CN108831183A - Managing system of car parking based on machine vision - Google Patents
Managing system of car parking based on machine vision Download PDFInfo
- Publication number
- CN108831183A CN108831183A CN201810587414.7A CN201810587414A CN108831183A CN 108831183 A CN108831183 A CN 108831183A CN 201810587414 A CN201810587414 A CN 201810587414A CN 108831183 A CN108831183 A CN 108831183A
- Authority
- CN
- China
- Prior art keywords
- parking
- information
- user
- parking stall
- vehicle
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- G—PHYSICS
- G08—SIGNALLING
- G08G—TRAFFIC CONTROL SYSTEMS
- G08G1/00—Traffic control systems for road vehicles
- G08G1/14—Traffic control systems for road vehicles indicating individual free spaces in parking areas
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/50—Context or environment of the image
- G06V20/52—Surveillance or monitoring of activities, e.g. for recognising suspicious objects
- G06V20/54—Surveillance or monitoring of activities, e.g. for recognising suspicious objects of traffic, e.g. cars on the road, trains or boats
-
- 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
- G08G1/0175—Detecting movement of traffic to be counted or controlled identifying vehicles by photographing vehicles, e.g. when violating traffic rules
-
- H—ELECTRICITY
- H04—ELECTRIC COMMUNICATION TECHNIQUE
- H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
- H04N7/00—Television systems
- H04N7/18—Closed-circuit television [CCTV] systems, i.e. systems in which the video signal is not broadcast
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/60—Type of objects
- G06V20/62—Text, e.g. of license plates, overlay texts or captions on TV images
- G06V20/625—License plates
Landscapes
- Physics & Mathematics (AREA)
- Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Signal Processing (AREA)
- Traffic Control Systems (AREA)
Abstract
This patent discloses a kind of managing system of car parking based on machine vision, which is characterized in that the system comprises：User's intelligent terminal；Stop field assembly；Server；The server includes data acquisition facility, data sending device, data processing equipment；The information transmitted for obtaining processing from user's intelligent terminal, and send user's instruction information to user's intelligent terminal and the parking field assembly.By intelligence obtain vehicle, user, parking lot information, and by the intelligence computation of server and intelligent predicting, guidance user rationally utilizes parking stall, improves parking efficiency, alleviate traffic pressure.
Description
Technical field
This patent belongs to information technology field, in particular to a kind of managing system of car parking based on machine vision.
Background technique
According to statistics, by, national vehicle guaranteeding organic quantity reaches 3.04 hundred million in by the end of June, 2017, and is in persistently to rise situation, phase
The construction speed in China parking lot is not caught up with but always than for, and the ratio between parking stall and motor vehicle is continuing to decline, out
Phenomena such as existing parking difficulty.
In addition, the information dissemination in parking lot is also insufficient, leads to driver other than the absolute quantity on parking stall is insufficient
Parking lot is on the way found, so-called " unnecessary traffic " is caused.Also, after vehicle enters parking lot, guided in parking lot
Deficiency, driver is inefficient in the case where being unfamiliar with parking lot environment, having no idea to stop rapidly to appropriate position,
Furthermore most parking lot still uses the form of manual toll collection, further affects the efficiency of parking, causes traffic
Pressure.
Summary of the invention
This patent is based on the above situation of the prior art and proposes, the technical problems to be solved by the patent is to provide
A kind of managing system of car parking based on machine vision, to improve the efficiency of parking by intelligentized solution.
In order to solve the above-mentioned technical problem, the technical solution of this patent offer includes：
A kind of managing system of car parking based on machine vision, which is characterized in that the system comprises：User's intelligent terminal
User's intelligent terminal can be the smart phone, tablet computer or other hand-held intelligent equipment of user；It can also be
Other are mounted in smart machine onboard, such as vehicle-mounted flat computer or vehicle intelligent system etc.；User's intelligence
On the one hand terminal is used to provide the information of user's vehicle to server, on the other hand the medium as interaction, receives server hair
The information and instruction sent, and provide the user with and use operation interface；Parking field assembly described in parking field assembly includes that parking stall monitoring is taken the photograph
Camera, vehicle identification video camera, parking stall fluorescent marker, parking Field-Indicator Lights；On the one hand the parking field assembly is used to complete
On the other hand information of vehicles and the guidance of parking position are driven into the acquisition of parking position and information of vehicles for identification
Vehicle travels in parking lot；Parking stall monitors video camera, shoots the photo on parking stall, and the photo on parking stall is sent to service
Device；The vehicle in parking lot is driven into or is driven out in vehicle identification video camera, shooting, and sends photo to server；Parking stall fluorescence mark
Note is arranged on parking stall, is placed within the scope of parking stall measure video camera shooting, for marked vehicle it is occupied with
No optical states；Stop Field-Indicator Lights, including LED dot matrix graphic display screen, display screen should be located at main fork crossing, show
The content shown includes information of vehicles and direction of travel information；Server described in server includes data acquisition facility, and data are sent
Device, data processing equipment；The information transmitted for obtaining processing from user's intelligent terminal, and user is indicated that information passes
Give user's intelligent terminal and the parking field assembly.
This patent by intelligence obtain vehicle, user, parking lot information, and pass through the intelligence computation and intelligence of server
Prediction, guidance user rationally utilize parking stall, improve parking efficiency, alleviate traffic pressure.
Detailed description of the invention
Fig. 1 is this patent system hardware structure figure
Fig. 2 is the image of obtained sequential single character after completing License Plate and Character segmentation in this patent
Fig. 3 is to be weighted the image obtained after summation to the picture pixels after binaryzation in this patent.
Specific embodiment
Specific implementation of the patent mode is described in detail with reference to the accompanying drawing, it should be pointed out that the specific reality
Applying mode is only the citing to this patent optimal technical scheme, can not be interpreted as the limitation to the scope of this patent.
Embodiment one
This specific embodiment provides a kind of managing system of car parking based on machine vision, as shown in Figure 1, the system
Including：
User's intelligent terminal
User's intelligent terminal can be the smart phone, tablet computer or other hand-held intelligent equipment of user；
It can also be other mounted in smart machine onboard, such as vehicle-mounted flat computer or vehicle intelligent system etc..It is described
On the one hand user's intelligent terminal is used to provide the information of user's vehicle to server, on the other hand the medium as interaction, receives
The information and instruction that server is sent, and provide the user with and use operation interface.
In the present embodiment, the information of user's vehicle, such as the class of vehicle can be arranged in user by the intelligent terminal
Type, license plate number etc.；On the other hand the parking space information from server can be received by the intelligent terminal, vehicle movement is believed
Breath and payment information etc., and corresponding operation is selected according to the information.In this embodiment, described to operate all
It can concentrate on realizing in the small routine of wechat, can use the wide general feature of audient of wechat and in this way convenient for obtaining phase
Geographical location information answered etc. realizes corresponding function.User can check the feelings of parking stall and vehicle in wechat small routine
Condition calls Baidu map API (Application Programming Interface, application programming interface) completion to stop
Navigation in parking lot equally uploads license board information when vehicle is driven out to parking lot to MySQL database by calling wechat little Cheng
The function that self-clocking is deducted fees can be completed in the payment API of sequence.
User manually selects target parking lot, and reserve parking stall after registration is good, after clicking reservation, mobile phone to
The calculation server in corresponding parking lot send reservation information (E.T.A can be estimated by the algorithm built in Baidu map,
Accuracy rate is very high) and information of vehicles (license plate number).Server is when outwardly sending the information of vacant parking stall, vacant parking stall information
The value of reservation number is subtracted, while using the parking stall quantitative value that camera in parking lot acquires in order to avoid field caused by pernicious reservation
The interior wasting of resources, parking lot only receive reservation of the E.T.A in 10 minutes, i.e. only user area near parking lot
It can just be reserved when in domain.User reaches parking lot before subscription time, and reservation is completed and terminated；User is more than subscription time
Target parking lot is not reached within 2 minutes also, reservation is cancelled.
In addition, the intelligent terminal also receives the information from the server, the inquiry of parking stall predictive information, vehicle are realized
Field navigation, and other operations.
Stop field assembly
On the one hand the parking field assembly is used to complete the acquisition of parking position information of park, on the other hand drive into for identification
The information of vehicles of parking position.Usually parking lot can be divided into three kinds in the prior art, and one is the curb parkings of ground
, one is underground either parking garages, and there are also one is stereo garage class parking lot, three kinds of parking lots can pass through
The acquisition that parking field assembly is arranged to realize parking space information He drive into information of vehicles.
In this embodiment, in order to accurately obtain the information on parking stall and vehicle, the parking field assembly includes
Photographic device, such as video camera either camera etc. realize the acquisition of information of vehicles, shooting by the shooting of photographic device
Then photo comprising vehicle license plate number realizes the extraction of vehicle license plate also by following algorithm：
Step 1: being partitioned into target license plate region by edge detection
Our license plate areas of " interested " are split by carrying out region operation (edge detection) to photo, at this moment
The length-width ratio for detecting this region, according to the standard in China, the preceding license plate ruler of common full-sized car, kart etc. in city
Very little all includes 7 characters for 440*140mm domestic standard license plate, and initial character is Chinese character, is secondly the English character of capitalization, finally
5 characters are that number is mixed with letter.
The edge detection includes obtaining the photo of the license plate near zone shot by video camera first, this step can
To be completed by the position that video camera is arranged, directly acquire by video camera transmission Lai picture file.
Then image is pre-processed.The pretreatment includes source figure being carried out gray processing, in order to carry out to image
Smoothing denoising.The medfilt2 two dimension median filter function processing image of MATLAB is recycled to remove noise.In real life,
Vehicle is often because environmental factor causes vehicle body to be covered with dust all over, characters on license plate is not that especially clearly, image can have noise, in
Value filtering replaces the pixel value of the noise with neighborhood intermediate value, here we set noise size parameter be 3*3 pixel matrix.
This operation can bring edge blurry, but the heart, the edge blurry of formation will not influence in the picture due to extracting license plate position
As a result.
After above-mentioned image preprocessing, then noise is removed by the method for mathematical morphology and fills up hole, in matlab
Middle concrete operation step is as follows, forms the circle structural body that radius is 1 using strel function to original image first, utilizes
Imdilate and imerode function carries out burn into expansion to image.Subtract each other the result after corroding and expanding, then convolutionTo enhance edge.By imjust function to enhance contrast.Due to initial license plate photo usually exist it is inclined
Image flame detection can be facilitated segmentation by Hough transformation by situation, the straight line in Hough transformation detection image first, and can be with
It is calculated relative to horizontal tilt angle, then keeps target area horizontal using imrotate function rotation image.
Coarse positioning is carried out to license plate after aforesaid operations.Since the texture of vehicle body is mainly horizontal edge, and characters on license plate
Texture be mainly vertical edge.Realize that edge-perpendicular detection can be to license plate coarse positioning using Sobel algorithm according to this characteristic.
Sobel algorithm is a discrete type difference operator, and to the gray approximation of the bright function of arithmograph image height, calculating process is
In formula：3*3 matrix is warp factor；It * is convolution；A is original image, and Gx is the image there are horizontal edge, Gy
It is the image there are vertical edge, G is the gradient intensity figure of original image.
Hole is filled using imfill function after processing, 500 connection is greater than using bwareaopen function selection pixel
Region；The bondingbox attribute of regionprops function is recycled to acquire the minimum rectangle including face domain, by judging rectangle
The ratio of width to height determine whether for character, if 0.5 error of the ratio of width to height and target is less than 0.05 (adjustable threshold value)
It is considered character, it is marked, such as desires to make money or profit and these rectangular areas are split to imcorp function and are stored in one temporarily
In file.
Step 2: reacquiring " sense with color positioning mode if the number of characters that region operation is split is less than 7
Interest " region
According to method described in step 1, although most of license plates can be positioned successfully, there are a large amount of vertical edges in processing
It is difficult to that license plate is accurately positioned when the image of edge, therefore in this embodiment, when the number of characters that step 1 obtains is less than 7
It is a, then the operation of step 2 is carried out, target area, i.e. license plate area are reacquired using color positioning mode.
Since most of China's license plate is yellow bottom, blue bottom and green bottom, this is provides item by color come positioning licence plate
Part.Color positioning mode in the step 2 includes：
Picture is converted into hsv color model first.H, S, V respectively represent tone, saturation degree and brightness, and wherein H is only
A component related with color one by one, therefore the threshold value of selected S and V screens H, positioning licence plate position.
It is split using imcorp function, the circle that radius is 1 is formed using strel function to the image after segmentation
Structural body carries out burn into expansion to image using imdilate and imerode function.Subtract each other the result after corroding and expanding,
Convolution againTo enhance edge.By imjust function to enhance contrast.Change detection straight line by Hough, obtains it
With horizontal angle, using imrotate function by image rotation to level.By image binaryzation, utilized after binary conversion treatment
Imfill function fills hole, and 500 connection region is greater than using bwareaopen function selection pixel；It recycles
The bondingbox attribute of regionprops function acquires the minimum rectangle including face domain, by judge the ratio of width to height of rectangle come
Determine whether to be considered character when if 0.5 error of the ratio of width to height and target is less than 0.05 (adjustable threshold value) for character,
It is marked, such as desires to make money or profit and these rectangular areas is split to imcorp function and are stored in a temporary file underedge.
Step 3: the character picture that processing obtains
After completing License Plate and Character segmentation, obtain by be sequential single character image, and binaryzation
Treated picture is not both matrix of different sizes according to the size of resolution ratio, and matrix element contains only 0 and 1, and matrix is unfolded
At an one-dimension array, the length of the array is pixel number contained by picture, and the mode that the picture of all characters is unfolded is answered identical.
A given picture belongs to the evidence of some optional network specific digit class in order to obtain, carries out to the picture pixels after binaryzation
Weighted sum (weight is defined as follows, for the matrix after the binaryzation of single character, element value is 1 in matrix pair
The element in weight matrix is answered to be positive, the correspondence weight matrix element that element value is 0 is negative) existing character photo is by same
Processing after label, each character corresponds to specific label.
Step 4: training BP neural network, identifies specific characters on license plate
This function is realized by training BP neural network, neural network input layer neuron number is pixel contained by picture
It is worth number, output layer has the Chinese character used in 65 neurons, with license plate, letter and number to correspond, and imports the word that label is crossed
Vector set after according with picture expansion is modified the weight distribution in network by the characteristic of BP neural network error back propagation and is reached
To trained purpose.
The basic principle of BP learning algorithm is the most fast descent method of gradient, i.e., keeps network overall error minimum by adjusting weight,
In the signal propagated forward stage, input signal is handled through input layer is finally transmitted to output layer processing through intermediate layer handles again；Accidentally
The signal value that output layer exports is obtained error compared with desired output signal value, if error is larger by poor back-propagation phase
Error signal is passed back middle layer until input layer, modifies weight coefficient using error signal in each layer neuron, it is laggard
Enter next round iteration, so circulation is until error is minimum, and real output signal value is close to desired output signal value.In present networks accidentally
Difference evaluation uses cross entropy.
In actual use, the temporary file that dividing processing before has stored is input in trained network, network is
Export a length be 65 vector " one-hot vectors ", such as (1,0,0,0 ...) only one of them numerical value be 1 its
Remaining is 0, and 1 corresponding affiliated character is the character that the character picture imported most possibly belongs to.
The parking field assembly further includes parking space information acquisition device, and parking space information acquisition device is directed to different parking lots
Type and different parking modes, such as stereo garage, parking space management system is usually provided in stereo garage,
The parking space information of stereoscopic parking lot can be obtained by obtaining the parking space information in the parking space management system.
But in the prior art, for common roadside either underground parking space, due to not establishing corresponding number
According to acquisition equipment, this needs to carry out other data acquisition equipment setting.In this embodiment, the parking space information obtains
Take device that can realize by photographic device and parking stall mark.
The photographic device can be the camera that parking area is arranged in, and be determined by way of shooting photo described
The image of parking stall, the substantial evidence for being then based on machine judge to determine whether parking stall is occupied.In this embodiment, described
Camera can flexibly set one or more, but the image for the parking area for needing to judge firstly the need of intake.
Due to factors such as light in parking lot, by identifying that vehicle goes to judge whether parking stall is occupied to have any problem,
Therefore in this embodiment, fluorized marking is also set on parking stall other than camera is set, in each parking stall
Specific location on setting fluorescent paint draw mark.Parking area captured by the camera described in this way be exactly include glimmering
The region of cursor will, when fluorized marking display when, that is, indicate the parking stall be it is empty, the vehicle is then shown when fluorized marking is occupied
Position is occupied.
Judge whether the parking stall in the parking stall image is occupied in present embodiment by BP neural network, it should
Process is completed by server, i.e., sends the image to server, server after the camera obtains parking stall image
In equipped with parking space state identification model, the parking space state identification model includes BP neural network.
Concrete processing procedure is as follows, first without vehicle when identified using a certain frame in video record, with vehicle
The treatment process of board is similar, by image gray processing, carries out median filtering to image and removes noise, then to the parking stall area in image
Domain is divided；Because camera angle will not change easily if fixed, in this embodiment can be pre-
Fixed standard divides picture, and every photo after making segmentation corresponds to a parking stall and contains a reflecting sign.
Because of the figure there is no the demand of the mark in segmented image, after the completion of segmentation after direct binary conversion treatment segmentation
Picture, reinflated is one-dimension array, and length is pixel number contained by picture.To there is picture when vehicle to carry out same operation, counted
Group stores for future use.
Because the position of icon is relatively fixed, does not need many sample and circulation many times goes modification middle layer
Weight, for training BP neural network, the vector (photo when with and without vehicle) being unfolded above and given corresponding picture it is defeated
It is 1,0 out, imports in BP neural network, a picture is calculated equally by way of the weighted sum of pixel belongs to us and think
The evidence (just for weight, remaining region weight is negative for definition icon overlay area) for the result wanted, passes through and calculates cross entropy calculating
Loss.Neural metwork training is completed after the circulation of limited times, modification weight, i.e., inputs a picture, neural network again
Output 1 or 0 represents a car or car free.
The input layer number of BP neural network is first prime number of one-dimension array, and middle layer neuron number is according to mapping net
Network existential theorem (Kolmogorov theorem), it is assumed that input layer has n neuron, then middle layer has 2n+1 neuron, output
Layer has a neuron, and output 1 or 0 is represented whether there is or not special reflective marker in this picture, i.e., whether there is or not vehicles.
In use, taking the image at the interval of set time to be identified, such as 0.1s, by image according to scheduled program point
After cutting, the processing such as binaryzation is carried out, vector is launched into, imported into trained network, network output 0 then indicates parking stall not
Occupied, output 1 is then occupied for parking stall.Result after identification can be used to the matrix model in modification parking field model, under
Wen Zhonghui mentions this model.
In order to facilitate the guidance to parking person, unit further includes indication LED display screen in field in the parking lot.In view of cloth
The problems such as line, reliability and energy consumption, flowing water lamp-based is unreliable in the light to guide article LED of road both sides setting flowing water shape
, therefore, in this embodiment, using LED dot matrix graphic display screen fixed in parking lot, display screen should be located at mainly
Fork crossing, the content of display is broadly divided into two parts, information of vehicles and direction of travel.The information of vehicles and direction of travel
It can be obtained by the server, it in this way can be pre- to guide vehicle to drive into the parking stall of the vehicle allocation according to server
Fixed parking stall.The information of vehicles can be transmitted from the server, be scrolled sequentially through broadcasting, renewal frequency according to admission
Depending on admission vehicle number in fixed time period, and the time for reaching each display screen should be demarcated on the spot, lead is estimated, to reach
To preferable display effect, carry out display reminding according to the time of calibration, to have avoided driver from missing information or having wondered letter in advance
It ceases, previous should be also shown on display screen, and (it is aobvious to run to some position before can't see display screen in vehicle for estimation lead
Show) and latter vehicle instruction information, but because with the information that should currently show by apparent level difference.The direction of travel
It can be shown by arrow, meet the reading habit of driver.
Server
As shown in Figure 1, the server is connected with user terminal and parking field assembly in present embodiment,
The service is wherein equipped with database, for collecting the data of the user terminal and the parking field assembly.
In addition to the as disclosed hereinabove server is used to collect the receipt of user terminal, and provide and user terminal
Except interaction, the server can be also used for collecting the data of parking field assembly, judge to transmit the information come in parking lot, and
It is interacted with parking field assembly, this is equally described above.
In addition, in this embodiment, the server further includes parking stall predicting unit, for predicting presumptive area
In parking stall quantity.
In the predicting unit, predicted using neural network.In the prior art, with the time be it is unique influence because
The prediction algorithm of element is insecure, because the vacant parking stall number in parking lot is also in city development level, parking lot present position
To surrounding traffic situation, weather condition, whether be that festivals or holidays etc. are related.
Therefore, in this embodiment, parking lot present position, weather, whether be to be quantified as festivals or holidays influencing
Factor is considered as the input variable of neural network.By mapping network existential theorem (Kolmogorov theorem), any company is given
Continuous function can be realized accurately with a three layer feedforward neural networks.
In the present system, we predict t+ using the vacant parking stall number of t-50, t-40, t-30, t-20, t-10, t moment
The vacant parking stall number at 10 moment, design neural network input layer have nine neurons, respectively the vacant vehicle at above-mentioned 6 moment of correspondence
Whether digit and position mentioned above weather, are three influence factors of festivals or holidays, and middle layer takes 19 neurons, output layer
There is a neuron to correspond to the vacant parking stall number at t+10 moment.Transmission function selects Sigmoid function
(can define the letter in above-mentioned formula) output layer uses purely linear function
F (x)=x (can define the letter in above-mentioned formula)
For the parking lot for thering is related data to record before, neural network can be trained using historical data, and
It is fitted for every month, influences (such as tourist season bring parking need to meet the parking demand bring of season property
It asks)
Parking lot without related data record can wouldn't open reservation parking service, using this parking management system
It is open-minded again after at least four months (quarterly one month), and the database that training uses is constantly updated, and can thus be made vacant
Parking stall prediction " can grow with each passing hour ".
It is above only the preferred technical solution of this patent, it is all that this patent is carried out under invention design
Modification, replacement and delete and should all bring the protection scope of this patent into.
Claims (2)
1. a kind of managing system of car parking based on machine vision, which is characterized in that the system comprises：
User's intelligent terminal
User's intelligent terminal can be the smart phone, tablet computer or other hand-held intelligent equipment of user；May be used also
To be other mounted in smart machine onboard, such as vehicle-mounted flat computer or vehicle intelligent system etc.；The user
On the one hand intelligent terminal is used to provide the information of user's vehicle to server, on the other hand the medium as interaction, receives service
The information and instruction that device is sent, and provide the user with and use operation interface；
Stop field assembly
The parking field assembly includes parking stall monitoring video camera, vehicle identification video camera, parking stall fluorescent marker, parking lot instruction
Lamp；On the one hand the parking field assembly is used to complete the acquisition of parking position and information of vehicles, on the other hand for identification
The information of vehicles and guidance vehicle for driving into parking position travel in parking lot；
Parking stall monitors video camera, shoots the photo on parking stall, and the photo on parking stall is sent to server；
The vehicle in parking lot is driven into or is driven out in vehicle identification video camera, shooting, and sends photo to server；
Parking stall fluorescent marker is arranged on parking stall, is placed within the scope of the parking stall measure video camera shooting, is used for
Optical states whether marked vehicle is occupied；
Stop Field-Indicator Lights, including LED dot matrix graphic display screen, display screen should be located at main fork crossing, the content of display
Including information of vehicles and direction of travel information；
Server
The server includes data acquisition facility, data sending device, data processing equipment；For obtaining processing from the use
The information of family intelligent terminal transmission, and send user's instruction information to user's intelligent terminal and the parking lot group
Part.
2. a kind of managing system of car parking based on machine vision, which is characterized in that the system comprises：
User's intelligent terminal
User's intelligent terminal can be the smart phone, tablet computer or other hand-held intelligent equipment of user；May be used also
To be other mounted in smart machine onboard, such as vehicle-mounted flat computer or vehicle intelligent system etc.；The user
On the one hand intelligent terminal is used to provide the information of user's vehicle to server, on the other hand the medium as interaction, receives service
The information and instruction that device is sent, and provide the user with and use operation interface；User's vehicle is arranged by the intelligent terminal in user
Information, the type including vehicle, license plate number；On the other hand the parking stall from server can be received by the intelligent terminal
Information, vehicular movement information and payment information etc., and corresponding behaviour is selected according to the information；The operation can collect
In the case where being realized in the small routine of wechat, parking stall and vehicle are checked in wechat small routine, call Baidu map API complete
At the navigation in parking lot, license board information is equally uploaded when vehicle is driven out to parking lot to MySQL database by calling wechat
The function that self-clocking is deducted fees can be completed in the payment API of small routine；User manually selects target parking after registration is good
, and parking stall is reserved, after clicking reservation, mobile phone sends reservation information and vehicle letter to the calculation server in corresponding parking lot
Breath；For server when outwardly sending the information of vacant parking stall, vacant parking stall information uses the vehicle of camera acquisition in parking lot
Digit magnitude subtracts the value of reservation number, while in order to avoid the wasting of resources in field caused by pernicious reservation, parking lot only receives in advance
Reservation of the arrival time in 10 minutes is counted, i.e., can just be reserved when only user is in the near zone of parking lot；User is reserving
Parking lot is reached before time, reservation is completed and terminated；User is more than not reach target parking lot also in subscription time 2 minutes,
Reservation is cancelled；
Stop field assembly
The parking field assembly includes parking stall monitoring video camera, vehicle identification video camera, parking stall fluorescent marker, parking lot instruction
Lamp；On the one hand the parking field assembly is used to complete the acquisition of parking position and information of vehicles, on the other hand for identification
The information of vehicles and guidance vehicle for driving into parking position travel in parking lot；
Parking stall monitors video camera, shoots the photo on parking stall, and the photo on parking stall is sent to server；
The vehicle in parking lot is driven into or is driven out in vehicle identification video camera, shooting, and sends photo to server；
Parking stall fluorescent marker is arranged on parking stall, is placed within the scope of the parking stall measure video camera shooting, is used for
Optical states whether marked vehicle is occupied；
Stop Field-Indicator Lights, including LED dot matrix graphic display screen, display screen should be located at main fork crossing, the content of display
Including information of vehicles and direction of travel information；
Server
The server includes data acquisition facility, data sending device, data processing equipment；
The data acquisition facility obtains the information that user's intelligent terminal and the parking field assembly are sent；
The data processing equipment, for handling and analyzing the data of parking field assembly and user's intelligent terminal；The data
Processing unit includes：Car license recognition unit, parking stall recognition unit, parking space intelligent predicting unit；
The Car license recognition unit includes：
Target license plate edges of regions detection segmentation module, is split license plate area by carrying out region operation to photo, this
When detect the length-width ratio in this region；The edge detection includes obtaining the license plate near zone shot by video camera first
Photo, obtain by video camera transmission Lai picture file；Then image is pre-processed, the pretreatment includes, by source
Figure carries out gray processing, in order to carry out smoothing denoising to image, recycles the medfilt2 two dimension median filter function of MATLAB
It handles image and removes noise, replace the pixel value of the noise with neighborhood intermediate value using median filtering；After above-mentioned image preprocessing, then
Removed by the method for mathematical morphology and noise and fill up hole, in matlab concrete operations include：First to original graph
As forming the circle structural body that radius is 1 using strel function, burn into is carried out to image using imdilate and imerode function
Then expansion subtracts each other the result after corroding and expanding, then convolutionTo enhance edge, by imjust function to enhance
Contrast；Image flame detection can be facilitated, and count by segmentation, the straight line in Hough transformation elder generation detection image by Hough transformation again
It is calculated relative to horizontal tilt angle, then keeps target area horizontal using imrotate function rotation image；Then again
Coarse positioning is carried out to license plate, Sobel algorithm is used to realize edge-perpendicular detection to license plate coarse positioning；Imfill is utilized after processing
Function fills hole；Divide including being greater than 500 connection region using bwareaopen function selection pixel；It recycles
The bondingbox attribute of regionprops function acquires the minimum rectangle including face domain, by judge the ratio of width to height of rectangle come
Determine whether to be considered character when if 0.5 error of the ratio of width to height and target is less than threshold value, and mark it for character；
Color locating module, the color locating module, which is used to detect segmentation module segmentation in the target license plate edges of regions, to be gone out
When the number of characters come is less than 7, license plate area is reacquired with color positioning mode；Color locating module first converts picture
For hsv color model, then selectes the threshold value of S and V and H is screened, obtain positioning licence plate position；Then by license plate
It is described to be split in position with imcorp function, the circle structure that radius is 1 is formed using strel function to the image after segmentation
Body carries out burn into expansion to image using imdilate and imerode function；Subtract each other the result after corroding and expanding, then rolls up
ProductTo enhance edge；By imjust function to enhance contrast, by Hough change detection straight line, obtain image with
Horizontal angle recycles imrotate function by image rotation to level；Then by image binaryzation, after binary conversion treatment
Hole is filled using imfill function, 500 connection region is greater than using bwareaopen function selection pixel；It recycles
The bondingbox attribute of regionprops function acquires the minimum rectangle including face domain, by judge the ratio of width to height of rectangle come
Determine whether to mark it if the threshold error of the ratio of width to height and target is considered character less than 0.05 for character, utilize
These rectangular areas are split and are stored in a temporary file underedge by imcorp function；
Character picture obtains module, handles and obtain character picture, completes to have obtained sequence after License Plate and Character segmentation
Single character image, and the picture after binary conversion treatment according to the size of resolution ratio is not both matrix of different sizes；
Matrix element contains only 0 and 1 and matrix is launched into an one-dimension array, and the length of the array is pixel number contained by picture, all words
The mode that the picture of symbol is unfolded is answered identical；A given picture belongs to the evidence of some optional network specific digit class in order to obtain, to two
Picture pixels after value are weighted the existing character photo label after similarly handling of summation, and each character is corresponding special
Fixed label；
BP neural network Recognition of License Plate Characters module realizes Recognition of License Plate Characters function by training BP neural network, nerve
Network input layer neuron number is pixel value number contained by picture, and output layer has Chinese character, the word used in 65 neurons, with license plate
Female and number corresponds, the vector set after importing the character picture expansion that label is crossed, reversed by BP neural network error
Weight distribution in the characteristic modification network of propagation achievees the purpose that trained；
The parking space intelligent predicting unit uses in the predicting unit for predicting the parking stall quantity in presumptive area
Neural network predicted, parking lot present position, weather, whether be festivals or holidays to be quantified as influence factor, is considered as mind
Input variable through network；The t+10 moment is predicted using the vacant parking stall number of t-50, t-40, t-30, t-20, t-10, t moment
Vacant parking stall number, design neural network input layer has nine neurons, respectively the vacant parking stall number at above-mentioned 6 moment of correspondence and
Whether position mentioned above weather, is three influence factors of festivals or holidays, and middle layer takes 19 neurons, and output layer has one
Neuron corresponds to the vacant parking stall number at t+10 moment；Transmission function selects Sigmoid function in the neural network Output layer uses purely linear function f (x)=x；It, can be with for the parking lot for thering is related data to record before
Neural network is trained using historical data, and is fitted for every month, to meet the parking demand of season property
Bring influences；
Parking stall recognition unit analyzes the parking stall picture that the parking field assembly transmission comes, and according to the parking stall fluorescent marker
The reflected optical information that whether is blocked judges whether it is empty parking space；
The data sending device, for the user exported in server to be indicated that information or operational order are sent to the use
Family intelligent terminal and parking field assembly.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810587414.7A CN108831183A (en) | 2018-06-06 | 2018-06-06 | Managing system of car parking based on machine vision |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810587414.7A CN108831183A (en) | 2018-06-06 | 2018-06-06 | Managing system of car parking based on machine vision |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108831183A true CN108831183A (en) | 2018-11-16 |
Family
ID=64144502
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810587414.7A Pending CN108831183A (en) | 2018-06-06 | 2018-06-06 | Managing system of car parking based on machine vision |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108831183A (en) |
Cited By (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110458912A (en) * | 2019-08-08 | 2019-11-15 | 金瓜子科技发展（北京）有限公司 | A kind of vehicle icon treating method and apparatus |
| CN110780671A (en) * | 2019-10-30 | 2020-02-11 | 华南理工大学 | Storage navigation intelligent vehicle scheduling method based on global vision |
| CN111383459A (en) * | 2020-03-31 | 2020-07-07 | 北京智芯原动科技有限公司 | Charging pile parking space management method and device based on videos |
| CN111881758A (en) * | 2020-06-29 | 2020-11-03 | 普瑞达建设有限公司 | Parking management method and system |
| CN112435499A (en) * | 2020-11-13 | 2021-03-02 | 华录易云科技有限公司 | Method, device and equipment for identifying parking space unbalance condition |
| CN113194334A (en) * | 2021-04-16 | 2021-07-30 | 厦门智瞳科技有限公司 | Image processing method and device for protecting privacy, terminal and storage medium |
| CN113570866A (en) * | 2021-09-24 | 2021-10-29 | 成都宜泊信息科技有限公司 | Parking lot management method and system, storage medium and electronic equipment |
| CN115025419A (en) * | 2022-04-20 | 2022-09-09 | 上海阔步实业有限公司 | Active fire fighting system and fire fighting control method for new energy vehicle |
| CN115383453A (en) * | 2022-08-31 | 2022-11-25 | 苏州小男孩智能科技有限公司 | Automatic blockage-measuring and nut-screwing machine for evaporator of air conditioner |
| CN115862377A (en) * | 2023-02-08 | 2023-03-28 | 山东科技大学 | Intelligent parking reservation method and system |
| CN117079494A (en) * | 2023-10-16 | 2023-11-17 | 广州技客信息科技有限公司 | Underground garage intelligent management system based on blockchain |
Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101872548A (en) * | 2009-04-23 | 2010-10-27 | 黄柏霞 | Truck space guiding system and method based on images |
| CN102005133A (en) * | 2010-11-04 | 2011-04-06 | 任杰 | Recognizable mark-based parking position detecting method |
| KR101343102B1 (en) * | 2012-07-13 | 2013-12-20 | 한국과학기술원 | System and method for managing of intelligent parking lot |
| CN105047008A (en) * | 2015-06-24 | 2015-11-11 | 吴建国 | System and method for reserving parking stall in stereo parking lot and parking car |
| CN105303881A (en) * | 2015-11-10 | 2016-02-03 | 北京蓝卡科技股份有限公司 | High-definition video parking place guiding system |
| CN105913685A (en) * | 2016-06-25 | 2016-08-31 | 上海大学 | Video surveillance-based carport recognition and intelligent guide method |
| CN106297399A (en) * | 2016-10-12 | 2017-01-04 | 江苏阅衡智能科技有限公司 | A kind of parking lot vehicle management system and parking lot vehicles management method |
| CN106910366A (en) * | 2015-12-21 | 2017-06-30 | 深圳市彩生活网络服务有限公司 | A kind of Intelligent parking lot management system and its implementation based on mobile terminal |
| CN107958611A (en) * | 2017-12-25 | 2018-04-24 | 北京无线电计量测试研究所 | A kind of intelligent interconnection shutdown system based on image recognition indoor positioning |
| CN107967817A (en) * | 2017-11-17 | 2018-04-27 | 张慧 | Intelligent managing system for parking lot and method based on multi-path camera deep learning |
-
2018
- 2018-06-06 CN CN201810587414.7A patent/CN108831183A/en active Pending
Patent Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101872548A (en) * | 2009-04-23 | 2010-10-27 | 黄柏霞 | Truck space guiding system and method based on images |
| CN102005133A (en) * | 2010-11-04 | 2011-04-06 | 任杰 | Recognizable mark-based parking position detecting method |
| KR101343102B1 (en) * | 2012-07-13 | 2013-12-20 | 한국과학기술원 | System and method for managing of intelligent parking lot |
| CN105047008A (en) * | 2015-06-24 | 2015-11-11 | 吴建国 | System and method for reserving parking stall in stereo parking lot and parking car |
| CN105303881A (en) * | 2015-11-10 | 2016-02-03 | 北京蓝卡科技股份有限公司 | High-definition video parking place guiding system |
| CN106910366A (en) * | 2015-12-21 | 2017-06-30 | 深圳市彩生活网络服务有限公司 | A kind of Intelligent parking lot management system and its implementation based on mobile terminal |
| CN105913685A (en) * | 2016-06-25 | 2016-08-31 | 上海大学 | Video surveillance-based carport recognition and intelligent guide method |
| CN106297399A (en) * | 2016-10-12 | 2017-01-04 | 江苏阅衡智能科技有限公司 | A kind of parking lot vehicle management system and parking lot vehicles management method |
| CN107967817A (en) * | 2017-11-17 | 2018-04-27 | 张慧 | Intelligent managing system for parking lot and method based on multi-path camera deep learning |
| CN107958611A (en) * | 2017-12-25 | 2018-04-24 | 北京无线电计量测试研究所 | A kind of intelligent interconnection shutdown system based on image recognition indoor positioning |
Non-Patent Citations (3)
| Title |
|---|
| 张冠华: "一种基于神经网络的车牌识别方法的实现", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
| 王强: "智能停车管理***的设计与实现", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
| 高广银 等: "基于BP 神经网络的停车诱导泊位预测", 《计算机***应用》 * |
Cited By (14)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110458912A (en) * | 2019-08-08 | 2019-11-15 | 金瓜子科技发展（北京）有限公司 | A kind of vehicle icon treating method and apparatus |
| CN110780671A (en) * | 2019-10-30 | 2020-02-11 | 华南理工大学 | Storage navigation intelligent vehicle scheduling method based on global vision |
| CN111383459A (en) * | 2020-03-31 | 2020-07-07 | 北京智芯原动科技有限公司 | Charging pile parking space management method and device based on videos |
| CN111383459B (en) * | 2020-03-31 | 2021-04-02 | 北京智芯原动科技有限公司 | Charging pile parking space management method and device based on videos |
| CN111881758A (en) * | 2020-06-29 | 2020-11-03 | 普瑞达建设有限公司 | Parking management method and system |
| CN112435499B (en) * | 2020-11-13 | 2022-07-01 | 华录易云科技有限公司 | Method, device and equipment for identifying parking space unbalance condition |
| CN112435499A (en) * | 2020-11-13 | 2021-03-02 | 华录易云科技有限公司 | Method, device and equipment for identifying parking space unbalance condition |
| CN113194334A (en) * | 2021-04-16 | 2021-07-30 | 厦门智瞳科技有限公司 | Image processing method and device for protecting privacy, terminal and storage medium |
| CN113570866A (en) * | 2021-09-24 | 2021-10-29 | 成都宜泊信息科技有限公司 | Parking lot management method and system, storage medium and electronic equipment |
| CN115025419A (en) * | 2022-04-20 | 2022-09-09 | 上海阔步实业有限公司 | Active fire fighting system and fire fighting control method for new energy vehicle |
| CN115383453A (en) * | 2022-08-31 | 2022-11-25 | 苏州小男孩智能科技有限公司 | Automatic blockage-measuring and nut-screwing machine for evaporator of air conditioner |
| CN115862377A (en) * | 2023-02-08 | 2023-03-28 | 山东科技大学 | Intelligent parking reservation method and system |
| CN117079494A (en) * | 2023-10-16 | 2023-11-17 | 广州技客信息科技有限公司 | Underground garage intelligent management system based on blockchain |
| CN117079494B (en) * | 2023-10-16 | 2024-01-09 | 广州技客信息科技有限公司 | Underground garage intelligent management system based on blockchain |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN108831183A (en) | Managing system of car parking based on machine vision | |
| CN110517487B (en) | Urban area traffic resource regulation and control method and system based on thermodynamic diagram change identification | |
| CN109145900A (en) | A kind of licence plate recognition method based on deep learning | |
| Lei et al. | Automated pavement distress detection and deterioration analysis using street view map | |
| CN108509954A (en) | A kind of more car plate dynamic identifying methods of real-time traffic scene | |
| CN106934374B (en) | Method and system for identifying traffic signboard in haze scene | |
| CN105354530A (en) | Vehicle body color identification method and apparatus | |
| Ren et al. | YOLOv5s-M: A deep learning network model for road pavement damage detection from urban street-view imagery | |
| CN102254152A (en) | License plate location method based on color change points and color density | |
| CN110047319A (en) | Parking position air navigation aid, electronic device and storage medium | |
| CN113160575A (en) | Traffic violation detection method and system for non-motor vehicles and drivers | |
| CN114332776B (en) | Non-motor vehicle occupant pedestrian lane detection method, system, device and storage medium | |
| Enríquez et al. | Existing approaches to smart parking: An overview | |
| CN112651293A (en) | Video detection method for road illegal stall setting event | |
| CN112712707A (en) | Vehicle carbon emission monitoring system and method | |
| CN113903008A (en) | Ramp exit vehicle violation identification method based on deep learning and trajectory tracking | |
| Li et al. | RoadID: a dedicated deep convolutional neural network for multipavement distress detection | |
| Yaseen et al. | Development of new Anpr dataset for automatic number plate detection and recognition in north of Iraq | |
| CN116597270A (en) | Road damage target detection method based on attention mechanism integrated learning network | |
| CN115761701A (en) | Laser radar point cloud data enhancement method, device, equipment and storage medium | |
| CN115204686A (en) | Urban street design quality evaluation method based on user perception experience | |
| Baduge et al. | Assessment of crack severity of asphalt pavements using deep learning algorithms and geospatial system | |
| CN114972731A (en) | Traffic light detection and identification method and device, moving tool and storage medium | |
| CN115294774A (en) | Non-motor vehicle road illegal parking detection method and device based on deep learning | |
| Rani et al. | Traffic sign detection and recognition using deep learning-based approach with haze removal for autonomous vehicle navigation |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20181116 |
|
| WD01 | Invention patent application deemed withdrawn after publication |