CN115394074A - Road monitoring vehicle detection system - Google Patents
Road monitoring vehicle detection system Download PDFInfo
- Publication number
- CN115394074A CN115394074A CN202210780512.9A CN202210780512A CN115394074A CN 115394074 A CN115394074 A CN 115394074A CN 202210780512 A CN202210780512 A CN 202210780512A CN 115394074 A CN115394074 A CN 115394074A
- Authority
- CN
- China
- Prior art keywords
- license plate
- network
- vehicle
- picture
- detection
- 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
- G08—SIGNALLING
- G08G—TRAFFIC CONTROL SYSTEMS
- G08G1/00—Traffic control systems for road vehicles
- G08G1/01—Detecting movement of traffic to be counted or controlled
- G08G1/0104—Measuring and analyzing of parameters relative to traffic conditions
- G08G1/0108—Measuring and analyzing of parameters relative to traffic conditions based on the source of data
-
- G—PHYSICS
- G08—SIGNALLING
- G08G—TRAFFIC CONTROL SYSTEMS
- G08G1/00—Traffic control systems for road vehicles
- G08G1/01—Detecting movement of traffic to be counted or controlled
- G08G1/0104—Measuring and analyzing of parameters relative to traffic conditions
- G08G1/0125—Traffic data processing
-
- G—PHYSICS
- G08—SIGNALLING
- G08G—TRAFFIC CONTROL SYSTEMS
- G08G1/00—Traffic control systems for road vehicles
- G08G1/01—Detecting movement of traffic to be counted or controlled
- G08G1/017—Detecting movement of traffic to be counted or controlled identifying vehicles
Landscapes
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Chemical & Material Sciences (AREA)
- Analytical Chemistry (AREA)
- Traffic Control Systems (AREA)
Abstract
A road monitoring vehicle detection system comprises a monitoring camera, a control box and network equipment; the control box is positioned beside the road where the monitoring camera is positioned; the monitoring camera is connected with network equipment in the control box, and transmits shot videos outwards through the network equipment. The detection device takes a computer as a core; the network equipment comprises a wireless and/or wired communication module and a network switch; the network switch is also connected with the memory; the monitoring camera and the detection module are connected with the network switch through a network cable; transmitting the video of the monitoring camera to a memory through a network switch; the detection module calls a required video stream from the memory and intercepts a single-frame picture; the detection module comprises a positioning unit and an identification unit; the positioning unit obtains the position information of the license plate; the recognition unit outputs the license plate characters. And the vehicle picture and the corresponding license plate character information acquired by the detection equipment are stored in a memory for being called by a management terminal.
Description
Technical Field
The invention relates to a road monitoring vehicle detection system, which detects vehicles and corresponding license plates through detection videos, identifies the vehicles and the corresponding license plates, and finally outputs identification results and vehicle images.
Background
The traditional road monitoring vehicle detection mainly identifies the license plate of the vehicle and corresponds to the illegal information of the vehicle. Because the traffic laws and regulations are updated and perfected, the types of the laws and regulations are more, and the behaviors of drivers and passengers are regulated besides the traditional vehicle law violation. It is required to detect that the vehicle picture corresponds to the license plate, and a technical support is provided for improving the supervision efficiency. Meanwhile, as the construction of the road monitoring system is completed, how to realize new functions by utilizing the existing hardware equipment with low cost and high efficiency is a problem to be solved.
Disclosure of Invention
In order to solve the above problems in the prior art, the invention provides a road monitoring vehicle detection system, which can be realized by adding a detection module on the basis of the existing road monitoring system, and can also be newly built. The technical scheme is as follows:
a road monitoring vehicle detection system comprises a monitoring camera, a control box and network equipment; the control box is positioned beside the monitoring camera; the monitoring camera is connected with network equipment in the control box and transmits shot video outwards through the network equipment, and the monitoring camera is characterized by also comprising a detection module, wherein the detection module takes a computer as a core;
the network equipment comprises a wireless and/or wired communication module and a network switch; the network switch is also connected with the memory;
the monitoring camera and the detection module are connected with the network switch through a network cable; transmitting the video of the monitoring camera to a memory through a network switch; the detection module calls a required video stream from the memory and intercepts a single-frame picture;
the detection module comprises a positioning unit and an identification unit;
in the positioning unit:
firstly, enhancing an intercepted single-frame picture;
then, a vehicle detection model is adopted to obtain a vehicle detection area from the enhanced picture;
then, cutting the original single-frame picture according to the vehicle detection area to obtain a picture of a vehicle part, and then enhancing the picture of the vehicle part;
finally, detecting the enhanced picture of the vehicle part by using a license plate detection model to obtain the position information of the license plate;
the identification unit is provided with:
firstly, according to the position information of a license plate, a license plate picture is captured from a picture of a vehicle part;
then recognizing characters, letters and numerical information in the license plate picture by using a license plate recognition model;
and finally, outputting the license plate characters according to the code of the license plate characters.
And the vehicle picture and the corresponding license plate character information acquired by the detection equipment are stored in a memory for being called by a management terminal.
The system strips the functions of vehicle license plate detection and license plate recognition from the original detection system, processes the video stream picture by adopting the neural network technology, obtains a result for the management terminal to call, reduces the computation of other parts of the system, and reduces the network transmission data amount between the management terminal and the field terminal.
Drawings
FIG. 1 is a schematic diagram of a road monitoring vehicle detection system.
Detailed Description
Referring to fig. 1, a road monitoring vehicle detecting system includes a monitoring camera, a control box and a network device; the control box is positioned beside the monitoring camera; the monitoring camera is connected with network equipment in the control box, and transmits shot videos outwards through the network equipment. The detection device is a computer system taking a CPU and a GPU as cores.
The network equipment comprises a wireless and/or wired communication module and a network switch; the network switch is also connected with the memory; the wireless and wired communication modules can adopt the existing mature communication technology, such as 4G, 5G or wired network transmission technology.
The monitoring camera and the detection module are connected with the network switch through a network cable; transmitting the video of the monitoring camera to a memory through a network switch; the detection module calls a required video stream from the memory and intercepts a single-frame picture;
the detection module comprises a positioning unit and an identification unit;
in the positioning unit:
firstly, enhancing an intercepted single-frame picture;
then, a vehicle detection model is adopted to obtain a vehicle detection area from the enhanced picture;
then, cutting the original single-frame picture according to the vehicle detection area to obtain a picture of a vehicle part, and then enhancing the picture of the vehicle part;
finally, detecting the enhanced picture of the vehicle part by using a license plate detection model to obtain the position information of the license plate;
the identification unit is provided with:
firstly, according to the position information of a license plate, a license plate picture is captured from a picture of a vehicle part;
then recognizing characters, letters and digital information in the license plate picture by using a license plate recognition model;
and finally, outputting the license plate characters according to the code of the license plate characters.
And the vehicle picture and the corresponding license plate character information acquired by the detection equipment are stored in a memory for being called by a management terminal.
Specifically, in this embodiment:
in the positioning unit:
1) And reading the video stream through opencv, and intercepting a single-frame picture.
2) Performing HSV enhancement on the image, wherein an H channel enhancement parameter is 0.014, an S channel enhancement parameter is 0.68, a V channel enhancement parameter is 0.36, processing the image to be 416 multiplied by 3, firstly scaling the image according to the ratio of the longest edge to 416, then supplementing 0 pixel in the short edge direction, and adopting the scaling mode can not change the original proportion of the image.
3) Training large-scale YOLOv5 vehicle detection network setup parameters using the detac dataset: the net depth parameter is 0.33, the net width parameter is 0.5, the learning rate learning _ rate is 0.01, the stochastic gradient descent momentum parameter momentum is 0.937, and the weight attenuation is 0.0005. The YOLOv5 vehicle detection loss function formula is set as:
L＝L car_conf +L car_loc
wherein L is car_conf For vehicle target confidence loss, L car_loc Being vehiclesThere is no loss of category since there is only a vehicle detection category.
TABLE 1 is the large-scale YOLOv5 network parameter information
4) Training small-scale YOLOv5 license plate detection network setting parameters by using a CCPD2020 dataset: the net depth parameter is 0.2, the net width parameter is 0.25, the learning rate learning _ rate is 0.01, the stochastic gradient descent momentum parameter momentum is 0.937, and the weight attenuation is 0.0005. The YOLOv5 license plate detection loss function formula is set as follows:
L＝L LP_conf +L LP_loc
wherein L is LP_conf Loss of confidence for license plate target, L LP_loc And (5) loss positioning is carried out on the license plate target.
TABLE 2 Small-Scale YOLOv5 network parameter information
5) And constructing a trained large-scale YOLOv5 neural network to detect the vehicle, and detecting the input image to obtain a vehicle detection ROI.
6) And (3) cutting the original input image by using the detection ROI to obtain vehicle parts in the image, and processing the vehicle pictures one by adopting the image processing method in the step 2).
7) And constructing a trained small-scale YOLOv5 neural network to detect the license plate, and detecting the vehicle pictures one by one to obtain the license plate detection coordinates of each vehicle.
8) And mapping the coordinates of the license plate detection back to the input picture to realize the vehicle license plate detection and positioning in the picture.
The positioning unit is designed to detect vehicles in the road video through the large-scale YOLOv5, and detect the license plates of the vehicles in the frame through the small-scale YOLOv5 by taking the detection frame as the ROI, so as to obtain the positioning information of the license plates. The YOLOv5 models with multiple scales are used for detecting the vehicle and the license plate respectively, so that the detection efficiency is greatly improved. In the middle positioning unit, the original image is cut and amplified through the detected vehicle ROI in a model cascade mode, license plate information loss is avoided, license plate detection precision and recall rate are improved, and the model has stronger robustness.
In the identification unit:
the design of the convolutional neural network of the license plate recognition model is as follows:
step 1) model architecture design
For the input license plate image, using ResNet18 as a main feature extraction network to extract features;
respectively sending the extracted features into a plurality of same classifiers; each classifier is composed of a three-layer feedforward neural network;
the output dimension of the 1 st classifier is 38, and the classifier is used for predicting the 1 st character in the license plate and is a subordinate province of the license plate;
the output dimension of the 2 nd classifier is 25, and the 2 nd classifier is used for predicting the 2 nd character in the license plate and is a membership area of the license plate;
the 3 rd to 8 th classifiers are used for predicting the 3 rd to 8 th characters in the license plate and are personal codes of the license plate;
2) Label making
The characters of the license plate are coded into three parts, namely province providing, regional ALPHABETS and personal ADS codes in sequence; during coding, the number of character bits corresponding to the personal code in the blue fuel license plate and the number of character bits corresponding to the personal code in the new energy license plate are set to be the same, and the difference part of the number of character bits is represented by 'n'; the blue fuel vehicle is also regarded as eight characters, and the last one is a space "
The license plate character code is expressed as:
PROVINCES = [ "Jing", "jin", "Shanghai", "Yu", "Ji", "Yu", "Chuan", "cloud", "Liao", "black", "Xiang", "Wan", "Shang", "Su", "Zhe", "gan", "Hu", "jin", "Gui", "Qin", "Gui", "Xie", "Zang", "shan", "Gan", "Qing", "Ning", "Xin", "Xian", "school" ]
ALPHABETS＝[‘A’,’B’,’C’,’D’,’E’,’F’,’G’,’H’,’J’,’K’,’L’,’M’,’N’,’P’,’Q’,’R’,’S’,’T’,’U’,’V’,’W’,’X’,’Y’,’Z’,]
ADS＝[’A’,’B’,’C’,’D’,’E’,’F’,’G’,’H’,’J’,’K’,’L’,’M’,’N’,’P’,’Q’,’R’,’S’,’T’,’U’,’V’,’W’,’X’,’Y’,’Z’,’0’,’1’,’2’,’3’,’4’,’5’,’6’,’7’,’8’,’9’,”]
3) Optimizing convolutional neural networks using loss functions
Using cross entropy as a loss function, the loss is the sum of the eight classifier losses;
4) Training network
Training by adopting a random gradient descent method, and loading a pre-training weight of ResNet18 on a coco data set in a transfer learning mode in order to facilitate convergence of a network;
the weight of the newly added classifier meets the Gaussian distribution random initialization, and the standard deviation is 0.01;
during each iteration, a batch of labeled training data is input into the network and the parameters are then updated.
Claims (5)
1. A road monitoring vehicle detection system comprises a monitoring camera, a control box and network equipment; the control box is positioned beside the road where the monitoring camera is positioned; the monitoring camera is connected with network equipment in the control box and transmits shot video outwards through the network equipment, and the monitoring camera is characterized by also comprising a detection module, wherein the detection module takes a computer as a core;
the network equipment comprises a wireless and/or wired communication module and a network switch; the network switch is also connected with the memory;
the monitoring camera and the detection module are connected with the network switch through a network cable; transmitting the video of the monitoring camera to a memory through a network switch; the detection module calls a required video stream from the memory and intercepts a single-frame picture;
the detection module comprises a positioning unit and an identification unit;
in the positioning unit:
firstly, enhancing an intercepted single-frame picture;
then, a vehicle detection model is adopted to obtain a vehicle detection area from the enhanced picture;
then, cutting the original single-frame picture according to the vehicle detection area to obtain a picture of a vehicle part, and then enhancing the picture of the vehicle part;
finally, detecting the enhanced picture of the vehicle part by using a license plate detection model to obtain the position information of the license plate;
the identification unit is provided with:
firstly, according to the position information of a license plate, a license plate picture is intercepted from a picture of a vehicle part;
then recognizing characters, letters and digital information in the license plate picture by using a license plate recognition model;
finally, outputting license plate characters according to the code of the license plate characters;
and the vehicle pictures and the corresponding license plate text information acquired by the detection equipment are stored in a memory for being called by a management terminal.
2. A road monitoring vehicle detection system as claimed in claim 1 wherein in the location unit: the vehicle detection model is a large-scale YOLOv5 network, and the license plate detection model is a small-scale YOLOv5 network; the identification unit is provided with: the license plate recognition model is a convolutional neural network.
3. The road monitoring vehicle detection system of claim 2, wherein during training of the vehicle detection model, the large-scale YOLOv5 vehicle detection network setup parameters are trained using the detac dataset; the loss function is the sum of the vehicle target confidence loss and the vehicle target localization loss.
4. The road monitoring vehicle detecting system of claim 2, wherein in the process of training the license plate detecting model, the CCPD2020 dataset is used for training the small-scale YOLOv5 license plate detecting network setting parameters: the loss function is the sum of the confidence loss of the license plate target and the positioning loss of the license plate target.
5. The road monitoring vehicle detection system of claim 1, wherein the license plate recognition model comprises:
firstly, for an input license plate image, using ResNet18 as a main feature extraction network to extract features;
then, the extracted features are respectively sent into 3 same classifiers;
each classifier is composed of a three-layer feedforward neural network;
the output dimension of the 1 st classifier is 38, and the 1 st character in the license plate is used for predicting the 1 st character which is the subordinate province of the license plate;
the output dimension of the 2 nd classifier is 25, and the 2 nd classifier is used for predicting the 2 nd character in the license plate and is a membership area of the license plate;
the output dimension of the 3 rd classifier is 35, which is used for predicting the rest characters in the license plate and is the personal code of the license plate;
the character code of the license plate is divided into provinces, regions and individuals; the province codes are called short for each province of the Chinese characters and called short for a special license plate; the regional codes are capital English letters; the personal codes are capital English letters, numbers and vacant lattice positions;
when the convolutional neural network is optimized by using a loss function, cross entropy is taken as the loss function, and the loss is the sum of losses of eight classifiers.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210780512.9A CN115394074A (en) | 2022-07-04 | 2022-07-04 | Road monitoring vehicle detection system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210780512.9A CN115394074A (en) | 2022-07-04 | 2022-07-04 | Road monitoring vehicle detection system |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115394074A true CN115394074A (en) | 2022-11-25 |
Family
ID=84116177
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210780512.9A Pending CN115394074A (en) | 2022-07-04 | 2022-07-04 | Road monitoring vehicle detection system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115394074A (en) |
Citations (17)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN201421655Y (en) * | 2009-05-26 | 2010-03-10 | 上海电信工程有限公司 | Vehicle monitoring system |
| CN101807346A (en) * | 2009-02-16 | 2010-08-18 | 汉王科技股份有限公司 | Automatic license plate identification system at urban checkpoint |
| CN102129775A (en) * | 2010-12-30 | 2011-07-20 | 上海安防电子有限公司 | Method and system for obtaining evidence by capturing vehicles at traffic crossing under panoramic video detection |
| CN103186982A (en) * | 2011-12-28 | 2013-07-03 | 南京理工大学常熟研究院有限公司 | Intelligent access system for public security |
| CN203722747U (en) * | 2014-02-19 | 2014-07-16 | 湖北科技学院 | Monitor signal acquisition device and monitor signal real-time alarm system |
| CN204856894U (en) * | 2015-07-30 | 2015-12-09 | 河南中天高新智能科技开发有限责任公司 | Bayonet socket vehicle monitoring system |
| CN107229929A (en) * | 2017-04-12 | 2017-10-03 | 西安电子科技大学 | A kind of license plate locating method based on R CNN |
| CN108388896A (en) * | 2018-02-09 | 2018-08-10 | 杭州雄迈集成电路技术有限公司 | A kind of licence plate recognition method based on dynamic time sequence convolutional neural networks |
| CN109271991A (en) * | 2018-09-06 | 2019-01-25 | 公安部交通管理科学研究所 | A kind of detection method of license plate based on deep learning |
| CN109543753A (en) * | 2018-11-23 | 2019-03-29 | 中山大学 | Licence plate recognition method based on adaptive fuzzy repair mechanism |
| CN109559520A (en) * | 2018-12-30 | 2019-04-02 | 广西数创智能科技有限公司 | A kind of intelligent traffic monitoring system |
| CN109657676A (en) * | 2018-12-06 | 2019-04-19 | 河池学院 | Licence plate recognition method and system based on convolutional neural networks |
| CN109697857A (en) * | 2019-01-18 | 2019-04-30 | 合肥米佑信息技术有限公司 | Intelligent traffic control system based on image recognition and neural network algorithm |
| CN110837807A (en) * | 2019-11-11 | 2020-02-25 | 内蒙古大学 | Identification method and system for fake-licensed vehicle |
| CN112232351A (en) * | 2020-11-09 | 2021-01-15 | 浙江工业职业技术学院 | License plate recognition system based on deep neural network |
| CN112966631A (en) * | 2021-03-19 | 2021-06-15 | 浪潮云信息技术股份公司 | License plate detection and identification system and method under unlimited security scene |
| CN114495022A (en) * | 2021-12-13 | 2022-05-13 | 江苏集萃未来城市应用技术研究所有限公司 | Vehicle license plate detection and positioning method based on multi-scale YOLOv5 nesting |
-
2022
- 2022-07-04 CN CN202210780512.9A patent/CN115394074A/en active Pending
Patent Citations (17)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101807346A (en) * | 2009-02-16 | 2010-08-18 | 汉王科技股份有限公司 | Automatic license plate identification system at urban checkpoint |
| CN201421655Y (en) * | 2009-05-26 | 2010-03-10 | 上海电信工程有限公司 | Vehicle monitoring system |
| CN102129775A (en) * | 2010-12-30 | 2011-07-20 | 上海安防电子有限公司 | Method and system for obtaining evidence by capturing vehicles at traffic crossing under panoramic video detection |
| CN103186982A (en) * | 2011-12-28 | 2013-07-03 | 南京理工大学常熟研究院有限公司 | Intelligent access system for public security |
| CN203722747U (en) * | 2014-02-19 | 2014-07-16 | 湖北科技学院 | Monitor signal acquisition device and monitor signal real-time alarm system |
| CN204856894U (en) * | 2015-07-30 | 2015-12-09 | 河南中天高新智能科技开发有限责任公司 | Bayonet socket vehicle monitoring system |
| CN107229929A (en) * | 2017-04-12 | 2017-10-03 | 西安电子科技大学 | A kind of license plate locating method based on R CNN |
| CN108388896A (en) * | 2018-02-09 | 2018-08-10 | 杭州雄迈集成电路技术有限公司 | A kind of licence plate recognition method based on dynamic time sequence convolutional neural networks |
| CN109271991A (en) * | 2018-09-06 | 2019-01-25 | 公安部交通管理科学研究所 | A kind of detection method of license plate based on deep learning |
| CN109543753A (en) * | 2018-11-23 | 2019-03-29 | 中山大学 | Licence plate recognition method based on adaptive fuzzy repair mechanism |
| CN109657676A (en) * | 2018-12-06 | 2019-04-19 | 河池学院 | Licence plate recognition method and system based on convolutional neural networks |
| CN109559520A (en) * | 2018-12-30 | 2019-04-02 | 广西数创智能科技有限公司 | A kind of intelligent traffic monitoring system |
| CN109697857A (en) * | 2019-01-18 | 2019-04-30 | 合肥米佑信息技术有限公司 | Intelligent traffic control system based on image recognition and neural network algorithm |
| CN110837807A (en) * | 2019-11-11 | 2020-02-25 | 内蒙古大学 | Identification method and system for fake-licensed vehicle |
| CN112232351A (en) * | 2020-11-09 | 2021-01-15 | 浙江工业职业技术学院 | License plate recognition system based on deep neural network |
| CN112966631A (en) * | 2021-03-19 | 2021-06-15 | 浪潮云信息技术股份公司 | License plate detection and identification system and method under unlimited security scene |
| CN114495022A (en) * | 2021-12-13 | 2022-05-13 | 江苏集萃未来城市应用技术研究所有限公司 | Vehicle license plate detection and positioning method based on multi-scale YOLOv5 nesting |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107194398B (en) | Vehicle damages recognition methods and the system at position | |
| CN107301385B (en) | Method and device for identifying shielded license plate | |
| KR101880004B1 (en) | Method and apparatus for identifying television channel information | |
| CN106529380A (en) | Image identification method and device | |
| EP3907660A1 (en) | Method, apparatus, electronic device, and storage medium for recognizing license plate | |
| CN111191663A (en) | License plate number recognition method and device, electronic equipment and storage medium | |
| CN106529446A (en) | Vehicle type identification method and system based on multi-block deep convolutional neural network | |
| CN112287912B (en) | Deep learning-based lane line detection method and device | |
| CN105512660A (en) | License number identification method and device | |
| CN111008632B (en) | License plate character segmentation method based on deep learning | |
| CN107292933B (en) | Vehicle color identification method based on BP neural network | |
| CN112580643A (en) | License plate recognition method and device based on deep learning and storage medium | |
| CN103413147A (en) | Vehicle license plate recognizing method and system | |
| CN109815948B (en) | Test paper segmentation algorithm under complex scene | |
| CN111062381B (en) | License plate position detection method based on deep learning | |
| CN101604381A (en) | License plate character recognition method based on multi-category support vector machines | |
| CN111353956B (en) | Image restoration method and device, computer equipment and storage medium | |
| CN113011338B (en) | Lane line detection method and system | |
| CN115394074A (en) | Road monitoring vehicle detection system | |
| CN112733851A (en) | License plate recognition method for optimizing grain warehouse truck based on convolutional neural network | |
| CN112036290A (en) | Complex scene character recognition method and system based on class mark coding representation | |
| CN114913515B (en) | End-to-end license plate recognition network construction method | |
| CN111985482B (en) | Multi-mode multi-scene automatic license plate recognition method based on deep learning | |
| Qu et al. | An improved character recognition algorithm for license plate based on BP neural network | |
| CN111613225A (en) | Method and system for automatically reporting road violation based on voice and image processing |
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 |