CN103186982A - Intelligent access system for public security - Google Patents

Abstract

The invention discloses a public security intelligent bayonet system, which includes a camera and auxiliary lighting unit, a vehicle detection unit, an image acquisition and processing unit, a transmission unit and a central management unit. The camera and auxiliary lighting unit outputs a video signal, and the vehicle detection unit outputs a vehicle inspection unit. signal, the output vehicle inspection signal and video signal are input to the signal input end of the image acquisition and processing unit, the signal output end of the image acquisition and processing unit outputs the license plate information, and the output license plate information is input to the signal input end of the central management unit, the center The management unit stores the license plate information into the database. The invention can realize functions of vehicle detection, automatic identification of vehicle license plate and dispatched vehicle check, vehicle speed measurement and overspeed alarm, traffic flow statistics, data transmission and remote maintenance, data retrieval, fault tolerance and anti-theft alarm.

Description

A public security smart checkpoint system

technical field

The invention belongs to the field of intelligent transportation and electronic police, in particular to a security intelligent checkpoint system.

Background technique

At present, there are quite a few cases related to motor vehicles among various criminal crimes, such as traffic hit-and-run, theft of motor vehicles and so on. If an automatic monitoring point is set up on the important roads in and out of the urban area and the key sections of the expressway to realize the 24-hour uninterrupted and all-weather automatic recording of the passing vehicles, that is, the installation of the road vehicle intelligent monitoring and recording system (commonly known as "smart bayonet" ), it can provide important clues and basis for local public security departments to detect cases, and can effectively improve the management level of public security traffic police departments on motor vehicles. In addition, the system also has the functions of speed measurement and traffic flow monitoring, which can realize the monitoring and recording of illegal speeding vehicles. The "blacklist" investigation is also a main function of the system.

The public security smart checkpoint system is developed using advanced computer image processing, fuzzy recognition, remote data access and other technologies. The system records the front physical image (license plate type and number) and the overall picture of the vehicle (color, vehicle length, etc.) of each motor vehicle passing through the monitored road section continuously and in real time. Identify and dynamically deploy and control vehicles, capture illegally driving vehicles such as speeding and retrograde, and generate data files to save information such as the time, speed, and vehicle type of the vehicle passing through the checkpoint monitoring point, or through the broadband network. The information of each monitoring point is transmitted to the remote management center for organic sharing. The current bayonet system often has a single function, lacks diversification, and cannot meet the actual needs. The recognition method of the license plate is often based on a single image, which cannot avoid accidental errors. The recognition rate of the system is often not high. It has a great impact on the system, and the system cannot quickly adapt to the change of the environment.

Contents of the invention

The purpose of the present invention is to provide a smart bayonet system, which can realize vehicle detection, automatic identification of vehicle license plate and control vehicle collection, vehicle speed measurement and overspeed alarm, traffic flow statistics, data transmission and remote maintenance, data retrieval, fault Fault tolerance and anti-theft alarm function.

The technical solution to realize the object of the present invention is: a public security smart checkpoint system, comprising:

Camera and auxiliary lighting unit: including high-definition cameras and fill lights, used to capture vehicle and license plate information around the clock, at least one close-up camera for each lane, and one panoramic camera for each driving direction;

Vehicle detection unit: including the detection host and the ground induction coil, which is used to complete the detection of the vehicle, the measurement of the vehicle speed, and the counting and statistics of the traffic flow; when the vehicle passes through the ground induction coil, the ground induction coil generates an induction signal and sends it To the detection host, the detection host measures the vehicle speed and counts the traffic flow according to these induction signals;

Image acquisition and processing unit: including industrial computer, video acquisition card and control module, the video acquisition card and control module are installed in the industrial control computer, the video acquisition card receives the video stream transmitted by the camera, and the control module includes the image acquisition and processing sub-module And the license plate automatic recognition sub-module, the video acquisition card converts the analog video image signal of the camera into a digital image, when the detection host detects that a vehicle passes by, the image acquisition sub-module intercepts a close-up in the video stream of the close-up camera At the same time, a panoramic picture is intercepted in the video stream of the panoramic camera, and the license plate information recognition of the close-up picture is completed by the license plate recognition sub-module;

Transmission unit: including remote network communication module, transmission equipment optical transceiver and transmission line optical fiber; used to transmit the video signal collected by the camera and auxiliary lighting unit and the vehicle inspection signal detected by the vehicle detection unit to the industrial control computer;

Central management unit: including database server, network switch and query workstation. The query workstation and database server are directly connected to the network switch. The central management unit adopts C/S and B/S structure, which is used to complete query information, release control, and alarm functions; database The server receives and stores the license plate information identified by the image acquisition and processing unit, and the close-up pictures and panoramic pictures intercepted by the image acquisition sub-module are stored in the database server;

The camera and auxiliary lighting unit outputs video signals, the vehicle detection unit outputs vehicle inspection signals, the output vehicle inspection signals and video signals are input to the signal input end of the image acquisition and processing unit, and the signal output end of the image acquisition and processing unit outputs license plate information, The output license plate information is input to the signal input terminal of the central management unit, and the central management unit stores the license plate information into the database.

Compared with the prior art, the present invention has significant advantages: the present invention adopts a license plate recognition algorithm based on intelligent decision-making and online learning, which not only improves the recognition rate of the system and the correctness and robustness of the recognition results, but also improves the rapidity of the system. Ability to adapt to new application environments. The invention adopts the remote transmission technology, adopts the watchdog and the integrated main board in the industrial control computer, which not only improves the reliability and maintainability of the system. The invention not only improves the accuracy of detection, but also can detect vehicle speed and reverse running by embedding two ground induction coils in each lane. The invention is provided with a central management unit, and internal users in the public security network can query information, issue real-time control information and alarm. The invention integrates multiple functions into one body and meets the diversified demands.

Description of drawings

Figure 1 is a topological structure diagram of the public security smart checkpoint system.

Figure 2 Front-end topology of the smart bayonet system

Figure 3 is a flow chart of the front-end work of the system.

Figure 4 is the frame diagram of the central management unit of the system.

Figure 5. Schematic diagram of the front-end installation of the bayonet system (same-direction dual lanes).

Fig. 6 is a system connection block diagram of the present invention.

Detailed ways

A public security smart checkpoint system, comprising:

Camera and auxiliary lighting unit: including high-definition cameras and fill lights, used to capture vehicle and license plate information around the clock, each lane is equipped with at least one close-up camera, and in order to clearly see the vehicle model, color, outline, loading status and road conditions , each driving direction is equipped with a panoramic camera, and CDM floodlights are used to supplement the light to reduce the driver's discomfort;

Vehicle detection unit: including the detection host and the ground sense coil, which is used to complete the vehicle detection, the measurement of the vehicle speed, and the counting and statistics of the traffic flow; the vehicle detection method adopts the coil detector, and two ground sense coils are buried in each lane , used to detect vehicles and vehicle speed, and can send alarm signals to overspeed vehicles and reverse driving vehicles; the detection host uses dial switches to set the detection sensitivity, and the system will never lose parameters when the system is powered off, with high reliability. When the vehicle passes through the ground induction coil, the ground induction coil generates induction signals and sends them to the detection host, and the detection host measures the vehicle speed and counts the traffic flow according to these induction signals; the vehicle detection unit embeds two ground induction coils in each lane, It is used to detect vehicles and vehicle speed, and can send alarm signals to overspeed vehicles and vehicles traveling in the opposite direction; the detection host uses a dial switch to set the detection sensitivity. :

Image acquisition and processing unit: use industrial computer, video capture card and corresponding control modules to complete image acquisition and processing, including license plate positioning, license plate segmentation, license plate recognition, and license plate recognition result decision-making. Among them, the license plate location adopts the license plate location algorithm based on feature saliency and fusion of multiple features to complete the license plate location, which is suitable for various complex background environments and different camera angles. The historical records (including recognition results, recognition credibility, track records, similarity records, etc.) under the system can make intelligent decisions on the recognition results; the video capture card and control module are installed in the industrial computer, and the video capture card is received by For the video stream transmitted by the camera, the control module includes an image acquisition and processing sub-module and an automatic license plate recognition sub-module. The video acquisition card converts the analog video image signal of the camera into a digital image. The submodule intercepts a close-up picture in the video stream of the close-up camera at this time, and intercepts a panoramic picture in the video stream of the panoramic camera at the same time, and the close-up picture completes the license plate information identification through the license plate recognition submodule;

Transmission unit: including remote network communication module, transmission equipment optical transceiver and transmission line optical fiber; used to transmit the video signal collected by the camera and auxiliary lighting unit and the vehicle inspection signal detected by the vehicle detection unit to the industrial control computer;

Central management unit: including database server, network switch and query workstation. The query workstation and database server are directly connected to the network switch. The central management unit adopts C/S and B/S structure, which is used to complete query information, release control, and alarm functions; database The server receives and stores the license plate information identified by the image acquisition and processing unit, and the close-up pictures and panoramic pictures captured by the image acquisition sub-module are stored in the database server; each user terminal has corresponding permissions, and is an internal user in the public security network. Implement hierarchical management.

The camera and auxiliary lighting unit outputs video signals, the vehicle detection unit outputs vehicle inspection signals, the output vehicle inspection signals and video signals are input to the signal input end of the image acquisition and processing unit, and the signal output end of the image acquisition and processing unit outputs license plate information, The output license plate information is input to the signal input terminal of the central management unit, and the central management unit stores the license plate information into the database, and provides functions such as query to the outside world, mainly completing query, control, and alarm functions.

The public security intelligent bayonet system of the present invention, the license plate information recognition method adopted by the image acquisition and processing unit, the specific steps are as follows:

The first step is to analyze the video stream, track the vehicle in the image, and capture a close-up picture;

The second step is to use a license plate location algorithm based on feature saliency and fusion of multiple features to complete the license plate location;

The third step is to use the character segmentation algorithm based on the connected domain to complete the segmentation of the license plate number characters. The specific steps are as follows:

3.1 Estimation of the character height of the license plate number;

1) Convert the license plate into black characters on a white background;

2) License plate vertical edge detection, first use the Sobel operator to complete the edge detection of the license plate, and then use the OTSU method to obtain the vertical mapping of the license plate, replace the edge pixels with black, and replace other pixels with white to find the vertical edge of the license plate;

3) Horizontal projection to find the upper boundary and lower boundary of the license plate number characters;

3.2 Estimation of the character width of the license plate number;

1) Use the OTSU algorithm to find the threshold and complete the binarization of the license plate image;

2) vertical projection, for blocks with a width greater than 6 pixels and smaller than the widest block, use the average of these blocks as an estimate of the character width;

3.3 Extracting and distinguishing character blocks through connected domains, including connected domain detection and connected domain inspection;

1) Connected domain detection algorithm;

First, a label count is created and initialized to 1, then the binary image is scanned from left to right and top to bottom;

1. For each pixel, check 8 - whether the northeast, north, northwest and west pixels of the connected domain are foreground pixels;

2. If there is no matching neighbor, the value of the area counter is assigned to the area, and the area counter is incremented by 1;

3. If there is only one matching neighbor, assign the pixel to the area, if multiple neighbors are in accordance and all belong to the same area, assign the pixel value to the area;

4. If multiple neighbors match and are in different regions, assign the pixel to any one of these regions, and record all these regions as equivalent regions;

5. Scan the image and assign the same area value to all equivalent areas;

6. Scanning stops until there are no unmarked pixels in the image;

2) Connected domain inspection, the rules and standards are as follows:

Rule 1: A connected domain is character-free if its upper boundary is smaller than the overall upper boundary, or if its upper boundary is greater than the overall lower boundary;

Rule 2: If the aspect ratio of a connected domain is greater than 0.8 or less than 0.3, the connected domain is assumed to contain no characters;

Rule 3: If the area of the connected domain is greater than one-sixth of the size of the overall image, further segmentation is required;

Rule 4: If the number of successfully detected connected domains exceeds 7 or falls below 3, a recheck is required;

Connected domains that do not meet the above criteria are considered to be character-free and are therefore deleted; the process is complete when all connected domains of a given license plate have been checked;

The fourth step is to use ORC to complete the identification of the license plate number;

The fifth step is to perform the above-mentioned real-time recognition on each frame of video image, and make a decision on the recognition result, that is, by calculating the number of observed frames, the stability of the recognition result, the trajectory stability, the speed stability, the average reliability and the similarity measurement value to obtain Comprehensive credibility evaluation of the license plate to determine whether to continue tracking the license plate, output a recognition result, or reject the result.

In the public security intelligent bayonet system of the present invention, the industrial control computer has a built-in hardware watchdog and adopts an integrated main board. (The system can be automatically restarted when it crashes, and the industrial computer adopts the current high computing speed and integrated motherboard, which makes the system process faster and reduces the adverse effects caused by plugging and unplugging multiple boards).

The present invention will be described in further detail below in conjunction with the accompanying drawings.

As shown in Figure 1, Figure 2, Figure 3 and Figure 4, the public security smart checkpoint system includes: vehicle detection unit, front-end camera and auxiliary lighting unit, image acquisition and processing unit, transmission unit and central management unit. The camera in the front-end system of this system has been in the real-time shooting state (25 frames per second), and the image is remotely transmitted to the video acquisition card in the industrial computer, and the video acquisition card converts the analog video image signal of the camera into a digital image, when the detection host When a vehicle is detected to pass by, the image acquisition sub-module intercepts a picture (close-up of the license plate) in the video stream of the close-up camera at this time, and intercepts a panoramic picture in the video stream of the panoramic camera at the same time, and the close-up picture is completed by the license plate recognition module. The control module records the time, location, vehicle speed, driving direction, body length and vehicle image of the vehicle passing the monitoring point. When the measured vehicle speed exceeds the speed limit value set by the system, it will immediately call the police on the spot and notify the on-duty personnel to pay attention , the above license plate number, vehicle image, speed measurement value and other images and data information can be uploaded to the central management unit through the transmission network after processing, and stored in the database server; the background system of the central management unit can accept the front-end alarm information and forward it to the duty workstation. Publish the blacklist, etc., and the background system processes the uploaded images and data information, and can perform queries (control query, illegal query, past query), statistics (traffic flow statistics, illegal statistics, vehicle statistics), etc., when it is necessary to deploy and control vehicles , can input deployment and control information, and transmit it to the front end of the system at the same time.

Front camera and auxiliary lighting unit: composed of panoramic camera, close-up camera and fill light. As shown in Figure 5, the installation of the front end of the bayonet system on the road with two lanes in the same direction requires two close-up cameras (one for each lane) and one panoramic camera, which are installed on the same "L" pole, and the fill light Mounts on another "L" rod.

Vehicle detection unit: the ground sense coil detects whether there is a vehicle passing by, sends the signal to the detection host, detects the time when the host vehicle passes the front and rear coils, judges whether the vehicle is driving in the opposite direction, and uses the time before and after the front and rear coils and the distance between the coils to calculate Vehicle speed: The detection host uses the time when the vehicle passes through a single coil, multiplied by the current vehicle speed to obtain the vehicle length information, and sends the result to the image acquisition and processing unit. This system adopts the model: JL810ZJ vehicle detector, built-in 6-way ground sense detection circuit, can connect 6 ground sense coils, the detection period of the ground sense coil is less than 1ms, and the coil sensitivity is 8 levels.

Image acquisition and processing unit: Image acquisition and processing are completed by an industrial control computer. The configuration of the industrial control computer is: CPU P4 2.8 or above, memory ≥ 512M, and a built-in video capture card. The system calculates that each vehicle stores at least one characteristic image and one panoramic image. The two pictures occupy about 80K of disk space, and the images of 1.5 million vehicles occupy about 144G of disk space, depending on the traffic flow of the monitoring point. , under normal circumstances, at least half a year of vehicle information can be stored. When the hard disk space is insufficient, it will automatically overwrite the vehicle information and pictures with the earliest recording time. The control module includes an image acquisition and processing submodule, a license plate automatic recognition submodule and a database. The license plate automatic recognition sub-module includes: license plate positioning, license plate segmentation, license plate recognition, and license plate recognition result decision-making.

License plate positioning module: This module is the basis of subsequent links, and its accuracy has a huge impact on the overall system performance. This system uses a license plate location algorithm based on feature saliency and fusion of multiple features to complete the license plate location, which is suitable for various complex background environments and different camera angles.

License plate segmentation module: using the character segmentation algorithm based on the connected domain, on the basis of the basic Unicom domain search algorithm, the detection of the Unicom domain is added, and 4 inspection standards are given.

License plate recognition module: This system uses ORC to complete the license plate recognition.

License plate recognition result decision-making module: This system can recognize each frame of video image in real time, so when a car passes through the field of view, this system will get several same or different recognition results. This requires a decision-making module for recognition results. Specifically, the decision-making module utilizes the historical records (including recognition results, recognition credibility, track records, similarity records, etc.) As a result, intelligent decision-making is carried out, and the comprehensive credibility evaluation of the license plate is obtained by calculating the number of observation frames, recognition result stability, trajectory stability, speed stability, average credibility and similarity, so as to decide whether to continue Track the license plate, output a recognition result, or reject the result. The final recognition result of a license plate is synthesized by analyzing the recognition results of all frames, intelligently classifying and voting them, and combining certain grammatical information. This method comprehensively utilizes the information of all frames, reduces the accidental errors caused by previous recognition methods based on a single image, and greatly improves the recognition rate of the system and the correctness and robustness of the recognition results.

Transmission unit: Each monitoring point is designed to install a pair of optical fiber transceivers (10/100M self-adaptive) or multiple video multiplexing optical transceivers.

Central management unit: The central management unit of this system mainly includes a database server, a network switch, and a query workstation. Each user terminal has corresponding authority, all are internal users in the public security network, implement hierarchical management, the whole system adopts C/S and B/S structure, and mainly completes query, deployment control, and alarm functions.

Claims (5)

1. A kind of public security smart card port system is characterized in that comprising:

   Shooting and floor light unit: comprise high-definition camera and light filling lamp, be used for round-the-clock vehicle and the license board information of catching, every track disposes a feature video camera at least, panoramic camera of each travel direction configuration;

   Vehicle detection unit: comprise and detect main frame and ground induction coil, be used for finishing the mensuration of vehicle detection, the speed of a motor vehicle and numeration and the statistics of vehicle flowrate; When vehicle passed through ground induction coil, ground induction coil produced induced signal and also sends it to the detection main frame, detected main frame and measured the speed of a motor vehicle and statistical vehicle flowrate according to these induced signals;

   Image acquisition and processing unit: comprise industrial computer, video frequency collection card and control module, video frequency collection card, control module is installed in the industrial computer, video frequency collection card receives by video camera and passes the video flowing of coming, control module comprises image acquisition and processing submodule and the automatic recognin module of car plate, convert the analog video image signals of video camera to digital picture by video frequency collection card, when detecting main frame and detected vehicle and pass through, the image acquisition submodule this moment of intercepting one width of cloth feature in the video flowing of feature video camera, intercept a Zhang Quanjing picture simultaneously in the video flowing of panoramic camera, feature is finished license board information identification through car plate recognin module;

   Transmission unit: comprise remote network communication module, transmission equipment fiber optic and transmission line optical fiber; Be used for make a video recording and the vision signal of floor light unit collection and the car test signal of vehicle detection unit detection are transferred to industrial computer;

   Central manage-ment unit: comprise database server, the network switch and inquiry workstation, inquiry workstation and database server direct interconnection network switch, central manage-ment unit adopts C/S and B/S structure, is used for finishing Query Information, deploys to ensure effective monitoring and control of illegal activities, warning function; The license board information that database server reception and memory image collection and processing unit identify, feature and the panoramic pictures of the intercepting of image acquisition submodule are stored in the database server;

   Shooting and floor light unit outputting video signal, vehicle detection unit output car test signal, the signal input part that the car test signal of output and vision signal are input to image acquisition and processing unit, the signal output part output license board information of image acquisition and processing unit, the license board information of output is input to the signal input part of central manage-ment unit, and central manage-ment unit is deposited into license board information in the database.
2. Public security smart card port system according to claim 1 is characterized in that: described shooting and floor light unit adopt ceramic golden halogen CDM floodlight light filling.
3. Public security smart card port system according to claim 1 is characterized in that: two ground induction coils are buried in each track, described vehicle detection unit underground, detect main frame and adopt toggle switch that detection sensitivity is set.
4. :

   Public security smart card port system according to claim 1 is characterized in that the collection of described image and the license board information recognition methods that processing unit adopts, and concrete steps are as follows:

   The first step is analyzed video flowing, the vehicle in the image is followed the tracks of the intercepting feature;

   In second step, employing is finished the car plate location based on the algorithm of locating license plate of vehicle of the various features fusion of characteristic remarkable;

   The 3rd step, adopt the cutting of finishing number-plate number character based on the character cutting algorithm of connected domain, concrete steps are as follows:

   3.1 number-plate number character height is estimated;

   1) car plate is converted into white gravoply, with black engraved characters;

   2) the car plate vertical edge detects, and at first uses the Sobel operator to finish rim detection to car plate, uses the OTSU method to obtain the vertical mapping of car plate then, and edge pixel replaces with black, and other pixels replace with white, find the vertical edge of car plate;

   3) horizontal projection finds coboundary and the lower boundary of number-plate number character;

   3.2 number-plate number character duration is estimated;

   1) use the OTSU algorithm to find out threshold value, finish the license plate image binaryzation;

   2) vertical projection greater than 6 pixels and less than the piece of the wideest piece, uses the mean value of these pieces as the estimated value of character duration for width;

   3.3 extract the differentiation character block by connected domain, comprise connected domain detection and connected domain and check two parts;

   1) connected domain detection algorithm;

   At first, a label counting is created and is initialized as 1, and then, binary picture is by from left to right, scanning from top to bottom;

   1. for each pixel, check whether northeast, north, northwest and the western pixel of 8 – connected domains is foreground pixels;

   2. if the neighbours that do not meet then compose the value of region-counter to this zone, this region-counter adds 1;

   3. if having only neighbours that meet, pixel is distributed to this zone, if a plurality of neighbours meet and all be the same area, pixel value is distributed to this zone;

   4. if a plurality of neighbours coupling and in zones of different is assigned in these zones any one to pixel, records all these zones and be equivalent region;

   5. scan image, and give identical regional value with the zone of all equivalences;

   Scanning stops, the pixel that does not indicate in image;

   2) connected domain inspection, rule criterion is as follows:

   The rule 1: if the coboundary of a connected domain less than overall coboundary, if or its coboundary greater than overall lower boundary, then this connected domain is not contain character;

   Rule 2: if the depth-width ratio of a connected domain greater than 0.8 or less than 0.3, then this connected domain is assumed that and does not contain character;

   Rule 3: if the area of this connected domain greater than the sixth of the size of general image, then needs further to cut apart;

   Rule 4: if the detected connected domain quantity of success surpasses 7 or be lower than 3, then need to reexamine;

   The connected domain that does not satisfy above-mentioned standard is regarded as not containing character, and is therefore deleted; After all connected domains of given car plate were examined, this process was finished;

   In the 4th step, adopt ORC to finish the identification of the number-plate number;

   The 5th step, every frame video image is carried out above-mentioned Real time identification, recognition result is made a strategic decision, namely obtain the comprehensive confidence level evaluation of this car plate by calculating observation frame number, recognition result stability, trajectory stability, speed stability, average confidence level and measuring similarity value, thereby determine it is to continue to follow the tracks of this car plate, still export recognition result, or refuse this result.
5. Public security smart card port system according to claim 1 is characterized in that described industrial computer built in hardware house dog, and adopts integrated mainboard.

Images