CN109635784A - Traffic sign recognition method based on improved convolutional neural networks - Google Patents
Traffic sign recognition method based on improved convolutional neural networks Download PDFInfo
- Publication number
- CN109635784A CN109635784A CN201910021739.3A CN201910021739A CN109635784A CN 109635784 A CN109635784 A CN 109635784A CN 201910021739 A CN201910021739 A CN 201910021739A CN 109635784 A CN109635784 A CN 109635784A
- Authority
- CN
- China
- Prior art keywords
- traffic sign
- convolutional neural
- neural networks
- model
- method based
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/50—Context or environment of the image
- G06V20/56—Context or environment of the image exterior to a vehicle by using sensors mounted on the vehicle
- G06V20/58—Recognition of moving objects or obstacles, e.g. vehicles or pedestrians; Recognition of traffic objects, e.g. traffic signs, traffic lights or roads
- G06V20/582—Recognition 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
- G06F18/2411—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on the proximity to a decision surface, e.g. support vector machines
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/44—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components
- G06V10/443—Local feature extraction by analysis of parts of the pattern, e.g. by detecting edges, contours, loops, corners, strokes or intersections; Connectivity analysis, e.g. of connected components by matching or filtering
- G06V10/449—Biologically inspired filters, e.g. difference of Gaussians [DoG] or Gabor filters
Abstract
The present invention relates to a kind of traffic sign recognition method based on improved convolutional neural networks, the Traffic Sign Images that this method is shot using the vehicle-mounted camera in practical environment construct convolutional neural networks model and identify to it as experiment basis.Pretreatment operation is done to Traffic Sign Images first, model is enable to be trained up, and increases the generalization ability of model；Secondly accelerate network training speed using batch method for normalizing (Batch Normalization, BN)；Classify finally by SVM classifier to the feature extracted.The Traffic Sign Images obtained in practical vehicle environment are identified using technical solution of the present invention, very high accuracy can be obtained, while the network model training time is shorter, additionally it is possible to identify traffic sign in a very short period of time.
Description
Technical field
The present invention relates to Traffic Sign Images identification field more particularly to a kind of friendships based on improved convolutional neural networks
Logical sign.
Background technique
With the rapid development of society, Global Auto quantity also sharply increases, and brings great pressure to road traffic,
Especially in big city, huge automobile quantity leads to traffic jam frequent occurrence, brings greatly not to people's life
Just.Further, since many place road conditions are more complicated, traffic thing often occurs due to the carelessness to traffic mark board for driver
Therefore the annual traffic accident in the whole world brings huge loss and casualties.Therefore, driving technology is assisted in recent years and is driven automatically
The technology of sailing receives the extensive concern of researchers at home and abroad.Wherein, automatic recognition of traffic signs technology is its important composition
Part, since automatic recognition of traffic signs technology is applied in actual environment, in this way for Traffic Sign Recognition system
System, does not require nothing more than high accuracy, it is also necessary to come out in a very short period of time to Traffic Sign Recognition, this is also researcher
The difficult point for needing to break through.
Convolutional neural networks achieve huge achievement in field of image recognition in recent years, relative to traditional image recognition
Method, convolutional neural networks can learn the feature to picture, do not need artificial design and suitably distinguish algorithm, design conjunction
The convolutional neural networks of reason can in a very short period of time identify traffic sign, and obtain very high discrimination.
Summary of the invention
The purpose of the present invention is to propose to it is a kind of the Traffic Sign Images acquired in practical environment can be carried out quickly,
Accurate classification method.
To achieve the above object, the present invention devise an improved convolutional neural networks model to Traffic Sign Images into
Row identification, key step are as follows:
Step 1, using the Traffic Sign Images acquired in practical environment as convolutional Neural net designed by the present invention
Network model training sample and test sample, and pretreatment operation is done to collected original traffic sign image.
Step 2, feature extraction is carried out to pretreated Traffic Sign Images using convolutional neural networks, wherein convolution
Neural network includes input layer, 3 convolutional layers and 3 pond layers alternately connection and 2 full articulamentum.
Step 3, the picture feature of extraction is input in SVM and is classified.
Preferably, in step 1, training sample and test sample pre-treatment step include:
Image enhancement operation is carried out to trained and test sample, enhances picture contrast.
To the processing of trained and test sample gray processing, simplify convolution operation, reduces convolutional neural networks model complexity.
Normalized is done to trained and test sample size.
Training sample is subjected to simple data set extension, model is allowed to be trained up.
Preferably, in step 2, convolutional neural networks structure includes:
First convolutional layer convolution kernel size is 5*5, and convolution kernel number is 60, second and third convolutional layer convolution kernel
Size is 3*3, and convolution kernel number is respectively 90 and 108.
The pond window size of three pond layers is 2*2, selects maximum pond.
Two full articulamentums all contain 1024 neurons.
Preferably, it is trained that steps are as follows to the convolutional neural networks model of above-mentioned design:
Step 1, the initial parameter of each layer of convolutional neural networks, including each layer convolution kernel size, convolution kernel number, pond are set
Change window size, pond step-length, convolutional layer weight and biasing and full articulamentum neuron number, completes the building of basic model；
Step 2, convolutional neural networks model training learning rate, target minimal error are set, allow maximum train epochs and
mini-batch；
Step 3, pretreated training data is input to the input layer of convolutional neural networks model, it is defeated to calculate network
Out；
Step 4, output error is calculated, weight is updated by back-propagation algorithm；
Step 5,1 is repeated the above steps to step 4, until output error is in the target minimal error of setting or training is secondary
Number is more than maximum train epochs.
It the use of mean value is 0, the gauss of distribution function that variance is 0.01 is random in the convolutional neural networks model of above-mentioned design
The weight in convolutional layer is initialized, is set to 0.Activation primitive uses ReLU function, accelerates network convergence rate.To accelerate net
Batch normalization layer (batch normalization, BN) is added in network training speed before activation primitive.And it is connecting entirely
Using Dropout strategy, can prevent convolutional neural networks model from over-fitting occur, detailed process is as follows:
1) convolutional layer weight initialization:
It the use of mean value is 0, the weight in gauss of distribution function random initializtion convolutional layer that variance is 0.01, Gaussian Profile
The formula of probability density function is as follows:
2) ReLU activation primitive formula is as follows:
F (x)=max (0, x)
3) batch normalization process is as follows:
4) Dropout prevents over-fitting strategy
The neuron of full articulamentum is set to 0 with 0.5 probability in the training process.
Designed convolutional neural networks model is trained by requirements above, then again by trained convolutional Neural
Network model identifies test data.
Compared with the conventional method, data set is carried out certain pretreatment operation by the present invention first, is then using convolution
Neural network carries out feature extraction to traffic sign picture, and finally the feature extracted is input in SVM classifier and is divided
Class.Network training speed can be greatly improved using above technical scheme, accelerates network convergence, and can obtain very high
Discrimination.
Detailed description of the invention
Fig. 1 is training dataset preprocessing process schematic diagram；
Fig. 2 is test data set preprocessing process schematic diagram；
Fig. 3 is training of the present invention and test process schematic diagram.
Specific embodiment
With reference to the accompanying drawings, the specific embodiment of the convolutional neural networks model designed the present invention is done furtherly
It is bright.
Fig. 1, Fig. 2 are respectively traffic sign training dataset and test data set preprocessing process, and Fig. 3 is convolutional Neural net
Network model training and identification process.
The invention mainly comprises two modules, Traffic Sign Images preprocessing module and convolutional neural networks module.Substantially
Steps are as follows:
Step 1, using the Traffic Sign Images acquired in practical environment as convolutional Neural net designed by the present invention
Network model training sample and test sample, and pretreatment operation is done to collected original traffic sign image.
Step 2, feature extraction is carried out to pretreated Traffic Sign Images using convolutional neural networks, wherein convolution
Neural network includes input layer, 3 convolutional layers and 3 pond layers alternately connection and 2 full articulamentum.
Step 3, the picture feature of extraction is input in SVM and is classified.
Traffic Sign Images preprocessing module can do pretreatment operation, mould to training dataset and test data set respectively
Collection selection Germany Traffic Sign Recognition database (the german traffic sign recognition that type needs
Benchmark, GTSRB) model is trained and is tested, GTSRB is a German road traffic sign detection data.In database
All Traffic Sign Images be all carried out by the vehicle-mounted camera in practical environment shooting extraction, meet practical need
It asks.Traffic Sign Images pretreatment mainly includes following content:
1) image enhancement operation is carried out to trained and test sample, histogram equalization processing is done to Traffic Sign Images,
Enhance picture contrast, keeps characteristics of image more obvious.
2) trained and test sample gray processing is handled, simplifies convolution operation, reduce convolutional neural networks model complexity.
3) normalized is done to trained and test sample size, dimension of picture size is unified for 48*48.
4) training sample is subjected to simple data set extension, there are mainly two types of modes, and first using center picture as origin
Counterclockwise and rotate clockwise 8 degree respectively, second respectively translates picture 4 pixels to the left and to the right, and the one of such picture
Side has the rest of 4 pixels, rest is removed, picture cross directional stretch is then returned 48*48 size.Not only increase in this way
The size for having added training dataset, allows model to be trained up, and can also increase the generalization ability of model.
Convolutional neural networks module mainly builds suitable convolutional neural networks model to extract pretreated friendship
Logical sign image feature, is then input to SVM classifier for the feature of extraction and classifies.It mainly include input layer, 3 convolution
Layer and 3 pond layers alternating connections, 2 full articulamentums and a SVM classifier.
It the use of mean value is 0, the gauss of distribution function that variance is 0.01 is initial at random in the convolutional neural networks model of design
Change the weight in convolutional layer, is set to 0.Activation primitive uses ReLU function, accelerates network convergence rate.To accelerate network instruction
Practice speed, batch normalization layer (batch normalization, BN) is added before activation primitive.And it is used in full connection
Dropout strategy, can prevent convolutional neural networks model from over-fitting occur, detailed process is as follows:
4) convolutional layer weight initialization:
It the use of mean value is 0, the weight in gauss of distribution function random initializtion convolutional layer that variance is 0.01, Gaussian Profile
The formula of probability density function is as follows:
5) ReLU activation primitive formula is as follows:
F (x)=max (0, x)
6) batch normalization process is as follows:
4) Dropout prevents over-fitting strategy
The neuron of full articulamentum is set to 0 with 0.5 probability in the training process.
A complete convolutional neural networks model is finally obtained, detail parameters are as shown in table 1: table 1
The convolutional neural networks model of design is trained that steps are as follows:
Step 1, the initial parameter of each layer of convolutional neural networks, including each layer convolution kernel size, convolution kernel number, pond are set
Change window size, pond step-length, convolutional layer weight and biasing and full articulamentum neuron number, completes the building of basic model；
Step 2, convolutional neural networks model training learning rate, target minimal error are set, allow maximum train epochs and
mini-batch；
Step 3, pretreated training data is input to the input layer of convolutional neural networks model, it is defeated to calculate network
Out；
Step 4, output error is calculated, weight is updated by back-propagation algorithm；
Step 5,1 is repeated the above steps to step 4, until output error is in the target minimal error of setting or training is secondary
Number is more than maximum train epochs.
The Traffic Sign Images that present invention experiment uses are all from practical vehicle environment, have very much practical reference price
Value, and compare and existing method, has done very big optimization on the training time, while also can reach on discrimination considerable
Effect has practical application value very much.
Claims (7)
1. a kind of traffic sign recognition method based on improved convolutional neural networks, it is characterised in that:
Pretreatment operation is done to the Traffic Sign Images acquired in practical environment first, then utilizes present invention design convolution
Neural network model carries out feature extraction to pretreated Traffic Sign Images, is finally input to the characteristics of image extracted
Last classification is carried out in SVM classifier, comprising the following steps:
Step 1, using the Traffic Sign Images acquired in practical environment as convolutional neural networks mould designed by the present invention
Type training sample and test sample, and pretreatment operation is done to collected original traffic sign image；
Step 2, feature extraction is carried out to pretreated Traffic Sign Images using convolutional neural networks, wherein convolutional Neural
Network includes input layer, 3 convolutional layers and 3 pond layers alternately connection and 2 full articulamentum；
Step 3, the picture feature of extraction is input in SVM and is classified.
2. a kind of traffic sign recognition method based on improved convolutional neural networks according to claim 1, feature
It is, in the step 1, training sample and test sample pretreatment include:
Image enhancement operation is carried out to trained and test sample, histogram equalization processing, enhancing figure are done to Traffic Sign Images
Piece contrast keeps characteristics of image more obvious；
To the processing of trained and test sample gray processing, simplify convolution operation, reduces convolutional neural networks model complexity；
Normalized is done to trained and test sample size, dimension of picture size is unified for 48*48；
Training sample is subjected to simple data set extension, there are mainly two types of modes, and first is inverse respectively by origin of center picture
Hour hands and 8 degree are rotated clockwise, second respectively translates picture 4 pixels to the left and to the right, and the side of such picture has 4
Rest is removed and picture cross directional stretch is then returned 48*48 size by the rest of pixel；Which not only adds training
The size of data set, allows model to be trained up, and can also increase the generalization ability of model.
3. a kind of traffic sign recognition method based on improved convolutional neural networks according to claim 1, feature
It is, in the step 2,3 layers of convolutional layer convolution kernel size are respectively 5*5,3*3 and 3*3, and step-length is all 1, are rolled up in third layer
To characteristic pattern edge supplement 1 when product.
4. a kind of traffic sign recognition method based on improved convolutional neural networks according to claim 1, feature
It is, in the step 2,3 layers of pond pond Hua Ceng window size is 2*2, and step-length is 2, all using maximum pond.
5. a kind of traffic sign recognition method based on improved convolutional neural networks according to claim 1, feature
It is, in the step 2, the activation primitive that convolutional layer uses is ReLU function.
6. a kind of traffic sign recognition method based on improved convolutional neural networks according to claim 1, feature
It is, in the step 2, three-layer coil lamination convolution kernel number is respectively 60,90 and 108.
7. a kind of traffic sign recognition method based on improved convolutional neural networks according to claim 1, feature
It is, network model training process is as follows:
Step 1, the initial parameter of each layer of convolutional neural networks, including each layer convolution kernel size, convolution kernel number, pond window are set
Mouth size, pond step-length, convolutional layer weight and biasing and full articulamentum neuron number, complete the building of basic model；
Step 2, convolutional neural networks model training learning rate, target minimal error are set, allows maximum train epochs and mini-
batch；
Step 3, pretreated training data is input to the input layer of convolutional neural networks model, calculates network output；
Step 4, output error is calculated, weight is updated by back-propagation algorithm；
Step 5,1 is repeated the above steps to step 4, until output error is in the target minimal error of setting or frequency of training is super
Cross maximum train epochs.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910021739.3A CN109635784A (en) | 2019-01-10 | 2019-01-10 | Traffic sign recognition method based on improved convolutional neural networks |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201910021739.3A CN109635784A (en) | 2019-01-10 | 2019-01-10 | Traffic sign recognition method based on improved convolutional neural networks |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109635784A true CN109635784A (en) | 2019-04-16 |
Family
ID=66061064
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201910021739.3A Pending CN109635784A (en) | 2019-01-10 | 2019-01-10 | Traffic sign recognition method based on improved convolutional neural networks |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109635784A (en) |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110197166A (en) * | 2019-06-04 | 2019-09-03 | 西安建筑科技大学 | A kind of car body loading condition identification device and method based on image recognition |
| CN110321803A (en) * | 2019-06-10 | 2019-10-11 | 南京邮电大学 | A kind of traffic sign recognition method based on SRCNN |
| CN111199217A (en) * | 2020-01-09 | 2020-05-26 | 上海应用技术大学 | Traffic sign identification method and system based on convolutional neural network |
| CN111274971A (en) * | 2020-01-21 | 2020-06-12 | 南京航空航天大学 | Traffic identification method based on color space fusion network and space transformation network |
| CN111291814A (en) * | 2020-02-15 | 2020-06-16 | 河北工业大学 | Crack identification algorithm based on convolution neural network and information entropy data fusion strategy |
| CN111325152A (en) * | 2020-02-19 | 2020-06-23 | 北京工业大学 | Deep learning-based traffic sign identification method |
| CN111767860A (en) * | 2020-06-30 | 2020-10-13 | 阳光学院 | Method and terminal for realizing image recognition through convolutional neural network |
| CN113221620A (en) * | 2021-01-29 | 2021-08-06 | 太原理工大学 | Multi-scale convolutional neural network-based traffic sign rapid identification method |
| CN113536942A (en) * | 2021-06-21 | 2021-10-22 | 上海赫千电子科技有限公司 | Road traffic sign recognition method based on neural network |
| CN115205637A (en) * | 2022-09-19 | 2022-10-18 | 山东世纪矿山机电有限公司 | Intelligent identification method for mine car materials |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108520212A (en) * | 2018-03-27 | 2018-09-11 | 东华大学 | Method for traffic sign detection based on improved convolutional neural networks |
| US20180285664A1 (en) * | 2017-02-09 | 2018-10-04 | SMR Patents S.à.r.l. | Method and device for identifying the signaling state of at least one signaling device |
| CN108985217A (en) * | 2018-07-10 | 2018-12-11 | 常州大学 | A kind of traffic sign recognition method and system based on deep space network |
| CN109087302A (en) * | 2018-08-06 | 2018-12-25 | 北京大恒普信医疗技术有限公司 | A kind of eye fundus image blood vessel segmentation method and apparatus |
-
2019
- 2019-01-10 CN CN201910021739.3A patent/CN109635784A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20180285664A1 (en) * | 2017-02-09 | 2018-10-04 | SMR Patents S.à.r.l. | Method and device for identifying the signaling state of at least one signaling device |
| CN108520212A (en) * | 2018-03-27 | 2018-09-11 | 东华大学 | Method for traffic sign detection based on improved convolutional neural networks |
| CN108985217A (en) * | 2018-07-10 | 2018-12-11 | 常州大学 | A kind of traffic sign recognition method and system based on deep space network |
| CN109087302A (en) * | 2018-08-06 | 2018-12-25 | 北京大恒普信医疗技术有限公司 | A kind of eye fundus image blood vessel segmentation method and apparatus |
Non-Patent Citations (2)
| Title |
|---|
| 王晓斌等: "基于优化卷积神经网络结构的交通标志识别", 《计算机应用》 * |
| 王猛: "基于深度学习的交通标志图像识别研究", 《中国优秀硕士学位论文全文数据库 (信息科技辑)》 * |
Cited By (16)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110197166B (en) * | 2019-06-04 | 2022-09-09 | 西安建筑科技大学 | Vehicle body loading state recognition device and method based on image recognition |
| CN110197166A (en) * | 2019-06-04 | 2019-09-03 | 西安建筑科技大学 | A kind of car body loading condition identification device and method based on image recognition |
| CN110321803A (en) * | 2019-06-10 | 2019-10-11 | 南京邮电大学 | A kind of traffic sign recognition method based on SRCNN |
| CN111199217A (en) * | 2020-01-09 | 2020-05-26 | 上海应用技术大学 | Traffic sign identification method and system based on convolutional neural network |
| CN111199217B (en) * | 2020-01-09 | 2023-03-28 | 上海应用技术大学 | Traffic sign identification method and system based on convolutional neural network |
| CN111274971A (en) * | 2020-01-21 | 2020-06-12 | 南京航空航天大学 | Traffic identification method based on color space fusion network and space transformation network |
| CN111291814A (en) * | 2020-02-15 | 2020-06-16 | 河北工业大学 | Crack identification algorithm based on convolution neural network and information entropy data fusion strategy |
| CN111291814B (en) * | 2020-02-15 | 2023-06-02 | 河北工业大学 | Crack identification algorithm based on convolutional neural network and information entropy data fusion strategy |
| CN111325152A (en) * | 2020-02-19 | 2020-06-23 | 北京工业大学 | Deep learning-based traffic sign identification method |
| CN111325152B (en) * | 2020-02-19 | 2023-09-26 | 北京工业大学 | Traffic sign recognition method based on deep learning |
| CN111767860A (en) * | 2020-06-30 | 2020-10-13 | 阳光学院 | Method and terminal for realizing image recognition through convolutional neural network |
| CN113221620A (en) * | 2021-01-29 | 2021-08-06 | 太原理工大学 | Multi-scale convolutional neural network-based traffic sign rapid identification method |
| CN113536942A (en) * | 2021-06-21 | 2021-10-22 | 上海赫千电子科技有限公司 | Road traffic sign recognition method based on neural network |
| CN113536942B (en) * | 2021-06-21 | 2024-04-12 | 上海赫千电子科技有限公司 | Road traffic sign recognition method based on neural network |
| CN115205637A (en) * | 2022-09-19 | 2022-10-18 | 山东世纪矿山机电有限公司 | Intelligent identification method for mine car materials |
| CN115205637B (en) * | 2022-09-19 | 2022-12-02 | 山东世纪矿山机电有限公司 | Intelligent identification method for mine car materials |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109635784A (en) | Traffic sign recognition method based on improved convolutional neural networks | |
| CN104537393B (en) | A kind of traffic sign recognition method based on multiresolution convolutional neural networks | |
| CN107886073B (en) | Fine-grained vehicle multi-attribute identification method based on convolutional neural network | |
| CN110033002B (en) | License plate detection method based on multitask cascade convolution neural network | |
| CN106407931B (en) | A kind of depth convolutional neural networks moving vehicle detection method | |
| CN107133974B (en) | Gaussian Background models the vehicle type classification method combined with Recognition with Recurrent Neural Network | |
| CN105956626A (en) | Deep learning based vehicle license plate position insensitive vehicle license plate recognition method | |
| CN103996018B (en) | Face identification method based on 4DLBP | |
| CN101944174B (en) | Identification method of characters of licence plate | |
| CN106096602A (en) | A kind of Chinese licence plate recognition method based on convolutional neural networks | |
| CN105809121A (en) | Multi-characteristic synergic traffic sign detection and identification method | |
| CN106650786A (en) | Image recognition method based on multi-column convolutional neural network fuzzy evaluation | |
| CN107729801A (en) | A kind of vehicle color identifying system based on multitask depth convolutional neural networks | |
| CN107545239A (en) | A kind of deck detection method matched based on Car license recognition with vehicle characteristics | |
| CN105354568A (en) | Convolutional neural network based vehicle logo identification method | |
| CN112270347A (en) | Medical waste classification detection method based on improved SSD | |
| CN110163069B (en) | Lane line detection method for driving assistance | |
| CN108918532A (en) | A kind of through street traffic sign breakage detection system and its detection method | |
| CN102915453B (en) | Real-time feedback and update vehicle detection method | |
| CN104408469A (en) | Firework identification method and firework identification system based on deep learning of image | |
| CN108875608A (en) | A kind of automobile traffic signal recognition method based on deep learning | |
| CN108520212A (en) | Method for traffic sign detection based on improved convolutional neural networks | |
| CN107092876A (en) | The low-light (level) model recognizing method combined based on Retinex with S SIFT features | |
| CN109886086B (en) | Pedestrian detection method based on HOG (histogram of oriented gradient) features and linear SVM (support vector machine) cascade classifier | |
| CN108647700A (en) | Multitask vehicle part identification model based on deep learning, method and system |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| WD01 | Invention patent application deemed withdrawn after publication | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20190416 |