CN111931768A - Vehicle identification method and system capable of self-adapting to sample distribution - Google Patents
Vehicle identification method and system capable of self-adapting to sample distribution Download PDFInfo
- Publication number
- CN111931768A CN111931768A CN202010818832.XA CN202010818832A CN111931768A CN 111931768 A CN111931768 A CN 111931768A CN 202010818832 A CN202010818832 A CN 202010818832A CN 111931768 A CN111931768 A CN 111931768A
- Authority
- CN
- China
- Prior art keywords
- vehicle
- identification
- image
- sampling
- network
- 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
Images
Classifications
-
- 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
-
- 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
-
- 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
- G06F18/2415—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches based on parametric or probabilistic models, e.g. based on likelihood ratio or false acceptance rate versus a false rejection rate
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/047—Probabilistic or stochastic networks
-
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V2201/00—Indexing scheme relating to image or video recognition or understanding
- G06V2201/08—Detecting or categorising vehicles
Abstract
The invention provides a vehicle identification method and system with adaptive sample distribution, which comprises the following steps: acquiring a vehicle image containing various vehicle types, setting sampling probability according to the number of vehicles corresponding to different vehicle types, and sampling the vehicle image to acquire a sampling image; respectively adopting different pre-trained branch networks to classify and recognize the vehicle image and the sampling image, and performing weighting processing on recognition results of the vehicle image and the sampling image to obtain a vehicle recognition result; the invention can effectively improve the accuracy of the whole vehicle type identification.
Description
Technical Field
The invention relates to the field of intelligent traffic, in particular to a vehicle identification method and system capable of self-adapting to sample distribution.
Background
With the rapid development of national economy, the increase of the number of vehicles brings great challenges to traffic supervision, and the intelligent traffic security system is considered to be the most effective solution for relieving traffic pressure. The vehicle identification technology is an important branch of the field of intelligent traffic security and protection, and has wide application prospects in various fields, such as standard traffic flow, a gate system, parking lot charge management, vehicle theft attack and the like. The vehicle identification technology is helpful for people to complete intelligent analysis of traffic monitoring data, and further promotes development of intelligent traffic.
Vehicle type recognition belongs to the fine recognition problem in vehicle recognition, however, under the real traffic condition, the quantity distribution among vehicle type categories is not balanced, and cars are many, and motorcycles and trucks are relatively few. The distribution of the number of samples between classes is greatly different, and a long-tail distribution mode is presented, namely, a small number of classes (head classes) have a large amount of data, and a large number of classes (tail classes) have only a small number of samples, which is a problem of long-tail distribution identification. When processing such visual data, the deep learning method is not enough to obtain excellent recognition accuracy, because the long-tail data distribution has an extreme class imbalance problem, so that the classifier training is difficult to obtain a good recognition effect. Therefore, it is urgent to find a vehicle type recognition method which is highly accurate and suitable for long-tailed distributed vehicle data.
Disclosure of Invention
In view of the problems in the prior art, the invention provides a vehicle identification method and system with adaptive sample distribution, which mainly solve the problem of low data identification rate of long tail distribution.
In order to achieve the above and other objects, the present invention adopts the following technical solutions.
A vehicle identification method of adaptive sample distribution comprises the following steps:
acquiring a vehicle image containing various vehicle types, setting sampling probability according to the number of vehicles corresponding to different vehicle types, and sampling the vehicle image to acquire a sampling image;
and respectively adopting different pre-trained branch networks to classify and recognize the vehicle image and the sampling image, and carrying out weighting processing on recognition results of the vehicle image and the sampling image to obtain a vehicle recognition result.
Optionally, the process of obtaining the vehicle identification result further includes:
acquiring the characteristics of the vehicle image as first characteristics, and performing classification and identification on the first characteristics to acquire a first classification result;
acquiring the characteristics of the sampling image as second characteristics, performing characteristic screening according to the first characteristics and the second characteristics, acquiring special characteristics of a vehicle corresponding to the type of the sampling vehicle, and performing classification and identification on the special characteristics to acquire a second classification result;
and weighting the first classification result and the second classification result to obtain a vehicle identification result.
Optionally, the sampling probability is set according to a proportion of the total number of vehicles in the vehicle image to the number of vehicles corresponding to each vehicle type; wherein the sampling probability is inversely proportional to the ratio.
Optionally, if the branch network for identifying the vehicle image is a first identification network, and the branch network for identifying the sampled image is a second identification network, the first identification network includes a feature layer and a classifier layer; the second identification network comprises a feature layer, a classifier layer and a feature migration layer for feature screening; the feature migration layer is connected with the outputs of the feature layer of the first identification network and the feature layer of the second identification network respectively.
Optionally, the feature layer of the first identification network adopts a MobileNet-V2 network structure; the classifier layer at least comprises two full-connection layers, and the number of the neurons in the last full-connection layer is equal to the number of vehicle types; and inputting the extracted features into the classifier layer for classification and identification by the last convolutional layer of the MobileNet-V2 network structure.
Optionally, the feature migration layer consists of a convolution layer with a convolution kernel of 1 x 1.
Optionally, setting adaptive parameters to obtain weights of different branch network identification results, and performing weighting processing according to the weights.
Optionally, when the branch network is trained, setting the adaptive parameter according to a ratio of a current update iteration number of the branch network to a total iteration number;
and after the training is finished, setting the self-adaptive parameter as a preset fixed value.
An adaptive sample distribution vehicle identification system comprising:
the reverse sampling module is used for acquiring vehicle images containing various vehicle types, setting sampling probability according to the number of vehicles corresponding to different vehicle types, sampling the vehicle images and acquiring sampling images;
and the recognition module is used for respectively adopting different pre-trained branch networks to carry out classification recognition on the vehicle image and the sampling image, and carrying out weighting processing on recognition results of the vehicle image and the sampling image to obtain a vehicle recognition result.
Optionally, the identification module comprises a first identification unit and a second identification unit; the first identification unit acquires the characteristics of the vehicle image as first characteristics, and carries out classification identification on the first characteristics to acquire a first classification result;
the second identification unit acquires the characteristics of the sampling image as second characteristics, performs characteristic screening according to the first characteristics and the second characteristics, acquires special characteristics of a vehicle corresponding to the type of the sampling vehicle, and performs classification identification on the special characteristics to acquire a second classification result;
and weighting the first classification result and the second classification result to obtain a vehicle identification result.
As described above, the vehicle identification method and system based on adaptive sample distribution according to the present invention have the following advantages.
The vehicle images and the sampling images are classified and recognized through different branch networks, and the results are weighted, so that the accuracy of overall vehicle type recognition can be effectively improved.
Drawings
FIG. 1 is a flow chart of a method for vehicle identification with adaptive sample distribution according to an embodiment of the invention.
FIG. 2 is a block diagram of an adaptive sample distribution vehicle identification system in accordance with an embodiment of the present invention.
Fig. 3 is a schematic flow chart of the branch network training according to an embodiment of the present invention.
Detailed Description
The embodiments of the present invention are described below with reference to specific embodiments, and other advantages and effects of the present invention will be easily understood by those skilled in the art from the disclosure of the present specification. The invention is capable of other and different embodiments and of being practiced or of being carried out in various ways, and its several details are capable of modification in various respects, all without departing from the spirit and scope of the present invention. It is to be noted that the features in the following embodiments and examples may be combined with each other without conflict.
It should be noted that the drawings provided in the following embodiments are only for illustrating the basic idea of the present invention, and the components related to the present invention are only shown in the drawings rather than drawn according to the number, shape and size of the components in actual implementation, and the type, quantity and proportion of the components in actual implementation may be changed freely, and the layout of the components may be more complicated.
Referring to fig. 1, the present invention provides a vehicle identification method with adaptive sample distribution, including steps S01-S02.
In step S01, a vehicle image including a plurality of vehicle type categories is acquired, and the vehicle image is sampled according to the number of vehicles corresponding to different vehicle type categories, and a sampled image is acquired:
in one embodiment, a vehicle image corresponding to the real-time traffic flow on the road surface can be acquired through a camera device installed in the urban road.
In one embodiment, the collected vehicle image may be preprocessed, for example, the vehicle image may be cut into a preset size image (224 × 224, etc.), and the gray-scale mean value of each color channel is subtracted to obtain a normalized standard vehicle image.
The distribution of vehicle types in the real-time traffic flow usually shows long-tail distribution, namely, more vehicles correspond to one vehicle type, and less other vehicle types have large distribution difference. Furthermore, the vehicle image can be reversely sampled, the sampling probability of the tail data in long tail distribution is increased, and the sampling probability of the head data is reduced, so that the identification is focused on the tail data. Specifically, when the inverse sampling is performed, the sampling probability of each vehicle type category is inversely proportional to the proportion of the vehicle type category in all vehicle type categories included in the vehicle image. I.e. the fewer the number of vehicles in a category, the higher the probability that the category is sampled.
Assuming a total of C categories, the number of vehicles in category i is NiAnd if the total number of vehicles corresponding to all vehicle type categories is N, the reverse sampling step is as follows:
(1) calculating the sampling probability p of class ii
Wherein the content of the first and second substances,
(2) according to the sampling probability piRandomly sampling a class
(3) And repeating the operation until all the classes are sampled to obtain a sampled image.
In step S02, the vehicle image and the sampled image are classified and identified by using different pre-trained branch networks, and the identification results are weighted to obtain a vehicle identification result:
referring to fig. 3, in an embodiment, when performing the branch network training, the vehicle image and the sampling image may be obtained according to the method of step S01 to respectively construct a vehicle image training sample set and a sampling image training sample set.
In one embodiment, the branch network comprises a first identification network (i.e. a conventional learning network branch) through which the vehicle image is classified and identified and a second identification network (i.e. a rebalance learning network branch) through which the sampled image is classified and identified. Wherein the first recognition network may include a feature layer and a classifier layer. And inputting the preprocessed vehicle image training sample into a feature layer of the first recognition network for feature extraction. The feature layer may employ a MobileNet-V2 network structure that includes deep separable convolutions and point convolutions. The deep separable convolution is used for feature extraction, and one convolution kernel of the deep separable convolution is responsible for one channel, and one channel is convoluted by only one convolution kernel. However, the number of feature maps after the depth separable convolution cannot be changed, and therefore, the increase and decrease of the feature maps are realized using the point convolution. After feature extraction by MobileNet V2, the feature of the last convolutional layer of MobileNet V2 was taken out and input into the classifier layer. The classifier layer at least comprises two full connection layers, and the number of neurons of the last full connection layer is the number of vehicle type categories. In another embodiment, feature extraction can be performed by using other network structures, and the MobileNet-V2 network structure is only used as an example.
In an embodiment, the second recognition network may include a feature layer, a feature migration layer, and a classifier layer. And inputting the sampled image training sample into a feature layer of a second recognition network for feature extraction. The feature layer of the second recognition network can be consistent with the network structure and parameters of the first recognition network, so that the time consumption of feature layer training is reduced. The feature migration layer receives an output of the feature layer of the second identified network and an output of the feature layer of the first identified network.
In an embodiment, the feature migration layer may be composed of convolution layers with convolution kernel 1 × 1, the number of output channels of the feature migration layer is equal to the number of input channels, different positions do not interfere with each other, the input feature map is not down-sampled, and the dimensionality of the original feature can be maintained. Specifically, the distance between the feature map extracted from the vehicle image and the feature map extracted from the sampled image can be judged through two feature migration layers, so that feature screening is performed, feature information which is not beneficial to tail data identification is filtered, and special features for tail data identification are obtained. And inputting the special features obtained by the feature migration layer into a classifier layer of the second recognition network. The classifier layer at least comprises a full connection layer, and the number of neurons of the last full connection layer is equal to the number of vehicle type categories.
In an embodiment, the classifier layer of the first recognition network and the classifier layer of the second recognition network may both employ a softmax function as the loss function. When weighting processing is performed, the total loss is obtained by performing weighted summation on the calculation results of the two loss functions, which is specifically expressed as follows:
L＝αL1+(1-α)L2
wherein L is1Denotes the softmax loss, L, of the first identified network2Representing the softmax loss of the second identified network. Alpha is a set adaptive parameter, is automatically generated according to the current training epoch (iteration number), and is mainly used for controlling the main focus of the network in different training stages, so that the two branches can simultaneously maintain respective learning states and mutually promote learning in the whole training process, and the expression is as follows
Where T denotes the epoch of the current training and T denotes the total epoch number. As t increases, the value of α gradually decreases. That is, the model initially favors the first recognition network in order to learn the generic features from the distribution of the vehicle images. With the increase of training times, the network focuses on the second recognition network gradually, so that the recognition rate of tail data is improved.
The total loss is utilized to adjust the whole network, on one hand, the network focuses on the accurate identification of the tail data, on the other hand, the weight updating of the first identification network is indirectly influenced by the loss of the second identification network, and the feature extraction result is further optimized.
The above steps are network training steps, when vehicle type recognition is performed after neural network training is completed, the preset value of the adaptive parameter can be set according to the importance degree of the first recognition network and the second recognition network, if the value of alpha is fixed to be 0.6, and the results of the two branches are connected one by one according to vehicle type categories (if the results of the same category of the first recognition network and the second recognition network are multiplied by one through element-wise operation), so as to further obtain the final vehicle type recognition result.
Referring to fig. 2, a vehicle identification system with adaptive sample distribution is further provided in this embodiment, for implementing the vehicle identification method with adaptive sample distribution described in the foregoing method embodiment. Since the technical principle of the embodiment of the apparatus is similar to that of the embodiment of the method, repeated description of the same technical details is omitted.
In one embodiment, the adaptive sample distribution vehicle identification system includes an inverse sampling module 10 and an identification module 11, wherein the inverse sampling module 10 is configured to assist in performing step S01 described in the foregoing method embodiment; the identification module 11 is used to assist in performing step S02 described in the previous method embodiment.
In one embodiment, the identification module comprises a first identification unit and a second identification unit; the first identification unit acquires the characteristics of the vehicle image as first characteristics, and carries out classification identification on the first characteristics to acquire a first classification result;
the second identification unit acquires the characteristics of the sampled image as second characteristics, performs characteristic screening according to the first characteristics and the second characteristics, acquires special characteristics of the vehicle corresponding to the type of the sampled vehicle, and performs classification and identification on the special characteristics to acquire a second classification result;
and weighting the first classification result and the second classification result to obtain a vehicle identification result.
In summary, according to the vehicle identification method and system based on adaptive sample distribution, the first identification network branch is used for learning the general characteristics of the vehicle image, the second identification network branch is used for emphasizing the identification effect of the tail data, and the characteristic description with excellent performance in the whole vehicle type range can be obtained by combining the first identification network branch and the second identification network branch, so that the accuracy of whole vehicle type identification is improved; and a lightweight deep convolution neural network model is integrated, and the model is further compressed, so that the overall operation speed is increased. Therefore, the invention effectively overcomes various defects in the prior art and has high industrial utilization value.
The foregoing embodiments are merely illustrative of the principles and utilities of the present invention and are not intended to limit the invention. Any person skilled in the art can modify or change the above-mentioned embodiments without departing from the spirit and scope of the present invention. Accordingly, it is intended that all equivalent modifications or changes which can be made by those skilled in the art without departing from the spirit and technical spirit of the present invention be covered by the claims of the present invention.
Claims (10)
1. A vehicle identification method of adaptive sample distribution is characterized by comprising the following steps:
acquiring a vehicle image containing various vehicle types, setting sampling probability according to the number of vehicles corresponding to different vehicle types, and sampling the vehicle image to acquire a sampling image;
and respectively adopting different pre-trained branch networks to classify and recognize the vehicle image and the sampling image, and carrying out weighting processing on recognition results of the vehicle image and the sampling image to obtain a vehicle recognition result.
2. The adaptive sample distribution vehicle identification method according to claim 1, wherein the process of obtaining the vehicle identification result further comprises:
acquiring the characteristics of the vehicle image as first characteristics, and performing classification and identification on the first characteristics to acquire a first classification result;
acquiring the characteristics of the sampling image as second characteristics, performing characteristic screening according to the first characteristics and the second characteristics, acquiring special characteristics of a vehicle corresponding to the type of the sampling vehicle, and performing classification and identification on the special characteristics to acquire a second classification result;
and weighting the first classification result and the second classification result to obtain a vehicle identification result.
3. The adaptive sample distribution vehicle identification method according to claim 1, wherein the sampling probability is set according to a ratio of the total number of vehicles in the vehicle image to the number of vehicles corresponding to each vehicle type; wherein the sampling probability is inversely proportional to the ratio.
4. The adaptive sample distribution vehicle identification method according to claim 2, wherein if the branch network for identifying the vehicle image is a first identification network and the branch network for identifying the sample image is a second identification network, the first identification network includes a feature layer and a classifier layer; the second identification network comprises a feature layer, a classifier layer and a feature migration layer for feature screening; the feature migration layer is connected with the outputs of the feature layer of the first identification network and the feature layer of the second identification network respectively.
5. The adaptive sample distribution vehicle identification method according to claim 4, wherein the feature layer of the first identification network adopts a MobileNet-V2 network structure; the classifier layer at least comprises two full-connection layers, and the number of the neurons in the last full-connection layer is equal to the number of vehicle types; and inputting the extracted features into the classifier layer for classification and identification by the last convolutional layer of the MobileNet-V2 network structure.
6. The adaptive sample distribution vehicle identification method of claim 4, wherein the feature migration layer is composed of convolution layers with a convolution kernel of 1 x 1.
7. The adaptive sample distribution vehicle identification method according to claim 1, wherein adaptive parameters are set to obtain weights of different branch network identification results, and weighting processing is performed according to the weights.
8. The adaptive sample distribution vehicle identification method according to claim 7, wherein the adaptive parameter is set according to a ratio of a current update iteration number of a branch network to a total iteration number when the branch network is trained;
and after the training is finished, setting the self-adaptive parameter as a preset fixed value.
9. An adaptive sample distribution vehicle identification system, comprising:
the reverse sampling module is used for acquiring vehicle images containing various vehicle types, setting sampling probability according to the number of vehicles corresponding to different vehicle types, sampling the vehicle images and acquiring sampling images;
and the recognition module is used for respectively adopting different pre-trained branch networks to carry out classification recognition on the vehicle image and the sampling image, and carrying out weighting processing on recognition results of the vehicle image and the sampling image to obtain a vehicle recognition result.
10. The adaptive sample distribution vehicle identification system of claim 9, wherein the identification module comprises a first identification unit and a second identification unit; the first identification unit acquires the characteristics of the vehicle image as first characteristics, and carries out classification identification on the first characteristics to acquire a first classification result;
the second identification unit acquires the characteristics of the sampling image as second characteristics, performs characteristic screening according to the first characteristics and the second characteristics, acquires special characteristics of a vehicle corresponding to the type of the sampling vehicle, and performs classification identification on the special characteristics to acquire a second classification result;
and weighting the first classification result and the second classification result to obtain a vehicle identification result.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010818832.XA CN111931768A (en) | 2020-08-14 | 2020-08-14 | Vehicle identification method and system capable of self-adapting to sample distribution |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202010818832.XA CN111931768A (en) | 2020-08-14 | 2020-08-14 | Vehicle identification method and system capable of self-adapting to sample distribution |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN111931768A true CN111931768A (en) | 2020-11-13 |
Family
ID=73311101
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202010818832.XA Pending CN111931768A (en) | 2020-08-14 | 2020-08-14 | Vehicle identification method and system capable of self-adapting to sample distribution |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111931768A (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113344479A (en) * | 2021-08-06 | 2021-09-03 | 首都师范大学 | Online classroom-oriented learning participation intelligent assessment method and device |
Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106874840A (en) * | 2016-12-30 | 2017-06-20 | 东软集团股份有限公司 | Vehicle information recognition method and device |
| CN107025459A (en) * | 2016-01-29 | 2017-08-08 | 中兴通讯股份有限公司 | A kind of model recognizing method and device |
| CN109214441A (en) * | 2018-08-23 | 2019-01-15 | 桂林电子科技大学 | A kind of fine granularity model recognition system and method |
| CN110147707A (en) * | 2018-10-25 | 2019-08-20 | 初速度（苏州）科技有限公司 | A kind of high-precision vehicle identification method and system |
| CN110717529A (en) * | 2019-09-25 | 2020-01-21 | 南京旷云科技有限公司 | Data sampling method and device |
| US20200074205A1 (en) * | 2017-06-28 | 2020-03-05 | Beijing Sensetime Technology Development Co., Ltd. | Methods and apparatuses for vehicle appearance feature recognition, methods and apparatuses for vehicle retrieval, storage medium, and electronic devices |
| CN110991349A (en) * | 2019-12-05 | 2020-04-10 | 中国科学院重庆绿色智能技术研究院 | Lightweight vehicle attribute identification method based on metric learning |
| CN111291841A (en) * | 2020-05-13 | 2020-06-16 | 腾讯科技（深圳）有限公司 | Image recognition model training method and device, computer equipment and storage medium |
| CN111310815A (en) * | 2020-02-07 | 2020-06-19 | 北京字节跳动网络技术有限公司 | Image recognition method and device, electronic equipment and storage medium |
| WO2020141683A1 (en) * | 2018-12-31 | 2020-07-09 | 렉스젠(주) | Image-based traffic information providing system and method thereof |
-
2020
- 2020-08-14 CN CN202010818832.XA patent/CN111931768A/en active Pending
Patent Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107025459A (en) * | 2016-01-29 | 2017-08-08 | 中兴通讯股份有限公司 | A kind of model recognizing method and device |
| CN106874840A (en) * | 2016-12-30 | 2017-06-20 | 东软集团股份有限公司 | Vehicle information recognition method and device |
| US20200074205A1 (en) * | 2017-06-28 | 2020-03-05 | Beijing Sensetime Technology Development Co., Ltd. | Methods and apparatuses for vehicle appearance feature recognition, methods and apparatuses for vehicle retrieval, storage medium, and electronic devices |
| CN109214441A (en) * | 2018-08-23 | 2019-01-15 | 桂林电子科技大学 | A kind of fine granularity model recognition system and method |
| CN110147707A (en) * | 2018-10-25 | 2019-08-20 | 初速度（苏州）科技有限公司 | A kind of high-precision vehicle identification method and system |
| WO2020141683A1 (en) * | 2018-12-31 | 2020-07-09 | 렉스젠(주) | Image-based traffic information providing system and method thereof |
| CN110717529A (en) * | 2019-09-25 | 2020-01-21 | 南京旷云科技有限公司 | Data sampling method and device |
| CN110991349A (en) * | 2019-12-05 | 2020-04-10 | 中国科学院重庆绿色智能技术研究院 | Lightweight vehicle attribute identification method based on metric learning |
| CN111310815A (en) * | 2020-02-07 | 2020-06-19 | 北京字节跳动网络技术有限公司 | Image recognition method and device, electronic equipment and storage medium |
| CN111291841A (en) * | 2020-05-13 | 2020-06-16 | 腾讯科技（深圳）有限公司 | Image recognition model training method and device, computer equipment and storage medium |
Non-Patent Citations (1)
| Title |
|---|
| 王召月等: "基于概率分布估计的私家车和出租车行程时间可变性度量", 《计算机应用与软件》 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN113344479A (en) * | 2021-08-06 | 2021-09-03 | 首都师范大学 | Online classroom-oriented learning participation intelligent assessment method and device |
| CN113344479B (en) * | 2021-08-06 | 2022-01-07 | 首都师范大学 | Online classroom-oriented learning participation intelligent assessment method and device |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106599773B (en) | Deep learning image identification method and system for intelligent driving and terminal equipment | |
| CN111914907B (en) | Hyperspectral image classification method based on deep learning space-spectrum combined network | |
| CN109902806B (en) | Method for determining target bounding box of noise image based on convolutional neural network | |
| CN107886073B (en) | Fine-grained vehicle multi-attribute identification method based on convolutional neural network | |
| CN111275711B (en) | Real-time image semantic segmentation method based on lightweight convolutional neural network model | |
| CN110163187B (en) | F-RCNN-based remote traffic sign detection and identification method | |
| CN105956532B (en) | A kind of traffic scene classification method based on multiple dimensioned convolutional neural networks | |
| CN111460919B (en) | Monocular vision road target detection and distance estimation method based on improved YOLOv3 | |
| CN110796009A (en) | Method and system for detecting marine vessel based on multi-scale convolution neural network model | |
| CN111310773A (en) | Efficient license plate positioning method of convolutional neural network | |
| CN108830254B (en) | Fine-grained vehicle type detection and identification method based on data balance strategy and intensive attention network | |
| CN104517103A (en) | Traffic sign classification method based on deep neural network | |
| CN112464911A (en) | Improved YOLOv 3-tiny-based traffic sign detection and identification method | |
| CN110532961B (en) | Semantic traffic light detection method based on multi-scale attention mechanism network model | |
| CN110991349B (en) | Lightweight vehicle attribute identification method based on metric learning | |
| CN110706239A (en) | Scene segmentation method fusing full convolution neural network and improved ASPP module | |
| CN110717493B (en) | License plate recognition method containing stacked characters based on deep learning | |
| CN111339935A (en) | Optical remote sensing picture classification method based on interpretable CNN image classification model | |
| CN112115871B (en) | High-low frequency interweaving edge characteristic enhancement method suitable for pedestrian target detection | |
| CN111738114A (en) | Vehicle target detection method based on anchor-free accurate sampling remote sensing image | |
| CN111540203B (en) | Method for adjusting green light passing time based on fast-RCNN | |
| CN109472733A (en) | Image latent writing analysis method based on convolutional neural networks | |
| CN111079543B (en) | Efficient vehicle color identification method based on deep learning | |
| CN113205107A (en) | Vehicle type recognition method based on improved high-efficiency network | |
| CN112084897A (en) | Rapid traffic large-scene vehicle target detection method of GS-SSD |
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 |
Application publication date: 20201113 |
|
| RJ01 | Rejection of invention patent application after publication |