CN107092855A - Vehicle part recognition methods and equipment, vehicle identification method and equipment - Google Patents
Vehicle part recognition methods and equipment, vehicle identification method and equipment Download PDFInfo
- Publication number
- CN107092855A CN107092855A CN201610091338.1A CN201610091338A CN107092855A CN 107092855 A CN107092855 A CN 107092855A CN 201610091338 A CN201610091338 A CN 201610091338A CN 107092855 A CN107092855 A CN 107092855A
- Authority
- CN
- China
- Prior art keywords
- vehicle
- cell
- image
- energy value
- vehicle part
- 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
- 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
- G06V20/584—Recognition of moving objects or obstacles, e.g. vehicles or pedestrians; Recognition of traffic objects, e.g. traffic signs, traffic lights or roads of vehicle lights or traffic lights
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V2201/00—Indexing scheme relating to image or video recognition or understanding
- G06V2201/08—Detecting or categorising vehicles
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Traffic Control Systems (AREA)
- Image Analysis (AREA)
Abstract
Present disclose provides a kind of method and apparatus for being used to recognize the vehicle part in image.This method includes：At least a portion of image is divided into multiple cells；For the presetted pixel in each cell, its energy value is calculated using edge detection algorithm, and obtains according to the energy value of the presetted pixel in the cell energy value of the cell；And the energy value of the cell in image determines whether include vehicle part in described image.In addition, the disclosure additionally provides vehicle identification method and relevant device based on this method.
Description
Technical field
The disclosure relates generally to field of image recognition, relates more specifically to vehicle part identification side
Method and equipment, vehicle peccancy determine method and apparatus and vehicle identification method and equipment.
Background technology
With the popularization of automobile in modern society, the real-time and non real-time management and control for automobile turns into
The problem that becomes more and more important.Most important one kind is directed to use with monitoring camera in management and control means,
Such as intelligent transportation bayonet socket, intelligent monitoring；Wherein intelligent transportation bayonet socket refer to above track,
Just to the front/rear portion of vehicle specially to shoot the camera of vehicle, intelligent monitoring refers in track
Side is tiltedly to vehicle specially to shoot the camera of vehicle.Existing intelligent transportation bayonet socket, intelligence
Monitoring generally uses vehicle detecting algorithm, to have detected when that vehicle passes through in video streaming.
Only determine when that vehicle passes through, follow-up identification could be carried out：For example vehicle, car plate,
Driver etc..
There is the technology that vehicle detection is carried out using vehicle part in the prior art, mainly led to
Machine learning is crossed to set up model picture is identified for vehicle part.But this technology
Recognition efficiency is than relatively low.With the popularization of high-definition camera, increasing intelligent transportation bayonet socket
Also high-definition camera is used.And how real-time robust is detected in the big image in different resolution of high definition
Vehicle is to influence one of key factor of whole vehicle identification system performance, to follow-up Tracking Recognition
Directly influence is produced Deng application.
For example, in Chinese patent application CN201410395827.7, it is proposed that one kind is based on
The vehicle identification method of part.This method is recognized according to the identification situation of 6 parts of vehicle
The presence or absence of vehicle.However, this method carries out multipart examining one by one in large-size images
Survey, efficiency is very low and robustness is bad.For mass image data can be got daily
For high-definition intelligent traffic block port, it is difficult to efficiently identify out vehicle therein.
The content of the invention
In order to solve the above problems there is provided the vehicle part recognition methods according to the disclosure and set
Standby, vehicle peccancy determines method and apparatus and vehicle identification method and equipment.
It is used to recognize the vehicle part in image there is provided a kind of according to the first aspect of the disclosure
Method.This method can include：At least a portion of described image is divided into multiple cells；
For the presetted pixel in each cell, its energy value is calculated using edge detection algorithm, and
The energy value of the cell is obtained according to the energy value of the presetted pixel in the cell；And root
Determine whether include vehicle part in described image according to the energy value of cell.
It is used to recognize the vehicle part in image there is provided a kind of according to the second aspect of the disclosure
Equipment.The equipment can include：Split module, for by least a portion of described image
It is divided into multiple cells；Energy computation module, for for the presetted pixel in each cell,
Its energy value is calculated using edge detection algorithm, and according to the energy of the presetted pixel in the cell
Value obtains the energy value of the cell；And vehicle part determining module, for according to unit
The energy value of lattice determines whether include vehicle part in described image.
According to the third aspect of the disclosure, there is provided a kind of method for recognizing the vehicle in image.
This method can include：The figure is recognized using the method according to disclosure first aspect
One or more of picture vehicle part；It is based upon one or more of vehicle parts in advance to divide
The weight matched somebody with somebody, calculates the decision content for indicating to whether there is vehicle in described image；And if institute
Decision content is stated more than or equal to threshold value, then by pair corresponding with one or more of vehicle parts
As being defined as vehicle.
According to the fourth aspect of the disclosure, there is provided a kind of equipment for recognizing the vehicle in image.
The equipment can include：Vehicle part identification module, for using described in disclosure first aspect
Method recognize one or more of described image vehicle part；Decision content computing module,
For being based upon the pre-assigned weight of one or more of vehicle parts, calculate described in indicating
It whether there is the decision content of vehicle in image；And vehicle determining module, if sentenced for described
Definite value is more than or equal to threshold value, then corresponding with one or more of vehicle parts object is true
It is set to vehicle.
By using the various methods and relevant device of the disclosure, can quick detection vehicle part,
Vehicle and/or situation violating the regulations, and change of the program to light, environment has very strong robustness,
And computational efficiency is higher, speed faster.
Brief description of the drawings
By illustrating preferred embodiment of the present disclosure below in conjunction with the accompanying drawings, the above-mentioned of the disclosure will be made
And other objects, features and advantages are clearer, wherein：
Fig. 1 is the example for showing the vehicle part recognition methods according to the first embodiment of the present disclosure
Flow chart.
Fig. 2 is the configuration for showing the vehicle part identification equipment according to the first embodiment of the present disclosure
Block diagram.
Fig. 3 is to show the example that method is determined according to the vehicle peccancy of the second embodiment of the present disclosure
Flow chart.
Fig. 4 is to show the configuration that equipment is determined according to the vehicle peccancy of the second embodiment of the present disclosure
Block diagram.
Fig. 5 is the example flow for showing the vehicle identification method according to the third embodiment of the present disclosure
Figure.
Fig. 6 is the configuration block diagram for showing the vehicle identification equipment according to the third embodiment of the present disclosure.
Embodiment
It is understood in advance that：Although the reality of one or more other embodiments of the present disclosure is provided below
It is existing, but any technology can essentially be used (either currently known is still existing)
To realize disclosed equipment and/or method.The disclosure should not be limited in any way by following theory
Bright illustrative realization including example design shown and described herein and including realizing, accompanying drawing
And technology, but this public affairs can be changed in the range of appended claims and its equivalent thing
Open.
Further it is to be noted that：Although describing some specific embodiments below, these are not represented
Specific embodiment is the minimum/optimal case for realizing the disclosure.In other words, these realities can be used
The some technical characteristics in example are applied as a complete technical scheme, it would however also be possible to employ these realities
Apply the skill that other unaccounted of equal value, replacements, alternative technique feature are combined with the disclosure in example
Art scheme is used as a complete technical scheme.Therefore, the protection domain of the disclosure is not limited to this
A little specific embodiments, but covering those skilled in the art can make according to the training centre of the disclosure
Various modifications, replacement, addition, deletion etc..
Before formally description embodiment of the disclosure, hereinafter may it will introduce first
The term used.Term " vehicle detection " is referred to by the side such as image procossing, machine learning
Method determines the position of vehicle in the picture, and it is widely used in traffic detection system or transport information
Acquisition system.Term " vehicle part " refer on vehicle position generally fix it is immutable and
Part with identifiability, for example：Car light, car plate, logo, tire, preceding cover, car
Top, case cover, car door, rearview mirror, side-view mirror etc..Term " cell " or " CELL "
Cell involved in image processing method is referred to, this method can divide the image into M*N
Individual cell, each cell is also referred to as a CELL.
Generally, this disclosure relates to which the predetermined fraction (or whole) in image is divided into M*N
Individual CELL, each CELL size can be designated as w*h.In one embodiment, it is assumed that
Interesting part in image is 640*480 resolution ratio, and fixed CELL size is 16*8,
That is w=16, h=8, then can be calculated M=640/16=40, N=480/8=80.Certainly, originally
Open not limited to this.Then, Sobel edges are all utilized to each pixel I (i, j) in CELL
Detection algorithm calculates pixel energy value, and by the pixel energy of all pixels in same CELL
Value is summed, referred to as CELL energy, and its calculation formula is as follows：
Wherein, EcellCELL energy value is represented, Sobel () represents Sobel rim detection letters
Number (operator).Then, all CELL of the energy value with more than preset value are indicated,
To determine whether CELL in flakes, if then thinking that the CELL in flakes is vehicle
Candidate.
Next, some embodiments of the present disclosure will be described in detail with reference to accompanying drawing.
First embodiment
Fig. 1 is to show the vehicle part recognition methods 100 according to the first embodiment of the present disclosure
Example flow diagram.Fig. 2 is to show to be set according to the identification of the vehicle part of the first embodiment of the present disclosure
Standby 200 configuration block diagram.In the present embodiment, method 100 can for example as shown in Figure 2
Performed by vehicle part identification equipment 200.More generally however, this method 100 can also
Performed by one or more of the other computing device, to realize same function.In addition, vehicle
In element identification device 200 included modules can be single hardware module (for example,
By field programmable gate array (FPGA), PLD (PLD), special integrated
The hardware module that circuit (ASIC) etc. is realized), single software module is (for example, by leading to
The software module realized with computing device software code), single firmware module (for example,
Special equipment with embedded code), and/or one or more of any combination in them.
For example, in certain embodiments, segmentation module 210, energy computation module 220 and car
Part determining module 230 can all be by the vehicle part identification equipment as all-purpose computer
The software module that 200 CPU (CPU) is performed, they can be used as software generation
Code is stored in the memory of vehicle part identification equipment 200, to cause to make when CPU is performed
It is changed into realize that vehicle part recognizes work(for the vehicle part identification equipment 200 of all-purpose computer
The special equipment of energy.
Next, being described in detail Fig. 1 and Fig. 2 is combined according to the first embodiment of the present disclosure
Vehicle part identifying schemes.
As shown in figure 1, for recognizing that the method 100 of the vehicle part in image can be wrapped at least
Include following some steps.
Step S110：At least a portion of image to be detected can be divided into by segmentation module 210
Multiple cells.In this example, at least a portion image can be in the middle of such as image
1/3 height region, i.e. area-of-interest.Area-of-interest is smaller, then execution efficiency is higher.
But, area-of-interest should travel whole possible paths comprising vehicle.Certainly, the disclosure is not
It is limited to this.Before at least a portion of segmentation figure picture, can determining unit lattice first size,
That is CELL size：W (width) and h (height).The size of each cell can phase
Together.The determination of the size plays a key effect for execution efficiency, the effect of subsequent processes.One
As for, w and h determination can be drawn by empirical statistics, while taking into account efficiency, effect
Compromise.
Step S120：It can calculate each single in the plurality of cell by energy computation module 220
The energy value of first lattice.More specifically, energy computation module 220 can be directed in each cell
Each pixel I (i, j) perform Sobel edge detection algorithms, to obtain the energy value of each pixel.
Then, for each cell, energy computation module 220 is by all pixels in the cell
Energy value is summed, to obtain the energy value of the cell, and its calculation formula is as follows：
Wherein, EcellThe energy value of the cell (CELL) is represented, Sobel () represents Sobel
Edge indicator function, also referred to as Sobel operators.In addition, in order that obtaining the energy of different units lattice
Value has a unified standard, it is therefore desirable to which the energy value of each cell is normalized
Processing.
Sobel edge detection algorithms are mainly used as the rim detection in image recognition.Technically,
It is a discrete type difference operator, for the gray approximation of computing brightness of image function.In figure
The operator is applied at any one pixel as in, corresponding gray scale vector or method will be all produced
Vector.Sobel operators are included for laterally (horizontal direction) and longitudinal (vertical direction)
Two matrixes, be respectively：
| -1 | 0 | +1 |
| -2 | 0 | +2 |
| -1 | 0 | +1 |
With
| +1 | +2 | +1 |
| 0 | 0 | 0 |
| -1 | -2 | -1 |
The gray value (brightness value) of they and image pixel and adjacent pixel is done into matrix multiplication,
The brightness difference approximation G of transverse direction and longitudinal direction can be drawn respectivelyxAnd Gy.It is then possible to logical
Below equation is crossed to obtain the gray value G (that is, the energy value of the pixel) of the pixel：
Certainly, disclosure not limited to this.Other edge detection algorithms can also be used to calculate picture
The energy value of element.It is for instance possible to use such as Prewitt operators, Roberts Cross operators
Etc. the energy value of each pixel in image is calculated, then by all pictures in same cell
The energy value summation of element, to determine the energy value of the cell.
In the step s 120, formula is passed through
After the energy value for having obtained the cell, it can also be normalized.
Step S130：Can be true according to the energy value of cell by vehicle part determining module 230
Determine whether include vehicle part in image.
In one embodiment, the quantity of Set cell that can be in image determines car
Part.Such as intelligent transportation bayonet socket, intelligent monitoring, the image of the monitor area photographed are every
One frame is carried out following operation, and multiple Set cells are there are in image once detecting
Target area, then can determine now have vehicle to pass through monitor area.I.e. step 130 is specifically included：
Step S130a：It can be determined to have by vehicle part determining module 230 and be higher than first threshold
Energy value all cells as Set cell, judge in image whether to include by connecting
Multiple Set cells composition target area, if then regarding the target area as vehicle
Part.In the present embodiment, first threshold can be 0.04.However, disclosure not limited to this,
Other threshold values can also be used.
Wherein it is possible to utilize the Set cell of " connection " to determine whether there is target area.
In one example, can by taking the Set cell for four connections being distributed into " # " font as an example
Determined with the bwlabel functions by using such as MATLAB softwares in binary picture
The number of four UNICOM's cells.So, it may be determined which CELL can form area in flakes
Domain, it is final it is interested be region in flakes CELL.
In addition, holding confusing image section to exclude some, in step S130a, such as
Fruit in the horizontal direction and/or vertically adjacent Set cell number be more than it is predetermined
Second Threshold and/or the 3rd threshold value, then can be defined as vehicle portion by multiple cells of the connection
Part.The Second Threshold and/or the 3rd threshold value can be determined according to the size of headstock in image.
For example, the unit that Second Threshold can in the horizontal direction be included with car plate in image or car light
The number of lattice is corresponding, and the 3rd threshold value can be with car plate in image or car light in vertical direction
Comprising cell number it is corresponding.So as to, in the case of identification car plate or car light,
It will can be foreclosed with the connected unit lattice that car plate or car light size are not consistent, further to carry
Rise recognition efficiency and the degree of accuracy.In addition, in certain embodiments, corresponding to car plate and/or car light
The height of target area can be the 1/2 of headstock height, and width can be headstock width
1/3。
In another embodiment, when very many cells have been divided in the picture, by
The problem of light, shooting angle, it is possible to which the vehicle part in the image photographed is corresponding
Set cell is not full communicating, but be spaced apart.I.e. step 130 is specifically included：
Step S130b：It can be determined to have by vehicle part determining module 230 and be higher than first threshold
Energy value all cells as Set cell, and judge in image whether to include target
Cell density is more than the target area of preset value, if it is regard the target area as vehicle
Part.In the present embodiment, first threshold can be 0.04.However, disclosure not limited to this,
Other threshold values can also be used.
In yet another embodiment, whether including in presumptive area that can be in image is pre-
The Set cell of fixed number amount, to determine whether there is vehicle part in presumptive area.Due to a lot
Monitoring probe, such as intelligent transportation bayonet socket, intelligent monitoring, are all directed towards fixed position and do not rotate
, therefore photograph in the picture by the vehicle of monitor area, the position of vehicle, towards all
It is fixed.Therefore the target area of vehicle, one can be determined by the picture analyzed in advance
When the quantity that denier detects the Set cell in target area is more than preset value, it is possible to it is determined that
Now there is vehicle to pass through monitor area.I.e. step 130 is specifically included：
Step S130c：It can be determined to have by vehicle part determining module 230 and be higher than first threshold
Energy value all cells as Set cell, and judge default target area in image
Whether the quantity of the Set cell in domain is more than preset value, if it is, in the target area
There is vehicle part.In the present embodiment, first threshold can be 0.04.However, the disclosure is not
It is limited to this, it would however also be possible to employ other threshold values.
The cell that energy value is more than into first threshold judges pre- in image as Set cell
If whether the quantity of the Set cell in region is more than preset value, if it is, the target area
There is vehicle part in domain., can be according to each part of headstock in the picture in step S130c
Size determines the size of target area.For example, car light height can be the 1/2 of headstock height,
Width can be the 1/3 of headstock width.And the size of car plate can be actual in image according to car plate
Size is determined.
So, by using the method 100 and relevant device 200 of the present embodiment, it becomes possible to
It is quick in image, robustly detect and position one or more vehicle parts.
Second embodiment
However, when a certain part of the vehicle in video image is identified, if the portion
Part is blocked or when partial occlusion or its unobvious feature, and some parts may be caused not deposit
.In order to tackle the problem, it is proposed that determined according to the vehicle peccancy of the second embodiment of the present disclosure
Scheme.
Fig. 3 is to show to determine method 300 according to the vehicle peccancy of the second embodiment of the present disclosure
Example flow diagram.Fig. 4 is to show to be determined to set according to the vehicle peccancy of the second embodiment of the present disclosure
Standby 400 configuration block diagram.In the present embodiment, method 300 can for example as shown in Figure 4
Vehicle peccancy is determined performed by equipment 400.More generally however, this method 300 can also
Performed by one or more of the other computing device, to realize same function.In addition, vehicle
It is violating the regulations determine modules included in equipment 400 can be single hardware module (for example,
By field programmable gate array (FPGA), PLD (PLD), special integrated
The hardware module that circuit (ASIC) etc. is realized), single software module is (for example, by leading to
The software module realized with computing device software code), single firmware module (for example,
With the special of embedded code), and/or one or more of any combination in them.
For example, in certain embodiments, vehicle part identification module 410, the first determining module
420 and second determining module 430 can all be by being determined as the vehicle peccancy of all-purpose computer
The software module that the CPU (CPU) of equipment 400 is performed, they can be as soft
Part code is stored in the memory that vehicle peccancy determines equipment 400, to make when CPU is performed
It must determine that equipment 400 is changed into realize that vehicle peccancy is true as the vehicle peccancy of all-purpose computer
Determine the special equipment of function.
Next, being described in detail Fig. 3 and Fig. 4 is combined according to the second embodiment of the present disclosure
Vehicle peccancy determines scheme.
As shown in figure 3, for determining that method 300 that whether vehicle in image break rules and regulations can be down to
Include following some steps less.
Step S310：The first car in image can be recognized by vehicle part identification module 410
Part.In the present embodiment, the first vehicle part can be according in aforementioned first embodiment
The vehicle part recognition methods of description is recognized.Certainly, disclosure not limited to this.Can also
The first vehicle part is recognized using other vehicle part recognition methods.For example, described above
Using other edge detection operators (for example, Prewitt operators, Roberts Cross operators)
Method.In the present embodiment, first it is detected that the first vehicle part can be left car light or the right side
Car light.
Step S320：It is then possible to by the first determining module 420 according at least include this first
It is predetermined between multiple vehicle parts including vehicle part and another second vehicle part
Position relationship, to determine to whether there is second vehicle part in image.In this example,
Two vehicle parts can be car plate.
For example, multiple parts of vehicle can be determined according to priori (for example, vehicle)
Between position relationship.More specifically, for example first it is detected that left car light, then according to left car
Position relationship (relative direction and/or relative distance) between lamp and other vehicle parts is determined
Right car light and car plate position whether there is corresponding part., can be with addition, in another example
First it is detected that right car light or car plate, it is then determined that whether other positions have corresponding component.
Step S330：If there is no the second vehicle part, then the vehicle in image can be determined
It is violating the regulations.For example, when the position of car plate in the picture is determined according to left car light and/or right car light,
If not finding car plate on the position, can be determined to be block number plate etc. it is similar disobey
Zhang Hangwei.Similarly, when being found that car plate first, then on the position that car light should occur
And be not detected by car light, then it can determine that the car car light is damaged, without car light etc., this is equally
Belong to violating the regulations.
In addition, determining the first vehicle portion using the cell method similar to first embodiment
In the case of part, same cell method can also be used to determine depositing for the second vehicle part
Property.If for example, the vehicle not formed on predetermined car plate position by connected unit lattice
Candidate, then it is assumed that block number plate or without number plate.
So, set by using the determination method violating the regulations according to the second embodiment of the present disclosure and correspondingly
It is standby, realize that a kind of efficiency is very high and the extraordinary detection scheme of robustness, can quickly examine
Whether rope occurs blocking the situation violating the regulations such as number plate or unlicensed car into video flowing.
3rd embodiment
It is probably the various different shapeds such as car, SUV, lorry to appear in the vehicle in video image
Number vehicle, then the layout of its each vehicle part be all not quite similar.Asked to solve this
Topic, it is proposed that according to the vehicle identification scheme of the third embodiment of the present disclosure.
Fig. 5 is the example for showing the vehicle identification method 500 according to the third embodiment of the present disclosure
Flow chart.Fig. 6 is to show the vehicle identification equipment 600 according to the third embodiment of the present disclosure
Configure block diagram.In the present embodiment, the vehicle identification that method 500 can for example as shown in Figure 6
Performed by equipment 600.More generally however, this method 500 can also be by one or more
Other computing devices are performed, to realize same function.In addition, vehicle identification equipment 500
In included modules can be single hardware module (for example, by field-programmable
Gate array (FPGA), PLD (PLD), application specific integrated circuit (ASIC)
Deng the hardware module realized), single software module is by general processor (for example, performed soft
The software module that part code is realized), single firmware module is (for example, with embedded code
It is special), and/or one or more of any combination in them.
For example, in certain embodiments, vehicle part identification module 610, decision content calculate mould
Block 620 and vehicle determining module 630 can all be by being set as the vehicle identification of all-purpose computer
The software module that standby 600 CPU (CPU) is performed, they can be used as software
Code is stored in the memory of vehicle identification equipment 600, using when CPU is performed so that as
The vehicle identification equipment 600 of all-purpose computer is changed into realize that the special of vehicle identification function sets
It is standby.
Next, being described in detail Fig. 5 and Fig. 6 is combined according to the third embodiment of the present disclosure
Vehicle identification scheme.
As shown in figure 5, for recognize the vehicle in image method 500 can at least include with
Under some steps.
Step S510：Each car in image can be recognized by vehicle part identification module 610
Part.In the present embodiment, these vehicle parts can be according in aforementioned first embodiment
The vehicle part recognition methods of description is recognized.Certainly, disclosure not limited to this.Can also
The first vehicle part is recognized using other vehicle part recognition methods.For example, described above
Using other edge detection operators (for example, Prewitt operators, Roberts Cross operators)
Method.In the present embodiment, these vehicle parts can include at least one of following：Left car light,
Right car light and/or car plate.
Step S520：One or more vehicles can be based upon by decision content computing module 620
The pre-assigned weight of part, calculates the decision content for indicating to whether there is vehicle in image.At this
In embodiment, decision content can be the weighted value sum of one or more vehicle parts.For example,
The weight of left car light can be 0.25, and the weight of right car light can be 0.25, and the weight of car plate
Can be 0.5.However, disclosure not limited to this, or these vehicle parts are assigned not
Same weight., can be relative to for example, in the case of it is determined that car light recognition correct rate is higher
Car plate improves the weight of car light, and vice versa.
Step S530：The feelings of threshold value can be more than or equal in decision content by vehicle determining module 630
The object corresponding with one or more vehicle parts is defined as vehicle under condition.In this implementation
In example, threshold value can be 0.5, as long as finding two car lights in the picture or finding one
Car plate, it is possible to determine there is vehicle in image.Certainly, disclosure not limited to this, but it is real
Any other appropriate threshold can be used on border.For example, higher in car light recognition correct rate
In the case of, threshold value can be adjusted downward to 0.25, with the case where detecting a car light, just
It is determined that vehicle is detected, while can be as described in second embodiment, by detecting
Car light determine that car plate whether there is.
Further, since relaxing for above-mentioned decision condition, may result in the rising of false drop rate.Cause
This, can also use the offline SVM classifier of headstock model training to further determine that by upper
Whether the vehicle that the method for stating is detected is genuine vehicle, or can be determined by manually appraising and deciding
Whether the correctness of detection to be, and is fed back, and adjusts the distribution of weighted value and/or the setting of threshold value,
To strengthen judgment accuracy.
, can by using the vehicle identification method and corresponding device according to the third embodiment of the present disclosure
Efficiently and accurately to recognize the vehicle in image for different automobile types.
In addition, according to the disclosure, some steps of each above-mentioned method can be performed individually or group
Close and perform, and can perform parallel or order execution, it is not limited to be shown in figure specific
Operation order.Although, should in addition, respectively describe three different embodiments above
It is appreciated that：Part and/or all technical characteristic in these embodiments can use appropriate mode
It is bonded to each other, to form another embodiment, the embodiment equally also falls into the scope of the present disclosure.
For example, the vehicle peccancy shown in above-mentioned Fig. 4 determines the vehicle part identification module 410 of equipment 400
Vehicle part identification module 610 with the vehicle identification equipment 600 shown in Fig. 6 can be exactly to scheme
Vehicle part identification equipment 200 shown in 2.
So far the disclosure is described combined preferred embodiment.It should be understood that ability
Field technique personnel in the case where not departing from spirit and scope of the present disclosure, can carry out it is various its
Its change, replacement and addition.Therefore, the scope of the present disclosure is not limited to above-mentioned particular implementation
Example, and should be defined by the appended claims.
Claims (9)
1. a kind of method for being used to recognize the vehicle part in image, it is characterised in that including：
At least a portion of image is divided into multiple cells；
For the presetted pixel in each cell, its energy value is calculated using edge detection algorithm,
And the energy value of the cell is obtained according to the energy value of the presetted pixel in the cell；And
The energy value of cell in image determines whether include vehicle part in described image.
2. according to the method described in claim 1, it is characterised in that the vehicle part is car
Lamp and/or car plate.
3. method according to claim 1 or 2, it is characterised in that described according to cell
Energy value determine whether specifically included in described image including vehicle part：Energy value is more than
Whether the cell of first threshold is judged in image including many by what is connected as Set cell
The target area of individual Set cell composition, if then regarding the target area as vehicle part.
4. method according to claim 1 or 2, it is characterised in that described according to cell
Energy value determine whether specifically included in described image including vehicle part：Energy value is more than
The cell of first threshold judges whether include having including many in image as Set cell
The target area of individual Set cell, if it is regard the target area as vehicle part.
5. method according to claim 1 or 2, it is characterised in that described according to cell
Energy value determine whether specifically included in described image including vehicle part：Energy value is more than
The cell of first threshold judges whether include object element in image as Set cell
Lattice density is more than the target area of preset value, if it is has vehicle part in the target area.
6. according to the method described in claim 1, it is characterised in that also include：
The position relationship between the vehicle part in described image is determined according to priori；
Determine have behind the target area of vehicle part in the picture, image is judged according to position relationship
In whether have other corresponding vehicle parts.
7. a kind of method for recognizing the vehicle in image, including：
One in described image is recognized using the method any one of claim 1~6
Or multiple vehicle parts；
It is based upon the pre-assigned weight of one or more of vehicle parts, calculates described in indicating
It whether there is the decision content of vehicle in image；And
If the decision content is more than or equal to threshold value, will be with one or more of vehicle parts
Corresponding object is defined as vehicle.
8. a kind of equipment for being used to recognize the vehicle part in image, it is characterised in that including：
Split module, at least a portion of image to be divided into multiple cells；
Energy computation module, for for the presetted pixel in each cell, being examined using edge
Method of determining and calculating calculates its energy value, and is somebody's turn to do according to the energy value of the presetted pixel in the cell
The energy value of cell；And
Vehicle part determining module, the energy value for the cell in image determines described
Whether include vehicle part in image.
9. a kind of equipment for recognizing the vehicle in image, including：
Vehicle part identification module, for utilizing the method any one of claim 1~6
To recognize one or more of described image vehicle part；
Decision content computing module, is allocated in advance for being based upon one or more of vehicle parts
Weight, calculate indicate described image in whether there is vehicle decision content；And
Vehicle determining module, if for the decision content be more than or equal to threshold value, will with it is described
The corresponding object of one or more vehicle parts is defined as vehicle.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610091338.1A CN107092855A (en) | 2016-02-18 | 2016-02-18 | Vehicle part recognition methods and equipment, vehicle identification method and equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610091338.1A CN107092855A (en) | 2016-02-18 | 2016-02-18 | Vehicle part recognition methods and equipment, vehicle identification method and equipment |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN107092855A true CN107092855A (en) | 2017-08-25 |
Family
ID=59648648
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610091338.1A Pending CN107092855A (en) | 2016-02-18 | 2016-02-18 | Vehicle part recognition methods and equipment, vehicle identification method and equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107092855A (en) |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110909781A (en) * | 2019-11-14 | 2020-03-24 | 长安大学 | Vehicle detection method based on vehicle rearview mirror |
| CN113158758A (en) * | 2021-02-07 | 2021-07-23 | 中国联合网络通信集团有限公司 | Method, system, equipment and storage medium for judging illegal use of vehicle license plate |
Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101251927A (en) * | 2008-04-01 | 2008-08-27 | 东南大学 | Vehicle detecting and tracing method based on video technique |
| CN101630361A (en) * | 2008-12-30 | 2010-01-20 | 北京邮电大学 | Plate number, body color and mark identification-based equipment and plate number, body color and mark identification-based method for identifying fake plate vehicles |
| CN102867416A (en) * | 2012-09-13 | 2013-01-09 | 中国科学院自动化研究所 | Vehicle part feature-based vehicle detection and tracking method |
| CN103324935A (en) * | 2013-06-27 | 2013-09-25 | 中山大学 | Method for vehicle positioning and region segmenting in image |
| CN103488973A (en) * | 2013-09-12 | 2014-01-01 | 上海依图网络科技有限公司 | Method and system for recognizing vehicle brand based on image |
| CN104200213A (en) * | 2014-08-12 | 2014-12-10 | 合肥工业大学 | Vehicle detection method based on multiple parts |
| CN104350510A (en) * | 2012-06-14 | 2015-02-11 | 国际商业机器公司 | Multi-cue object detection and analysis |
| CN104792792A (en) * | 2015-04-27 | 2015-07-22 | 武汉武大卓越科技有限责任公司 | Stepwise-refinement pavement crack detection method |
| CN105260701A (en) * | 2015-09-14 | 2016-01-20 | 中电海康集团有限公司 | Front vehicle detection method applied to complex scene |
-
2016
- 2016-02-18 CN CN201610091338.1A patent/CN107092855A/en active Pending
Patent Citations (9)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101251927A (en) * | 2008-04-01 | 2008-08-27 | 东南大学 | Vehicle detecting and tracing method based on video technique |
| CN101630361A (en) * | 2008-12-30 | 2010-01-20 | 北京邮电大学 | Plate number, body color and mark identification-based equipment and plate number, body color and mark identification-based method for identifying fake plate vehicles |
| CN104350510A (en) * | 2012-06-14 | 2015-02-11 | 国际商业机器公司 | Multi-cue object detection and analysis |
| CN102867416A (en) * | 2012-09-13 | 2013-01-09 | 中国科学院自动化研究所 | Vehicle part feature-based vehicle detection and tracking method |
| CN103324935A (en) * | 2013-06-27 | 2013-09-25 | 中山大学 | Method for vehicle positioning and region segmenting in image |
| CN103488973A (en) * | 2013-09-12 | 2014-01-01 | 上海依图网络科技有限公司 | Method and system for recognizing vehicle brand based on image |
| CN104200213A (en) * | 2014-08-12 | 2014-12-10 | 合肥工业大学 | Vehicle detection method based on multiple parts |
| CN104792792A (en) * | 2015-04-27 | 2015-07-22 | 武汉武大卓越科技有限责任公司 | Stepwise-refinement pavement crack detection method |
| CN105260701A (en) * | 2015-09-14 | 2016-01-20 | 中电海康集团有限公司 | Front vehicle detection method applied to complex scene |
Non-Patent Citations (1)
| Title |
|---|
| 王呈鹏等: "基于空间关联特征分布的目标跟踪", 《计算机工程与设计》 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110909781A (en) * | 2019-11-14 | 2020-03-24 | 长安大学 | Vehicle detection method based on vehicle rearview mirror |
| CN113158758A (en) * | 2021-02-07 | 2021-07-23 | 中国联合网络通信集团有限公司 | Method, system, equipment and storage medium for judging illegal use of vehicle license plate |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| US10025998B1 (en) | Object detection using candidate object alignment | |
| US8902053B2 (en) | Method and system for lane departure warning | |
| US9552524B2 (en) | System and method for detecting seat belt violations from front view vehicle images | |
| US11284062B2 (en) | Automotive display validation | |
| US9070023B2 (en) | System and method of alerting a driver that visual perception of pedestrian may be difficult | |
| US20220004818A1 (en) | Systems and Methods for Evaluating Perception System Quality | |
| CN105374013B (en) | Image visibility restorative procedure and image processing apparatus based on double dark channel priors | |
| US20140177925A1 (en) | License plate recognition system and method | |
| US10878259B2 (en) | Vehicle detecting method, nighttime vehicle detecting method based on dynamic light intensity and system thereof | |
| TWI640964B (en) | Image-based vehicle counting and classification system | |
| CN104408475B (en) | A kind of licence plate recognition method and car license recognition equipment | |
| CN107563299B (en) | Pedestrian detection method using RecNN to fuse context information | |
| CN107004137B (en) | System and method for object detection | |
| Kim et al. | Autonomous vehicle detection system using visible and infrared camera | |
| CN103530640A (en) | Unlicensed vehicle detection method based on AdaBoost and SVM (support vector machine) | |
| CN105740751A (en) | Object detection and identification method and system | |
| CN109313699A (en) | The device and method for carrying out target identification for the input picture to a vehicle | |
| CN104115203A (en) | Three-dimensional object detection device | |
| Mammeri et al. | North-American speed limit sign detection and recognition for smart cars | |
| EP3726421A2 (en) | Recognition method and apparatus for false detection of an abandoned object and image processing device | |
| CN104361366B (en) | A kind of licence plate recognition method and car license recognition equipment | |
| Choi et al. | Fog detection for de-fogging of road driving images | |
| Skodras et al. | Rear lights vehicle detection for collision avoidance | |
| US10115028B2 (en) | Method and device for classifying an object in an image | |
| CN107092855A (en) | Vehicle part recognition methods and equipment, vehicle identification method and equipment |
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 |