN 110458136A
Authority
CN
China
Prior art keywords
traffic sign
trained
neural networks
impulsive neural
sign images
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
Application number
CN201910764738.8A
Other languages
Chinese (zh)
Other versions
CN110458136B (en
Inventor
刘怡俊
李逸聪
叶武剑
刘文杰
张子文
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Guangdong University of Technology
Original Assignee
Guangdong University of Technology
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Guangdong University of Technology filed Critical Guangdong University of Technology
Priority to CN201910764738.8A priority Critical patent/CN110458136B/en
Publication of CN110458136A publication Critical patent/CN110458136A/en
Application granted granted Critical
Publication of CN110458136B publication Critical patent/CN110458136B/en
Active legal-status Critical Current
Anticipated expiration legal-status Critical

Links

Classifications

    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06FELECTRIC DIGITAL DATA PROCESSING
    • G06F18/00Pattern recognition
    • G06F18/20Analysing
    • G06F18/21Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
    • G06F18/213Feature extraction, e.g. by transforming the feature space; Summarisation; Mappings, e.g. subspace methods
    • GPHYSICS
    • G06COMPUTING; CALCULATING OR COUNTING
    • G06VIMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00Scenes; Scene-specific elements
    • G06V20/50Context or environment of the image
    • G06V20/56Context or environment of the image exterior to a vehicle by using sensors mounted on the vehicle
    • G06V20/58Recognition of moving objects or obstacles, e.g. vehicles or pedestrians; Recognition of traffic objects, e.g. traffic signs, traffic lights or roads
    • G06V20/582Recognition of moving objects or obstacles, e.g. vehicles or pedestrians; Recognition of traffic objects, e.g. traffic signs, traffic lights or roads of traffic signs

Landscapes

  • Engineering & Computer Science (AREA)
  • Theoretical Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Computer Vision & Pattern Recognition (AREA)
  • Data Mining & Analysis (AREA)
  • General Physics & Mathematics (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Evolutionary Biology (AREA)
  • Evolutionary Computation (AREA)
  • Bioinformatics & Computational Biology (AREA)
  • General Engineering & Computer Science (AREA)
  • Bioinformatics & Cheminformatics (AREA)
  • Artificial Intelligence (AREA)
  • Multimedia (AREA)
  • Image Analysis (AREA)

Abstract

This application discloses a kind of traffic sign recognition method, device and equipment, and wherein method includes: to obtain Traffic Sign Images to be identified;Traffic Sign Images to be identified are input in trained deepness belief network model and carry out feature extraction, obtain the corresponding first eigenvector of Traffic Sign Images to be identified;First eigenvector is converted into the first pulse train;First pulse train is input in trained impulsive neural networks, the recognition result of trained impulsive neural networks output is obtained.The application carries out Traffic Sign Recognition in such a way that deepness belief network model is combined with impulsive neural networks, it does not need to carry out manual features extraction, greatly reduce manual intervention, improve recognition speed, the advantages of by making full use of deepness belief network model and impulsive neural networks, recognition result is improved, solves that existing Traffic Sign Recognition accuracy is low, slow-footed technical problem.

Description

A kind of traffic sign recognition method, device and equipment
Technical field
This application involves image identification technical field more particularly to a kind of traffic sign recognition methods, device and equipment.
Background technique
Traffic sign is the important sources that information is obtained in vehicle travel process, identification pair accurate to traffic sign, quick In ensure traffic safety, traffic order, improve traffic efficiency important in inhibiting, also unmanned ground to what is currently risen Study carefully significant.
Traditional traffic sign recognition method is known using the methods of images match, the combination of feature extraction and classifying device Not, human intervention is more, low, slow-footed problem that there are recognition accuracies.
Summary of the invention
This application provides a kind of traffic sign recognition method, device and equipment, know for solving existing traffic sign Other accuracy is low, slow-footed technical problem.
In view of this, the application first aspect provides a kind of traffic sign recognition method, comprising:
Obtain Traffic Sign Images to be identified;
The Traffic Sign Images to be identified are input in trained deepness belief network model and carry out feature extraction, Obtain the corresponding first eigenvector of the Traffic Sign Images to be identified;
The first eigenvector is converted into the first pulse train;
First pulse train is input in trained impulsive neural networks, the trained pulse mind is obtained The recognition result exported through network.
Preferably, further includes:
Obtain traffic indication map image set to be trained;
The Traffic Sign Images to be trained that the Traffic Sign Images to be trained are concentrated are input to trained depth letter It reads and carries out feature extraction in network model, obtain the corresponding second feature vector of the Traffic Sign Images to be trained;
The second feature vector is converted based on time lag phase code, obtains the second pulse train;
Second pulse train is input in impulsive neural networks, the impulsive neural networks are trained;
The impulsive neural networks are calculated to the recognition accuracy of the Traffic Sign Images to be trained, when the identification is quasi- When true rate is higher than threshold value, training is completed, and obtains trained impulsive neural networks.
Preferably, the Traffic Sign Images to be trained that the Traffic Sign Images to be trained are concentrated are input to training Carry out feature extraction in good deepness belief network model, obtain the corresponding second feature of the Traffic Sign Images to be trained to Amount, before further include:
The Traffic Sign Images to be trained are pre-processed.
Preferably, described pre-process includes:
Size normalized is carried out to the Traffic Sign Images to be trained based on bilinear interpolation algorithm, obtains normalizing Change treated Traffic Sign Images to be trained.
Preferably, described that second pulse train is input in impulsive neural networks, to the impulsive neural networks It is trained, comprising:
Second pulse train is input in impulsive neural networks, based on three pulse STDP in conjunction with threshold value plasticity Learning method the impulsive neural networks are trained.
The application second aspect provides a kind of Traffic Sign Recognition device, comprising:
First image collection module, for obtaining Traffic Sign Images to be identified;
Fisrt feature extraction module, for the Traffic Sign Images to be identified to be input to trained depth conviction net Feature extraction is carried out in network model, obtains the corresponding first eigenvector of the Traffic Sign Images to be identified;
First conversion module, for the first eigenvector to be converted into the first pulse train;
Identification module, for first pulse train to be input in trained impulsive neural networks, described in acquisition The recognition result of trained impulsive neural networks output.
Preferably, further includes:
Second image collection module, for obtaining traffic indication map image set to be trained;
Second feature extraction module, the Traffic Sign Images to be trained for concentrating the Traffic Sign Images to be trained It is input in trained deepness belief network model and carries out feature extraction, it is corresponding to obtain the Traffic Sign Images to be trained Second feature vector;
Second conversion module obtains second for converting based on time lag phase code to the second feature vector Pulse train;
Training module, for second pulse train to be input in impulsive neural networks, to the pulse nerve net Network is trained;
Computing module is accurate to the identification of the Traffic Sign Images to be trained for calculating the impulsive neural networks Rate, when the recognition accuracy is higher than threshold value, training is completed, and obtains trained impulsive neural networks.
Preferably, further includes:
Preprocessing module, for being pre-processed to the Traffic Sign Images to be trained.
The application third aspect provides a kind of Traffic Sign Recognition equipment characterized by comprising the equipment includes Processor and memory;
Said program code is transferred to the processor for storing program code by the memory;
The processor is used for according to the described in any item traffic marks of instruction execution first aspect in said program code Will recognition methods.
The application fourth aspect provides a kind of computer readable storage medium, which is characterized in that described computer-readable Storage medium is for storing program code, and said program code is for executing the described in any item Traffic Sign Recognitions of first aspect Method.
As can be seen from the above technical solutions, the embodiment of the present application has the advantage that
In the application, a kind of traffic sign recognition method is provided, comprising: obtain Traffic Sign Images to be identified;It will be to Identification Traffic Sign Images, which are input in trained deepness belief network model, carries out feature extraction, obtains traffic mark to be identified The corresponding first eigenvector of will image;First eigenvector is converted into the first pulse train;First pulse train is inputted Into trained impulsive neural networks, the recognition result of trained impulsive neural networks output is obtained.The application utilizes instruction The deepness belief network model perfected carries out feature extraction to Traffic Sign Images, does not need to carry out manual features extraction, pass through Deepness belief network model carries out Feature Dimension Reduction and feature selecting to the Traffic Sign Images of input, greatly reduces artificial dry In advance, high-level feature is extracted from original Traffic Sign Images, and has screened out the noise information of redundancy, helps to mention The recognition result of high succeeding impulse neural network, in such a way that deepness belief network model is combined with impulsive neural networks into Row Traffic Sign Recognition, improves recognition speed, by making full use of the excellent of deepness belief network model and impulsive neural networks Point, improves recognition result, solves that existing Traffic Sign Recognition accuracy is low, slow-footed technical problem.
Detailed description of the invention
Fig. 1 is a kind of flow diagram of one embodiment of traffic sign recognition method provided by the present application;
Fig. 2 is a kind of flow diagram of another embodiment of traffic sign recognition method provided by the present application;
Fig. 3 is a kind of structural schematic diagram of one embodiment of Traffic Sign Recognition device provided by the present application.
Specific embodiment
In order to make those skilled in the art more fully understand application scheme, below in conjunction in the embodiment of the present application Attached drawing, the technical scheme in the embodiment of the application is clearly and completely described, it is clear that described embodiment is only this Apply for a part of the embodiment, instead of all the embodiments.Based on the embodiment in the application, those of ordinary skill in the art exist Every other embodiment obtained under the premise of creative work is not made, shall fall in the protection scope of this application.
In order to make it easy to understand, referring to Fig. 1, a kind of one embodiment of traffic sign recognition method provided by the present application, Include:
Step 101, Traffic Sign Images to be identified are obtained.
It should be noted that having undesirable image due to that may exist in the image got, that is, exist not Image containing traffic sign can sieve the Traffic Sign Images of acquisition in order not to have an impact to the result of identification Choosing, screens out the Traffic Sign Images of some images not comprising traffic sign and some smudgy Chu, will meet the requirements Traffic Sign Images as the Traffic Sign Images to be identified finally got.
Step 102, Traffic Sign Images to be identified are input in trained deepness belief network model and carry out feature It extracts, obtains the corresponding first eigenvector of Traffic Sign Images to be identified.
It should be noted that deepness belief network model (deep belief network, DBN) in the present embodiment by Two layer depth Boltzmann machine models (deep Boltzmann machine, DBM) and two layers of DBN model fusion are constituted, wherein DBM model and DBN model are with limited Boltzmann machine (Restricted Boltzmann Machine, RBM) for basic structure At unit, the difference is that, DBM interlayer is undirected connection, and DBN interlayer is directed connection.
Low layer in deepness belief network model is using two layers strong of DBM of information reduction degree to Traffic Sign Images to be identified Preliminary dimensionality reduction is carried out, after the obtaining denoising and higher feature of integrity degree, using obtained feature as the defeated of two layers DBN Enter, higher feature is then extracted by two layers of DBN.By carrying out unsupervised training respectively to DBM and DBN and having supervision micro- It adjusts, finally obtains trained deepness belief network model, by trained deepness belief network model, extract friendship to be identified The high-level characteristic of logical sign image, helps to improve the recognition result of succeeding impulse neural network.
Step 103, first eigenvector is converted into the first pulse train.
It should be noted that the input of impulsive neural networks is indicated with pulse train, so needing special by the first of extraction Sign vector is converted into the first pulse train, so as to mutually be fitted with subsequent using impulsive neural networks progress Traffic Sign Recognition It answers, impulsive neural networks is preferably identified.
Step 104, the first pulse train is input in trained impulsive neural networks, obtains trained pulse mind The recognition result exported through network.
It should be noted that traditional artificial neural network is that pulse to biological neuron is provided frequency and encoded, The simulation in the generally given section of the output of neuron, operational capability and biological authenticity are weaker than impulsive neural networks, and adopt It carries out Traffic Sign Recognition with trained impulsive neural networks to be conducive to improve recognition result, therefore, the present embodiment is using instruction The impulsive neural networks perfected carry out Traffic Sign Recognition.
It has been found that the method combined in the prior art using images match, feature extraction and classifying device, manual intervention It is more, low, slow-footed problem that there are recognition accuracies.It therefore, is solution these problems existing in the prior art, applicant It is proposed the traffic sign recognition method provided in the embodiment of the present application, this method has reached following technical effect:
Traffic sign recognition method provided by the embodiments of the present application, by utilizing trained deepness belief network model pair Traffic Sign Images carry out feature extraction, do not need to carry out manual features extraction, by deepness belief network model to input Traffic Sign Images carry out Feature Dimension Reduction and feature selecting, manual intervention are greatly reduced, from original Traffic Sign Images High-level feature is extracted, and has screened out the noise information of redundancy, helps to improve the identification of succeeding impulse neural network As a result, carrying out Traffic Sign Recognition in such a way that deepness belief network model is combined with impulsive neural networks, knowledge is improved Other speed the advantages of by making full use of deepness belief network model and impulsive neural networks, improves recognition result, solves Existing Traffic Sign Recognition accuracy is low, slow-footed technical problem.
In order to make it easy to understand, referring to Fig. 2, a kind of another implementation of traffic sign recognition method provided by the present application Example, comprising:
Step 201, traffic indication map image set to be trained is obtained.
It should be noted that the Traffic Sign Images to be trained of the traffic indication map image set to be trained in the present embodiment come from In German Traffic Sign Recognition database (German traffic sign recognition benchmark, GTSRB).
Step 202, the Traffic Sign Images to be trained that trained Traffic Sign Images are concentrated are treated to be pre-processed.
It should be noted that deepness belief network carries out feature extraction for convenience, trained traffic indication map can be treated Traffic Sign Images to be trained in image set carry out size normalized, can treat trained friendship using bilinear interpolation algorithm Logical sign image carries out size normalization, and the Traffic Sign Images to be trained after normalization are onesize, such as size is 48 × 48 or 28 × 28 pixels.
There are the Traffic Sign Images of obscure portions, large area watermark in GTSRB database, it can be screened, is sieved Low-quality Traffic Sign Images are fallen in choosing, leave the Traffic Sign Images of high quality, are conducive to deepness belief network and have extracted The characteristic information of benefit, to improve the recognition result of impulsive neural networks.
Step 203, pretreated Traffic Sign Images to be trained are input to trained deepness belief network model Middle carry out feature extraction obtains the corresponding second feature vector of Traffic Sign Images to be trained.
Step 204, second feature vector is converted based on time lag phase code, obtains the second pulse train.
It should be noted that common coding mode has time lag coding, phase code, encoded in the present embodiment using time lag The mode combined with phase code encodes the second feature vector of extraction, i.e. time lag phase code, using time lag phase Better pulse train can be generated in position coding, helps to improve the Traffic Sign Recognition result of impulsive neural networks.
Step 205, the second pulse train is input in impulsive neural networks, impulsive neural networks is trained.
It should be noted that the impulsive neural networks in the present embodiment use LIF (Leaky Integrate-and- Fire) neuron, wherein first layer is competition layer, is made of multiple LIF neurons, uses side in competition layer between neuron The mode of inhibition achievees the purpose that competitive study, lateral inhibition process specifically: acts whenever a neuron bursts out one Current potential, which can reset to original state immediately, and enter refractory period, and other all neurons reset to tranquillization film electricity Position, into the phase of inhibition.Whenever there is neuron to provide pulse, all neurons can all be resetted, then be renewed competition, by competing It strives to learn, to be trained to impulsive neural networks;The second layer of impulsive neural networks is output layer, and output layer output is every The similarity value of a classification, similarity value is smaller, illustrates that similarity is higher, and the label of the smallest class of similarity value is final knowledge Not as a result, the specific steps of the calculating of similarity value are as follows:
Assuming that input picture matrix is I=xij∈Rn×n, input picture matrix is standardized, it may be assumed that
x′ij=(xij-xmin)/(xmax-xmin)
Wherein, xmax、xminMaximum, minimum pixel value, x ' in respectively IijFor the input picture after standardization.
Assuming that the neuron that label is L is denoted as respectivelyWherein,For m-th that label is L Neuron, MLFor the number of neuron, L=0,1 ..., 9;Label is m-th of neuron of L to the corresponding pulse of input picture The pulse number of sequence granting isThe corresponding receptive field weight matrix of m-th of neuron that label is L isBy label For the M of LLThe pulse number of a neuron is multiplied accumulating with receptive field weight, and obtaining label is the neuron of L to input picture Reconstructed image, it may be assumed that
Assuming that RL=rij∈Rn×n, it is equally standardized, the reconstructed image r after obtaining standardization ′ij, the similarity value S of the reconstructed image after normalized treated input picture and standardizationL, specific similarity It is as follows to be worth calculation formula:
Obtain 10 similarity numerical value S0,S1,…,S9, compare S0,S1..., S9Size, the smallest class of similarity value Label is final recognition result, it is assumed that S6Minimum, then classification representated by label 6 is final recognition result.
Impulsive neural networks are trained using learning method of the three pulse STDP in conjunction with threshold value plasticity, three pulses STDP learning method is for cynapse, and threshold value plasticity learning method is the threshold potential for neuron, using three arteries and veins Learning method of the STDP in conjunction with threshold value plasticity is rushed, so that the granting frequency of neuron is limited.
Step 206, the recognition accuracy that impulsive neural networks treat trained Traffic Sign Images is calculated, recognition accuracy is worked as When higher than threshold value, training is completed, and obtains trained impulsive neural networks.
It should be noted that recognition accuracy passes through the Traffic Sign Images number to be trained correctly identified and needs to be instructed The ratio calculation for practicing picture number obtains, when recognition accuracy is higher than preset threshold value, then it is assumed that training is completed, and is stopped Training, obtains trained impulsive neural networks.
Step 207, Traffic Sign Images to be identified are obtained.
Step 208, Traffic Sign Images to be identified are input in trained deepness belief network model and carry out feature It extracts, obtains the corresponding first eigenvector of Traffic Sign Images to be identified.
Step 209, first eigenvector is converted into the first pulse train.
Step 210, the first pulse train is input in trained impulsive neural networks, obtains trained pulse mind The recognition result exported through network.
It should be noted that step 101 of the step 207 into step 210 and a upper embodiment in the embodiment of the present application It is consistent to step 104, it is no longer repeated herein.
In order to make it easy to understand, referring to Fig. 3, a kind of one embodiment of Traffic Sign Recognition device provided by the invention, Include:
First image collection module 301, for obtaining Traffic Sign Images to be identified;
Fisrt feature extraction module 302, for Traffic Sign Images to be identified to be input to trained depth conviction net Feature extraction is carried out in network model, obtains the corresponding first eigenvector of Traffic Sign Images to be identified;
First conversion module 303, for first eigenvector to be converted into the first pulse train;
Identification module 304, for the first pulse train to be input in trained impulsive neural networks, acquisition is trained Impulsive neural networks output recognition result.
Further, further includes:
Second image collection module 305, for obtaining traffic indication map image set to be trained;
Second feature extraction module 306, the Traffic Sign Images to be trained for concentrating Traffic Sign Images to be trained It is input in trained deepness belief network model and carries out feature extraction, obtain Traffic Sign Images to be trained corresponding second Feature vector;
Second conversion module 307 obtains the second arteries and veins for converting using time lag phase code to second feature vector Rush sequence;
Training module 308 carries out impulsive neural networks for the second pulse train to be input in impulsive neural networks Training;
Computing module 309 treats the recognition accuracy of trained Traffic Sign Images for calculating impulsive neural networks, works as knowledge When other accuracy rate is higher than threshold value, training is completed, and obtains trained impulsive neural networks.
Further, further includes:
Preprocessing module 310 is pre-processed for treating trained Traffic Sign Images.
In several embodiments provided herein, it should be understood that disclosed device and method can pass through it Its mode is realized.For example, the apparatus embodiments described above are merely exemplary, for example, the division of the unit, only Only a kind of logical function partition, there may be another division manner in actual implementation, such as multiple units or components can be tied Another system is closed or is desirably integrated into, or some features can be ignored or not executed.Another point, it is shown or discussed Mutual coupling, direct-coupling or communication connection can be through some interfaces, the INDIRECT COUPLING or logical of device or unit Letter connection can be electrical property, mechanical or other forms.
The unit as illustrated by the separation member may or may not be physically separated, aobvious as unit The component shown may or may not be physical unit, it can and it is in one place, or may be distributed over multiple In network unit.It can select some or all of unit therein according to the actual needs to realize the mesh of this embodiment scheme 's.
It, can also be in addition, each functional unit in each embodiment of the application can integrate in one processing unit It is that each unit physically exists alone, can also be integrated in one unit with two or more units.Above-mentioned integrated list Member both can take the form of hardware realization, can also realize in the form of software functional units.
If the integrated unit is realized in the form of SFU software functional unit and sells or use as independent product When, it can store in a computer readable storage medium.Based on this understanding, the technical solution of the application is substantially The all or part of the part that contributes to existing technology or the technical solution can be in the form of software products in other words It embodies, which is stored in a storage medium, including some instructions are to pass through a computer Equipment (can be personal computer, server or the network equipment etc.) executes the complete of each embodiment the method for the application Portion or part steps.And storage medium above-mentioned includes: USB flash disk, mobile hard disk, read-only memory (full name in English: Read-Only Memory, english abbreviation: ROM), random access memory (full name in English: Random Access Memory, english abbreviation: RAM), the various media that can store program code such as magnetic or disk.
The above, above embodiments are only to illustrate the technical solution of the application, rather than its limitations;Although referring to before Embodiment is stated the application is described in detail, those skilled in the art should understand that: it still can be to preceding Technical solution documented by each embodiment is stated to modify or equivalent replacement of some of the technical features;And these It modifies or replaces, the spirit and scope of each embodiment technical solution of the application that it does not separate the essence of the corresponding technical solution.

Claims (10)

1. a kind of traffic sign recognition method characterized by comprising
Obtain Traffic Sign Images to be identified;
The Traffic Sign Images to be identified are input in trained deepness belief network model and carry out feature extraction, are obtained The corresponding first eigenvector of the Traffic Sign Images to be identified;
The first eigenvector is converted into the first pulse train;
First pulse train is input in trained impulsive neural networks, the trained pulse nerve net is obtained The recognition result of network output.
2. traffic sign recognition method according to claim 1, which is characterized in that further include:
Obtain traffic indication map image set to be trained;
The Traffic Sign Images to be trained that the Traffic Sign Images to be trained are concentrated are input to trained depth conviction net Feature extraction is carried out in network model, obtains the corresponding second feature vector of the Traffic Sign Images to be trained;
The second feature vector is converted based on time lag phase code, obtains the second pulse train;
Second pulse train is input in impulsive neural networks, the impulsive neural networks are trained;
The impulsive neural networks are calculated to the recognition accuracy of the Traffic Sign Images to be trained, when the recognition accuracy When higher than threshold value, training is completed, and obtains trained impulsive neural networks.
3. traffic sign recognition method according to claim 2, which is characterized in that described by the traffic sign to be trained Traffic Sign Images to be trained in image set are input in trained deepness belief network model and carry out feature extraction, obtain The corresponding second feature vector of the Traffic Sign Images to be trained, before further include:
The Traffic Sign Images to be trained are pre-processed.
4. traffic sign recognition method according to claim 3, which is characterized in that the pretreatment includes:
Size normalized is carried out to the Traffic Sign Images to be trained based on bilinear interpolation algorithm, is obtained at normalization Traffic Sign Images to be trained after reason.
5. traffic sign recognition method according to claim 2, which is characterized in that described that second pulse train is defeated Enter into impulsive neural networks, the impulsive neural networks be trained, comprising:
Second pulse train is input in impulsive neural networks, based on three pulse STDP in conjunction with threshold value plasticity Learning method is trained the impulsive neural networks.
6. a kind of Traffic Sign Recognition device characterized by comprising
First image collection module, for obtaining Traffic Sign Images to be identified;
Fisrt feature extraction module, for the Traffic Sign Images to be identified to be input to trained deepness belief network mould Feature extraction is carried out in type, obtains the corresponding first eigenvector of the Traffic Sign Images to be identified;
First conversion module, for the first eigenvector to be converted into the first pulse train;
Identification module obtains the training for first pulse train to be input in trained impulsive neural networks The recognition result of good impulsive neural networks output.
7. Traffic Sign Recognition device according to claim 6, which is characterized in that further include:
Second image collection module, for obtaining traffic indication map image set to be trained;
Second feature extraction module, the Traffic Sign Images to be trained for concentrating the Traffic Sign Images to be trained input Feature extraction is carried out into trained deepness belief network model, obtains the Traffic Sign Images corresponding second to be trained Feature vector;
Second conversion module obtains the second pulse for converting based on time lag phase code to the second feature vector Sequence;
Training module, for second pulse train to be input in impulsive neural networks, to the impulsive neural networks into Row training;
Computing module, for calculating the impulsive neural networks to the recognition accuracy of the Traffic Sign Images to be trained, when When the recognition accuracy is higher than threshold value, training is completed, and obtains trained impulsive neural networks.
8. a kind of Traffic Sign Recognition device, which is characterized in that further include:
Preprocessing module, for being pre-processed to the Traffic Sign Images to be trained.
9. a kind of Traffic Sign Recognition equipment characterized by comprising the equipment includes processor and memory;
Said program code is transferred to the processor for storing program code by the memory;
The processor is used for according to the described in any item traffic signs of instruction execution claim 1-5 in said program code Recognition methods.
10. a kind of computer readable storage medium, which is characterized in that the computer readable storage medium is for storing program generation Code, said program code require traffic sign recognition method described in 1-5 any one for perform claim.
CN201910764738.8A 2019-08-19 2019-08-19 Traffic sign identification method, device and equipment Active CN110458136B (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201910764738.8A CN110458136B (en) 2019-08-19 2019-08-19 Traffic sign identification method, device and equipment

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201910764738.8A CN110458136B (en) 2019-08-19 2019-08-19 Traffic sign identification method, device and equipment

Publications (2)

Publication Number Publication Date
CN110458136A true CN110458136A (en) 2019-11-15
CN110458136B CN110458136B (en) 2022-07-12

Family

ID=68487609

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201910764738.8A Active CN110458136B (en) 2019-08-19 2019-08-19 Traffic sign identification method, device and equipment

Country Status (1)

Country Link
CN (1) CN110458136B (en)

Cited By (1)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113395331A (en) * 2021-05-25 2021-09-14 郑州信大捷安信息技术股份有限公司 Safety traffic sign error surveying method and system based on Internet of vehicles

Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN102262728A (en) * 2011-07-28 2011-11-30 电子科技大学 Road traffic sign identification method
CN103235937A (en) * 2013-04-27 2013-08-07 武汉大学 Pulse-coupled neural network-based traffic sign identification method
US20130304683A1 (en) * 2010-01-19 2013-11-14 James Ting-Ho Lo Artificial Neural Networks based on a Low-Order Model of Biological Neural Networks
US20130325777A1 (en) * 2012-06-04 2013-12-05 Csaba Petre Spiking neuron network apparatus and methods
CN109522448A (en) * 2018-10-18 2019-03-26 天津大学 A method of robustness speech Gender Classification is carried out based on CRBM and SNN
CN109816026A (en) * 2019-01-29 2019-05-28 清华大学 The fusion structure and method of convolutional neural networks and impulsive neural networks

Patent Citations (6)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
US20130304683A1 (en) * 2010-01-19 2013-11-14 James Ting-Ho Lo Artificial Neural Networks based on a Low-Order Model of Biological Neural Networks
CN102262728A (en) * 2011-07-28 2011-11-30 电子科技大学 Road traffic sign identification method
US20130325777A1 (en) * 2012-06-04 2013-12-05 Csaba Petre Spiking neuron network apparatus and methods
CN103235937A (en) * 2013-04-27 2013-08-07 武汉大学 Pulse-coupled neural network-based traffic sign identification method
CN109522448A (en) * 2018-10-18 2019-03-26 天津大学 A method of robustness speech Gender Classification is carried out based on CRBM and SNN
CN109816026A (en) * 2019-01-29 2019-05-28 清华大学 The fusion structure and method of convolutional neural networks and impulsive neural networks

Non-Patent Citations (2)

* Cited by examiner, † Cited by third party
Title
YONGHONG XIE,ET.AL: "A handwritten numeral recognition method based on STDP based with unsupervised learning", 《2017 INTERNATIONAL CONFERENCE ON WIRELESS COMMUNICATIONS, SIGNAL PROCESSING AND NETWORKING (WISPNET)》 *
黄铁军等: "类脑机的思想与体系结构综述", 《计算机研究与发展》 *

Cited By (2)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113395331A (en) * 2021-05-25 2021-09-14 郑州信大捷安信息技术股份有限公司 Safety traffic sign error surveying method and system based on Internet of vehicles
CN113395331B (en) * 2021-05-25 2022-03-15 郑州信大捷安信息技术股份有限公司 Safety traffic sign error surveying method and system based on Internet of vehicles

Also Published As

Publication number Publication date
CN110458136B (en) 2022-07-12

Similar Documents

Publication Publication Date Title
CN111325099B (en) Sign language identification method and system based on double-current space-time diagram convolutional neural network
DeVries et al. Improved regularization of convolutional neural networks with cutout
CN110458844B (en) Semantic segmentation method for low-illumination scene
CN104268568B (en) Activity recognition method based on Independent subspace network
CN108133188A (en) A kind of Activity recognition method based on motion history image and convolutional neural networks
CN108875807A (en) A kind of Image Description Methods multiple dimensioned based on more attentions
CN110462680A (en) System and method for improving image texture
CN107145893A (en) A kind of image recognition algorithm and system based on convolution depth network
CN110378208B (en) Behavior identification method based on deep residual error network
CN111832516B (en) Video behavior recognition method based on unsupervised video representation learning
CN113239784B (en) Pedestrian re-identification system and method based on space sequence feature learning
CN105512680A (en) Multi-view SAR image target recognition method based on depth neural network
CN106372581A (en) Method for constructing and training human face identification feature extraction network
CN104268593A (en) Multiple-sparse-representation face recognition method for solving small sample size problem
Elleuch et al. Optimization of DBN using regularization methods applied for recognizing Arabic handwritten script
CN110175248B (en) Face image retrieval method and device based on deep learning and Hash coding
CN104298974A (en) Human body behavior recognition method based on depth video sequence
CN106022363A (en) Method for recognizing Chinese characters in natural scene
CN109255381A (en) A kind of image classification method based on the sparse adaptive depth network of second order VLAD
CN110232361A (en) Human body behavior intension recognizing method and system based on the dense network of three-dimensional residual error
CN105404865A (en) Probability state restricted Boltzmann machine cascade based face detection method
Sethy et al. Off-line Odia handwritten numeral recognition using neural network: a comparative analysis
CN114092930A (en) Character recognition method and system
CN117373111A (en) AutoHOINet-based human-object interaction detection method
Teow A minimal convolutional neural network for handwritten digit recognition

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