CN106909937A - Traffic lights recognition methods, control method for vehicle, device and vehicle - Google Patents
Traffic lights recognition methods, control method for vehicle, device and vehicle Download PDFInfo
- Publication number
- CN106909937A CN106909937A CN201710070814.6A CN201710070814A CN106909937A CN 106909937 A CN106909937 A CN 106909937A CN 201710070814 A CN201710070814 A CN 201710070814A CN 106909937 A CN106909937 A CN 106909937A
- Authority
- CN
- China
- Prior art keywords
- region
- traffic lights
- signal lamp
- image
- candidate
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
- G06F18/241—Classification techniques relating to the classification model, e.g. parametric or non-parametric approaches
-
- 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
- G06V10/25—Determination of region of interest [ROI] or a volume of interest [VOI]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/50—Extraction of image or video features by performing operations within image blocks; by using histograms, e.g. histogram of oriented gradients [HoG]; by summing image-intensity values; Projection analysis
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/56—Extraction of image or video features relating to colour
Abstract
This disclosure relates to a kind of traffic lights recognition methods, control method for vehicle, device and vehicle, the traffic lights recognition methods includes：The Chinese herbaceous peony image of Real-time Collection vehicle；Determine the area-of-interest of Chinese herbaceous peony image；The image of the area-of-interest is transformed into HSI spaces by rgb space, to obtain the HSI component maps of the area-of-interest；S components to the HSI component maps carry out threshold segmentation, to be partitioned into traffic lights region；The traffic lights region is classified using default grader, obtains the information of traffic lights.By the technical scheme of the disclosure, signal lamp recognition effect is not good under can solving the problems, such as the complex illumination environment such as night, improves adaptability of the image processing algorithm to photoenvironment, improves the accuracy of signal lamp identification；Additionally, the traffic lights region obtained after to segmentation is classified, it is possible to achieve the identification to the signal lamp of the plurality of classes such as circular signal lamp, arrow-shaped signal lamp.
Description
Technical field
This disclosure relates to technical field of vehicle control, in particular it relates to a kind of traffic lights recognition methods, wagon control
Method, device and vehicle.
Background technology
With the high speed development of car social economy, transport need increasingly increases, and observes road by people and drives the tradition of vehicle
The limitation of wagon control mode is increasingly apparent, in order to comply with the new trend of advanced information society's vehicle development, the intelligence of vehicle
Driving technology is arisen at the historic moment.The identification of traffic lights, plays the role of important in the intelligent driving technology of vehicle.
Recognized in the signal lamp of correlation technique, mostly both for the knowledge of simple background or the signal lamp of single type
Not, i.e., according to inherent characteristicses such as luminous, black backboard, the color fixations of Chinese herbaceous peony image and signal lamp for collecting, by color
Or brightness carries out region segmentation and matching to obtain the status information of traffic lights.But, this mode is in external illumination
Environment is dark, too small area of the signal lamp in Chinese herbaceous peony image, hypertelorism, imaging be not complete or the complicated bar such as signal lamp dimness
Easily there is identification mistake under part and the problems such as recognizing difficulty easily occurs in the signal lamp for arrowhead form.
The content of the invention
In order to overcome problem present in correlation technique, the disclosure to provide a kind of traffic lights recognition methods, vehicle control
Method processed, device and vehicle.
In a first aspect, the disclosure provides a kind of traffic lights recognition methods, including：
Alternatively, the Chinese herbaceous peony image of Real-time Collection vehicle；Determine the area-of-interest of the Chinese herbaceous peony image；The sense is emerging
The image in interesting region is transformed into HSI spaces by rgb space, to obtain the HSI component maps of the area-of-interest；To the HSI
The S components of component map carry out threshold segmentation, to be partitioned into traffic lights region；Using default grader to described
Traffic lights region is classified, and obtains the information of traffic lights.
Alternatively, the step of area-of-interest of the determination Chinese herbaceous peony image includes：Obtain the vehicle and traffic
The distance of signal lamp；According to the image collecting device of the distance, the default setting height(from bottom) of traffic lights and the vehicle
Parameter, determine the area-of-interest of the Chinese herbaceous peony image.
Alternatively, the image by the area-of-interest is transformed into HSI spaces by rgb space, to obtain the sense
The step of HSI component maps in interest region, includes：
The image of the area-of-interest is changed using below equation：
I=max
R, G and B are respectively the R component of the pixel of area-of-interest, G components and B component, max be in R, G and B most
Big value, min is the minimum value in R, G and B.
Alternatively, the S components to the HSI component maps carry out threshold segmentation, to be partitioned into traffic letter
The step of signal lamp region, includes：Pixel of the S components horizontal gradient more than predetermined gradient threshold value is filtered out from the HSI component maps
Point is used as candidate point；Mask window is set up by search source point of the candidate point；When the variance of the S components of the mask window surpasses
When crossing default variance threshold values, dynamic threshold segmentation is carried out to the mask window, generate bianry image；To in the bianry image
Connected region carry out shape constraining, extract candidate signal lamp region；Black backboard is carried out to the candidate signal lamp region
Constraint, extracts traffic lights region.
Alternatively, the S components to the HSI component maps carry out threshold segmentation, to be partitioned into traffic letter
The step of signal lamp region, also includes：Obtain the size in the traffic lights region；According to the chi in the traffic lights region
Very little, the size to the mask window is updated.
Alternatively, the connected region in the bianry image carries out shape constraining, extracts candidate signal lamp area
The step of domain, includes：Region growth is carried out to the connected region in the bianry image according to the default rule that increases；Increase from region
The aspect ratio that minimum enclosed rectangle frame is extracted in connected region after length is less than pre-set aspect ratio threshold value and minimum enclosed rectangle
The size of frame meets the connected region of pre-set dimension scope, used as candidate's connected region；Using the Hu, bending moment does not extract described
The characteristic value of candidate's connected region；Obtain the geneva of the characteristic value of candidate's connected region and the characteristic value of signal lamp template away from
From；The connected region that the mahalanobis distance meets preset threshold range is extracted from candidate's connected region, as described
Candidate signal lamp region.
Alternatively, it is described that black backboard constraint is carried out to the candidate signal lamp region, extract traffic lights region
The step of include：Extend rule according to default, the candidate signal lamp region is extended；Obtain the candidate signal after extending
I component projection in lamp region；The candidate signal lamp region that I component projection meets default constraints is believed as traffic
Signal lamp region.
Alternatively, described information includes color state and classification；
It is described the traffic lights region is classified using default grader, obtain the information of traffic lights
Step includes：Extract the color histogram feature and histograms of oriented gradients feature in the traffic lights region；Using first
Default grader is classified to the color histogram feature, obtains the color state of traffic lights；It is default using second
Grader is classified to the histograms of oriented gradients feature, obtains the classification of traffic lights.
Second aspect, the disclosure provides a kind of control method for vehicle, including：Traffic lights are obtained according to the above method
Information；According to the Lane detection result of the vehicle for getting, the traveling lane and travel direction of the vehicle are determined；Root
According to the information of the traveling lane, the travel direction and the traffic lights, determine that the corresponding target of the vehicle is handed over
Ventilating signal lamp information；According to the target traffic lights information, thermoacoustic prime engine is carried out to the vehicle.
The third aspect, the disclosure provides a kind of traffic signals light identifier, including：Image capture module, for real-time
The Chinese herbaceous peony image of collection vehicle；Area-of-interest determining module, the area-of-interest for determining the Chinese herbaceous peony image；Image turns
Mold changing block, for the image of the area-of-interest to be transformed into HSI spaces by rgb space, to obtain the area-of-interest
HSI component maps；Image segmentation module, threshold segmentation is carried out for the S components to the HSI component maps, to divide
Cut out traffic lights region；Sort module, for being classified to the traffic lights region using default grader, is obtained
Take the information of traffic lights.
Alternatively, the area-of-interest determining module includes：Apart from acquisition submodule, for obtaining the vehicle and handing over
The distance of ventilating signal lamp；Area-of-interest determination sub-module, for according to the distance, the default setting height(from bottom) of traffic lights
And the parameter of the image collecting device of the vehicle, determine the area-of-interest of the Chinese herbaceous peony image.
Alternatively, described image modular converter includes：
Image transform subblock, for being changed the image of the area-of-interest using below equation：
I=max
R, G and B are respectively the R component of the pixel of area-of-interest, G components and B component, max be in R, G and B most
Big value, min is the minimum value in R, G and B.
Alternatively, described image segmentation module includes：Candidate point screens submodule, for being sieved from the HSI component maps
Pixel of the S components horizontal gradient more than predetermined gradient threshold value is selected as candidate point；Mask window setting up submodule, for
The candidate point sets up mask window for search source point；Segmentation submodule, the variance for the S components when the mask window surpasses
When crossing default variance threshold values, dynamic threshold segmentation is carried out to the mask window, generate bianry image；Shape constraining submodule,
For carrying out shape constraining to the connected region in the bianry image, candidate signal lamp region is extracted；Black backboard is constrained
Submodule, for carrying out black backboard constraint to the candidate signal lamp region, extracts traffic lights region.
Alternatively, the segmentation module also includes：Size acquisition submodule, the chi for obtaining the traffic lights
It is very little；Mask window updates submodule, and for the size according to the traffic lights region, the size to the mask window is entered
Row updates.
Alternatively, the shape constraining submodule includes：Region increases submodule, for increasing regular to institute according to default
The connected region stated in bianry image carries out region growth；Candidate's connected region extracting sub-module, after increasing from region
The aspect ratio of minimum enclosed rectangle frame is extracted in connected region less than pre-set aspect ratio threshold value and the chi of minimum enclosed rectangle frame
The very little connected region for meeting pre-set dimension scope, as candidate's connected region；Hu invariant moment features value extracting sub-modules, for profit
With the Hu, bending moment does not extract the characteristic value of candidate's connected region；Mahalanobis distance acquisition submodule, for obtaining the time
Select the mahalanobis distance of the characteristic value of connected region and the characteristic value of signal lamp template；Candidate signal lamp extracted region submodule, uses
In the connected region that the mahalanobis distance meets preset threshold range is extracted from candidate's connected region, as the time
Select signal lamp region.
Alternatively, the black backboard constraint submodule includes：Region extends submodule, for being advised according to default extension
Then, the candidate signal lamp region is extended；Projection acquisition submodule, for obtaining the candidate signal lamp region after extending
Interior I component projection；Candidate signal lamp extracted region submodule, for I component projection to be met into default constraints
Candidate signal lamp region is used as traffic lights region.
Alternatively, described information includes color state and classification；
The sort module includes：Feature extraction submodule, the color histogram for extracting the traffic lights region
Figure feature and histograms of oriented gradients feature；Color state acquisition submodule, for using default first grader to the face
Color Histogram feature carries out classification based training, obtains the color state of traffic lights；Classification acquisition submodule, for using default
Second grader carries out classification based training to the histograms of oriented gradients feature, obtains the classification of traffic lights.
Fourth aspect, the disclosure provides a kind of controller of vehicle, including：Signal information acquisition module, for passing through
Above-mentioned traffic signals light identifier obtains the information of traffic lights；Traveling lane and travel direction acquisition module, for root
According to the Lane detection result of the vehicle for getting, the traveling lane and travel direction of the vehicle are obtained；Target traffic
Signal lamp determining module, for the information according to the traveling lane, the travel direction and the traffic lights, it is determined that
The corresponding target traffic lights information of the vehicle；Thermoacoustic prime engine module, for according to the target traffic lights information,
Thermoacoustic prime engine is carried out to the vehicle.
5th aspect, the disclosure provides a kind of vehicle, including above-mentioned traffic signals light identifier.
By the technical scheme of the embodiment of the present disclosure, turned by RGB color by the Chinese herbaceous peony image of the vehicle for collecting
HSI spaces are turned to, and the S components of the HSI component maps to being obtained after conversion carry out threshold segmentation, can solve complexity
The not good problem of signal lamp recognition effect under photoenvironment, improves adaptability of the image processing algorithm to illumination and environment etc.,
Improve the accuracy of signal lamp identification；Additionally, the traffic lights region obtained after to segmentation is classified, Ke Yishi
The identification of existing polytype signal lamp, obtains the information of traffic lights.
Other feature and advantage of the disclosure will be described in detail in subsequent specific embodiment part.
Brief description of the drawings
Accompanying drawing is, for providing further understanding of the disclosure, and to constitute the part of specification, with following tool
Body implementation method is used to explain the disclosure together, but does not constitute limitation of this disclosure.In the accompanying drawings：
Fig. 1 is the flow chart of the traffic lights recognition methods of the embodiment of the disclosure one；
Fig. 2 is the schematic view of the mounting position of the image collecting device of the vehicle of the embodiment of the disclosure one；
Fig. 3 is the flow chart of the HSI image local dynamic threshold segmentation methods of the embodiment of the disclosure one；
Fig. 4 is the mask window schematic diagram of the embodiment of the disclosure one；
Fig. 5 is the flow chart of the connected region shape constraining method of the embodiment of the disclosure one；
Fig. 6 is the flow chart of the candidate signal lamp region black backboard constrained procedure of the embodiment of the disclosure one；
Fig. 7 A to Fig. 7 C are the schematic diagrames of the candidate signal lamp region extension of the embodiment of the disclosure one；
Fig. 8 is the flow chart of the HSI image local dynamic threshold segmentation methods of another embodiment of the disclosure；
Fig. 9 is the flow chart of the traffic lights territorial classification method of the embodiment of the disclosure one；
Figure 10 is the flow chart of the control method for vehicle of the embodiment of the disclosure one；
Figure 11 is the block diagram of the traffic signals light identifier of the embodiment of the disclosure one；
Figure 12 is the block diagram of the traffic signals light identifier of another embodiment of the disclosure；
Figure 13 is the block diagram of the controller of vehicle of the embodiment of the disclosure one；
Figure 14 is the block diagram for traffic signals light identifier of the embodiment of the disclosure one.
Specific embodiment
It is described in detail below in conjunction with accompanying drawing specific embodiment of this disclosure.It should be appreciated that this place is retouched
The specific embodiment stated is merely to illustrate and explains the disclosure, is not limited to the disclosure.
Fig. 1 is the flow chart of the traffic lights recognition methods of the embodiment of the disclosure one.Reference picture 1, the method include with
Lower step：
In step s 11, the Chinese herbaceous peony image of Real-time Collection vehicle.
In embodiment of the disclosure, can be by setting image collecting device (such as camera), Real-time Collection on vehicle
The Chinese herbaceous peony image of vehicle.Alternatively, image collecting device is mountable to vehicle front, as shown in Fig. 2 wherein l, d, h are respectively
Image collecting device is with respect to world coordinates OvXvYvZvPosition, can by measure or demarcate obtain；OcXcYcZcIt is image collector
Put coordinate system；θ,φ is respectively the angle of pitch, deflection and the angle of heel of image collecting device installation.
It should be noted that the parameter (such as position and setting angle) of image collecting device, can be carried out according to actual conditions
Adjustment, to obtain accurately Chinese herbaceous peony image.
In step s 12, the area-of-interest of Chinese herbaceous peony image is determined.
The top of Chinese herbaceous peony image is generally present in due to traffic lights, in order to exclude billboard, road sign, street lamp in environment
Deng interference, the region for being likely to occur traffic lights can be chosen from Chinese herbaceous peony image as area-of-interest, will be except interested
Other parts beyond region cut away.Thus, it is possible to save the process time to Chinese herbaceous peony image, traffic lights identification is improved
Efficiency, it is ensured that traffic lights identification real-time.
In one embodiment, it is contemplated that vehicle in the process of moving, distance, the upper figure of vehicle of vehicle and traffic lights
As harvester parameter and traffic lights setting height(from bottom) can be to traffic lights in Chinese herbaceous peony image image space
Influence is produced, therefore can be according to the distance of vehicle and traffic lights, the setting height(from bottom) of traffic lights and image collector
The parameter (such as position of image collecting device, setting angle and focal length) put and the area-of-interest to determine Chinese herbaceous peony image,
As shown in formula (1) and formula (2).
Wherein, R is tied to the rotation relationship matrix of image collecting device coordinate system for world coordinates；(cx, cy) in image
The heart, is the intersection point of optical axis and the plane of delineation；(fx, fy) it is the focal length of image collecting device；L, d, h are that image collecting device is relative
The position of world coordinates center origin；X arrives the distance between signal lamp for image collecting device；Y is image collecting device to letter
Lateral separation between signal lamp；Z is the setting height(from bottom) of signal lamp；θ,φ be respectively image collecting device installation the angle of pitch,
Deflection and angle of heel.
In another embodiment, it is contemplated that the position of signal lamp and the lateral attitude of vehicle may differ by it is very big, its
Image space in Chinese herbaceous peony image is likely to be at picture centre, left side or right side, therefore it is determined that only needing meter during area-of-interest
Its altitude range in Chinese herbaceous peony image is calculated, width range is not limited.Due to the deflection that image collecting device is installed
It is smaller with the lengthwise position influence that angle of heel is imaged on signal lamp, it is negligible when hunting zone is substantially determined, therefore search
Rope scope can be determined by formula (3).
Thus, adopted with the image of the distance of traffic lights, the setting height(from bottom) of traffic lights and vehicle by vehicle
The area-of-interest of the parameter determination Chinese herbaceous peony image of acquisition means, can reduce the interference of other false signal lamps in Chinese herbaceous peony image, contracting
Small image procossing scope, reduces the workload of signal lamp identification.
In one embodiment, it is further to reduce the workload that signal lamp is recognized, can sets when vehicle and signal lamp
The just area-of-interest of determination Chinese herbaceous peony image and after being carried out in area-of-interest when distance is less than or equal to predeterminable range threshold value
Continuous traffic lights identification work.
Vehicle can be obtained with the distance of traffic lights by GPS positioning device, and predeterminable range threshold value can be according to vehicle
Actual conditions be configured, for example can be 80m.
In step s 13, the image of area-of-interest is transformed into HSI spaces by rgb space, to obtain area-of-interest
HSI component maps.
Tri- component correlations of R, G and B in rgb space are higher, influenceed larger by illumination, therefore can be by region of interest
The image in domain is transformed into HSI spaces by rgb space, the HSI component maps of area-of-interest is obtained, using the H of HSI component maps
(Hue, tone), S (Saturation, saturation degree) and three separable characteristics of component of I (Intensity, brightness), can disappear
Except the influence of illumination variation.
In one embodiment, in order to reduce the influence of the factors such as track both sides leaf, blue sky, and in order to exclude not
May be the interference such as billboard of red, green, yellow three kinds of colors, the conversion formula such as formula (5) in HSI spaces, formula (6), formula (7) and formula
(8) shown in.
I=max (7)
Wherein, R, G and B are respectively the R component of the pixel of area-of-interest, G components and B component；Max is in R, G and B
Maximum；Min is the minimum value in R, G and B.
In step S14, the S components to HSI component maps carry out threshold segmentation, to be partitioned into traffic lights
Region.
Because traffic lights have autonomous luminous solid color, and there is black backboard around it, thus it is in Chinese herbaceous peony
The S components of the imaging region (traffic lights region) in image are higher, in addition in view of the luminous intensity and pure color of signal lamp
The factor such as the different and extraneous strong illumination of luminous intensity of the luminescence unit such as advertisement influence, the S in traffic lights region
Maximum in the not whole HSI images of component, therefore threshold segmentation can be carried out to the S components of HSI images, by S
The local highlight regions of the projecting pixel of component split, and obtain traffic lights region.
Referring to Fig. 3, in an embodiment of the disclosure, the S components to HSI component maps carry out threshold segmentation,
The method for being partitioned into traffic lights region is comprised the following steps：
In step S31, pixel of the S components horizontal gradient more than predetermined gradient threshold value is filtered out from HSI component maps
As candidate point.
Due to there is black backboard around traffic lights, therefore S component value of the black backboard in HSI images is smaller,
And S component value of the traffic lights in HSI images is larger (brighter areas in HSI images).
In one embodiment, the S components that can calculate often capable pixel in HSI component maps or so neighbor pixel are poor
Value (i.e. the transverse gradients value of pixel), if the S component transverse gradients values of current pixel point are smaller, i.e., less than predetermined gradient threshold value
THH, then can not possibly for black backboard to the transitional region of signal lamp, the pixel is screened out；If the S components of current pixel point are horizontal
It is larger to Grad, i.e., more than or equal to predetermined gradient threshold value THH, then the pixel may be mistake of the black backboard to signal lamp
Region is crossed, thus can be using the pixel as candidate point.Wherein, shown in the S components horizontal gradient of pixel such as formula (9).
Hx(x, y)=S (x-1, y)-S (x+1, y) (9)
Wherein, (x-1, y) is the S component values of the pixel on the left of pixel (x, y) to S, and (x+1 y) is pixel (x, y) to S
The S component values of the pixel on right side.
In step s 32, mask window is set up by search source point of candidate point.
Because candidate point may be transitional region of the black backboard to traffic lights, therefore can be search source with candidate point
Point sets up the mask window that a size is w × w, and the mask window is the region of doubtful traffic lights, as shown in figure 4,
Wherein, P0It is candidate point.
The size of mask window can according to the actual size of signal lamp and the parameter setting of image collecting device, when initial,
Desirable w=11.
In step S33, when the variance of the S components of mask window exceedes default variance threshold values, mask window is carried out
Dynamic threshold segmentation, generates bianry image.
Mask window include may be traffic lights light-emitting zone and may be black backboard non-luminous region, and
S component of the traffic lights in HSI images is larger, and S component of the black backboard in HSI images is smaller, if mask window
Include traffic lights and black backboard, the variance of its S component is larger, i.e. the fluctuation of S components is larger；Conversely, then mask windows
The variance of the S components of mouth is smaller, i.e. the fluctuation of S components is smaller.
Therefore, can by the variance of S components screen out can not possibly the mask window comprising traffic lights, retain S components
Variance exceedes the mask window of default variance threshold values and carries out dynamic threshold segmentation to it, obtains only comprising 0 (prospect) and 255
(background) two binary images of pixel value.
In one embodiment, variance threshold values can set according to formula (10), and segmentation threshold can set according to formula (11).
Tsigma=mean/3 (10)
Tbina=mean+k × sigma (11)
Wherein, TsigmaIt is variance threshold values；TbinaIt is segmentation threshold；K is weight coefficient, in the disclosure, can choose k=1/
8；Mean is the average of the S components of mask window；Sigma is the variance of the S components of mask window.
If there are multiple candidate points in HSI images, after window mask process is completed to a candidate point, then use
Same method carries out window mask process to other candidate points successively, the bianry image until obtaining all mask windows.
In step S34, shape constraining is carried out to the connected region in bianry image, extract candidate signal lamp region.
After carrying out connected component analysis to bianry image, the connected region in bianry image is can extract out.In an implementation
In example, bianry image can be scanned from top to bottom by row using the connected component analysis mode of 4 neighborhoods, find out binary map
Foreground pixel point (i.e. pixel value is 255 pixel) as in, and foreground pixel point is connected, obtain connected region.
Because traffic lights have some substantially stationary shape facilities, such as size (such as longitudinal ruler of traffic lights
The feature such as very little, lateral dimension and aspect ratio) there is relatively stable scope, traffic lights typically have circular signal lamp and arrow
The major class of shape signal lamp two and traffic lights all have the position phase of a black backboard and each color signal lamp in backboard
To fixation, therefore the connected region in bianry image can be filtered according to the substantially stationary feature of signal lamp, screening out can not
Can be the connected region of traffic lights region (such as car light, street lamp, pavement reflecting and billboard are reflective), and retain remaining
Connected region is used as candidate's traffic lights region.
Referring to Fig. 5, in one embodiment, the connected region after increasing to region carries out shape constraining, extracts candidate
The method in signal lamp region is comprised the following steps：
In step S341, region growth is carried out to the connected region in bianry image according to the default rule that increases.
If when in connected region comprising light sources such as traffic lights, the influence of illumination and green light source may when being imaged
Cause connected region over-exposed, pixel performance is white, so as to cause connected region imperfect, it is therefore desirable to connected region
Interior I component carries out region growth, to obtain complete connected region.
In one embodiment, presetting growth rule is：The I component average I_mean in connected region is calculated first, and
With this average to increase threshold value；If the I component value of a pixel in the connected region meets I-I_mean, by the pixel
Point includes connected region, otherwise it is assumed that the pixel belongs to background.
In step S342, the aspect ratio that minimum enclosed rectangle frame is extracted from the connected region after the growth of region is less than
The size of pre-set aspect ratio threshold value and minimum enclosed rectangle frame meets the connected region of pre-set dimension scope, used as candidate connected region
Domain.
Constraint such as formula (13) to the aspect ratio of the minimum enclosed rectangle frame of connected region is shown.
Max (w, h) ＜ THwh·min(w,h) (13)
Wherein, w is the lateral dimension of the minimum enclosed rectangle frame in candidate signal lamp region；H is candidate signal lamp region
The longitudinal size of minimum enclosed rectangle frame；THwhIt is pre-set aspect ratio threshold value.
In one embodiment, it is contemplated that the traffic lights gathered in natural scene have visual angle change, can
Aspect ratio thresholds are suitably increased, for example THwh=2.Thus, it is possible to avoid missing real traffic lights, while also reaching
Screen out the purpose of non-traffic lights.
Size to the minimum enclosed rectangle frame of connected region enters row constraint, as shown in formula (14).
Wherein, Min and Max are respectively default minimum size threshold and default full-size threshold value.
Row constraint is entered by the aspect ratio and size of the minimum enclosed rectangle frame to connected region, doubtful traffic can be obtained
Candidate's connected region of signal lamp, thus, it is possible to delete the excessive billboard area of area and the too small connected region of area (is made an uproar
Point), the quantity in candidate signal lamp region is reduced, so as to reduce the workload of follow-up cognitive phase, improve traffic lights identification
Speed.
In step S343, using Hu not bending moment extract candidate's connected region characteristic value.
In step S344, the mahalanobis distance of the characteristic value of candidate's connected region and the characteristic value of signal lamp template is obtained.
In step S345, candidate's connection that mahalanobis distance meets preset threshold range is extracted from candidate's connected region
Region, as candidate signal lamp region.
Because imaging of the traffic lights in Chinese herbaceous peony image has one with standard traffic signal lamp image in practical application
Determine difference, and traffic lights have circular signal lamp and the major class of arrow-shaped signal lamp two, therefore circle can be stored in Sample Storehouse
Signal lamp template and arrow-shaped signal lamp template, similarity measurement is carried out by candidate signal lamp region and each signal lamp template.
For a candidate signal lamp region, final similitude is candidate signal lamp region and each signal lamp mould
Maximum in plate result of the comparison.If the certain threshold value of similar sexual satisfaction, it is believed that candidate signal lamp region and similitude
The corresponding signal lamp template of maximum belongs to a kind of traffic lights together, then retain the candidate signal lamp region, otherwise it is believed that waiting
Select signal lamp region for non-traffic lights, then screen out the candidate signal lamp region.
In one embodiment, can extract candidate signal lamp region Hu not bending moment as candidate signal lamp region spy
Levy, and be compared with the feature of each signal lamp template in Sample Storehouse, and candidate signal lamp area is characterized using mahalanobis distance
The similitude of domain and standard traffic signal lamp image, mahalanobis distance is bigger, then the phase of candidate signal lamp region and signal lamp template
It is bigger like property；Conversely, candidate signal lamp region is smaller with the similitude of signal lamp template.
Thus, matched with signal lamp template by by candidate signal lamp region, judge candidate's connected region whether be
One kind in circular signal lamp and arrowhead-shaped signal lamp, so as to further screen out non-traffic lights.
In step s 35, black backboard constraint is carried out to candidate signal lamp region, extracts traffic lights region.
After shape constraining is carried out to candidate's connected region, can determine whether candidate's connected region is traffic signals substantially
Lamp region.But for circular signal lamp, easily by automobile tail light and some other similar light interference, therefore to candidate
After connected region carries out shape constraining, also need to further confirm that whether candidate's connected region is real traffic lights area
Domain.
According to the design standard in international standard for traffic lights, each traffic lights has a black to carry on the back
Plate.Black backboard generally has two kinds of horizontal version and a portrait, and during common traffic lights normal work, synchronization only one of which is handed over
Ventilating signal lamp lights, and the position of three kinds of signal lamps of color in black backboard is fixed, and three kinds of chis of the signal lamp of color
It is very little identical.
This feature is different from the lamp source that other are disturbed, therefore carries out black to candidate signal lamp region using this feature
Backboard is constrained, and identifies the black backboard of signal lamp, there will be no candidate signal lamp region (the i.e. traffic non-signal of black backboard
Lamp region) filter out.
Referring to Fig. 6, in one embodiment, black backboard constraint is carried out to candidate signal lamp region, extract traffic letter
The method of signal lamp is comprised the following steps：
In step S351, rule is extended according to default, candidate signal lamp region is extended.
In one embodiment, candidate can be connected according to the color in candidate signal lamp region and the size of black backboard
Region is extended according to preset rules, obtains candidate's black backplane region.
For example, red eye is located at the top or the leftmost side of black backboard, therefore red color range is located at for H values
Candidate signal lamp region 71, can downwards or right side extend two regions of unit-sized (with candidate signal lamp region most
Small boundary rectangle frame is a unit, similarly hereinafter), as shown in Figure 7 A；Greensignal light is located at the bottom or most right of black backboard
Side, therefore the candidate signal lamp region 72 of green fields is located at for H values, can be upward or left by the candidate signal lamp region
Side extends two regions of unit-sized, as shown in Figure 7 B；Amber lamp is located at the middle part of black backboard, is located at for H values
The candidate signal lamp region 73 of yellow, can be by the candidate signal lamp region to up and down or each side extending a list
The region of position size, as seen in figure 7 c.
In step S352, the I component projection in the candidate signal lamp region after extending is obtained.
Because signal lamp is actively luminous, if the candidate signal lamp region (candidate's black backplane region) after extending is if comprising letter
Signal lamp region, then the I component in signal lamp region is larger, and the I component in the non-signal lamp region of candidate's black backboard is smaller, therefore
Can be judged according to the I component of candidate signal lamp region (the candidate's black backplane region) each several part after extension.
In one embodiment, the I component in the candidate signal lamp region after extension can be projected by its I component,
Number according to its signal lamp I component average being less than is judged.
I component projection includes floor projection and upright projection.By calculating level in the candidate signal lamp region after extending
Direction is less than the number of signal lamp I component average in often going, obtain the I component floor projection in the region；Similarly, the area can be obtained
The I component upright projection in domain.
In step S353, floor projection and gray scale vertical projection are met the candidate signal lamp region of default constraints
As traffic lights region.
In one embodiment, default constraints can be I component floor projection or upright projection in there is trough (pole
Small value).The traffic lights region in candidate signal lamp region after extension, its corresponding I component is equal less than signal lamp I component
The pixel number of value is less；Conversely, I component is more more than the pixel number of signal lamp I component average, therefore thrown in level
Can there is trough (minimum) in shadow or upright projection, wave trough position be traffic lights where position.
Thus, traffic lights are recognized by carrying out the constraint of black backboard to candidate signal lamp region, friendship can be improved
The accuracy and certainty of ventilating signal lamp identification.
Threshold segmentation is carried out by the S components to HSI component maps, signal under complex illumination environment can be solved
The not good problem of lamp recognition effect, and the method split compared to global threshold, reduce the workload of image procossing, improve
The speed of signal lamp identification.
Referring to Fig. 8, in another embodiment of the present disclosure, the S components to HSI component maps carry out threshold point
Cut, the method for being partitioned into traffic lights is further comprising the steps of：
In step S36, the size in traffic lights region is obtained.
In step S37, according to the size in traffic lights region, the size to mask window is updated.
In the process of moving, with the change of vehicle and traffic lights distance, traffic lights are in Chinese herbaceous peony image for vehicle
In size also change therewith, be difficult to meet the requirement of different size traffic lights using the mask window of fixed dimension.Cause
This, can be updated according to the size in the traffic lights region of current frame image, the mask window size to next two field picture.
In one embodiment, the size in traffic lights region can use the width and altimeter of its minimum external world's rectangle frame
Levy, the size (w × w) of mask window can be updated according to w=max (w, h)+1, wherein, w and h are respectively current frame image
The width and height of the minimum enclosed rectangle frame in middle traffic lights region.
By the technical scheme of the present embodiment, the chi of the size real-time adjustment mask window according to traffic lights region
It is very little, various sizes of traffic lights region is adapted to, improvement causes signal lamp recognition speed slow because mask window is excessive, or
Person because mask window is too small cause signal lamp recognize it is imperfect and caused by missing inspection problem.
With continued reference to Fig. 1, in step S15, traffic lights region is classified using default grader, obtain and hand over
The information of ventilating signal lamp.
Behind the traffic lights region extracted by Threshold segmentation, also need further to obtain the letter of traffic lights
Breath, including traffic lights color state and classification (such as circular signal lamp, straight trip arrow, left-hand rotation arrow and right-hand rotation arrow
Head).
Referring to Fig. 9, in one embodiment, traffic lights region is classified using default grader, obtain and hand over
The method of the information of ventilating signal lamp is comprised the following steps：
In step S151, the color histogram feature and histograms of oriented gradients feature in traffic lights region are extracted.
Traffic lights have different classifications, by taking motor vehicle signal lamp as an example, including circular signal lamp, straight trip arrow,
Four kinds of classifications of left-hand rotation arrow and right-hand rotation arrow, and its color state under the conditions of different photoenvironments occurs difference, because
This needs further to process traffic lights region after traffic lights region is extracted by image segmentation, also, obtains and hands over
The color state and classification of ventilating signal lamp, to carry out thermoacoustic prime engine to vehicle.
In one embodiment, can be straight by extracting the color histogram feature and direction gradient in traffic lights region
Scheme (Histogram of Oriented Gradient, HOG) feature to obtain the color state of traffic lights in side.
In step S152, color histogram feature is classified using the first default grader, obtain traffic signals
The color state of lamp.
In one embodiment, the first default grader is SVM classifier, can be handed over by gathering three kinds of colors of red, yellow, and green
Multiple samples of the ventilating signal lamp under different photoenvironments, and it is special that its color histogram is extracted after sample is normalized
Study is levied and is trained to obtain.The default classification of color histogram feature feeding first in the traffic lights region that will be extracted
Classification and Identification is carried out in device, the color state of traffic lights can be obtained.
In step S153, histograms of oriented gradients feature is classified using the second default grader, obtain traffic
The classification of signal lamp.
Second default grader can be SVM multi classifiers, can be by gathering circular signal lamp, straight trip arrow, left-hand rotation arrow
The signal lamp sample of the plurality of classes such as head and right-hand rotation arrow, extracts its HOG feature and is instructed after samples normalization is processed
Practice study to obtain.After the color state for getting traffic lights, by the default grader of traffic lights region feeding second
In classified, the classification of traffic lights can be obtained.
Thus, it is possible to the identification of the traffic lights of plurality of classes under realizing complex illumination environment.
By the technical scheme of the embodiment of the present disclosure, signal lamp recognition effect is not good under can solving complex illumination environment
Problem, improves adaptability of the image processing algorithm to photoenvironment, improves the accuracy of signal lamp identification；Additionally, passing through
Classify in traffic lights region to being obtained after segmentation, it is possible to achieve various to circular signal lamp, arrow-shaped signal lamp etc.
The identification of the signal lamp of classification.
Figure 10 is the flow chart of the control method for vehicle of the embodiment of the disclosure one, and the method can apply to wagon control dress
Put, the controller of vehicle can be included but is not limited to for example：Vehicle, other-end equipment etc..Reference picture 10, the method includes
Following steps：
In step S101, the information of traffic lights is obtained.
The information of traffic lights can be got by the traffic lights recognition methods shown in above-mentioned Fig. 1 to Fig. 9, herein
Repeat no more.
In step s 102, the Lane detection result according to the vehicle for getting, determines the traveling lane and row of vehicle
Sail direction.
In step s 103, according to traveling lane, travel direction and the information of the traffic lights for getting, car is determined
Corresponding target traffic lights information.
In step S104, according to the information of target traffic lights, thermoacoustic prime engine is carried out to vehicle.
Driving is entered according to the Chinese herbaceous peony image for collecting and uses diatom recognizer, and combine the map letter got from GPS
Breath, can get the recognition result of lane line, so that it is determined that the traveling lane that is currently located of vehicle and travel direction.From Chinese herbaceous peony figure
The traffic lights extracted as in potentially include the traffic lights in multiple tracks, the Travel vehicle that can be currently located according to vehicle
Road and travel direction, determine the target traffic lights that vehicle is referred to, and generate according to the information of target traffic lights
Corresponding control strategy carries out thermoacoustic prime engine to vehicle, reasonably plans the driving path of vehicle.
Figure 11 is the block diagram of the traffic signals light identifier of the embodiment of the disclosure one.Reference picture 11, the device 1100 is wrapped
Include：Image capture module 1101, area-of-interest determining module 1102, image conversion module 1103, image segmentation module 1104
And sort module 1105.
The image capture module 1101 is used for the Chinese herbaceous peony image of Real-time Collection vehicle；
The area-of-interest determining module 1102 is used to determine the area-of-interest of the Chinese herbaceous peony image；
The image conversion module 1103 is used to for the image of the area-of-interest to be transformed into HSI spaces by rgb space, with
Obtain the HSI component maps of the area-of-interest；
The image segmentation module 1104 is used to carry out threshold segmentation to the S components of the HSI component maps, to divide
Cut out traffic lights region；
The sort module 1105 is used to classify the traffic lights region using default grader, obtains traffic
The information of signal lamp.
Alternatively, reference picture 12, area-of-interest determining module 1102 includes：
Apart from acquisition submodule 1121, the distance for obtaining the vehicle and traffic lights；
Area-of-interest determination sub-module 1122, for according to the distance, the default setting height(from bottom) of traffic lights with
And the parameter of the image collecting device of the vehicle, determine the area-of-interest of the Chinese herbaceous peony image.
Alternatively, reference picture 12, image conversion module 1103 includes：
Image transform subblock 1131, for being changed the image of the area-of-interest using below equation：
I=max
R, G and B are respectively the R component of the pixel of area-of-interest, G components and B component, max be in R, G and B most
Big value, min is the minimum value in R, G and B.
Alternatively, reference picture 12, segmentation module 1104 includes：
Candidate point screens submodule 1141, for filtering out S components horizontal gradient from the HSI component maps more than default
The pixel of Grads threshold is used as candidate point；
Mask window setting up submodule 1142, for setting up mask window by search source point of the candidate point；
Segmentation submodule 1143, when the variance for the S components when the mask window exceedes default variance threshold values, to institute
Stating mask window carries out dynamic threshold segmentation, generates bianry image；
Shape constraining submodule 1144, for carrying out shape constraining to the connected region in the bianry image, extracts
Candidate signal lamp region；
Black backboard constrains submodule 1145, for carrying out black backboard constraint to the candidate signal lamp region, extracts
Go out traffic lights region.
Alternatively, reference picture 12, segmentation module 1104 also includes：
Size acquisition submodule 1146, the size for obtaining the traffic lights；
Mask window updates submodule 1147, for the size according to the traffic lights region, to the mask windows
The size of mouth is updated.
Alternatively, described information includes color state and classification；
Sort module 1105 includes：
Feature extraction submodule 1151, the color histogram for extracting the traffic lights candidate signal lamp region is special
Seek peace histograms of oriented gradients feature；
Color state acquisition submodule 1152, for being carried out to the color histogram feature using default first grader
Classification based training, obtains the color state of traffic lights；
Classification acquisition submodule 1153, for being carried out to the histograms of oriented gradients feature using default second grader
Classification based training, obtains the classification of the traffic lights.
Alternatively, the shape constraining submodule includes：
Region increases submodule, for carrying out region to the connected region in the bianry image according to the default rule that increases
Increase；Candidate's connected region extracting sub-module, for increasing from region after connected region in extract minimum enclosed rectangle frame
Aspect ratio the connected region of pre-set dimension scope is met less than the size of pre-set aspect ratio threshold value and minimum enclosed rectangle frame, make
It is candidate's connected region；Hu invariant moment features value extracting sub-modules, for bending moment not to extract candidate's connection using the Hu
The characteristic value in region；Mahalanobis distance acquisition submodule, characteristic value and signal lamp template for obtaining candidate's connected region
Characteristic value mahalanobis distance；Candidate signal lamp extracted region submodule, for extracting institute from candidate's connected region
The connected region that mahalanobis distance meets preset threshold range is stated, as the candidate signal lamp region.
Alternatively, the black backboard constraint submodule includes：Region extends submodule, for being advised according to default extension
Then, the candidate signal lamp region is extended；Projection acquisition submodule, for obtaining the candidate signal lamp region after extending
Interior I component projection；Candidate signal lamp extracted region submodule, for I component projection to be met into default constraints
Candidate signal lamp region is used as traffic lights region.
Figure 13 is the block diagram of the controller of vehicle of the embodiment of the disclosure one.Reference picture 13, the device 1300 includes：Signal
Lamp data obtaining module 1301, traveling lane and travel direction acquisition module 1302, target traffic lights determining module 1303
With thermoacoustic prime engine module 1304.
The signal information acquisition module 1301 is used to obtain traffic lights by above-mentioned traffic signals light identifier
Information；
The traveling lane and travel direction acquisition module 1302 are used for the Lane detection according to the vehicle for getting
As a result, the traveling lane and travel direction of the vehicle are obtained；
The target traffic lights determining module 1303 is used for according to the traveling lane, the travel direction and described
The information of traffic lights, determines the corresponding target traffic lights information of the vehicle；
The thermoacoustic prime engine module 1304 is used for according to the target traffic lights information, and driving control is carried out to the vehicle
System.
It should be noted that the controller of vehicle can be included but is not limited to for example：Vehicle, other-end equipment etc..
On the device in above example, wherein modules perform the concrete mode of operation in method
It has been described in detail in embodiment, explanation will be not set forth in detail herein.
Figure 14 is the block diagram for traffic signals light identifier 1400 of the embodiment of the disclosure one.For example, device 1400
May be provided in vehicle.Reference picture 14, device 1400 includes：Electronic control unit 1401, image collecting device 1402, treatment
Device 1403, brake system 1404, steering wheel angle sensor 1405, wheel speed sensors 1406, engine system 1407 and CAN are total
Line 1408.
Image collecting device 1402 can be used to perform the Real-time Collection of above-mentioned Chinese herbaceous peony image.Processor 1403 can be used to perform
The identification of traffic lights, and lane line identification.Electronic control unit 1401 can be according to the information of traffic lights and car
The recognition result of diatom performs thermoacoustic prime engine.
The disclosure also provides a kind of vehicle, including above-mentioned traffic signals light identifier.
Describe the preferred embodiment of the disclosure in detail above in association with accompanying drawing, but, the disclosure is not limited to above-mentioned reality
The detail in mode is applied, in the range of the technology design of the disclosure, various letters can be carried out with technical scheme of this disclosure
Monotropic type, these simple variants belong to the protection domain of the disclosure.
It is further to note that each particular technique feature described in above-mentioned specific embodiment, in not lance
In the case of shield, can be combined by any suitable means.In order to avoid unnecessary repetition, the disclosure to it is various can
The combination of energy is no longer separately illustrated.
Additionally, can also be combined between a variety of implementation methods of the disclosure, as long as it is without prejudice to originally
Disclosed thought, it should equally be considered as disclosure disclosure of that.
Claims (19)
1. a kind of traffic lights recognition methods, it is characterised in that including：
The Chinese herbaceous peony image of Real-time Collection vehicle；
Determine the area-of-interest of the Chinese herbaceous peony image；
The image of the area-of-interest is transformed into HSI spaces by rgb space, to obtain HSI points of the area-of-interest
Spirogram；
S components to the HSI component maps carry out threshold segmentation, to be partitioned into traffic lights region；
The traffic lights region is classified using default grader, obtains the information of traffic lights.
2. method according to claim 1, it is characterised in that the step of the area-of-interest of the determination Chinese herbaceous peony image
Suddenly include：
Obtain the distance of the vehicle and traffic lights；
The parameter of the image collecting device according to the distance, the default setting height(from bottom) of traffic lights and the vehicle, really
The area-of-interest of the fixed Chinese herbaceous peony image.
3. method according to claim 1, it is characterised in that the image by the area-of-interest is by rgb space
HSI spaces are transformed into, are included the step of with the HSI component maps for obtaining the area-of-interest：
The image of the area-of-interest is changed using below equation：
I=max
R, G and B are respectively the R component of the pixel of area-of-interest, G components and B component, and max is the maximum in R, G and B,
Min is the minimum value in R, G and B.
4. method according to claim 1, it is characterised in that the S components to the HSI component maps carry out local dynamic
State Threshold segmentation, includes the step of to be partitioned into traffic lights region：
Pixel of the S components horizontal gradient more than predetermined gradient threshold value is filtered out from the HSI component maps as candidate point；
Mask window is set up by search source point of the candidate point；
When the variance of the S components of the mask window exceedes default variance threshold values, dynamic threshold is carried out to the mask window
Segmentation, generates bianry image；
Shape constraining is carried out to the connected region in the bianry image, candidate signal lamp region is extracted；
Black backboard constraint is carried out to the candidate signal lamp region, traffic lights region is extracted.
5. method according to claim 4, it is characterised in that the S components to the HSI component maps carry out local dynamic
State Threshold segmentation, also includes the step of to be partitioned into traffic lights region：
Obtain the size in the traffic lights region；
According to the size in the traffic lights region, the size to the mask window is updated.
6. method according to claim 4, it is characterised in that the connected region in the bianry image carries out shape
Shape is constrained, and is included the step of extract candidate signal lamp region：
Region growth is carried out to the connected region in the bianry image according to the default rule that increases；
From region increase after connected region in extract the aspect ratio of minimum enclosed rectangle frame less than pre-set aspect ratio threshold value and
The size of minimum enclosed rectangle frame meets the connected region of pre-set dimension scope, used as candidate's connected region；
Using the Hu, bending moment does not extract the characteristic value of candidate's connected region；
Obtain the mahalanobis distance of the characteristic value of candidate's connected region and the characteristic value of signal lamp template；
The connected region that the mahalanobis distance meets preset threshold range is extracted from candidate's connected region, as described
Candidate signal lamp region.
7. method according to claim 4, it is characterised in that described that black backboard is carried out to the candidate signal lamp region
Constraint, includes the step of extract traffic lights region：
Extend rule according to default, the candidate signal lamp region is extended；
Obtain the I component projection in the candidate signal lamp region after extending；
Using the candidate signal lamp region of the I component default constraints of projection satisfaction as traffic lights region.
8. method according to claim 1, it is characterised in that described information includes color state and classification；
It is described the traffic lights region is classified using default grader, obtain traffic lights information the step of
Including：
Extract the color histogram feature and histograms of oriented gradients feature in the traffic lights region；
The color histogram feature is classified using the first default grader, obtains the color state of traffic lights；
The histograms of oriented gradients feature is classified using the second default grader, obtains the classification of traffic lights.
9. a kind of control method for vehicle, it is characterised in that methods described includes：
Method according to claim any one of 1-8 obtains the information of traffic lights；
According to the Lane detection result of the vehicle for getting, the traveling lane and travel direction of the vehicle are determined；
According to the information of the traveling lane, the travel direction and the traffic lights, determine that the vehicle is corresponding
Target traffic lights information；
According to the target traffic lights information, thermoacoustic prime engine is carried out to the vehicle.
10. a kind of traffic signals light identifier, it is characterised in that including：
Image capture module, for the Chinese herbaceous peony image of Real-time Collection vehicle；
Area-of-interest determining module, the area-of-interest for determining the Chinese herbaceous peony image；
Image conversion module, it is described to obtain for the image of the area-of-interest to be transformed into HSI spaces by rgb space
The HSI component maps of area-of-interest；
Image segmentation module, carries out threshold segmentation, to be partitioned into traffic for the S components to the HSI component maps
Signal lamp region；
Sort module, for classifying to the traffic lights region using default grader, obtains traffic lights
Information.
11. devices according to claim 10, it is characterised in that the area-of-interest determining module includes：
Apart from acquisition submodule, the distance for obtaining the vehicle and traffic lights；
Area-of-interest determination sub-module, for according to the distance, the default setting height(from bottom) of traffic lights and the car
Image collecting device parameter, determine the area-of-interest of the Chinese herbaceous peony image.
12. devices according to claim 10, it is characterised in that described image modular converter includes：
Image transform subblock, for being changed the image of the area-of-interest using below equation：
I=max
R, G and B are respectively the R component of the pixel of area-of-interest, G components and B component, and max is the maximum in R, G and B,
Min is the minimum value in R, G and B.
13. devices according to claim 10, it is characterised in that described image segmentation module includes：
Candidate point screens submodule, for filtering out S components horizontal gradient from the HSI component maps more than predetermined gradient threshold value
Pixel as candidate point；
Mask window setting up submodule, for setting up mask window by search source point of the candidate point；
Segmentation submodule, when the variance for the S components when the mask window exceedes default variance threshold values, to the mask windows
Mouth carries out dynamic threshold segmentation, generates bianry image；
Shape constraining submodule, for carrying out shape constraining to the connected region in the bianry image, extracts candidate signal
Lamp region；
Black backboard constrains submodule, for carrying out black backboard constraint to the candidate signal lamp region, extracts traffic letter
Signal lamp region.
14. devices according to claim 13, it is characterised in that the segmentation module also includes：
Size acquisition submodule, the size for obtaining the traffic lights；
Mask window updates submodule, for the size according to the traffic lights region, to the size of the mask window
It is updated.
15. devices according to claim 13, it is characterised in that the shape constraining submodule includes：
Region increases submodule, for carrying out region increasing to the connected region in the bianry image according to the default rule that increases
It is long；
Candidate's connected region extracting sub-module, for increasing from region after connected region in extract minimum enclosed rectangle frame
Aspect ratio meets the connected region of pre-set dimension scope less than the size of pre-set aspect ratio threshold value and minimum enclosed rectangle frame, as
Candidate's connected region；
Hu invariant moment features value extracting sub-modules, for bending moment not to extract the feature of candidate's connected region using the Hu
Value；
Mahalanobis distance acquisition submodule, for obtaining the characteristic value of candidate's connected region and the characteristic value of signal lamp template
Mahalanobis distance；
Candidate signal lamp extracted region submodule, meets pre- for extracting the mahalanobis distance from candidate's connected region
If the connected region of threshold range, as the candidate signal lamp region.
16. devices according to claim 13, it is characterised in that the black backboard constraint submodule includes：
Region extends submodule, for extending rule according to default, the candidate signal lamp region is extended；
Projection acquisition submodule, for obtaining the I component projection in the candidate signal lamp region after extending；
Candidate signal lamp extracted region submodule, the candidate signal lamp for I component projection to be met default constraints
Region is used as traffic lights region.
17. devices according to claim 10, it is characterised in that described information includes color state and classification；
The sort module includes：
Feature extraction submodule, color histogram feature and histograms of oriented gradients for extracting the traffic lights region
Feature；
Color state acquisition submodule, for carrying out classification instruction to the color histogram feature using default first grader
Practice, obtain the color state of traffic lights；
Classification acquisition submodule, for carrying out classification instruction to the histograms of oriented gradients feature using default second grader
Practice, obtain the classification of traffic lights.
18. a kind of controller of vehicle, it is characterised in that described device includes：
Signal information acquisition module, for obtaining traffic lights by the device described in claim any one of 10-17
Information；
Traveling lane and travel direction acquisition module, for the Lane detection result according to the vehicle for getting, obtain
The traveling lane and travel direction of the vehicle；
Target traffic lights determining module, for according to the traveling lane, the travel direction and the traffic signals
The information of lamp, determines the corresponding target traffic lights information of the vehicle；
Thermoacoustic prime engine module, for according to the target traffic lights information, thermoacoustic prime engine being carried out to the vehicle.
19. a kind of vehicles, it is characterised in that including the traffic signals light identifier described in above-mentioned any one of 10-17.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710070814.6A CN106909937B (en) | 2017-02-09 | 2017-02-09 | Traffic signal lamp identification method, vehicle control method and device and vehicle |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201710070814.6A CN106909937B (en) | 2017-02-09 | 2017-02-09 | Traffic signal lamp identification method, vehicle control method and device and vehicle |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106909937A true CN106909937A (en) | 2017-06-30 |
| CN106909937B CN106909937B (en) | 2020-05-19 |
Family
ID=59209204
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201710070814.6A Active CN106909937B (en) | 2017-02-09 | 2017-02-09 | Traffic signal lamp identification method, vehicle control method and device and vehicle |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106909937B (en) |
Cited By (28)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107563301A (en) * | 2017-08-09 | 2018-01-09 | 上海炬宏信息技术有限公司 | Red signal detection method based on image processing techniques |
| CN107644538A (en) * | 2017-11-01 | 2018-01-30 | 广州汽车集团股份有限公司 | The recognition methods of traffic lights and device |
| CN107704853A (en) * | 2017-11-24 | 2018-02-16 | 重庆邮电大学 | A kind of recognition methods of the traffic lights based on multi-categorizer |
| CN108304813A (en) * | 2018-02-08 | 2018-07-20 | 长安大学 | A kind of method of intelligent recognition circle traffic signal light condition |
| GB2559005A (en) * | 2016-11-23 | 2018-07-25 | Ford Global Tech Llc | Traffic light detection and classification using computer vision and deep learning |
| CN108446668A (en) * | 2018-04-10 | 2018-08-24 | 吉林大学 | Traffic lights detection recognition method and system based on unmanned platform |
| CN109035289A (en) * | 2018-07-27 | 2018-12-18 | 重庆师范大学 | Purple soil image segmentation extracting method based on Chebyshev inequality H threshold value |
| CN109215364A (en) * | 2018-11-19 | 2019-01-15 | 长沙智能驾驶研究院有限公司 | Traffic signals recognition methods, system, equipment and storage medium |
| CN109389079A (en) * | 2018-09-30 | 2019-02-26 | 无锡职业技术学院 | A kind of traffic lights recognition methods |
| CN109711379A (en) * | 2019-01-02 | 2019-05-03 | 电子科技大学 | A kind of complex environment traffic lights candidate region is extracted and recognition methods |
| CN110287828A (en) * | 2019-06-11 | 2019-09-27 | 北京三快在线科技有限公司 | Detection method, device and the electronic equipment of signal lamp |
| CN110473414A (en) * | 2019-08-20 | 2019-11-19 | 北京经纬恒润科技有限公司 | A kind of vehicle running path determines method, apparatus and system |
| WO2020001223A1 (en) * | 2018-06-29 | 2020-01-02 | 北京市商汤科技开发有限公司 | Method and device for traffic signal detection and intelligent driving, vehicle, and electronic device |
| CN110675635A (en) * | 2019-10-09 | 2020-01-10 | 北京百度网讯科技有限公司 | Method and device for acquiring external parameters of camera, electronic equipment and storage medium |
| CN110837789A (en) * | 2019-10-31 | 2020-02-25 | 北京奇艺世纪科技有限公司 | Method and device for detecting object, electronic equipment and medium |
| CN112215759A (en) * | 2019-07-11 | 2021-01-12 | 浙江宇视科技有限公司 | Signal lamp color enhancement method and device and electronic equipment |
| CN112581534A (en) * | 2020-12-24 | 2021-03-30 | 济南博观智能科技有限公司 | Signal lamp repositioning method and device, electronic equipment and storage medium |
| CN112598674A (en) * | 2020-12-03 | 2021-04-02 | 北京汽车研究总院有限公司 | Image processing method and device for vehicle and vehicle |
| CN112926483A (en) * | 2021-03-11 | 2021-06-08 | 中国工商银行股份有限公司 | Standard cabinet state indicator lamp identification monitoring method, device and system |
| CN113311905A (en) * | 2020-02-26 | 2021-08-27 | 初速度（苏州）科技有限公司 | Data processing system |
| WO2021183376A1 (en) * | 2020-03-13 | 2021-09-16 | Argo AI, LLC | Using rasterization to identify traffic signal devices |
| WO2021183381A1 (en) * | 2020-03-13 | 2021-09-16 | Argo AI, LLC | Bulb mask representation for traffic light classification |
| CN113468911A (en) * | 2020-03-30 | 2021-10-01 | 上海高德威智能交通系统有限公司 | Vehicle-mounted red light running detection method and device, electronic equipment and storage medium |
| CN114241438A (en) * | 2021-11-26 | 2022-03-25 | 电子科技大学 | Traffic signal lamp rapid and accurate identification method based on prior information |
| CN114783192A (en) * | 2022-03-24 | 2022-07-22 | 杭州海康威视数字技术股份有限公司 | Signal lamp color processing method and device |
| US11704912B2 (en) | 2020-06-16 | 2023-07-18 | Ford Global Technologies, Llc | Label-free performance evaluator for traffic light classifier system |
| CN116704015A (en) * | 2023-08-07 | 2023-09-05 | 中国科学院合肥物质科学研究院 | Noise facula image self-adaptive window preprocessing centroid algorithm and system |
| CN117542220A (en) * | 2024-01-10 | 2024-02-09 | 深圳市拓安科技有限公司 | Driving safety induction method and system applied to severe weather |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102117546A (en) * | 2011-03-10 | 2011-07-06 | 上海交通大学 | On-vehicle traffic light assisting device |
| CN103612565A (en) * | 2013-12-02 | 2014-03-05 | 北京汽车研究总院有限公司 | Method and device for preventing vehicle speeding and vehicle |
| CN104050447A (en) * | 2014-06-05 | 2014-09-17 | 奇瑞汽车股份有限公司 | Traffic light identification method and device |
-
2017
- 2017-02-09 CN CN201710070814.6A patent/CN106909937B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102117546A (en) * | 2011-03-10 | 2011-07-06 | 上海交通大学 | On-vehicle traffic light assisting device |
| CN103612565A (en) * | 2013-12-02 | 2014-03-05 | 北京汽车研究总院有限公司 | Method and device for preventing vehicle speeding and vehicle |
| CN104050447A (en) * | 2014-06-05 | 2014-09-17 | 奇瑞汽车股份有限公司 | Traffic light identification method and device |
Non-Patent Citations (4)
| Title |
|---|
| 周刊: "智能车载红外视觉预警系统中的图像特征提取技术研究", 《万方在线公开：WWW.WANFANGDATA.COM.CN/DETAILS/DETAIL.DO?_TYPE=DEGREE*ID=Y1886183》 * |
| 孟江曼: ""基于视觉的交通信号灯检测与识别系统的设计与实现"", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
| 张倩: "基于支持向量机的多特征交通标志识别的研究与实现", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
| 朱梦超: "视频交通信号灯识别算法研究", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Cited By (46)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| GB2559005A (en) * | 2016-11-23 | 2018-07-25 | Ford Global Tech Llc | Traffic light detection and classification using computer vision and deep learning |
| US10185881B2 (en) | 2016-11-23 | 2019-01-22 | Ford Global Technologies, Llc | Traffic-light detection and classification using computer vision and deep learning |
| CN107563301A (en) * | 2017-08-09 | 2018-01-09 | 上海炬宏信息技术有限公司 | Red signal detection method based on image processing techniques |
| CN107644538A (en) * | 2017-11-01 | 2018-01-30 | 广州汽车集团股份有限公司 | The recognition methods of traffic lights and device |
| CN107704853A (en) * | 2017-11-24 | 2018-02-16 | 重庆邮电大学 | A kind of recognition methods of the traffic lights based on multi-categorizer |
| CN108304813A (en) * | 2018-02-08 | 2018-07-20 | 长安大学 | A kind of method of intelligent recognition circle traffic signal light condition |
| CN108304813B (en) * | 2018-02-08 | 2021-07-23 | 长安大学 | Method for intelligently identifying state of round traffic signal lamp |
| CN108446668A (en) * | 2018-04-10 | 2018-08-24 | 吉林大学 | Traffic lights detection recognition method and system based on unmanned platform |
| KR20200128145A (en) | 2018-06-29 | 2020-11-11 | 베이징 센스타임 테크놀로지 디벨롭먼트 컴퍼니 리미티드 | Methods and devices, vehicles, and electronic devices for traffic light detection and intelligent driving |
| CN110660254A (en) * | 2018-06-29 | 2020-01-07 | 北京市商汤科技开发有限公司 | Traffic signal lamp detection and intelligent driving method and device, vehicle and electronic equipment |
| CN110660254B (en) * | 2018-06-29 | 2022-04-08 | 北京市商汤科技开发有限公司 | Traffic signal lamp detection and intelligent driving method and device, vehicle and electronic equipment |
| WO2020001223A1 (en) * | 2018-06-29 | 2020-01-02 | 北京市商汤科技开发有限公司 | Method and device for traffic signal detection and intelligent driving, vehicle, and electronic device |
| CN109035289A (en) * | 2018-07-27 | 2018-12-18 | 重庆师范大学 | Purple soil image segmentation extracting method based on Chebyshev inequality H threshold value |
| CN109035289B (en) * | 2018-07-27 | 2021-11-12 | 重庆师范大学 | Purple soil image segmentation and extraction method based on Chebyshev inequality H threshold |
| CN109389079A (en) * | 2018-09-30 | 2019-02-26 | 无锡职业技术学院 | A kind of traffic lights recognition methods |
| CN109389079B (en) * | 2018-09-30 | 2022-02-15 | 无锡职业技术学院 | Traffic signal lamp identification method |
| CN109215364A (en) * | 2018-11-19 | 2019-01-15 | 长沙智能驾驶研究院有限公司 | Traffic signals recognition methods, system, equipment and storage medium |
| CN109215364B (en) * | 2018-11-19 | 2020-08-18 | 长沙智能驾驶研究院有限公司 | Traffic signal recognition method, system, device and storage medium |
| CN109711379A (en) * | 2019-01-02 | 2019-05-03 | 电子科技大学 | A kind of complex environment traffic lights candidate region is extracted and recognition methods |
| CN110287828A (en) * | 2019-06-11 | 2019-09-27 | 北京三快在线科技有限公司 | Detection method, device and the electronic equipment of signal lamp |
| CN112215759A (en) * | 2019-07-11 | 2021-01-12 | 浙江宇视科技有限公司 | Signal lamp color enhancement method and device and electronic equipment |
| CN112215759B (en) * | 2019-07-11 | 2024-04-09 | 浙江宇视科技有限公司 | Signal lamp color enhancement method and device and electronic equipment |
| CN110473414A (en) * | 2019-08-20 | 2019-11-19 | 北京经纬恒润科技有限公司 | A kind of vehicle running path determines method, apparatus and system |
| CN110473414B (en) * | 2019-08-20 | 2021-03-23 | 北京经纬恒润科技股份有限公司 | Vehicle driving path determining method, device and system |
| CN110675635B (en) * | 2019-10-09 | 2021-08-03 | 北京百度网讯科技有限公司 | Method and device for acquiring external parameters of camera, electronic equipment and storage medium |
| CN110675635A (en) * | 2019-10-09 | 2020-01-10 | 北京百度网讯科技有限公司 | Method and device for acquiring external parameters of camera, electronic equipment and storage medium |
| CN110837789A (en) * | 2019-10-31 | 2020-02-25 | 北京奇艺世纪科技有限公司 | Method and device for detecting object, electronic equipment and medium |
| CN113311905A (en) * | 2020-02-26 | 2021-08-27 | 初速度（苏州）科技有限公司 | Data processing system |
| CN113311905B (en) * | 2020-02-26 | 2022-06-24 | 魔门塔(苏州)科技有限公司 | Data processing system |
| WO2021183376A1 (en) * | 2020-03-13 | 2021-09-16 | Argo AI, LLC | Using rasterization to identify traffic signal devices |
| WO2021183381A1 (en) * | 2020-03-13 | 2021-09-16 | Argo AI, LLC | Bulb mask representation for traffic light classification |
| US11436842B2 (en) | 2020-03-13 | 2022-09-06 | Argo AI, LLC | Bulb mask representation for traffic light classification |
| US11210571B2 (en) | 2020-03-13 | 2021-12-28 | Argo AI, LLC | Using rasterization to identify traffic signal devices |
| US11670094B2 (en) | 2020-03-13 | 2023-06-06 | Ford Global Technologies, Llc | Using rasterization to identify traffic signal devices |
| CN113468911A (en) * | 2020-03-30 | 2021-10-01 | 上海高德威智能交通系统有限公司 | Vehicle-mounted red light running detection method and device, electronic equipment and storage medium |
| CN113468911B (en) * | 2020-03-30 | 2022-12-23 | 上海高德威智能交通系统有限公司 | Vehicle-mounted red light running detection method and device, electronic equipment and storage medium |
| US11704912B2 (en) | 2020-06-16 | 2023-07-18 | Ford Global Technologies, Llc | Label-free performance evaluator for traffic light classifier system |
| CN112598674A (en) * | 2020-12-03 | 2021-04-02 | 北京汽车研究总院有限公司 | Image processing method and device for vehicle and vehicle |
| CN112581534A (en) * | 2020-12-24 | 2021-03-30 | 济南博观智能科技有限公司 | Signal lamp repositioning method and device, electronic equipment and storage medium |
| CN112926483A (en) * | 2021-03-11 | 2021-06-08 | 中国工商银行股份有限公司 | Standard cabinet state indicator lamp identification monitoring method, device and system |
| CN114241438A (en) * | 2021-11-26 | 2022-03-25 | 电子科技大学 | Traffic signal lamp rapid and accurate identification method based on prior information |
| CN114783192A (en) * | 2022-03-24 | 2022-07-22 | 杭州海康威视数字技术股份有限公司 | Signal lamp color processing method and device |
| CN116704015A (en) * | 2023-08-07 | 2023-09-05 | 中国科学院合肥物质科学研究院 | Noise facula image self-adaptive window preprocessing centroid algorithm and system |
| CN116704015B (en) * | 2023-08-07 | 2023-11-14 | 中国科学院合肥物质科学研究院 | Method and system for realizing noise facula image self-adaptive window preprocessing centroid |
| CN117542220A (en) * | 2024-01-10 | 2024-02-09 | 深圳市拓安科技有限公司 | Driving safety induction method and system applied to severe weather |
| CN117542220B (en) * | 2024-01-10 | 2024-04-16 | 深圳市拓安科技有限公司 | Driving safety induction method and system applied to severe weather |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106909937B (en) | 2020-05-19 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106909937A (en) | Traffic lights recognition methods, control method for vehicle, device and vehicle | |
| CN111582216B (en) | Unmanned vehicle-mounted traffic signal lamp identification system and method | |
| CN110197589B (en) | Deep learning-based red light violation detection method | |
| CN105913041B (en) | It is a kind of based on the signal lamp recognition methods demarcated in advance | |
| Fang et al. | Road-sign detection and tracking | |
| CN110619750B (en) | Intelligent aerial photography identification method and system for illegal parking vehicle | |
| CN105844257B (en) | The early warning system and method for road sign are missed based on machine vision travelling in fog day | |
| Cheng et al. | Lane detection with moving vehicles in the traffic scenes | |
| KR100912746B1 (en) | Method for traffic sign detection | |
| EP1930863B1 (en) | Detecting and recognizing traffic signs | |
| CN103824081B (en) | Method for detecting rapid robustness traffic signs on outdoor bad illumination condition | |
| CN108564030A (en) | Classifier training method and apparatus towards vehicle-mounted thermal imaging pedestrian detection | |
| CN109389046B (en) | All-weather object identification and lane line detection method for automatic driving | |
| CN107704853A (en) | A kind of recognition methods of the traffic lights based on multi-categorizer | |
| CN108734189A (en) | Vehicle License Plate Recognition System based on atmospherical scattering model and deep learning under thick fog weather | |
| CN103366156A (en) | Road structure detection and tracking | |
| CN106570494A (en) | Traffic signal lamp recognition method and device based on convolution neural network | |
| CN106709412B (en) | Traffic sign detection method and device | |
| CN102163278B (en) | Illegal vehicle intruding detection method for bus lane | |
| CN106919939B (en) | A kind of traffic signboard tracks and identifies method and system | |
| Kim et al. | Effective traffic lights recognition method for real time driving assistance systemin the daytime | |
| CN110188606A (en) | Lane recognition method, device and electronic equipment based on high light spectrum image-forming | |
| CN112712057A (en) | Traffic signal identification method and device, electronic equipment and storage medium | |
| CN109784214A (en) | A kind of detection device and method of railroad track foreign matter | |
| CN109255279A (en) | A kind of method and system of road traffic sign detection identification |
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 |