CN107085696A - A kind of vehicle location and type identifier method based on bayonet socket image - Google Patents
A kind of vehicle location and type identifier method based on bayonet socket image Download PDFInfo
- Publication number
- CN107085696A CN107085696A CN201610900576.2A CN201610900576A CN107085696A CN 107085696 A CN107085696 A CN 107085696A CN 201610900576 A CN201610900576 A CN 201610900576A CN 107085696 A CN107085696 A CN 107085696A
- Authority
- CN
- China
- Prior art keywords
- vehicle
- layer
- neural networks
- convolutional neural
- bayonet socket
- 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
- G06V20/584—Recognition of moving objects or obstacles, e.g. vehicles or pedestrians; Recognition of traffic objects, e.g. traffic signs, traffic lights or roads of vehicle lights or traffic lights
-
- 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
- G06F18/2413—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on distances to training or reference patterns
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
- G06N3/088—Non-supervised learning, e.g. competitive learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V30/00—Character recognition; Recognising digital ink; Document-oriented image-based pattern recognition
- G06V30/10—Character recognition
- G06V30/19—Recognition using electronic means
- G06V30/192—Recognition using electronic means using simultaneous comparisons or correlations of the image signals with a plurality of references
- G06V30/194—References adjustable by an adaptive method, e.g. learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/60—Type of objects
- G06V20/62—Text, e.g. of license plates, overlay texts or captions on TV images
- G06V20/625—License plates
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Artificial Intelligence (AREA)
- Multimedia (AREA)
- Evolutionary Computation (AREA)
- Life Sciences & Earth Sciences (AREA)
- General Engineering & Computer Science (AREA)
- Computational Linguistics (AREA)
- Molecular Biology (AREA)
- Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Bioinformatics & Computational Biology (AREA)
- General Health & Medical Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Evolutionary Biology (AREA)
- Databases & Information Systems (AREA)
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
The invention provides a kind of vehicle location based on bayonet socket image and type identifier method, utilize bayonet socket image pattern collection, the positioning of vehicle in the convolutional neural networks of the detectable vehicle model of training, the image shot to bayonet system and model are identified, and step includes:Sample collection and mark;The design of convolutional neural networks;The training of convolutional neural networks;Vehicle location and type identifier.The present invention can detect that identification vehicle model accurately position and recognize in the bayonet socket image of various complex environments, in terms of intelligent transportation, vehicle Flow Detection and illegal tracking.
Description
Technical field
The present invention relates to computer vision target identification technology field, specifically a kind of vehicle location based on bayonet socket image
And type identifier method, available for intelligent transportation field
Background technology
Intelligent transportation is the developing direction of future transportation system, and it improves conevying efficiency, alleviates traffic congestion, is ensured
Traffic safety, reduces energy resource consumption and environmental pollution.Vehicle cab recognition is as the important branch in intelligent transportation system, in car
Flow detection, traffic behavior monitoring, the illegal tracking of red light and criminal investigation aspect etc. of deciding a case have application prospect widely.
Current vehicle and vehicle model recognition methods mainly has:It is electromagnetic induction coil detection method, radar method of identification, infrared
Line method of identification, Car license recognition method and the recognition methods based on video image.In foregoing several model recognizing methods, electromagnetism sense
Answer Coil Detector method, radar method of identification, infrared ray method of identification and Car license recognition method studies in China relatively more, and existing maturation mostly
Application, but these methods need to increase corresponding hardware device, such as ground induction coil, radar, and can only detect cart, dolly
Or the basic parameter such as car plate, car plate color, testing result precision is relatively low, it is impossible to which the model to vehicle is identified.With near
The development of year computer vision technique, because its use cost is low, the features such as obtaining more detailed information utilizes computer
The method that vehicle and vehicle model are identified vision technique is also slowly put forward by people.
Due to there is a situation where that similarity is larger between vehicle model number, the vehicle of different model, and highway scene
Complexity, the shooting quality of bayonet socket camera differs, and these all bring very to the vehicle location based on bayonet socket image and type identifier
Big problem, many existing technologies all can only obtain preferable effect for specific scene, therefore how design a robust
Property it is high, the high method of accuracy rate is still one and is worth the problem of challenge the vehicle model in bayonet socket image to be identified.
The content of the invention
The purpose of the present invention is the deficiency for existing vehicle model recognition methods, is proposed a kind of based on bayonet socket image
Vehicle location and type identifier method, using application of the convolutional neural networks in terms of target detection, to solve in bayonet socket image
Vehicle is positioned at identification problem.
In order to achieve the above object, the technical solution adopted in the present invention is:
A kind of vehicle location and type identifier method based on bayonet socket image, using bayonet socket image pattern collection, training can be examined
The convolutional neural networks of vehicle model are surveyed, the positioning of the vehicle in the image shot to bayonet system and model are identified, wrapped
Include following steps:
(1) sample collection and mark:The source images of collection vehicle, are marked out in source images from highway bayonet system
Comprising each vehicle position and model, using source images and markup information as convolutional neural networks training sample set;
(2) design of convolutional neural networks:Designed for the convolutional neural networks of fixation and recognition vehicle, by source images and mark
The input layer that information inputs network is noted, feature, the output layer of network are extracted in the intermediate layer of network using convolutional layer and full articulamentum
The bounding box and type of vehicle in picture are respectively obtained by two output branchs using the feature extracted;
(3) training of convolutional neural networks:The parameter of convolutional neural networks is changed using source images and markup information
Generation training, the function of making it possess positioning and identification type of vehicle;
(4) vehicle location and type identifier:According to the convolutional neural networks trained, in the image shot to bayonet system
Vehicle position is positioned, and judges the model belonging to it.
Preferably, in the step (1), the collection of sample and annotation process are as follows:
(1) sample collection of big data:According to the model of driving vehicle on highway at this stage, according to every kind of model at least
The requirement of 100 sample source images, different time is being gathered from different geographical, the bayonet system of the highway of different road conditions
The source images sample of section;
(2) the sample mark of big data:The license plate number obtained automatically according to bayonet socket, car is obtained using the database of vehicle administration office
Model;The method selected by artificial frame obtains the upper left corner of vehicle location and bottom right angular coordinate in source images, by same
All vehicle locations included and classification information in source images are all recorded in an XML document;To each vehicle source
Image is split using selective search, and the coordinate that the segmentation of generation is fast is all stored in a ss.mat.
Preferably, in the step (2), the design process of convolutional neural networks is as follows:
(1) input layer of convolutional neural networks is whole network input data:The data of training stage input include image
Data, area-of-interest bounding box, the target bounding box and its label of generic of demarcation;The input layer input of detection-phase
Data include view data and area-of-interest bounding box;
(2) convolutional layer and full articulamentum of feature is are extracted in the intermediate layer of convolutional neural networks, and the structure of network is eight layers
Network structure, first layer and the second layer are that 96 core sizes are the convolutional layer that 11*11 and 256 core size is 5*5 respectively, each
It is used for the pond layer of dimensionality reduction provided with one behind layer;Third layer, the 4th layer, layer 5 be core size be 3*3 convolutional layer, volume
Product nuclear volume is 384 respectively, 384 and 256, provided with a ROI ponds layer, the ROI ponds layer wherein after layer 5 convolutional layer
To realize operation of the convolutional layer on whole pictures above, and the result of convolution is mapped to each area-of-interest, together
When produce a fixed size output be used as full articulamentum one, full articulamentum two input;
(3) the full articulamentum combination of the last branched structure of convolutional neural networks:
Be first full articulamentum one, full articulamentum two using the output of ROI ponds layer be used as input, produce one 4096
For characteristic vector;
Then this characteristic vector is separately input to the full articulamentum three and full articulamentum four of two branched structures:Quan Lian
The classification that layer three is used to judge target is connect, its length exported is N+1, and wherein N is the sum for including vehicle classification, and 1 is background
Numerical value;Full articulamentum four is used to export the value after the bounding box adjustment of target, and its length is (N+1) * 4, and wherein N is to include car
The sum of type classification, 1 is the numerical value of background, and 4 be the numerical value of the transverse and longitudinal coordinate of bounding box upper left lower-right most point.
Preferably, in the step (3), the training process of convolutional neural networks is as follows:
The cut zone produced using the vehicle data and selective search collected is to convolutional neural networks
Carry out supervised learning, each iteration takes two vehicle pictures, using demarcation interested area of vehicle bounding box with
The cut zone that selective search are automatically generated, which is asked, to be handed over and compares, and will be handed over and is used as positive sample than the cut zone more than 50%
This, the label of generic is the classification of the bounding box of demarcation, will hand over and is used as negative sample than the cut zone for 10%~50%
This, classification is that label is 0, represents negative sample;Error back propagation is adjusted by each layer of convolution kernel by continuous iteration
Weights so that each layer of convolution mask can reach extraction characteristics of image, recognize type of vehicle, position the mesh of bounding box
's.
Preferably, it is as follows the step of vehicle location and its type identifier in the step (4):
For bayonet socket image to be detected, several are extracted on image first with selective search method
Possible cut zone, then the coordinate of image and segmentation is inputted into the convolutional neural networks trained, convolutional layer is entirely being schemed
After upper extraction feature, Feature Mapping to cut zone is judged vehicle model by ROI ponds layer in full articulamentum using feature,
Adjustment vehicle bounding box obtains the final information of vehicles detected, finally, non-maxima suppression is used to all bounding boxs
Method, exclude unnecessary bounding box.
A kind of vehicle location and type identifier method based on bayonet socket image that the present invention is provided, using convolutional neural networks
, the vehicle in highway bayonet socket image is accurately positioned and recognized its model, method mainly includes initial data and prepares, instructs
Practice data generation, neutral net design, the training of neutral net and vehicle location and recognize six steps, initial data is prepared as whole
Individual recognition methods provides necessary data;Training data life will be produced using ready data more can be used for training convolutional
The positive and negative sample data of neutral net;Neutral net is mainly designed to design the nerve net available for fixation and recognition vehicle
Network, different from traditional neutral net, present networks use ROI ponds layer, it may be unnecessary to limit the size of network output image, and
The type identifier that is positioned at of vehicle can once be completed;Design and ready data just can be used after network to neutral net
Weights are trained, and can reach requirement;The network trained can split generation by picture to be detected and to it
Candidate region inputs neutral net, and forward-propagating once obtains vehicle location and type identifier result, reuses non-maximum suppression
Make and obtained result is screened, obtain last result.The present invention can detect identification in the bayonet socket image of various complex environments
Vehicle model accurately position and recognize, in terms of intelligent transportation, vehicle Flow Detection and illegal tracking.
Compared with existing technology, the present invention has the beneficial effect that:
(1) sample used in the present invention is the image that true road conditions under bayonet is shot, and background complicated variety is high, can fill
Divide the vehicle characteristics extracted under any state, greatly increase the robustness of the grader trained.
(2) using the method for convolutional neural networks, but it is not traditional convolutional neural networks, the network is subjected to multiple
Input, once exports the positional information and type information of vehicle, and centre can enable the network receive difference using ROI ponds layer
The picture of size, not in the picture input of limitation fixed size, makes network have scale invariability, adds identification as input
Effect.
(3) identification of the present invention to the car direct picture in bayonet socket image has very high discrimination, and can be simultaneously
Identify multiple vehicles in same bayonet socket picture, efficiency high.
Brief description of the drawings
Fig. 1 is flow chart of the invention
Fig. 2 is convolutional neural networks structure chart used in the present invention
Embodiment
As shown in figure 1, a kind of vehicle location and type identifier method based on bayonet socket image, comprise the following steps:(1) no
The image of different time sections is copied in same region, the highway bayonet system of different road conditions, included in picture every is marked
The position of one vehicle and model, using source images and markup information as convolutional neural networks training sample set detailed process such as
Under:
(a) the model number of normal sport car is about 2000 kinds on highway at this stage, according to wanting for every kind of at least 100 samples
Ask, collect 200,000 pictures in a balanced way as far as possible in the bayonet system of national highway and be used as sample.
(b) license plate number obtained automatically according to bayonet socket, the model of vehicle, then artificial frame are obtained using the database of vehicle administration office
The method of choosing obtains the upper left corner of vehicle location and bottom right angular coordinate in picture, by all cars included in same pictures
Position and classification information are all recorded in an XML document.Each vehicle pictures are divided using selective search
Cut, the coordinate that the segmentation of generation is fast is all stored in a ss.mat.
(2) convolutional neural networks of fixation and recognition vehicle are designed for, the input of network is source images and mark text, net
Feature is extracted in the intermediate layer of network using convolutional layer and full articulamentum, and the output layer of last network utilizes the feature extracted by two
Individual output branch respectively obtains the bounding box and type of vehicle in picture.The specific form of network is as follows:
(a) input layer is whole network input data, and the data of training stage input include view data, area-of-interest
Bounding box, the target bounding box and its label of generic of demarcation, the data of the input layer input of detection-phase include image
Data and area-of-interest bounding box.
(b) convolutional layer and full articulamentum of feature is are extracted in intermediate layer, and the structure of network is eight traditional layer network structures,
First layer and the second layer are that 96 core sizes are behind the convolutional layer that 11*11 and 256 core size is 5*5, each layer respectively
Be used for the pond layer of dimensionality reduction with one, third layer, the 4th layer, layer 5 be convolutional layer that core size is 3*3, convolution nuclear volume divides
Be not 384,384 and 256.Immediately being a crucial ROI ponds layer after the 5th convolutional layer, it can be convolution above
Layer can be operated directly on whole pictures, and the result of convolution is mapped into each area-of-interest in this layer, together
When, the output for producing a fixed size is used as the input of full articulamentum one and full articulamentum two.
(c) network be finally branched structure the combination of full articulamentum, be two full articulamentums one first and connect entirely
Layer two is connect using the output of ROI ponds layer as input, one 4096 characteristic vector for being is produced, then by this characteristic vector point
The full articulamentum three and full articulamentum four of two branched structures are not input to, and full articulamentum three is used for the classification for judging target, its
The length of output is N+1, and wherein N is the sum for including vehicle classification, and 1 is the numerical value of background;Full articulamentum four is used to export mesh
Value after the adjustment of target bounding box, its length is (N+1) * 4, and wherein N is the sum for including vehicle classification, and 1 is the numerical value of background,
4 be the numerical value of the transverse and longitudinal coordinate of bounding box upper left lower-right most point.
(3) training is iterated to the parameter of convolutional neural networks using ready sample set, be it possess positioning and
Recognize the function of type of vehicle.Training process is as follows:
Carry out having prison to network using ready vehicle data and selective the search cut zone produced
Educational inspector is practised, and each iteration is taken two vehicle pictures, automatically generated using the vehicle bounding box and selective search of demarcation
Cut zone ask and hand over and compare, will hand over and than the cut zone more than 50% as positive sample, the label of generic is demarcation
Bounding box classification, will hand over and than the cut zone for 10%~50% as negative sample, classification is that label is 0, represent negative
Sample;By continuous iteration by error back propagation come the weights of the convolution kernel that adjusts each layer so that each layer of convolution
Template can reach extraction characteristics of image, recognize type of vehicle, position the purpose of bounding box.
(4) finally using the convolutional neural networks trained to the vehicle location in bayonet socket picture and its step of type identifier
It is rapid as follows:
For picture to be detected, several are extracted on picture first with selective search method may
Cut zone, then by picture and segmentation coordinate input into convolutional neural networks, convolutional layer extracts feature on whole picture
Afterwards, Feature Mapping to cut zone is judged vehicle model, adjustment vehicle is surrounded by ROI ponds layer in full articulamentum using feature
Box obtains the final information of vehicles detected, finally, the method for using all bounding boxs non-maxima suppression, excludes many
Remaining bounding box rectangle frame.
The unique distinction of the present invention embodies:
1. use the convolutional neural networks for including ROI ponds so that positioning and type identification to vehicle in bayonet socket picture
Can once it complete, speed;
2. used in sample data derive from each real bayonet socket image, background is complicated, and illumination is complicated, vehicle model
It is numerous, therefore the grader trained can be competent at the Detection task of most scenes.
Claims (8)
1. a kind of vehicle location and type identifier method based on bayonet socket image, it is characterised in that:Using bayonet socket image pattern collection,
The positioning of vehicle in the convolutional neural networks of the detectable vehicle model of training, the image shot to bayonet system and model are carried out
Identification, comprises the following steps:
(1)Sample collection and mark:The source images of collection vehicle, mark out in source images and include from highway bayonet system
Each vehicle position and model, using source images and markup information as convolutional neural networks training sample set;
(2)The design of convolutional neural networks:Designed for the convolutional neural networks of fixation and recognition vehicle, source images and mark are believed
Feature is extracted in the input layer of breath input network, the intermediate layer of network using convolutional layer and full articulamentum, and the output layer of network is utilized
The feature extracted respectively obtains the bounding box and type of vehicle in picture by two output branchs;
(3)The training of convolutional neural networks:Instruction is iterated to the parameter of convolutional neural networks using source images and markup information
Practice, the function of making it possess positioning and identification type of vehicle;
(4)Vehicle location and type identifier:According to the convolutional neural networks trained, vehicle in the image shot to bayonet system
Position is positioned, and judges the model belonging to it.
2. a kind of vehicle location and type identifier method based on bayonet socket image according to claim 1, it is characterised in that
The step(1)Middle sample collection process is as follows:
The sample collection of big data:According to the model of driving vehicle on highway at this stage, according at least 100 samples of every kind of model
The requirement of source images, is gathering the source figure of different time sections from different geographical, the bayonet system of the highway of different road conditions
Decent.
3. a kind of vehicle location and type identifier method based on bayonet socket image according to claim 1, it is characterised in that
The step(1)Middle sample annotation process is as follows:
The sample mark of big data:The license plate number obtained automatically according to bayonet socket, the type of vehicle is obtained using the database of vehicle administration office
Number;The method selected by artificial frame obtains the upper left corner of vehicle location and bottom right angular coordinate in source images, by same Zhang Yuan's image
In all vehicle locations included and classification information all record in an XML document;Each vehicle source images are made
Split with selective search, the coordinate that the segmentation of generation is fast is all stored in a ss.mat.
4. a kind of vehicle location and type identifier method based on bayonet socket image according to claim 1, it is characterised in that
The step(2)In, the input layer design process of convolutional neural networks is as follows:
The input layer of convolutional neural networks is whole network input data:The data of training stage input include view data, sense
Interest region bounding box, the target bounding box and its label of generic of demarcation;The data of the input layer input of detection-phase
Including view data and area-of-interest bounding box.
5. a kind of vehicle location and type identifier method based on bayonet socket image according to claim 1, it is characterised in that
The step(2)In, the intermediate layer design process of convolutional neural networks is as follows:
The convolutional layer and full articulamentum of feature is are extracted in the intermediate layer of convolutional neural networks, and the structure of network is eight layer network knots
Structure, first layer and the second layer are after 96 core sizes are the convolutional layer that 11*11 and 256 core size is 5*5, each layer respectively
Face is used for the pond layer of dimensionality reduction provided with one;Third layer, the 4th layer, layer 5 be core size be 3*3 convolutional layer, convolution check figure
Amount is 384 respectively, 384 and 256, provided with a ROI ponds layer wherein after layer 5 convolutional layer, the ROI ponds layer is used to reality
Operation of the convolutional layer on whole pictures before now, and the result of convolution is mapped to each area-of-interest, produce simultaneously
The output of one fixed size as full articulamentum one, full articulamentum two input.
6. a kind of vehicle location and type identifier method based on bayonet socket image according to claim 1, it is characterised in that
The step(2)In, the output layer of convolutional neural networks is combined using the full articulamentum of branched structure, and its design process is as follows:
Be first full articulamentum one, full articulamentum two using the output of ROI ponds layer be used as input, produce one 4096
Characteristic vector;
Then this characteristic vector is separately input to the full articulamentum three and full articulamentum four of two branched structures:Full articulamentum
Three classifications for judging target, its length exported is N+1, and wherein N is the sum for including vehicle classification, and 1 is the number of background
Value;Full articulamentum four is used to export the value after the bounding box adjustment of target, and its length is (N+1) * 4, and wherein N is to include vehicle class
Other sum, 1 is the numerical value of background, and 4 be the numerical value of the transverse and longitudinal coordinate of bounding box upper left lower-right most point.
7. a kind of vehicle location and type identifier method based on bayonet socket image according to claim 1, it is characterised in that
The step(3)In, the training process of convolutional neural networks is as follows:
The cut zone produced using the vehicle data and selective search collected is carried out to convolutional neural networks
Supervised learning, each iteration takes two vehicle pictures, utilizes the interested area of vehicle bounding box and selective of demarcation
The cut zone that search is automatically generated, which is asked, to be handed over and compares, and will be handed over and is used as positive sample, generic than the cut zone more than 50%
Label for demarcation bounding box classification, will hand over and than for 10% ~ 50% cut zone as negative sample, classification is that label is
0, represent negative sample;By continuous iteration by error back propagation come the weights of the convolution kernel that adjusts each layer so that each
The convolution mask of layer can reach extraction characteristics of image, recognize type of vehicle, position the purpose of bounding box.
8. a kind of vehicle location and type identifier method based on bayonet socket image according to claim 1, it is characterised in that
The step(4)In, it is as follows the step of vehicle location and its type identifier:
For bayonet socket image to be detected, several are extracted on image first with selective search method may
Cut zone, then by image and segmentation coordinate input into the convolutional neural networks trained, convolutional layer is on the entire image
Extract after feature, Feature Mapping to cut zone is judged vehicle model by ROI ponds layer in full articulamentum using feature, adjustment
Vehicle bounding box obtains the final information of vehicles detected, finally, uses all bounding boxs the side of non-maxima suppression
Method, excludes unnecessary bounding box.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610900576.2A CN107085696A (en) | 2016-10-15 | 2016-10-15 | A kind of vehicle location and type identifier method based on bayonet socket image |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201610900576.2A CN107085696A (en) | 2016-10-15 | 2016-10-15 | A kind of vehicle location and type identifier method based on bayonet socket image |
Publications (1)
Publication Number | Publication Date |
---|---|
CN107085696A true CN107085696A (en) | 2017-08-22 |
Family
ID=59614096
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201610900576.2A Pending CN107085696A (en) | 2016-10-15 | 2016-10-15 | A kind of vehicle location and type identifier method based on bayonet socket image |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN107085696A (en) |
Cited By (21)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN107657815A (en) * | 2017-10-26 | 2018-02-02 | 成都九洲电子信息系统股份有限公司 | A kind of efficient vehicle image positioning identifying method |
CN108021909A (en) * | 2017-12-28 | 2018-05-11 | 北京悦畅科技有限公司 | A kind of parking lot vehicle Forecasting Methodology and device |
CN108154149A (en) * | 2017-12-08 | 2018-06-12 | 济南中维世纪科技有限公司 | Licence plate recognition method based on deep learning network share |
CN108171203A (en) * | 2018-01-17 | 2018-06-15 | 百度在线网络技术(北京)有限公司 | For identifying the method and apparatus of vehicle |
CN108182807A (en) * | 2017-12-27 | 2018-06-19 | 天津智芯视界科技有限公司 | A kind of generation method of car identifier |
CN109165541A (en) * | 2018-05-30 | 2019-01-08 | 北京飞鸿云际科技有限公司 | Coding method for vehicle component in intelligent recognition rail traffic vehicles image |
CN109284407A (en) * | 2018-08-21 | 2019-01-29 | 芜湖启迪睿视信息技术有限公司 | Device for training automatic labeling data set of intelligent sales counter |
CN109472264A (en) * | 2018-11-09 | 2019-03-15 | 北京字节跳动网络技术有限公司 | Method and apparatus for generating object detection model |
CN109508637A (en) * | 2018-10-10 | 2019-03-22 | 广州鹰瞰信息科技有限公司 | Embedded real-time vehicle detection method and system |
CN109948416A (en) * | 2018-12-31 | 2019-06-28 | 上海眼控科技股份有限公司 | A kind of illegal occupancy bus zone automatic auditing method based on deep learning |
CN109993032A (en) * | 2017-12-29 | 2019-07-09 | 杭州海康威视数字技术股份有限公司 | A kind of shared bicycle target identification method, device and camera |
WO2019134298A1 (en) * | 2018-01-02 | 2019-07-11 | 武汉斗鱼网络科技有限公司 | Image processing method, appapatus, electronic device and storage medium |
CN110047076A (en) * | 2019-03-29 | 2019-07-23 | 腾讯科技(深圳)有限公司 | A kind of processing method of image information, device and storage medium |
CN110135318A (en) * | 2019-05-08 | 2019-08-16 | 佳都新太科技股份有限公司 | Cross determination method, apparatus, equipment and the storage medium of vehicle record |
CN110222593A (en) * | 2019-05-18 | 2019-09-10 | 四川弘和通讯有限公司 | A kind of vehicle real-time detection method based on small-scale neural network |
CN111095265A (en) * | 2017-09-07 | 2020-05-01 | 奥迪股份公司 | Method for evaluating an optical representation in a vehicle environment and vehicle |
CN111160183A (en) * | 2019-12-20 | 2020-05-15 | 华瑞新智科技(北京)有限公司 | Method and device for detecting red light running of vehicle |
CN111950550A (en) * | 2020-08-13 | 2020-11-17 | 王宗尧 | Vehicle frame number identification system based on deep convolutional neural network |
CN112085018A (en) * | 2020-08-27 | 2020-12-15 | 南京佑驾科技有限公司 | License plate recognition system based on neural network |
CN112947872A (en) * | 2019-12-10 | 2021-06-11 | 上海品奇数码科技有限公司 | Method for placing detection equipment on display screen based on multilayer convolutional neural network |
CN113378992A (en) * | 2021-07-07 | 2021-09-10 | 山东建筑大学 | Vehicle positioning method and system based on position identification |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105488517A (en) * | 2015-11-30 | 2016-04-13 | 杭州全实鹰科技有限公司 | Vehicle brand model identification method based on deep learning |
CN105574543A (en) * | 2015-12-16 | 2016-05-11 | 武汉烽火众智数字技术有限责任公司 | Vehicle brand and model identifying method and system based on deep learning |
-
2016
- 2016-10-15 CN CN201610900576.2A patent/CN107085696A/en active Pending
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN105488517A (en) * | 2015-11-30 | 2016-04-13 | 杭州全实鹰科技有限公司 | Vehicle brand model identification method based on deep learning |
CN105574543A (en) * | 2015-12-16 | 2016-05-11 | 武汉烽火众智数字技术有限责任公司 | Vehicle brand and model identifying method and system based on deep learning |
Non-Patent Citations (4)
Title |
---|
HELLO~AGAIN: "Fast RCNN训练自己的数据集(2修改读写接口)", 《博客园HTTPS://WWW.CNBLOGS.COM/LOUYIHANG-LOVES-BAIYAN/P/4903231.HTML》 * |
ROSS GIRSHICK: "Fast R-CNN", 《2015 IEEE INTERNATIONAL CONFERENCE ON COMPUTER VISION》 * |
ZHIMEI ZHANG ET AL: "Vision-based vehicle detecting and counting for traffic flow analysis", 《2016 INTERNATIONAL JOINT CONFERENCE ON NEURAL NETWORKS (IJCNN)》 * |
凌继宝: "《机械工业、图形符号、文字代号实用手册》", 30 November 1994, 机械工业出版社 * |
Cited By (28)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN111095265B (en) * | 2017-09-07 | 2023-09-05 | 奥迪股份公司 | Method for evaluating an optical representation in a vehicle environment and vehicle |
CN111095265A (en) * | 2017-09-07 | 2020-05-01 | 奥迪股份公司 | Method for evaluating an optical representation in a vehicle environment and vehicle |
CN107657815A (en) * | 2017-10-26 | 2018-02-02 | 成都九洲电子信息系统股份有限公司 | A kind of efficient vehicle image positioning identifying method |
CN108154149A (en) * | 2017-12-08 | 2018-06-12 | 济南中维世纪科技有限公司 | Licence plate recognition method based on deep learning network share |
CN108154149B (en) * | 2017-12-08 | 2021-12-10 | 济南中维世纪科技有限公司 | License plate recognition method based on deep learning network sharing |
CN108182807A (en) * | 2017-12-27 | 2018-06-19 | 天津智芯视界科技有限公司 | A kind of generation method of car identifier |
CN108021909A (en) * | 2017-12-28 | 2018-05-11 | 北京悦畅科技有限公司 | A kind of parking lot vehicle Forecasting Methodology and device |
CN109993032A (en) * | 2017-12-29 | 2019-07-09 | 杭州海康威视数字技术股份有限公司 | A kind of shared bicycle target identification method, device and camera |
CN109993032B (en) * | 2017-12-29 | 2021-09-17 | 杭州海康威视数字技术股份有限公司 | Shared bicycle target identification method and device and camera |
WO2019134298A1 (en) * | 2018-01-02 | 2019-07-11 | 武汉斗鱼网络科技有限公司 | Image processing method, appapatus, electronic device and storage medium |
CN108171203B (en) * | 2018-01-17 | 2020-04-17 | 百度在线网络技术(北京)有限公司 | Method and device for identifying vehicle |
CN108171203A (en) * | 2018-01-17 | 2018-06-15 | 百度在线网络技术(北京)有限公司 | For identifying the method and apparatus of vehicle |
CN109165541A (en) * | 2018-05-30 | 2019-01-08 | 北京飞鸿云际科技有限公司 | Coding method for vehicle component in intelligent recognition rail traffic vehicles image |
CN109165541B (en) * | 2018-05-30 | 2023-05-02 | 北京飞鸿云际科技有限公司 | Encoding method for intelligently identifying vehicle parts in rail transit vehicle image |
CN109284407A (en) * | 2018-08-21 | 2019-01-29 | 芜湖启迪睿视信息技术有限公司 | Device for training automatic labeling data set of intelligent sales counter |
CN109508637A (en) * | 2018-10-10 | 2019-03-22 | 广州鹰瞰信息科技有限公司 | Embedded real-time vehicle detection method and system |
CN109472264A (en) * | 2018-11-09 | 2019-03-15 | 北京字节跳动网络技术有限公司 | Method and apparatus for generating object detection model |
CN109472264B (en) * | 2018-11-09 | 2020-10-27 | 北京字节跳动网络技术有限公司 | Method and apparatus for generating an object detection model |
CN109948416A (en) * | 2018-12-31 | 2019-06-28 | 上海眼控科技股份有限公司 | A kind of illegal occupancy bus zone automatic auditing method based on deep learning |
CN110047076A (en) * | 2019-03-29 | 2019-07-23 | 腾讯科技(深圳)有限公司 | A kind of processing method of image information, device and storage medium |
CN110135318A (en) * | 2019-05-08 | 2019-08-16 | 佳都新太科技股份有限公司 | Cross determination method, apparatus, equipment and the storage medium of vehicle record |
CN110222593A (en) * | 2019-05-18 | 2019-09-10 | 四川弘和通讯有限公司 | A kind of vehicle real-time detection method based on small-scale neural network |
CN112947872A (en) * | 2019-12-10 | 2021-06-11 | 上海品奇数码科技有限公司 | Method for placing detection equipment on display screen based on multilayer convolutional neural network |
CN111160183A (en) * | 2019-12-20 | 2020-05-15 | 华瑞新智科技(北京)有限公司 | Method and device for detecting red light running of vehicle |
CN111950550A (en) * | 2020-08-13 | 2020-11-17 | 王宗尧 | Vehicle frame number identification system based on deep convolutional neural network |
CN112085018A (en) * | 2020-08-27 | 2020-12-15 | 南京佑驾科技有限公司 | License plate recognition system based on neural network |
CN113378992A (en) * | 2021-07-07 | 2021-09-10 | 山东建筑大学 | Vehicle positioning method and system based on position identification |
CN113378992B (en) * | 2021-07-07 | 2023-11-21 | 山东建筑大学 | Vehicle positioning method and system based on position identification |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107085696A (en) | A kind of vehicle location and type identifier method based on bayonet socket image | |
Luo et al. | Traffic sign recognition using a multi-task convolutional neural network | |
CN105574550B (en) | A kind of vehicle identification method and device | |
CN104166841B (en) | The quick detection recognition methods of pedestrian or vehicle is specified in a kind of video surveillance network | |
CN105069746B (en) | Video real-time face replacement method and its system based on local affine invariant and color transfer technology | |
CN108509954A (en) | A kind of more car plate dynamic identifying methods of real-time traffic scene | |
CN109447169A (en) | The training method of image processing method and its model, device and electronic system | |
CN111079584A (en) | Rapid vehicle detection method based on improved YOLOv3 | |
CN104268528B (en) | A kind of crowd massing method for detecting area and device | |
CN110245644A (en) | A kind of unmanned plane image transmission tower lodging knowledge method for distinguishing based on deep learning | |
CN109359696A (en) | A kind of vehicle money recognition methods, system and storage medium | |
CN109948416A (en) | A kind of illegal occupancy bus zone automatic auditing method based on deep learning | |
CN108491797A (en) | A kind of vehicle image precise search method based on big data | |
CN107945153A (en) | A kind of road surface crack detection method based on deep learning | |
CN106909886B (en) | A kind of high-precision method for traffic sign detection and system based on deep learning | |
CN109766769A (en) | A kind of road target detection recognition method based on monocular vision and deep learning | |
CN104504377B (en) | A kind of passenger on public transport degree of crowding identifying system and method | |
CN106448184A (en) | Identifying method of Vehicles and exit of vehicles | |
CN106610969A (en) | Multimodal information-based video content auditing system and method | |
CN109033950A (en) | Vehicle based on multiple features fusion cascade deep model, which is disobeyed, stops detection method | |
CN108154102A (en) | A kind of traffic sign recognition method | |
CN105654066A (en) | Vehicle identification method and device | |
CN111274886B (en) | Deep learning-based pedestrian red light running illegal behavior analysis method and system | |
CN107092883A (en) | Object identification method for tracing | |
CN109902676A (en) | A kind of separated based on dynamic background stops detection algorithm |
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 | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20170822 |