CN107665603A - A kind of real-time detection method for judging parking stall and taking - Google Patents
A kind of real-time detection method for judging parking stall and taking Download PDFInfo
- Publication number
- CN107665603A CN107665603A CN201710795417.5A CN201710795417A CN107665603A CN 107665603 A CN107665603 A CN 107665603A CN 201710795417 A CN201710795417 A CN 201710795417A CN 107665603 A CN107665603 A CN 107665603A
- Authority
- CN
- China
- Prior art keywords
- parking stall
- vehicle
- carnet
- layer
- depth map
- 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.)
- Granted
Links
- 238000000034 method Methods 0.000 title claims abstract description 25
- 238000011897 real-time detection Methods 0.000 title claims abstract description 21
- 238000001514 detection method Methods 0.000 claims abstract description 33
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 13
- 230000011218 segmentation Effects 0.000 claims abstract description 12
- 238000012544 monitoring process Methods 0.000 claims abstract description 9
- 238000013461 design Methods 0.000 claims abstract description 7
- 239000007787 solid Substances 0.000 claims abstract description 7
- 230000004927 fusion Effects 0.000 claims abstract description 5
- 238000003475 lamination Methods 0.000 claims description 8
- 230000000877 morphologic effect Effects 0.000 claims description 6
- 238000012360 testing method Methods 0.000 claims description 4
- 230000004913 activation Effects 0.000 claims description 3
- 230000010339 dilation Effects 0.000 claims description 3
- 230000003628 erosive effect Effects 0.000 claims description 3
- 239000000284 extract Substances 0.000 claims description 3
- 238000000605 extraction Methods 0.000 claims description 3
- 238000003709 image segmentation Methods 0.000 claims description 3
- 238000007689 inspection Methods 0.000 claims description 3
- 238000012216 screening Methods 0.000 claims description 3
- 230000000007 visual effect Effects 0.000 claims description 3
- 238000010606 normalization Methods 0.000 claims description 2
- 238000011161 development Methods 0.000 abstract description 4
- 230000018109 developmental process Effects 0.000 abstract description 4
- 230000008901 benefit Effects 0.000 abstract description 3
- 238000005516 engineering process Methods 0.000 description 9
- 230000008859 change Effects 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 230000002708 enhancing effect Effects 0.000 description 2
- 238000012545 processing Methods 0.000 description 2
- 230000009466 transformation Effects 0.000 description 2
- 230000009471 action Effects 0.000 description 1
- 238000004458 analytical method Methods 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000003384 imaging method Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000006698 induction Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000001537 neural effect Effects 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 238000013519 translation Methods 0.000 description 1
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
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T5/00—Image enhancement or restoration
- G06T5/20—Image enhancement or restoration using local operators
- G06T5/30—Erosion or dilatation, e.g. thinning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/50—Depth or shape recovery
- G06T7/55—Depth or shape recovery from multiple images
- G06T7/593—Depth or shape recovery from multiple images from stereo images
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
- G06V10/30—Noise filtering
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/50—Context or environment of the image
- G06V20/56—Context or environment of the image exterior to a vehicle by using sensors mounted on the vehicle
- G06V20/58—Recognition of moving objects or obstacles, e.g. vehicles or pedestrians; Recognition of traffic objects, e.g. traffic signs, traffic lights or roads
- G06V20/586—Recognition of moving objects or obstacles, e.g. vehicles or pedestrians; Recognition of traffic objects, e.g. traffic signs, traffic lights or roads of parking space
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/10—Image acquisition modality
- G06T2207/10004—Still image; Photographic image
- G06T2207/10012—Stereo images
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/30—Subject of image; Context of image processing
- G06T2207/30248—Vehicle exterior or interior
- G06T2207/30252—Vehicle exterior; Vicinity of vehicle
- G06T2207/30264—Parking
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- General Health & Medical Sciences (AREA)
- Computing Systems (AREA)
- Computational Linguistics (AREA)
- Data Mining & Analysis (AREA)
- Evolutionary Computation (AREA)
- Biomedical Technology (AREA)
- Molecular Biology (AREA)
- Biophysics (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Life Sciences & Earth Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Health & Medical Sciences (AREA)
- Traffic Control Systems (AREA)
Abstract
The invention discloses a kind of real-time detection method for judging parking stall and taking, belong to technical field of machine vision, and in particular to a kind of depth map by vertical view merges the detection methods of convolutional neural networks to judge real-time detection method that whether parking stall is correctly taken.Its specifically include build binocular solid camera, collect vehicle sample and non-vehicle sample, design and train vehicle detection model based on convolutional neural networks CarNet, artificial demarcation monitoring parking stall in camera image position, obtain depth map under monitoring scene, the steps such as vehicle region detects vehicle using CarNet detection models simultaneously, fusion depth map and convolutional neural networks result obtain reliable vehicle detection block and whether judgement parking stall takes tentatively obtained to depth map segmentation.The semi-outdoors such as gas station, underground parking or parking garage are the method can be widely used in, can accurately export parking stall occupancy situation, there is the advantages that implementation cost is low, and operating efficiency is high, automatic capability is strong, there is vast potential for future development.
Description
Technical field
The invention belongs to technical field of machine vision, and in particular to a kind of real-time detection method for judging parking stall and taking.
Background technology
In recent years, with the fast development of economy and the increasingly raising of living standards of the people, the volume of holding per capita of automobile
Rise year by year.Entering the quickening of family's paces along with automobile, resultant static traffic problem becomes increasingly conspicuous, and as system
About Domestic Automotive Industry or even the outstanding problem of the city national economic development.The problem of there is currently is mainly shown as following
Aspect：First, programming and distribution are unreasonable；Second, land use index control is inadequate；Third, the parking position negligible amounts of cell；Four
It is that public building appertaining parking is very few, road road occupying phenomenon is serious.The hair for the automobile economy that case above has a strong impact on
Exhibition, therefore, the problem of the efficiency of operation for improving parking lot become the key for solving parking difficulty.
With increasing rapidly for motor vehicle number, reasonably can effectively solve storing cycle using parking lot management
Problem, road surface occupation rate is reduced, so as to alleviate road traffic problem.Chinese vehicle ownership now breaks through 300,000,000, specification vehicle
Park, keep good parking order, rationally have using parking stall resource, real-time display parking stall occupancy situation for daily life
Very obvious action.But traditional parking lot the more is built the more big, the efficiency of labor management is scarce more and more lower, therefore intelligent parking
By higher and higher concern in the industry, research direction is also constantly widened deeply for field.
Often there are many sub-modules in Intelligent parking lot management system, and parking stall measure module occupies in the entire system
Important status, the effect of parking stall measure, directly affect the quality of whole system.The parking stall measure in parking lot is parking
One of key technology of field management information system is also the basis of parking guidance, in parking management information system, parking space information
It is the most important information of bottom, the parking stall occupancy situation in parking lot can be monitored in real time by parking stall measure, and be sent to
Into the database of administrative center, so as to realize timely issue and management to parking stall occupied information, branch is provided for vehicle guidance
Hold.
With the rapid development of information technology, especially computer vision and artificial intelligence technology achieve it is breakthrough enter
Exhibition so that can should much be completed by the work that manpower is completed by computer.With manual type, ground induction coil, ultrasonic wave
Technology for detection parking stall is compared, and the parking stall measure technology based on video has unique advantage, thus causes the extensive concern of people.
Parking stall measure technology based on video, is analyzed and processed by the video collected to front end camera, can be detected in real time
The occupied situation in parking stall.In addition, multiple parking stalls can be monitored simultaneously using a camera, parking lot pipe is greatlyd save
Manage cost.
Therefore, video frequency car position detection is generally used for the management system of large parking lot parking stall, can both facilitate car owner to utilize
The minimum time finds empty parking space, can also reduce cost, reduce crowded parking lot road, raising parking stall utilization ratio and enhancing
The security to park cars.For example, to alleviating company, community parking hardly possible, improve gas station parking stall and take efficiency, carry out road prison
Control, improving storing cycle management has highly important effect.Carrying out parking stall management using the means of Computer Image Processing has
Very high accuracy rate, the workload of people will be greatly reduced.
The content of the invention
It is an object of the invention to provide a kind of depth map by vertical view merge the detection methods of convolutional neural networks come
Judge the real-time detection method whether parking stall is correctly taken.
The object of the present invention is achieved like this：
The invention discloses a kind of real-time detection method for judging parking stall and taking, cromogram is extracted by binocular solid camera
With depth map, vehicle rectangle frame is obtained using depth image segmentation method, is carried out with the CarNet depth convolutional neural networks of design
Vehicle detection, then fusion depth map segmentation result and CarNet testing results in cromogram are combined judgement, it is final to determine
Whether selected parking stall is occupied, and its concrete implementation step includes：
(1) binocular solid camera is placed in directly over monitoring parking stall, straight down, camera overlooks monitoring one to camera angle
Individual or multiple parking stalls, the parking stall for ensureing to need to monitor is all in the visual field of camera；
(2) vehicle sample and non-vehicle sample are collected, all samples are all normalized；
(3) design is applied to the convolutional neural networks CarNet under specific environment, and detects vehicle with CarNet networks；
(4) parking stall region is manually demarcated in binocular stereo camera left image, and marks parking stall number successively；
(5) left and right two width coloured image is obtained using binocular stereo camera, depth map is obtained by calculating parallax；
(6) artificial set depth threshold value, the depth graph region for the artificial setting height that is above the ground level is extracted；
(7) isolated pixel point is eliminated to the depth graph region morphological erosion that segmentation obtains, filled up with morphological dilations
Cavity, the too small noise section of area is filtered out by area screening, extracts depth map segmentation region minimum enclosed rectangle frame, so as to
Obtain vehicle rectangle frame R1；
(8) stereo camera left color image is directed to, vehicle rectangle frame R is obtained using CarNet detection models2；
(9) the result R obtained according to CarNet detection models2With reference to the rectangle frame R of step (7) depth map segmentation1Joined
Close and judge, obtained output result is the rectangle frame R of vehiclef；
(10) parking stall seizure condition is judged with the parking stall region accounting manually set by vehicle detection frame, accounting exceedes people
Work given threshold is then judged as that parking stall is occupied.
For a kind of real-time detection method for judging parking stall and taking, the vehicle sample collected in described step (2) and non-
Vehicle sample is applied to specific environment scene, and all samples use the top view after normalization.
For a kind of real-time detection method for judging parking stall and taking, the convolutional neural networks CarNet described in step (3)
Characteristic layer is extracted in detection model based on preceding 4 layers of convolutional layer, rear 3 layers of convolutional layer group carries out multiple scale detecting, its concrete structure
Including：
In (3.1) level 1 volume laminations, convolution kernel size is 11 × 11, step-length 4, and output characteristic figure port number is 16；The
In level 2 volume lamination, convolution kernel size is 3 × 3, step-length 1, and output characteristic figure port number is 20；In 3rd layer of convolutional layer, convolution
Core size is 3 × 3, step-length 1, and output characteristic figure port number is 30；In 4th layer of convolutional layer, convolution kernel size is 3 × 3, step-length
For 1, output characteristic figure port number is 48；
(3.2) a maximum pond layer, pond layer convolution kernel chi are all connected in first 4 layers of volume basic unit after every layer of convolutional layer
Very little is 3 × 3, step-length 2, and the activation primitive in each convolutional layer uses Leaky-ReLU functions
Wherein, α is a small constant；
(3.3) after the 4th layer of convolution, 3 layers of characteristic pattern convolutional layer group are added, and it is special in the output of this 3 layers of convolutional layer groups
Multiple dimensioned return is carried out on sign figure to detect；K encirclement frame of the various sizes of each position generation different scale of characteristic pattern, is calculated
Each encirclement frame belongs to car, the possibility of the class of background two, i.e. score value；4 skews relative to actual object frame are calculated simultaneously
Value, these encirclement frames are finely tuned by Bounding Box recurrence deviants；
(3.4) 3 layers of characteristic pattern convolutional layer group structure of addition are：5th layer of convolutional layer group is formed by 2 layers, level 2 volume product core chi
Very little is respectively 3 × 3 and 1 × 1, and step-length is all 1, and output characteristic figure port number is all 96；6th layer of convolutional layer group is also formed by 2 layers,
Level 2 volume product core size is respectively 1 × 1 and 3 × 3, and step-length is respectively 1 and 2, and output characteristic figure port number is respectively 24 and 48；7th
Layer convolutional layer group is also made up of level 2 volume lamination, and level 2 volume product core size is respectively 1 × 1 and 3 × 3, and step-length is respectively 1 and 2, is exported
Characteristic pattern port number is respectively 24 and 48.
For a kind of real-time detection method for judging parking stall and taking, set depth figure threshold value T in described step (6)d, carry
The image-region for the artificial setting height that is above the ground level is taken, as depth value d in depth mapp≤Td, retain current pixel, work as depth map
Middle depth value dp≥Td, delete current pixel and be arranged to 0.
For a kind of real-time detection method for judging parking stall and taking, utilization depth map segmentation described in step (9) with
The specific implementation step that CarNet detection models carry out result judgement includes：
(9.1) if step (7) and step (8) detect vehicle, and the two car rectangle frame coincidence factors detected are more than and set
Determine threshold value, then it is same car to judge two candidate frames, and it is the final detection block R of vehicle to take the common factors of two rectangle framesf；
(9.2) if step (7) is determined as that the rectangle frame coincidence factor of car is less than given threshold with step (8), then it is assumed that detection
Unstable result, without determining whether；
(9.3) if to being not detected by vehicle in same parking stall step (7), and vehicle is detected in step (8), then judged
CarNet model inspection false-alarms；If corresponding parking stall detects vehicle in step (7), and is not detected by vehicle in step (8), then
It is determined as that tall and big non-car object enters monitor area.
For a kind of real-time detection method for judging parking stall and taking, a certain delimitation parking stall is directed in described step (10),
Rectangle frame R based on vehiclefOverlapped with the area for delimiting parking stall in advance than threshold value TaTo judge parking stall occupancy situation, if actual face
It is β that product, which overlaps ratio, as β >=TaWhen, judge that the parking stall is occupied；If β≤Ta, then return to step (5).
The beneficial effects of the present invention are：
A kind of real-time detection method for judging parking stall and taking disclosed by the invention, cromogram is overlooked by using the width of left and right two
Depth map is obtained, depth map is split to obtain the detection method of vehicle rectangle frame fusion convolutional neural networks, can be automatic
Detection identification is carried out to vehicle, judges whether vehicle is correctly parked；
Video can be gathered using the camera of setting in real time by the parking stall measure technology of view-based access control model and divided
Analysis is handled, and detects the occupied situation in parking stall in real time；In addition, multiple parking stalls can be monitored simultaneously using a camera, greatly
Save parking lot management cost greatly；
Simultaneously as the advantage of vision parking stall measure technology itself, vision parking stall measure are generally used for large parking lot car
The management system of position, both can facilitate car owner to find empty parking space using the minimum time, can also reduce cost, reduce parking lot road
Road is crowded, improves parking stall utilization ratio and security that enhancing parks cars, difficult to alleviating company, community parking, improve plus
Petrol station parking stall takes efficiency, carries out road monitoring, improving storing cycle management has highly important effect；
In addition, carrying out parking stall management using the means of Computer Image Processing has very high accuracy rate, solve traditional
Though labor management cost is high but efficiency is low, manages the drawbacks of poor, the workload of people is greatly reduced.
Brief description of the drawings
Fig. 1 is the schematic flow sheet that the real-time detection method that parking stall takes is judged in the present invention；
Fig. 2 is binocular stereo vision schematic diagram in the present invention.
Embodiment
The present invention is described further below in conjunction with the accompanying drawings.
With reference to Fig. 1, the invention discloses a kind of real-time detection method for judging parking stall and taking, carried by binocular solid camera
Cromogram and depth map are taken, vehicle rectangle frame is obtained using depth image segmentation method, with the CarNet depth convolutional Neurals of design
Network carries out vehicle detection, then fusion depth map segmentation result and CarNet testing results in cromogram are combined into judgement,
It is final to determine whether selected parking stall is occupied, and its concrete implementation step includes：
(1) binocular solid camera is placed in directly over monitoring parking stall, straight down, camera overlooks monitoring one to camera angle
Individual or multiple parking stalls, the parking stall for ensureing to need to monitor is all in the visual field of camera.
(2) vehicle sample and non-vehicle sample are collected, all samples are all normalized；
Due to being monitored using depression angle, carrying out the sample that classification uses needs to use top view.Using putting up
Camera collection vehicle overlook sample.Upset, translation transformation, change of scale and rotation transformation at random are carried out to vehicle sample to increase
Add sample size, by all vehicle samples by be sized normalize size；Change of scale is carried out to non-vehicle sample, will be all
Non-vehicle sample by be sized normalize size；Vehicle sample presses 1 with non-vehicle sample:2 ratio obtains.
(3) design is applied to the convolutional neural networks CarNet detection models under specific environment, and is examined with CarNet networks
Measuring car；Characteristic layer, rear 3 layers of convolutional layer group are extracted in convolutional neural networks CarNet detection models based on preceding 4 layers of convolutional layer
Multiple scale detecting is carried out, its concrete structure includes：
In (3.1) level 1 volume laminations, convolution kernel size is 11 × 11, step-length 4, and output characteristic figure port number is 16；The
In level 2 volume lamination, convolution kernel size is 3 × 3, step-length 1, and output characteristic figure port number is 20；In 3rd layer of convolutional layer, convolution
Core size is 3 × 3, step-length 1, and output characteristic figure port number is 30；In 4th layer of convolutional layer, convolution kernel size is 3 × 3, step-length
For 1, output characteristic figure port number is 48；
(3.2) a maximum pond layer, pond layer convolution kernel chi are all connected in first 4 layers of volume basic unit after every layer of convolutional layer
Very little is 3 × 3, step-length 2, and the activation primitive in each convolutional layer uses Leaky-ReLU functions
Wherein, α is a small constant；
(3.3) after the 4th layer of convolution, 3 layers of characteristic pattern convolutional layer group are added, and it is special in the output of this 3 layers of convolutional layer groups
Multiple dimensioned return is carried out on sign figure to detect；K encirclement frame of the various sizes of each position generation different scale of characteristic pattern, is calculated
Each encirclement frame belongs to car, the possibility of the class of background two, i.e. score value；4 skews relative to actual object frame are calculated simultaneously
Value, these encirclement frames are finely tuned by Bounding Box recurrence deviants；
(3.4) 3 layers of characteristic pattern convolutional layer group structure of addition are：5th layer of convolutional layer group is formed by 2 layers, level 2 volume product core chi
Very little is respectively 3 × 3 and 1 × 1, and step-length is all 1, and output characteristic figure port number is all 96；6th layer of convolutional layer group is also formed by 2 layers,
Level 2 volume product core size is respectively 1 × 1 and 3 × 3, and step-length is respectively 1 and 2, and output characteristic figure port number is respectively 24 and 48；7th
Layer convolutional layer group is also made up of level 2 volume lamination, and level 2 volume product core size is respectively 1 × 1 and 3 × 3, and step-length is respectively 1 and 2, is exported
Characteristic pattern port number is respectively 24 and 48.
(4) it is artificial that parking stall region is demarcated in binocular stereo camera left image with quadrangle, parking stall number is marked successively, together
When calculate each quadrangle parking stall area.
(5) left and right two width RGB color image p is obtained using binocular stereo camera1And p2, depth is obtained by calculating parallax
Degree figure, is represented with eight-digit binary number figure；The RGB image P obtained with t1And P2Exemplified by；
With reference to Fig. 2, C11And C12Represent two cameras of binocular camera.Parallax range between them represents with T, π1With
π2Represent the imaging plane of two cameras, f is focal length, and the distance of any point p to two photocentre line in space is Z, and P exists
The projection of two image planes is respectively p1And pr, p1And prRespectively with C11And C12It is x for the coordinate in the coordinate system of origin1
And xr, parallax is
D=x1-xr,
According to the principle of similar triangles in figure, we can obtain
Two formula can show that the actual grade of object and the relation of parallax are more than
(6) artificial set depth threshold value Td, extraction is above the ground level the depth graph region of artificial setting height, when in depth map
Depth value dp≤Td, retain current pixel, as depth value d in depth mapp≥Td, delete current pixel and be arranged to 0.
(7) isolated pixel point is eliminated to the depth graph region morphological erosion that segmentation obtains, filled up with morphological dilations
Cavity, the too small noise section of area is filtered out by area screening, extracts depth map segmentation region minimum enclosed rectangle frame, so as to
Obtain vehicle rectangle frame R1。
(8) stereo camera left color image is directed to, vehicle rectangle frame R is obtained using CarNet detection models2。
(9) the result R obtained according to CarNet detection models2With reference to the rectangle frame R of step (7) depth map segmentation1Joined
Close and judge, obtained output result is the rectangle frame R of vehiclef, it, which implements step, includes：
(9.1) if step (7) and step (8) detect vehicle, and the two car rectangle frame coincidence factors detected are more than and set
Determine threshold value, then it is same car to judge two candidate frames, and it is the final detection block R of vehicle to take the common factors of two rectangle framesf；
(9.2) if step (7) is determined as that the rectangle frame coincidence factor of car is less than given threshold with step (8), then it is assumed that detection
Unstable result, without determining whether；
(9.3) if to being not detected by vehicle in same parking stall step (7), and vehicle is detected in step (8), then judged
CarNet model inspection false-alarms；If corresponding parking stall detects vehicle in step (7), and is not detected by vehicle in step (8), then
It is determined as that tall and big non-car object enters monitor area.
(10) parking stall seizure condition is judged with the parking stall region accounting manually set by vehicle detection frame, accounting exceedes people
Work given threshold is then judged as that parking stall is occupied；For a certain delimitation parking stall, the rectangle frame R based on vehiclefWith delimiting car in advance
The area of position is overlapped than threshold value TaTo judge parking stall occupancy situation, if it is β that real area, which overlaps ratio, as β >=TaWhen, judge the car
Position is occupied；If β≤Ta, then return to step (5).
The preferred embodiments of the present invention are the foregoing is only, are not intended to limit the invention, for the skill of this area
For art personnel, the present invention can have various modifications and variations.Within the spirit and principles of the invention, that is made any repaiies
Change, equivalent substitution, improvement etc., should be included in the scope of the protection.
Claims (6)
1. it is a kind of judge parking stall take real-time detection method, it is characterised in that by binocular solid camera extract cromogram with
Depth map, vehicle rectangle frame is obtained using depth image segmentation method, enter driving with the CarNet depth convolutional neural networks of design
Detection, then CarNet testing results judgements will be combined in fusion depth map segmentation result and cromogram, it is final to determine to select
Whether occupied determine parking stall, its concrete implementation step includes：
(1) by binocular solid camera be placed in monitoring parking stall directly over, camera angle straight down, camera overlook monitoring one or
The multiple parking stalls of person, the parking stall for ensureing to need to monitor is all in the visual field of camera；
(2) vehicle sample and non-vehicle sample are collected, all samples are all normalized；
(3) design is applied to the convolutional neural networks CarNet under specific environment, and detects vehicle with CarNet networks；
(4) parking stall region is manually demarcated in binocular stereo camera left image, and marks parking stall number successively；
(5) left and right two width coloured image is obtained using binocular stereo camera, depth map is obtained by calculating parallax；
(6) artificial set depth threshold value, the depth graph region for the artificial setting height that is above the ground level is extracted；
(7) isolated pixel point is eliminated to the obtained depth graph region morphological erosion of segmentation, with morphological dilations filling cavity,
The too small noise section of area, extraction depth map segmentation region minimum enclosed rectangle frame, so as to obtain are filtered out by area screening
Vehicle rectangle frame R1；
(8) stereo camera left color image is directed to, vehicle rectangle frame R is obtained using CarNet detection models2；
(9) the result R obtained according to CarNet detection models2With reference to the rectangle frame R of step (7) depth map segmentation1Combine sentencing
Disconnected, obtained output result is the rectangle frame R of vehiclef；
(10) parking stall seizure condition is judged with the parking stall region accounting manually set by vehicle detection frame, accounting, which exceedes, manually to be set
Determine threshold value and be then judged as that parking stall is occupied.
A kind of 2. real-time detection method for judging parking stall and taking according to claim 1, it is characterised in that：Described step
(2) the vehicle sample and non-vehicle sample collected in are applied to specific environment scene, after all samples are using normalization
Top view.
3. a kind of real-time detection method for judging parking stall and taking according to claim 1, it is characterised in that in step (3)
Characteristic layer is extracted in described convolutional neural networks CarNet detection models based on preceding 4 layers of convolutional layer, rear 3 layers of convolutional layer group is entered
Row multiple scale detecting, its concrete structure include：
In (3.1) level 1 volume laminations, convolution kernel size is 11 × 11, step-length 4, and output characteristic figure port number is 16；2nd layer
In convolutional layer, convolution kernel size is 3 × 3, step-length 1, and output characteristic figure port number is 20；In 3rd layer of convolutional layer, convolution kernel chi
Very little is 3 × 3, step-length 1, and output characteristic figure port number is 30；In 4th layer of convolutional layer, convolution kernel size is 3 × 3, step-length 1,
Output characteristic figure port number is 48；
(3.2) a maximum pond layer is all connected in first 4 layers of volume basic unit after every layer of convolutional layer, pond layer convolution kernel size is
3 × 3, step-length 2, the activation primitive in each convolutional layer uses Leaky-ReLU functions
<mfenced open = "{" close = "">
<mtable>
<mtr>
<mtd>
<mi>f</mi>
<mo>(</mo>
<mi>x</mi>
<mo>)</mo>
<mo>=</mo>
<mi>&alpha;</mi>
<mi>x</mi>
<mo>,</mo>
<mo>(</mo>
<mi>x</mi>
<mo><</mo>
<mn>0</mn>
<mo>)</mo>
</mtd>
</mtr>
<mtr>
<mtd>
<mi>f</mi>
<mo>(</mo>
<mi>x</mi>
<mo>)</mo>
<mo>=</mo>
<mi>x</mi>
<mo>,</mo>
<mo>(</mo>
<mi>x</mi>
<mo>&GreaterEqual;</mo>
<mn>0</mn>
<mo>)</mo>
</mtd>
</mtr>
</mtable>
</mfenced>
Wherein, α is a small constant；
(3.3) after the 4th layer of convolution, 3 layers of characteristic pattern convolutional layer group are added, and in the output characteristic figure of this 3 layers of convolutional layer groups
It is upper to carry out multiple dimensioned recurrence detection；K encirclement frame of the various sizes of each position generation different scale of characteristic pattern, is calculated each
Encirclement frame belongs to car, the possibility of the class of background two, i.e. score value；4 deviants relative to actual object frame are calculated simultaneously, are led to
Bounding Box recurrence deviants are crossed to finely tune these encirclement frames；
(3.4) 3 layers of characteristic pattern convolutional layer group structure of addition are：5th layer of convolutional layer group is formed by 2 layers, level 2 volume product core size point
Not Wei 3 × 3 and 1 × 1, step-length all be 1, output characteristic figure port number all be 96；6th layer of convolutional layer group is also formed by 2 layers, 2 layers
Convolution kernel size is respectively 1 × 1 and 3 × 3, and step-length is respectively 1 and 2, and output characteristic figure port number is respectively 24 and 48；7th layer
Convolutional layer group is also made up of level 2 volume lamination, and level 2 volume product core size is respectively 1 × 1 and 3 × 3, and step-length is respectively 1 and 2, and output is special
Sign figure port number is respectively 24 and 48.
A kind of 4. real-time detection method for judging parking stall and taking according to claim 1, it is characterised in that：Described step
(6) set depth figure threshold value T ind, extraction is above the ground level the image-region of artificial setting height, as depth value d in depth mapp≤
Td, retain current pixel, as depth value d in depth mapp≥Td, delete current pixel and be arranged to 0.
5. a kind of real-time detection method for judging parking stall and taking according to claim 1, it is characterised in that in step (9)
The described specific implementation step that result judgement is carried out using depth map segmentation and CarNet detection models is included：
(9.1) if step (7) detects vehicle with step (8), and the two car rectangle frame coincidence factors detected are more than setting threshold
Value, then it is same car to judge two candidate frames, and it is the final detection block R of vehicle to take the common factors of two rectangle framesf；
(9.2) if step (7) is determined as that the rectangle frame coincidence factor of car is less than given threshold with step (8), then it is assumed that testing result
It is unstable, without determining whether；
(9.3) if to being not detected by vehicle in same parking stall step (7), and vehicle is detected in step (8), then judges CarNet
Model inspection false-alarm；If corresponding parking stall detects vehicle in step (7), and vehicle is not detected by step (8), then it is determined as
Tall and big non-car object enters monitor area.
A kind of 6. real-time detection method for judging parking stall and taking according to claim 1, it is characterised in that：Described step
(10) a certain delimitation parking stall, the rectangle frame R based on vehicle are directed infOverlapped with the area for delimiting parking stall in advance than threshold value TaTo sentence
Parking stall occupancy situation is determined, if it is β that real area, which overlaps ratio, as β >=TaWhen, judge that the parking stall is occupied；If β≤Ta, then step is returned
Suddenly (5).
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710795417.5A CN107665603B (en) | 2017-09-06 | 2017-09-06 | Real-time detection method for judging parking space occupation |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710795417.5A CN107665603B (en) | 2017-09-06 | 2017-09-06 | Real-time detection method for judging parking space occupation |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107665603A true CN107665603A (en) | 2018-02-06 |
| CN107665603B CN107665603B (en) | 2020-06-16 |
Family
ID=61097316
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710795417.5A Active CN107665603B (en) | 2017-09-06 | 2017-09-06 | Real-time detection method for judging parking space occupation |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107665603B (en) |
Cited By (27)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108491828A (en) * | 2018-04-20 | 2018-09-04 | 济南浪潮高新科技投资发展有限公司 | A kind of parking site detecting system and method for the pairwise similarity PVAnet based on level |
| CN108847052A (en) * | 2018-08-21 | 2018-11-20 | 北京深瞐科技有限公司 | Parking position determines method, apparatus, system and computer-readable medium |
| CN109559545A (en) * | 2018-11-09 | 2019-04-02 | 歌尔股份有限公司 | A kind of parking stall sharing method, system and device |
| CN109631896A (en) * | 2018-07-23 | 2019-04-16 | 同济大学 | A kind of parking lot autonomous parking localization method based on vehicle vision and motion information |
| CN109726637A (en) * | 2018-12-03 | 2019-05-07 | 初速度（苏州）科技有限公司 | The parking site detecting system and method in parking stall region are indicated with resilient polygonal |
| CN109741625A (en) * | 2019-02-15 | 2019-05-10 | 江苏镭博智能科技有限公司 | A kind of prison bat identification equipment and its prison bat recognition methods for parking lot |
| CN109817013A (en) * | 2018-12-19 | 2019-05-28 | 新大陆数字技术股份有限公司 | Parking stall state identification method and device based on video flowing |
| CN110135394A (en) * | 2019-05-27 | 2019-08-16 | 杭州亚美利嘉科技有限公司 | Region occupancy and licence plate recognition method and device |
| CN110299028A (en) * | 2019-07-31 | 2019-10-01 | 深圳市捷顺科技实业股份有限公司 | Method, apparatus, equipment and the readable storage medium storing program for executing of line detection are got in a kind of parking |
| CN110322681A (en) * | 2018-03-30 | 2019-10-11 | 杭州海康威视数字技术股份有限公司 | The method and apparatus for detecting parking space state |
| CN110348297A (en) * | 2019-05-31 | 2019-10-18 | 纵目科技（上海）股份有限公司 | A kind of detection method, system, terminal and the storage medium of parking systems for identification |
| CN110490117A (en) * | 2019-08-14 | 2019-11-22 | 智慧互通科技有限公司 | A kind of Parking based on image depth information determines method and system |
| CN111079540A (en) * | 2019-11-19 | 2020-04-28 | 北航航空航天产业研究院丹阳有限公司 | Target characteristic-based layered reconfigurable vehicle-mounted video target detection method |
| CN111191546A (en) * | 2019-12-20 | 2020-05-22 | 广西柳州联耕科技有限公司 | Intelligent product assembling method based on machine vision recognition |
| CN111383457A (en) * | 2018-12-30 | 2020-07-07 | 浙江宇视科技有限公司 | Parking space state detection method and device, equipment and storage medium |
| EP3690728A1 (en) * | 2019-01-31 | 2020-08-05 | StradVision, Inc. | Method and device for detecting parking area using semantic segmentation in automatic parking system |
| CN111554123A (en) * | 2020-05-19 | 2020-08-18 | 北京爱笔科技有限公司 | Control method and device of parking space control component |
| CN111932933A (en) * | 2020-08-05 | 2020-11-13 | 杭州像素元科技有限公司 | Urban intelligent parking space detection method and equipment and readable storage medium |
| CN112634650A (en) * | 2020-12-18 | 2021-04-09 | 中标慧安信息技术股份有限公司 | Parking lot management method and system based on audio and video monitoring |
| CN112669377A (en) * | 2019-10-15 | 2021-04-16 | 杭州海康威视数字技术股份有限公司 | Parking space detection method and device, electronic equipment and storage medium |
| CN113235998A (en) * | 2021-05-18 | 2021-08-10 | 东风汽车集团股份有限公司 | Lifting type hydraulic three-dimensional parking garage and working method thereof |
| CN113762272A (en) * | 2021-09-10 | 2021-12-07 | 北京精英路通科技有限公司 | Road information determination method and device and electronic equipment |
| CN113807209A (en) * | 2021-08-30 | 2021-12-17 | 阿波罗智联(北京)科技有限公司 | Parking space detection method and device, electronic equipment and storage medium |
| CN114067598A (en) * | 2020-07-30 | 2022-02-18 | 上汽通用汽车有限公司 | Method and device for determining idle parking space |
| CN114179788A (en) * | 2021-12-29 | 2022-03-15 | 纵目科技（上海）股份有限公司 | Automatic parking method, system, computer readable storage medium and vehicle terminal |
| CN114255584A (en) * | 2021-12-20 | 2022-03-29 | 济南博观智能科技有限公司 | Positioning method and system for parking vehicle, storage medium and electronic equipment |
| CN114694124A (en) * | 2022-05-31 | 2022-07-01 | 成都国星宇航科技股份有限公司 | Parking space state detection method and device and storage medium |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101187984A (en) * | 2007-12-05 | 2008-05-28 | 北京中星微电子有限公司 | An image detection method and device |
| CN105225482A (en) * | 2015-09-02 | 2016-01-06 | 上海大学 | Based on vehicle detecting system and the method for binocular stereo vision |
| EP3147799A1 (en) * | 2015-09-22 | 2017-03-29 | Xerox Corporation | Similarity-based detection of prominent objects using deep cnn pooling layers as features |
| CN106650708A (en) * | 2017-01-19 | 2017-05-10 | 南京航空航天大学 | Visual detection method and system for automatic driving obstacles |
| CN107067813A (en) * | 2017-06-14 | 2017-08-18 | 大连海事大学 | A kind of parking stall bootstrap technique and system based on image procossing and pattern-recognition |
-
2017
- 2017-09-06 CN CN201710795417.5A patent/CN107665603B/en active Active
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101187984A (en) * | 2007-12-05 | 2008-05-28 | 北京中星微电子有限公司 | An image detection method and device |
| CN105225482A (en) * | 2015-09-02 | 2016-01-06 | 上海大学 | Based on vehicle detecting system and the method for binocular stereo vision |
| EP3147799A1 (en) * | 2015-09-22 | 2017-03-29 | Xerox Corporation | Similarity-based detection of prominent objects using deep cnn pooling layers as features |
| CN106650708A (en) * | 2017-01-19 | 2017-05-10 | 南京航空航天大学 | Visual detection method and system for automatic driving obstacles |
| CN107067813A (en) * | 2017-06-14 | 2017-08-18 | 大连海事大学 | A kind of parking stall bootstrap technique and system based on image procossing and pattern-recognition |
Cited By (36)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110322681A (en) * | 2018-03-30 | 2019-10-11 | 杭州海康威视数字技术股份有限公司 | The method and apparatus for detecting parking space state |
| CN110322681B (en) * | 2018-03-30 | 2021-03-26 | 杭州海康威视数字技术股份有限公司 | Method and device for detecting parking space state |
| CN108491828A (en) * | 2018-04-20 | 2018-09-04 | 济南浪潮高新科技投资发展有限公司 | A kind of parking site detecting system and method for the pairwise similarity PVAnet based on level |
| CN109631896A (en) * | 2018-07-23 | 2019-04-16 | 同济大学 | A kind of parking lot autonomous parking localization method based on vehicle vision and motion information |
| CN108847052A (en) * | 2018-08-21 | 2018-11-20 | 北京深瞐科技有限公司 | Parking position determines method, apparatus, system and computer-readable medium |
| CN108847052B (en) * | 2018-08-21 | 2020-06-05 | 北京深瞐科技有限公司 | Parking position determining method, device, system and computer readable medium |
| CN109559545A (en) * | 2018-11-09 | 2019-04-02 | 歌尔股份有限公司 | A kind of parking stall sharing method, system and device |
| CN109726637A (en) * | 2018-12-03 | 2019-05-07 | 初速度（苏州）科技有限公司 | The parking site detecting system and method in parking stall region are indicated with resilient polygonal |
| CN109817013A (en) * | 2018-12-19 | 2019-05-28 | 新大陆数字技术股份有限公司 | Parking stall state identification method and device based on video flowing |
| CN111383457A (en) * | 2018-12-30 | 2020-07-07 | 浙江宇视科技有限公司 | Parking space state detection method and device, equipment and storage medium |
| CN111383457B (en) * | 2018-12-30 | 2021-02-09 | 浙江宇视科技有限公司 | Parking space state detection method and device, equipment and storage medium |
| CN111507158A (en) * | 2019-01-31 | 2020-08-07 | 斯特拉德视觉公司 | Method and device for detecting parking area by semantic segmentation |
| US10796434B1 (en) | 2019-01-31 | 2020-10-06 | Stradvision, Inc | Method and device for detecting parking area using semantic segmentation in automatic parking system |
| CN111507158B (en) * | 2019-01-31 | 2023-11-24 | 斯特拉德视觉公司 | Method and device for detecting parking area by using semantic segmentation |
| EP3690728A1 (en) * | 2019-01-31 | 2020-08-05 | StradVision, Inc. | Method and device for detecting parking area using semantic segmentation in automatic parking system |
| CN109741625A (en) * | 2019-02-15 | 2019-05-10 | 江苏镭博智能科技有限公司 | A kind of prison bat identification equipment and its prison bat recognition methods for parking lot |
| CN110135394A (en) * | 2019-05-27 | 2019-08-16 | 杭州亚美利嘉科技有限公司 | Region occupancy and licence plate recognition method and device |
| CN110348297A (en) * | 2019-05-31 | 2019-10-18 | 纵目科技（上海）股份有限公司 | A kind of detection method, system, terminal and the storage medium of parking systems for identification |
| CN110348297B (en) * | 2019-05-31 | 2023-12-26 | 纵目科技（上海）股份有限公司 | Detection method, system, terminal and storage medium for identifying stereo garage |
| CN110299028B (en) * | 2019-07-31 | 2022-06-14 | 深圳市捷顺科技实业股份有限公司 | Parking line crossing detection method, device, equipment and readable storage medium |
| CN110299028A (en) * | 2019-07-31 | 2019-10-01 | 深圳市捷顺科技实业股份有限公司 | Method, apparatus, equipment and the readable storage medium storing program for executing of line detection are got in a kind of parking |
| CN110490117A (en) * | 2019-08-14 | 2019-11-22 | 智慧互通科技有限公司 | A kind of Parking based on image depth information determines method and system |
| CN112669377A (en) * | 2019-10-15 | 2021-04-16 | 杭州海康威视数字技术股份有限公司 | Parking space detection method and device, electronic equipment and storage medium |
| CN112669377B (en) * | 2019-10-15 | 2023-09-05 | 杭州海康威视数字技术股份有限公司 | Parking space detection method and device, electronic equipment and storage medium |
| CN111079540A (en) * | 2019-11-19 | 2020-04-28 | 北航航空航天产业研究院丹阳有限公司 | Target characteristic-based layered reconfigurable vehicle-mounted video target detection method |
| CN111191546A (en) * | 2019-12-20 | 2020-05-22 | 广西柳州联耕科技有限公司 | Intelligent product assembling method based on machine vision recognition |
| CN111554123A (en) * | 2020-05-19 | 2020-08-18 | 北京爱笔科技有限公司 | Control method and device of parking space control component |
| CN114067598A (en) * | 2020-07-30 | 2022-02-18 | 上汽通用汽车有限公司 | Method and device for determining idle parking space |
| CN111932933A (en) * | 2020-08-05 | 2020-11-13 | 杭州像素元科技有限公司 | Urban intelligent parking space detection method and equipment and readable storage medium |
| CN112634650A (en) * | 2020-12-18 | 2021-04-09 | 中标慧安信息技术股份有限公司 | Parking lot management method and system based on audio and video monitoring |
| CN113235998A (en) * | 2021-05-18 | 2021-08-10 | 东风汽车集团股份有限公司 | Lifting type hydraulic three-dimensional parking garage and working method thereof |
| CN113807209A (en) * | 2021-08-30 | 2021-12-17 | 阿波罗智联(北京)科技有限公司 | Parking space detection method and device, electronic equipment and storage medium |
| CN113762272A (en) * | 2021-09-10 | 2021-12-07 | 北京精英路通科技有限公司 | Road information determination method and device and electronic equipment |
| CN114255584A (en) * | 2021-12-20 | 2022-03-29 | 济南博观智能科技有限公司 | Positioning method and system for parking vehicle, storage medium and electronic equipment |
| CN114179788A (en) * | 2021-12-29 | 2022-03-15 | 纵目科技（上海）股份有限公司 | Automatic parking method, system, computer readable storage medium and vehicle terminal |
| CN114694124A (en) * | 2022-05-31 | 2022-07-01 | 成都国星宇航科技股份有限公司 | Parking space state detection method and device and storage medium |
Also Published As
| Publication number | Publication date |
|---|---|
| CN107665603B (en) | 2020-06-16 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107665603A (en) | A kind of real-time detection method for judging parking stall and taking | |
| CN106373426B (en) | Parking stall based on computer vision and violation road occupation for parking monitoring method | |
| CN103985250B (en) | The holographic road traffic state vision inspection apparatus of lightweight | |
| WO2022141911A1 (en) | Roadside sensing unit-based method for quick recognition of dynamic target point cloud and point cloud segmentation | |
| CN110097109A (en) | A kind of road environment obstacle detection system and method based on deep learning | |
| CN108509954A (en) | A kind of more car plate dynamic identifying methods of real-time traffic scene | |
| CN109033950A (en) | Vehicle based on multiple features fusion cascade deep model, which is disobeyed, stops detection method | |
| CN102867417B (en) | Taxi anti-forgery system and taxi anti-forgery method | |
| CN108491797A (en) | A kind of vehicle image precise search method based on big data | |
| CN105913685A (en) | Video surveillance-based carport recognition and intelligent guide method | |
| CN105118305B (en) | Motor pool outlet vehicle management platform | |
| CN106326893A (en) | Vehicle color recognition method based on area discrimination | |
| CN106127137A (en) | A kind of target detection recognizer based on 3D trajectory analysis | |
| CN102663357A (en) | Color characteristic-based detection algorithm for stall at parking lot | |
| CN101807352A (en) | Method for detecting parking stalls on basis of fuzzy pattern recognition | |
| CN102024148A (en) | Method for identifying green mark of taxi | |
| CN109993138A (en) | A kind of car plate detection and recognition methods and device | |
| CN109935080A (en) | The monitoring system and method that a kind of vehicle flowrate on traffic route calculates in real time | |
| CN103164958B (en) | Method and system for vehicle monitoring | |
| CN109993163A (en) | A kind of non-rating plate identifying system and its recognition methods based on artificial intelligence | |
| CN110490150A (en) | A kind of automatic auditing system of picture violating the regulations and method based on vehicle retrieval | |
| CN109272482A (en) | A kind of urban road crossing vehicle queue detection system based on sequence image | |
| CN110610153A (en) | Lane recognition method and system for automatic driving | |
| CN107195182A (en) | A kind of dynamic high speed overload remediation detecting system | |
| CN106778540A (en) | Parking detection is accurately based on the parking event detecting method of background double layer |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |