CN107886033A - Identify the method, apparatus and vehicle of circular traffic lights - Google Patents
Identify the method, apparatus and vehicle of circular traffic lights Download PDFInfo
- Publication number
- CN107886033A CN107886033A CN201610874308.8A CN201610874308A CN107886033A CN 107886033 A CN107886033 A CN 107886033A CN 201610874308 A CN201610874308 A CN 201610874308A CN 107886033 A CN107886033 A CN 107886033A
- Authority
- CN
- China
- Prior art keywords
- circular
- image
- area
- traffic lights
- target
- 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
- 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
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, apparatus and vehicle for identifying circular traffic lights.The image with depth information that this method is collected based on 3D cameras, circular traffic lights is identified.Because the Depth Imaging principle of 3D cameras is not influenceed by natural lighting, so during circular traffic lights is identified, can be more accurate for the identification under the different light conditions such as daytime, night.Also, because depth information is directly exported by 3D cameras, without extra process, therefore the complexity of image procossing can be reduced to a certain extent, lift recognition efficiency.In addition, output of the 3D cameras to depth information and colouring information is almost consistent on a timeline, so that when in combination, circular traffic lights is identified for both, recognition result is more accurate.Without collecting sample and machine learning, the process for identifying circular traffic lights is simplified, improves the efficiency for identifying circular traffic lights and the degree of accuracy.
Description
Technical field
This disclosure relates to technical field of image processing, in particular it relates to a kind of method, apparatus for identifying circular traffic lights and
Vehicle.
Background technology
With the continuous development of scientific technology, vehicle drive intellectuality has gradually become reality from the imagination, based on friendship
The detection of ventilating signal lamp, recognition and tracking algorithm, provided a great help for the even unpiloted realization of motor vehicle driving.
Current traffic lights recognition methods is all to obtain image, reuse mode matching using single 2D cameras
Identified with image processing techniques such as machine learning to realize.Want to obtain accurate recognition result, except to image quality requirements
Outside especially high, also very harsh is required to the complexity of algorithm, while recognition result is also unsatisfactory, main reason is that 2D takes the photograph
As the picture quality of image that head collects in itself has certain limitation, and picture quality influenceed by external interference it is bigger, these
Recognition result will be produced directly influences.
The content of the invention
The purpose of the disclosure is to provide a kind of method, apparatus and vehicle for identifying circular traffic lights, circular to simplify identification
The process of traffic lights, improve the efficiency for identifying circular traffic lights and the degree of accuracy.
To achieve these goals, the disclosure provides a kind of method for identifying circular traffic lights, and methods described includes：
It is partitioned into the image with depth information collected from 3D cameras special with red, green or yellow any color
The circular target region of property；
The mesh for meeting predetermined depth threshold range is extracted in the image with depth information collected from 3D cameras
Mark depth image；
The circular target region and the target depth image are contrasted, to orient square from the target depth image
Shape background frame；
Between the area in the circular target region included according to the area of the rectangular background frame and the rectangular background frame
Magnitude relationship, determine that the circular target region that the rectangular background frame includes is circular traffic lights；
With reference to the color characteristics of the circular traffic lights, the type of the circular traffic lights is determined.
Alternatively, the face in the circular target region included according to the area of the rectangular background frame and the rectangular background frame
Magnitude relationship between product, determine that the circular target region that the rectangular background frame includes is circular traffic lights, including：
Determine the area ratio in the circular target region that the area of the rectangular background frame and the rectangular background frame include；
If ratio meets preset range, the rectangular background frame is circular traffic lights background frame, the rectangular background frame
Comprising circular target region be circular traffic lights.
Alternatively, it is partitioned into the image with depth information collected from 3D cameras with red, green or yellow any
The circular target region of kind color characteristics, including：
It is partitioned into the image with depth information collected from 3D cameras special with red, green or yellow any color
The target area of property；
According to the length-width ratio of the target area, Uncircular field is filtered out；
Judge whether the Hu features of the target area after filtering out match with the Hu features of circular traffic lights template；
The region that the match is successful is extracted from the target area after described filter out.
Alternatively, methods described also includes：
The image with depth information collected to 3D cameras carries out histogram equalization；
Noise reduction process is carried out to the image after histogram equalization；
Morphological scale-space is carried out to the image after noise reduction process；
It is partitioned into the image with depth information collected from 3D cameras with red, green or yellow any face
The target area of color characteristic, including：
The image with depth information collected to the 3D cameras carries out color segmentation；
The image after Morphological scale-space and the image after color segmentation are contrasted, to obtain the target area.
Alternatively, the circular target region and the target depth image are contrasted, with from the target depth image
Rectangular background frame is oriented, including：
The circular target region and the target depth image are contrasted, is included with being extracted from the target depth image
The background area in the circular target region；
According to the rectangular degree of the background area, rectangular background frame is oriented from background area.
The disclosure also provides a kind of device for identifying circular traffic lights, and described device includes：
Circular target region segmentation module, for being partitioned into the image with depth information that is collected from 3D cameras
Circular target region with red, green or yellow any color characteristics；
Target depth image zooming-out module, for being extracted in the image with depth information that is collected from 3D cameras
Meet the target depth image of predetermined depth threshold range；
Rectangular background frame locating module, for contrasting the circular target region and the target depth image, with from institute
State and rectangular background frame is oriented in target depth image；
Circular traffic lights determining module, included for the area according to the rectangular background frame and the rectangular background frame
Magnitude relationship between the area in circular target region, determine that the circular target region that the rectangular background frame includes is handed over to be circular
Logical lamp；
Circular traffic light kind determining module, for the color characteristics with reference to the circular traffic lights, determine the circle
The type of traffic lights.
Alternatively, the circular traffic lights determining module includes：
Area is than determination sub-module, circle that the area for determining the rectangular background frame includes with the rectangular background frame
The area ratio of shape target area；
Circular traffic lights determination sub-module, if meeting preset range for ratio, the rectangular background frame is handed over to be circular
Logical lamp background frame, the circular target region that the rectangular background frame includes is circular traffic lights.
Alternatively, the circular target region segmentation module includes：Target area determination sub-module, for from 3D cameras
It is less than predetermined threshold value and special with red, green or yellow any color that area is partitioned into the image with depth information collected
The target area of property；
Submodule is filtered out, for the length-width ratio according to the target area, filters out Uncircular field；
Judging submodule, the Hu features of the Hu features and circular traffic lights template of the target area after being filtered out for judgement are
No matching；
First extracting sub-module, for extracting the region that the match is successful from the target area after described filter out.
Alternatively, described device also includes：Histogram equalization module, there is depth for what is collected to 3D cameras
The image of information carries out histogram equalization；
Noise reduction module, for carrying out noise reduction process to the image after histogram equalization；
Morphological scale-space module, for carrying out Morphological scale-space to the image after noise reduction process；
The target area determination sub-module includes：
Split submodule, the image with depth information for being collected to the 3D cameras carries out color segmentation；
Submodule is contrasted, for contrasting the image after Morphological scale-space and the image after color segmentation, to obtain
The target area.
Alternatively, the rectangular background frame locating module includes：
Background area extracting sub-module, for contrasting the circular target region and the target depth image, with from institute
State the background area that extraction includes the circular target region in target depth image；
Rectangular background confines bit submodule, for the rectangular degree according to the background area, is oriented from background area
Rectangular background frame.
The disclosure additionally provides a kind of vehicle, and the vehicle includes：
3D cameras, for gathering the image with depth information；And
The device of the circular traffic lights of identification provided according to the disclosure.
In the disclosure, circular traffic lights is identified for the image with depth information that is collected based on 3D cameras.
Because the Depth Imaging principle of 3D cameras is not influenceed by natural lighting, so during circular traffic lights is identified, for
Identification under the different light conditions such as daytime, night can be more accurate.Also, because depth information is directly defeated by 3D cameras
Go out, without extra process, therefore the complexity of image procossing can be reduced to a certain extent, lift recognition efficiency.In addition,
Output of the 3D cameras to depth information and colouring information is almost consistent on a timeline, so that combining, both are right
When circular traffic lights is identified, recognition result is more accurate.Without collecting sample and machine learning, it is circular to simplify identification
The process of traffic lights, improve the efficiency for identifying circular traffic lights and the degree of accuracy.
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 a part for constitution instruction, with following tool
Body embodiment is used to explain the disclosure together, but does not form the limitation to the disclosure.In the accompanying drawings：
Fig. 1 is a kind of flow chart of the method for identification circle traffic lights according to an exemplary embodiment.
Fig. 2 is the schematic diagram of the progress Hu characteristic matchings according to an exemplary embodiment.
Fig. 3 is the circular schematic diagram determined according to the method for Hu characteristic matchings.
Fig. 4 is the schematic diagram of the extraction target depth image according to an exemplary embodiment.
Fig. 5 is a kind of schematic diagram of the device of identification circle traffic lights according to an exemplary embodiment.
Embodiment
The embodiment of the disclosure is described in detail below in conjunction with accompanying drawing.It should be appreciated that this place is retouched
The embodiment stated is merely to illustrate and explained the disclosure, is not limited to the disclosure.
In correlation technique, circular traffic lights is identified based on the image that 2D cameras collect, due to 2D cameras
The picture quality of the image collected in itself has certain limitation, and picture quality influenceed by external interference it is bigger, along with calculation
The influence of method complexity, cause to identify that the efficiency of circular traffic lights and the degree of accuracy be not high.To solve the technical problem, the disclosure carries
For a kind of method, apparatus and vehicle for identifying circular traffic lights, to simplify the process for identifying circular traffic lights, it is circular to improve identification
The efficiency of traffic lights and the degree of accuracy.The method, apparatus and vehicle of the circular traffic lights of identification provided separately below the disclosure are entered
Row explanation.
Fig. 1 is refer to, Fig. 1 is a kind of flow of the method for identification circle traffic lights according to an exemplary embodiment
Figure.As shown in figure 1, this method comprises the following steps：
Step S11：It is partitioned into the image with depth information collected from 3D cameras with red, green or yellow any
The circular target region of kind color characteristics；
Step S12：Extracted in the image with depth information collected from 3D cameras and meet predetermined depth threshold value
The target depth image of scope；
Step S13：The circular target region and the target depth image are contrasted, with from the target depth image
Orient rectangular background frame；
Step S14：The circular target region included according to the area of the rectangular background frame and the rectangular background frame
Magnitude relationship between area, determine that the circular target region that the rectangular background frame includes is circular traffic lights；
Step S15：With reference to the color characteristics of the circular traffic lights, the type of the circular traffic lights is determined.
The disclosure proposes to collect the image with depth information based on 3D cameras, circular traffic lights is identified.
Because the Depth Imaging principle of 3D cameras is not influenceed by natural lighting, so during circular traffic lights is identified, for
Identification under the different light conditions such as daytime, night can be more accurate.Also, because depth information is directly defeated by 3D cameras
Go out, without extra process, therefore the complexity of image procossing can be reduced to a certain extent, lift recognition efficiency.In addition,
Output of the 3D cameras to depth information and colouring information is almost consistent on a timeline, so that combining, both are right
When circular traffic lights is identified, recognition result is more accurate.
In actual applications, 3D cameras may be mounted on the vehicle body of automobile, and a kind of possible mounting means is：Will figure
As harvester is on the front windshield of vehicle body and relative with room mirror.So, during automobile is advanced,
Coloured image and depth image, and Real time identification circle traffic lights can be gathered in real time by 3D cameras, think that driver advises
Sliding bicycle route and road speed provide reference, and ensure the traffic safety of driver.
Alternatively, the image that the angle that 3D cameras rotate up or down can collect in real time according to 3D cameras enters
Rower is determined.In order in image processing process, it is possible to reduce the region of image procossing and data volume, moreover it is possible to reduce a part its
Its light source (such as：Partial automobile taillight etc.) influence that brings.
In the disclosure, the whole image that can be collected to 3D cameras is handled, and can also be chosen 3D cameras and be adopted
The region that the position of the whole image collected is on the upper side handled (such as：Whole image it is upperRegion), to reduce image procossing
Region and data volume.
The processing of the image with depth information collected to 3D cameras includes processing to colouring information and right
The processing of depth information, both are two relatively independent processes, and therefore, both execution sequences in no particular order, can be successively
Perform or perform parallel.
On the one hand, by performing step S11, the color letter in the image with depth information collected to 3D cameras
Breath is handled.Including：
It is partitioned into the image with depth information collected from 3D cameras special with red, green or yellow any color
The target area of property；
According to the length-width ratio of the target area, Uncircular field is filtered out；
Judge whether the Hu features of the target area after filtering out match with the Hu features of circular traffic lights template；
The region that the match is successful is extracted from the target area after described filter out.
Wherein, obtaining to target area has a variety of possible embodiments, illustrates separately below.
Obtaining the first possible embodiment of target area is：Image segmentation is carried out using Lab color spaces.First
The image with depth information that D cameras are collected is transformed into Lab color spaces, is then possible to circular traffic lights
Light color a values and b values be threshold value, in threshold range is probably then circular traffic lights, can thus be taken the photograph 3D
The image with depth information collected as head is divided into two parts：It is probably the region and background area of circular traffic lights,
Wherein, it may be possible to which a values and b values of the pixel in the region of circular traffic lights are in above-mentioned threshold range.
Illustratively, all possible light color of circular traffic lights is red, green and yellow.Wherein, green threshold range
For：- 50 ＜ a ＜ -8 and 15 ＜ b ＜ 80；Red threshold range is：15 ＜ a ＜ 110 and 15 ＜ b ＜ 60；The threshold of yellow
Value scope is：1 ＜ a ＜ 16 and 25 ＜ b ＜ 60.The a values of red, green and yellow three kinds of colors and the scope of b values are how determined, can be joined
Correlation technique is examined, is just repeated no more herein.
Obtaining second of possible embodiment of target area is：To reduce the illumination conditions such as sunshine to image procossing
As a result influence, color segmentation is carried out using the tone value of pixel.With the color of all possible light color of circular traffic lights
Tone pitch is threshold value, and in threshold range is probably then circular traffic lights, thus can be by the background frame area of circular traffic lights
Domain is divided into two parts：It is probably the region and background area of circular traffic lights, wherein, it may be possible in the region of circular traffic lights
Pixel tone value in above-mentioned threshold range.
Illustratively, all possible light color of circular traffic is red, green and yellow.Wherein, red tone value scope
It is less than 6 or more than 244；The tone value scope of green is between 81 to 130, and the tone value scope of yellow is between 21 to 46.
The tone value scope of red, green and yellow three kinds of colors is how determined, correlation technique is referred to, just repeats no more herein.
There is independence in view of H (tone) in HSV color spaces and V (lightness), 3D cameras can be collected
Image with depth information is transformed into HSV color spaces, and then obtains target area using above-mentioned tone value scope.
Alternatively, the third the possible embodiment for obtaining target area comprises the following steps：
The image with depth information collected to 3D cameras carries out histogram equalization；
Noise reduction process is carried out to the image after histogram equalization；
Morphological scale-space is carried out to the image after noise reduction process；
The image with depth information collected to the 3D cameras carries out color segmentation；
The image after Morphological scale-space and the image after color segmentation are contrasted, to obtain the target area.
In the disclosure, in order to which the colouring information in the image with depth information that is collected to 3D cameras increases
By force, and then the color of circular traffic lights is protruded, the image with depth information that can be collected to 3D cameras carries out Nogata
Figure equalization.
After histogram equalization, the colouring information in the image with depth information that 3D cameras collect is more
It is abundant, it is possible that being exaggerated the noise in image, it is possible to which the image after histogram equalization is carried out at noise reduction
Reason.Or, it is contemplated that the image with depth information that 3D cameras collect contains noise in itself, can also be to 3D cameras
The image with depth information collected carries out noise reduction process.
To retain the general characteristic for the image with depth information that 3D cameras collect, Gaussian smoothing side can be used
Method carries out denoising, and after noise reduction process, the general characteristic for the image with depth information that 3D cameras collect preserves
It is more complete.
It is smaller in view of the area of circular traffic lights and larger with the aberration in other regions of surrounding, it is possible to by 3D
The image with depth information that camera collects, the image after histogram equalization or the image after noise reduction process point
Be not converted to gray-scale map first, Morphological scale-space, such as carnival hat operation or top cap operation are then carried out to obtained gray-scale map
(TopHat), to filter out the same or like bulk region of color and dark region, brighter pocket is retained.So,
A large amount of background informations in colour information must all be filtered out by abundant, can be that amount of calculation is saved in follow-up color segmentation operation.
After Morphological scale-space, OTSU (maximum variance between clusters) can be used to calculate binary-state threshold, then used
The binary-state threshold that calculates carries out binary conversion treatment, and then will likely be that the region disconnecting of circular traffic lights comes out.
In actual applications, the step of the step of the step of above-mentioned histogram equalization, noise reduction process and binary conversion treatment
Suddenly it is optional, an execution can be selected, both execution are selected, all perform or do not perform all, if execution can basis
Recognition efficiency and the demand of the degree of accuracy are selected.
The above-mentioned image with depth information collected to 3D cameras carries out color segmentation, may be referred to obtain target
The first embodiment or second of embodiment in region, and other possible embodiments, no matter using which kind of implementation
Mode, the region that all possible light color of circular traffic lights is covered there can be depth letter from what 3D cameras collected
Separated in the image of breath.
To further determine that the region where circular traffic lights, exclude to be unlikely to be the region of circular traffic lights, can be right
Than the image after Morphological scale-space and the image after color segmentation, color and area are all met to the spy of circular traffic lights
Property region (i.e. target area) separated from the image after Morphological scale-space, exclude area coinciding but color and do not meet
The region of circular traffic lights.Then, the image comprising the region separated and after Morphological scale-space is carried out at binaryzation
Reason.
Certainly, if also having carried out binary conversion treatment to the image after Morphological scale-space, contrast through binary conversion treatment
Image afterwards and the image after color segmentation, to isolate color from the image after binary conversion treatment and area all accords with
Close the region (i.e. target area) of the characteristic of circular traffic lights.
After target area is obtained, according to the shape facility of circular traffic lights (such as：Length-width ratio) target area is carried out
Further screening.The length-width ratio of circular traffic lights should be ideally 1: 1, if the length of target area and wide ratio difference
Very big, then it is not circular traffic lights that can determine target area.In actual applications, length-width ratio threshold range, example can be set
Such as：0.7 to 1.4, if the length-width ratio of target area exceeds the threshold range, then it is assumed that the region is not circular lamp, can be filtered out.
After being filtered out to target area, meet shape facility the target area of circular traffic lights, using Hu features
The method of matching is further screened, to extract circular target region.Fig. 2 is refer to, Fig. 2 is according to an exemplary implementation
The schematic diagram of the progress Hu characteristic matchings exemplified.Matching process is as follows：
First, the Hu features of circular traffic lights template are calculated, as shown in Fig. 2 the figure that numbering is (a) is circular traffic lights
Template.
Then, the target area of circular traffic lights is met for shape facility, is confined with minimum enclosed rectangle.Calculate
Go out the boundary value up and down in the region, in this, as the border confined, and then obtain the minimum enclosed rectangle in the region.Such as figure
Shown in 2, numbering is (b) and the figure of (c) is respectively the minimum external square for the target area that shape facility meets circular traffic lights
Shape.
Then, calculate the Hu features of minimum enclosed rectangle, and with the Hu features of circular traffic lights template match, root
Determine whether the target area that shape facility meets circular traffic lights is circular according to matching degree.Refer to Fig. 3, Fig. 3 be according to
The circular schematic diagram that the method for Hu characteristic matchings is determined.
Above-mentioned is the color letter in the image with depth information collected to 3D cameras by performing step S11
Breath is handled, and then determines the process in circular target region.Explanation performs step S12 process below, i.e., to 3D cameras
Depth information in the image with depth information collected is handled, and then determines the process of target depth image.
Using 3D camera collection images, the depth information of image, i.e., each pixel in image can be obtained
There is a range information, suitably screened according to predetermined depth threshold range, some interference regions can be excluded.And this is default
A kind of possible setting means of depth threshold scope is：According to《GB 14886-2006 road traffic signal lamps are set and installation
Specification》, too near or too remote target area is then interference region in image, in proper range (as in the range of 50~200m)
The target depth image that region then extracts for needs.It is alternatively possible to the pixel that predetermined depth threshold range will be met
White is arranged to, and black will be arranged to beyond the pixel of predetermined depth threshold range.
Alternatively, for ease of distinguishing the different objects that 3D cameras photograph, preset range can be classified.It refer to
Fig. 4, Fig. 4 are the schematic diagram of the extraction target depth image according to an exemplary embodiment.Illustratively, 5~100m is selected
For predetermined depth threshold range, depth value is arranged to white in 5~100m pixel, and depth value is exceeded into 5~100m
Pixel be arranged to black.Between 5~100m by pre-determined distance (such as：5m, 10m etc.) unit be classified, and will
The depth of same stages is placed in same gray level that (in Fig. 4, we are represented with white, but are represented not with numeral
With the region of depth value), as 3 in Fig. 4,4 represent two front and rear cars.Wherein, pre-determined distance can be according to the length of vehicle
Degree is configured, and in order to comparison diagram 3 and Fig. 4, then excludes the interference that car light is brought.
Alternatively, it is contemplated that some tiny protrusion (such as traffic lights backgrounds in the target depth image extracted be present
Connection between frame and traffic lights lamp stand etc.) so that noise jamming be present around traffic lights background frame.To eliminate noise jamming,
So that the profile of traffic lights background frame is smoothened, morphology opening operation processing can be carried out to target depth image, to remove
Tiny protrusion.As shown in figure 4, traffic lights background frame divides with traffic lights lamp stand after representing the processing of morphology opening operation with No. X
From removal traffic lights lamp stand part.
It is determined that after circular target region and determination target depth image, step S13 and step S14 is performed.
Wherein, step S13 includes：The circular target region and the target depth image are contrasted, with from the target
Extraction includes the background area in the circular target region in depth image；According to the rectangular degree of the background area, from background
Rectangular background frame is oriented in region.
Circular target region is corresponding into target depth image, if in target depth image not with circular target area
Region corresponding to domain, then it is assumed that be interfering object (such as：Road sign, without car of driving lamp etc.), after exclusive PCR object, obtain
To background area.As shown in figure 4, the region in Fig. 4 marked as 5 and 6 can be excluded.
Next the rectangular degree R of background area is calculated, usual R value is between 0~1.When object is rectangle, R is obtained
Maximum 1.Rectangular degree threshold value (such as 0.9) can be set, i.e. the rectangular degree of background area is more than set rectangular degree threshold value
When, then it is assumed that background area is rectangular background frame.It can be excluded for the region for being unsatisfactory for rectangular degree threshold value, such as Fig. 4 acceptances of the bid
Number be 3 region.
After rectangular background frame is oriented, step S14 can be performed.Step S14 includes：Determine the rectangular background frame
The area ratio in the circular target region that area includes with the rectangular background frame；If ratio meets preset range, the square
Shape background frame is circular traffic lights background frame, and the circular target region that the rectangular background frame includes is circular traffic lights.
On the one hand, the area A of the minimum enclosed rectangle in circular target region is calculatedMER, on the other hand calculate rectangular background frame
Area AO, then calculate AOWith AMERThan for K, being represented with equation below：
Have 3 or 4 circular traffic lights in view of normal condition in circular traffic lights background frame, thus K should 3~
Between 4.5.Street lamp (region in such as Fig. 4 marked as 1) can be so excluded, automobile is (as such as the area in Fig. 3 marked as 4 and 5
Domain, because the area in the region in Fig. 3 marked as 4 and 5 is smaller relative to the area in the region in Fig. 4 marked as 4, it is impossible to be
Circular traffic lights) etc..
Therefore, preset range can be arranged to 3~4.5, if ratio is within a preset range, rectangular background frame is circle
Shape traffic lights background frame, meanwhile, the circular target region of rectangular background inframe is circular traffic lights., both can be with using this method
Circular traffic lights background frame is oriented, circular traffic lights can be oriented again.
After circular traffic lights is oriented, the result of color combining segmentation, you can it is determined that the light color of circular traffic lights,
And then determine the type of circular traffic lights.
The disclosure also provides a kind of device for identifying circular traffic lights.Fig. 5 is refer to, Fig. 5 is according to an exemplary implementation
A kind of schematic diagram of the device of the identification circle traffic lights exemplified.As shown in figure 5, the device 500 includes：
Circular target region segmentation module 501, in the image with depth information that is collected from 3D cameras points
Cut out the circular target region with red, green or yellow any color characteristics；
Target depth image zooming-out module 502, for being carried in the image with depth information that is collected from 3D cameras
Take out the target depth image for meeting predetermined depth threshold range；
Rectangular background frame locating module 503, for contrasting the circular target region and the target depth image, with from
Rectangular background frame is oriented in the target depth image；
Circular traffic lights determining module 504, for the area according to the rectangular background frame and the rectangular background frame bag
Magnitude relationship between the area in the circular target region contained, determine that the circular target region that the rectangular background frame includes is circle
Shape traffic lights；
Circular traffic light kind determining module 505, for the color characteristics with reference to the circular traffic lights, determine the circle
The type of shape traffic lights.
Alternatively, the circular traffic lights determining module includes：
Area is than determination sub-module, circle that the area for determining the rectangular background frame includes with the rectangular background frame
The area ratio of shape target area；
Circular traffic lights determination sub-module, if meeting preset range for ratio, the rectangular background frame is handed over to be circular
Logical lamp background frame, the circular target region that the rectangular background frame includes is circular traffic lights.
Alternatively, the circular target region segmentation module includes：Target area determination sub-module, for from 3D cameras
It is less than predetermined threshold value and special with red, green or yellow any color that area is partitioned into the image with depth information collected
The target area of property；
Submodule is filtered out, for the length-width ratio according to the target area, filters out Uncircular field；
Judging submodule, the Hu features of the Hu features and circular traffic lights template of the target area after being filtered out for judgement are
No matching；
First extracting sub-module, for extracting the region that the match is successful from the target area after described filter out.
Alternatively, described device also includes：Histogram equalization module, there is depth for what is collected to 3D cameras
The image of information carries out histogram equalization；
Noise reduction module, for carrying out noise reduction process to the image after histogram equalization；
Morphological scale-space module, for carrying out Morphological scale-space to the image after noise reduction process；
The target area determination sub-module includes：
Split submodule, the image with depth information for being collected to the 3D cameras carries out color segmentation；
Submodule is contrasted, for contrasting the image after Morphological scale-space and the image after color segmentation, to obtain
The target area.
Alternatively, the rectangular background frame locating module includes：
Background area extracting sub-module, for contrasting the circular target region and the target depth image, with from institute
State the background area that extraction includes the circular target region in target depth image；
Rectangular background confines bit submodule, for the rectangular degree according to the background area, is oriented from background area
Rectangular background frame.
On the device in above-described embodiment, wherein modules and unit perform the concrete mode of operation relevant
It is described in detail in the embodiment of this method, explanation will be not set forth in detail herein.
In addition, the present invention also provides a kind of vehicle, the vehicle can include 3D cameras, have depth information for gathering
Image；And the device of the circular traffic lights of identification provided according to the disclosure.
The preferred embodiment of the disclosure is described in detail above in association with accompanying drawing, still, 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, a variety of letters can be carried out to the technical scheme of the 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 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 no longer separately illustrates.
In addition, it can also be combined between a variety of embodiments 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 (11)
- A kind of 1. method for identifying circular traffic lights, it is characterised in that methods described includes：It is partitioned into the image with depth information collected from 3D cameras with red, green or yellow any color characteristics Circular target region；Extracted in the image with depth information collected from 3D cameras and meet that the target of predetermined depth threshold range is deep Spend image；The circular target region and the target depth image are contrasted, to orient the rectangle back of the body from the target depth image Scape frame；It is big between the area in the circular target region included according to the area of the rectangular background frame and the rectangular background frame Small relation, determine that the circular target region that the rectangular background frame includes is circular traffic lights；With reference to the color characteristics of the circular traffic lights, the type of the circular traffic lights is determined.
- 2. according to the method for claim 1, it is characterised in that carried on the back according to the area of the rectangular background frame and the rectangle Magnitude relationship between the area in the circular target region that scape frame includes, determine the circular target area that the rectangular background frame includes Domain is circular traffic lights, including：Determine the area ratio in the circular target region that the area of the rectangular background frame and the rectangular background frame include；If ratio meets preset range, the rectangular background frame is circular traffic lights background frame, and the rectangular background frame includes Circular target region be circular traffic lights.
- 3. according to the method for claim 1, it is characterised in that the image with depth information collected from 3D cameras In be partitioned into the circular target region with red, green or yellow any color characteristics, including：It is partitioned into the image with depth information collected from 3D cameras with red, green or yellow any color characteristics Target area；According to the length-width ratio of the target area, Uncircular field is filtered out；Judge whether the Hu features of the target area after filtering out match with the Hu features of circular traffic lights template；The region that the match is successful is extracted from the target area after described filter out.
- 4. according to the method for claim 3, it is characterised in that methods described also includes：The image with depth information collected to 3D cameras carries out histogram equalization；Noise reduction process is carried out to the image after histogram equalization；Morphological scale-space is carried out to the image after noise reduction process；It is partitioned into the image with depth information collected from 3D cameras special with red, green or yellow any color The target area of property, including：The image with depth information collected to the 3D cameras carries out color segmentation；The image after Morphological scale-space and the image after color segmentation are contrasted, to obtain the target area.
- 5. according to the method for claim 1, it is characterised in that contrast the circular target region and the target depth figure Picture, to orient rectangular background frame from the target depth image, including：The circular target region and the target depth image are contrasted, to be extracted from the target depth image comprising described The background area in circular target region；According to the rectangular degree of the background area, rectangular background frame is oriented from background area.
- 6. a kind of device for identifying circular traffic lights, it is characterised in that described device includes：Circular target region segmentation module, for be partitioned into the image with depth information that is collected from 3D cameras with The circular target region of red, green or yellow any color characteristics；Target depth image zooming-out module, for extracting satisfaction in the image with depth information that is collected from 3D cameras The target depth image of predetermined depth threshold range；Rectangular background frame locating module, for contrasting the circular target region and the target depth image, with from the mesh Rectangular background frame is oriented in mark depth image；Circular traffic lights determining module, the circle included for the area according to the rectangular background frame and the rectangular background frame Magnitude relationship between the area of target area, determine that the circular target region that the rectangular background frame includes is circular traffic Lamp；Circular traffic light kind determining module, for the color characteristics with reference to the circular traffic lights, determine the circular traffic The type of lamp.
- 7. device according to claim 6, it is characterised in that the circular traffic lights determining module includes：Area is than determination sub-module, circular mesh that the area for determining the rectangular background frame includes with the rectangular background frame Mark the area ratio in region；Circular traffic lights determination sub-module, if meeting preset range for ratio, the rectangular background frame is circular traffic lights Background frame, the circular target region that the rectangular background frame includes are circular traffic lights.
- 8. device according to claim 6, it is characterised in that the circular target region segmentation module includes：Target area determination sub-module is small for being partitioned into area in the image with depth information that is collected from 3D cameras In predetermined threshold value and with the target area of red, green or yellow any color characteristics；Submodule is filtered out, for the length-width ratio according to the target area, filters out Uncircular field；Judging submodule, for judge the target area after filtering out Hu features and circular traffic lights template Hu features whether Match somebody with somebody；First extracting sub-module, for extracting the region that the match is successful from the target area after described filter out.
- 9. device according to claim 8, it is characterised in that described device also includes：Histogram equalization module, the image with depth information for being collected to 3D cameras carry out histogram equalization Change；Noise reduction module, for carrying out noise reduction process to the image after histogram equalization；Morphological scale-space module, for carrying out Morphological scale-space to the image after noise reduction process；The target area determination sub-module includes：Split submodule, the image with depth information for being collected to the 3D cameras carries out color segmentation；Submodule is contrasted, for contrasting the image after Morphological scale-space and the image after color segmentation, with described in acquisition Target area.
- 10. device according to claim 6, it is characterised in that the rectangular background frame locating module includes：Background area extracting sub-module, for contrasting the circular target region and the target depth image, with from the mesh Extraction includes the background area in the circular target region in mark depth image；Rectangular background confines bit submodule, and for the rectangular degree according to the background area, rectangle is oriented from background area Background frame.
- 11. a kind of vehicle, it is characterised in that the vehicle includes：3D cameras, for gathering the image with depth information；AndThe device of the circular traffic lights of identification according to claim any one of 6-10.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610874308.8A CN107886033B (en) | 2016-09-30 | 2016-09-30 | Method and device for identifying circular traffic light and vehicle |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610874308.8A CN107886033B (en) | 2016-09-30 | 2016-09-30 | Method and device for identifying circular traffic light and vehicle |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN107886033A true CN107886033A (en) | 2018-04-06 |
| CN107886033B CN107886033B (en) | 2021-04-20 |
Family
ID=61769598
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610874308.8A Active CN107886033B (en) | 2016-09-30 | 2016-09-30 | Method and device for identifying circular traffic light and vehicle |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN107886033B (en) |
Cited By (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111445511A (en) * | 2020-03-24 | 2020-07-24 | 杭州东信北邮信息技术有限公司 | Method for detecting circle in image |
| CN111967370A (en) * | 2020-08-12 | 2020-11-20 | 广州小鹏车联网科技有限公司 | Traffic light identification method and device |
| CN112201117A (en) * | 2020-09-29 | 2021-01-08 | 深圳市优必选科技股份有限公司 | Logic board identification method and device and terminal equipment |
| CN112528794A (en) * | 2020-12-03 | 2021-03-19 | 北京百度网讯科技有限公司 | Signal lamp fault identification method and device and road side equipment |
| CN113761967A (en) * | 2020-06-01 | 2021-12-07 | 中移(苏州)软件技术有限公司 | Identification method and device |
Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101395645A (en) * | 2006-03-06 | 2009-03-25 | 丰田自动车株式会社 | Image processing system and method |
| CN102176287A (en) * | 2011-02-28 | 2011-09-07 | 无锡中星微电子有限公司 | Traffic signal lamp identifying system and method |
| US20110229023A1 (en) * | 2002-11-01 | 2011-09-22 | Tenebraex Corporation | Technique for enabling color blind persons to distinguish between various colors |
| CN102354457A (en) * | 2011-10-24 | 2012-02-15 | 复旦大学 | General Hough transformation-based method for detecting position of traffic signal lamp |
| CN103177256A (en) * | 2013-04-02 | 2013-06-26 | 上海理工大学 | Method for identifying display state of traffic signal lamp |
| CN103489324A (en) * | 2013-09-22 | 2014-01-01 | 北京联合大学 | Real-time dynamic traffic light detection identification method based on unmanned driving |
| CN104021378A (en) * | 2014-06-07 | 2014-09-03 | 北京联合大学 | Real-time traffic light recognition method based on space-time correlation and priori knowledge |
| CN104050447A (en) * | 2014-06-05 | 2014-09-17 | 奇瑞汽车股份有限公司 | Traffic light identification method and device |
| CN104766046A (en) * | 2015-02-06 | 2015-07-08 | 哈尔滨工业大学深圳研究生院 | Detection and recognition algorithm conducted by means of traffic sign color and shape features |
| CN105913041A (en) * | 2016-04-27 | 2016-08-31 | 浙江工业大学 | Pre-marked signal lights based identification method |
-
2016
- 2016-09-30 CN CN201610874308.8A patent/CN107886033B/en active Active
Patent Citations (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20110229023A1 (en) * | 2002-11-01 | 2011-09-22 | Tenebraex Corporation | Technique for enabling color blind persons to distinguish between various colors |
| CN101395645A (en) * | 2006-03-06 | 2009-03-25 | 丰田自动车株式会社 | Image processing system and method |
| CN102176287A (en) * | 2011-02-28 | 2011-09-07 | 无锡中星微电子有限公司 | Traffic signal lamp identifying system and method |
| CN102354457A (en) * | 2011-10-24 | 2012-02-15 | 复旦大学 | General Hough transformation-based method for detecting position of traffic signal lamp |
| CN103177256A (en) * | 2013-04-02 | 2013-06-26 | 上海理工大学 | Method for identifying display state of traffic signal lamp |
| CN103489324A (en) * | 2013-09-22 | 2014-01-01 | 北京联合大学 | Real-time dynamic traffic light detection identification method based on unmanned driving |
| CN104050447A (en) * | 2014-06-05 | 2014-09-17 | 奇瑞汽车股份有限公司 | Traffic light identification method and device |
| CN104021378A (en) * | 2014-06-07 | 2014-09-03 | 北京联合大学 | Real-time traffic light recognition method based on space-time correlation and priori knowledge |
| CN104766046A (en) * | 2015-02-06 | 2015-07-08 | 哈尔滨工业大学深圳研究生院 | Detection and recognition algorithm conducted by means of traffic sign color and shape features |
| CN105913041A (en) * | 2016-04-27 | 2016-08-31 | 浙江工业大学 | Pre-marked signal lights based identification method |
Non-Patent Citations (3)
| Title |
|---|
| 徐成等: ""基于Lab色彩空间和模板匹配的实时交通灯识别算法"", 《计算机应用》 * |
| 武莹等: ""基于图像处理的交通信号灯识别方法"", 《交通信息与安全》 * |
| 胡良梅等: ""融合颜色信息与深度信息的运动目标检测方法"", 《电子与信息学报》 * |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111445511A (en) * | 2020-03-24 | 2020-07-24 | 杭州东信北邮信息技术有限公司 | Method for detecting circle in image |
| CN111445511B (en) * | 2020-03-24 | 2022-08-05 | 杭州东信北邮信息技术有限公司 | Method for detecting circle in image |
| CN113761967A (en) * | 2020-06-01 | 2021-12-07 | 中移(苏州)软件技术有限公司 | Identification method and device |
| CN111967370A (en) * | 2020-08-12 | 2020-11-20 | 广州小鹏车联网科技有限公司 | Traffic light identification method and device |
| CN111967370B (en) * | 2020-08-12 | 2021-12-07 | 广州小鹏自动驾驶科技有限公司 | Traffic light identification method and device |
| CN112201117A (en) * | 2020-09-29 | 2021-01-08 | 深圳市优必选科技股份有限公司 | Logic board identification method and device and terminal equipment |
| CN112528794A (en) * | 2020-12-03 | 2021-03-19 | 北京百度网讯科技有限公司 | Signal lamp fault identification method and device and road side equipment |
Also Published As
| Publication number | Publication date |
|---|---|
| CN107886033B (en) | 2021-04-20 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107886033A (en) | Identify the method, apparatus and vehicle of circular traffic lights | |
| CN102314600B (en) | Shadow removal in image captured by vehicle-based camera for clear path detection | |
| CN103020623B (en) | Method for traffic sign detection and road traffic sign detection equipment | |
| CN102509098B (en) | Fisheye image vehicle identification method | |
| US10496891B2 (en) | Driver assistance system and method for object detection and notification | |
| US9911051B2 (en) | Method and apparatus for confirmation of relevant white inner circle in environment of circular traffic sign recognition | |
| EP2026313B1 (en) | A method and a system for the recognition of traffic signs with supplementary panels | |
| CN106909937A (en) | Traffic lights recognition methods, control method for vehicle, device and vehicle | |
| CN108765443A (en) | A kind of mark enhancing processing method of adaptive color Threshold segmentation | |
| CN106651872A (en) | Prewitt operator-based pavement crack recognition method and system | |
| EP2575077A2 (en) | Road sign detecting method and road sign detecting apparatus | |
| CN107886034B (en) | Driving reminding method and device and vehicle | |
| JP2011216051A (en) | Program and device for discriminating traffic light | |
| CN103324935B (en) | Vehicle is carried out the method and system of location and region segmentation by a kind of image | |
| CN104097565B (en) | A kind of automobile dimming-distance light lamp control method and device | |
| CN106682597A (en) | Method for identifying traffic lights | |
| CN110688907A (en) | Method and device for identifying object based on road light source at night | |
| CN107563301A (en) | Red signal detection method based on image processing techniques | |
| CN105023452A (en) | Multipath traffic signal lamp signal acquisition method and device | |
| CN107766802A (en) | A kind of motor vehicle front row driver and crew do not detain the self-adapting detecting method of safety belt | |
| CN107886035B (en) | Method and device for identifying arrow traffic lights and vehicle | |
| CN113989771A (en) | Traffic signal lamp identification method based on digital image processing | |
| KR20140022197A (en) | Lane detection method and lane departure warning system using same | |
| CN109800693B (en) | Night vehicle detection method based on color channel mixing characteristics | |
| CN110388985B (en) | Method for determining color of street sign and image processing system |
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 |