CN109624666A - A kind of anti-glare method and system of automobile intelligent - Google Patents
A kind of anti-glare method and system of automobile intelligent Download PDFInfo
- Publication number
- CN109624666A CN109624666A CN201811600490.3A CN201811600490A CN109624666A CN 109624666 A CN109624666 A CN 109624666A CN 201811600490 A CN201811600490 A CN 201811600490A CN 109624666 A CN109624666 A CN 109624666A
- Authority
- CN
- China
- Prior art keywords
- light
- vehicle
- light source
- information
- driver
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Pending
Links
Classifications
-
- B—PERFORMING OPERATIONS; TRANSPORTING
- B60—VEHICLES IN GENERAL
- B60J—WINDOWS, WINDSCREENS, NON-FIXED ROOFS, DOORS, OR SIMILAR DEVICES FOR VEHICLES; REMOVABLE EXTERNAL PROTECTIVE COVERINGS SPECIALLY ADAPTED FOR VEHICLES
- B60J3/00—Antiglare equipment associated with windows or windscreens; Sun visors for vehicles
Landscapes
- Engineering & Computer Science (AREA)
- Mechanical Engineering (AREA)
- Traffic Control Systems (AREA)
Abstract
The invention discloses a kind of anti-glare method and system of automobile intelligent, system of the present invention is controlled by the way that different directions are entered with the difference of light of driver's eyes, the unexpected strong light on opposite is made not influence the vision of driver；System of the present invention is obtaining existing intense light source and is adjusting the light transmittance of the corresponding part of windshield, existing intense light source is made to will not influence the sight of driver；System of the present invention simultaneously, the light source that may issue strong light is predicted using the existing object detected, and possible light source is predicted in the position at next moment, and makes preventative adjusting, the unexpected strong light for being likely to occur opposite will not influence the sight of driver.
Description
Technical field
The present invention relates to a kind of anti-glare method and system of automobile intelligent, are identified more particularly, to a kind of by detection device
External environment brightness and the strong light of single-point and the intelligent automobile windshield auxiliary equipment for carrying out predictive adjusting.
Background technique
(brightness is lower than 10- in very dark environment2cd/m2When), such as without the night of light-illuminating, the cone cell of human eye
Photosensitization is lost, visual performance is replaced by rhabdocyte, and human eye loses sensibility colored ability, is only capable of distinguishing white and ash
Color, if occurring strong light in short-term at this moment, cone cell can directly work, and rod cell is because of internal rhodopsin strong
It is decomposed without playing a leading role rapidly under the irradiation of light, if strong light disappears, the rhodopsin in rod cell can be gradually recovered,
This time can be long, so cone cell works and is converted to that the time that rhabdocyte works is long, and here it is nights
In eyes by after the strong optical flare in opposite a period of time can not see thing around clearly the reason of, according to statistics, night in traffic accident data
Traffic accident accounts for 60% or more, and in the non-artificial factor for leading to night traffic accident, " headlight " is dazzling at high-incidence factor.In recent years, I
It is related with high beam to account for 30% to 40% and in rising trend in the murderous accident of state's night traffic accident；Needle
To the above problem, application No. is " 201210560307.8 ", " 201410048552.X ", " 201480064155.X ",
" 201510848696.8 ", " 201720490336.X ", " 201810955087.6 ", " 201710284333.5 " patent of invention mention
Solution is supplied, but there are the following problems for these schemes:
1, application No. is the patent of invention described device of " 201210560307.8 ", driver is driven using this device
Automobile, when encountering opposite automobile big light opening at night, application No. is the patent of invention described device meetings of " 201210560307.8 "
Front windshield is all become into black, driver is made to cannot see that other parts do not issue by force in addition to the headlight for vehicles of opposite
The part of light, is easy to appear accident.
2, application No. is the patent of invention described device of " 201410048552.X ", the light that this device compares outside car is strong
Degree is adjusted, be not for eyes entering light illumination variation and be adjusted, in night request for utilization number be
The patent of invention described device of " 201410048552.X ", opposite headlight open when, according to the light velocity and application No. is
The detection of the patent of invention described device of " 201410048552.X ", calculating speed, it is believed that interior outer brightness simultaneously increases,
Inside and outside light intensity difference is almost without this has the patent of invention described device application No. is " 201410048552.X " not
Reaction.
3, application No. is the patent of invention described devices of " 201480064155.X ", encounter opposite direction in detection driver's eyes
Just acted after common action when headlight is dazzling, driver be possible to eyes encounter opposite headlight it is dazzling when do not act, only
There is eye closing to close, application No. is the patent of invention described devices of " 201480064155.X " not to work；In addition, acting as
Used time driver is also because have occurred and that dazzling situation cannot see front and occur dangerous.
It 4, is camera on spectacle frame, at data application No. is the patent of invention described device of " 201510848696.8 "
Manage unit etc. because device is arranged on mirror holder, the size of mirror holder because different drivers needs different size, change in size,
The glasses that driver inherently wears degree are also possible to, even family-sized car, it is also desirable to while it being adapted to the size of two people,
Application No. is the patent of invention described devices of " 201510848696.8 " to be inconvenient to be adapted to；Because application No. is
" 201510848696.8 " patent of invention described device can will lead to mirror holder skew on spectacle frame, lead to Shen than heavier
Please number for " 201510848696.8 " patent of invention described device detect a position have error, thus impact effect.
5, application No. is the patent of invention described devices of " 201810955087.6 ", detect external light intensity beyond threshold values i.e.
Act, if application No. is the patent of invention described device of " 201810955087.6 " at night opposite headlight according to meeting after coming
By front windshield whole blackening, this will lead to the part that should not be seen and does not see, it should be observed that part also can't see
Problem, so impracticable；Application No. is the patent of invention described device of " 201810955087.6 ", there is also windshield colors
The slow problem of change procedure, so the patent of invention described device application No. is " 201810955087.6 " is impracticable.
6, application No. is non-whole screen controls all inside the patent of invention of " 201710284333.5 " and foregoing invention patent
Patent of invention, the light stream changed before and after opposite car light is not predicted, causes any variation of opposed vehicle that can all make
For eyes by repeatedly dazzling, i.e., opponent vehicle headlight, which causes dazzling, device to be adjusted, makes headlight according to extensive less than eyes, human eye
It answers, opposed vehicle is mobile, headlight shines eyes again, dazzling again.
7, application No. is the patents of invention of " 201710284333.5 " and foregoing invention patent described device, do not carry out pre-
The property surveyed protection just carries out the position of its headlight when opposed vehicle is not turned on headlight that is, when detecting opposite vehicle
Preconditioning, above-mentioned apparatus is passively adjusted when opening the headlight of opposed vehicle suddenly, and driver has occurred and that when adjusting
It is dazzling, anti-effect is not played, only post, so application No. is the patents of invention of " 201710284333.5 " and upper
It states that patent of invention described device is all impracticable, does not play anti-glare effect.
Above-mentioned all devices are all a kind of post measures, i.e., in opposite vehicle strong illumination to this vehicle driver's eye
Generated on eyeball it is dazzling in the case where, then adjust corresponding device, but driver's eyes have occurred and that dazzling situation at this time,
So above-mentioned patent is not very practical.
Summary of the invention
For existing product, the patent dazzling and vision Whiteout caused by the headlight of driving at night opposite, cause to drive
There is the problems such as dangerous situation is dealt with improperly in member or surrounding people, and the invention discloses a kind of anti-glare methods of automobile intelligent
And system, the present invention are controlled by the light prediction and difference that different directions are entered with driver's eyes, the vehicle for travelling opposite
The strong light occurred when unexpected switch-on distance light does not influence the vision of driver, the vehicular high beam lamp for travelling opposite from the distant to the near when
The strong light occurred does not influence the vision of driver.
In order to achieve the above-mentioned object of the invention, the technical solution adopted by the present invention are as follows: a kind of anti-glare method of automobile intelligent and
System includes information collection component point, control section, light difference control section, driver head's state detection portion, each portion
Divide specific as follows:
1, the information collection component point may issue strong light within the scope of pilot's line of vision for collecting, in front of detecting vehicle
Object (target), collection mode may is that
1) using shooting, radar detecting data object identification method collect object information, identify front automobile,
The object information of the possible intense light source such as battery truck, tricycle, information can have light source type, position, distance, the intensity of light source, know
Other method can be the artificial intelligence approaches such as artificial neural network, and process is:
It extracts clarification of objective (ORB, Hist, HOG, SIFT etc.)；
The corresponding classifier of training；
Sliding window search；
It repeats and wrong report is filtered.
Recognition methods is also possible to RCNN, SPP-NET, Fast-RCNN, Faster-RCNN, YOLO, SSD even depth
Algorithm is practised to identify target.
2) using this parking stall of one or several kinds of data sources such as CAN, acceleration transducer, gyroscope, GPS, Beidou
It sets, velocity information；
3) vehicle front information of road surface is identified using the data of shooting, radar detecting；
2, the control section, for controlling the other parts in this system, functional module is light source identification module, distance
Calculating position prediction module, eye position computing module, light difference control module, specific as follows:
Light source identification module, the information point being collected into according to the information collection component are identified and are had been sent from or may
The light source of strong light is issued, the information process of identification is: identifying existing headlight for vehicles information, received further according to the information collection component point
Automobile, the battery truck, tricycle target collected, according to vehicle, vehicle dead reckoning headlight where position；Identify existing automobile
Headlight informational function can also be realized by the information collection component point；Distance can be deleted directly beyond object, the light source of setting threshold values.
Distance calculates and position prediction module, calculates it in the possibility at next time point to the object that may issue strong light
Position, distance, and predict that object may issue the position at part next time point of strong light, distance, object may be sent out
The part size of strong light out；The velocity interval that opposite object is considered when prediction, the maximum dipping and heaving obtained according to vehicle tire
Amplitude and range, turning rate range relevant to speed, this vehicle speed, this vehicle angle of turn, this front side traffic information,
The range of next time point object and the range of light emitting source are obtained, i.e., the position of next time point same light emitting source may
Have it is multiple, prediction Kalman filtering can be used as prediction technique；Prediction can also be according to the difference Horn- of picture interframe
The methods of Schunck, Lucas-Kanade calculate its light stream；Its possible motion profile letter is obtained using curve-fitting method afterwards
Number, the position of next time point target is predicted with lopcus function；Or after calculating light stream using affine transformation predict it is next
The position of a time point target.
Eye position computing module, the calculated driver of data provided with driver head's state detection portion
The position of eyes and the light source identification module, the distance calculate and position prediction in object luminous component size, away from
From, position, with a distance from next time point, the position range at next time point, next time point object luminous component
Size information calculates the point for needing to control on light difference control section.
Light difference control module, according to the eye position calculated result, the information of existing light emitting source, next time
Possible light emitting source and the possible position range of light emitting source, the light transmission capacity needed in control area is adjusted.
3, light difference control section enters the every of eyes light for controlling under the control of the control module
The light transmittance in one region shows image in this region while cover region entering light.
4, driver head's state detection portion, for detect driver head state and two positions, driven
The position of human eye under the person's of sailing current head state.
In order to guarantee effect of the invention, the region that light difference control section covers appropriate can expand, and expand
The region that light difference control module needs to control light transmission capacity can be handled with dilation operation greatly, result that treated is for hiding
Lid.
Beneficial effects of the present invention are as follows:
1, opposite automobile, battery truck, tricycle, camera that tests the speed etc. in driving can be avoided to be likely to occur to the greatest extent
Light source dazzling situation caused by driver of strong light is not only suitable for day driving, is also suitable for night driving.
2, it not will cause Qiang Guangxiang when opposed vehicle occurs blocking, turning, jolt by the prediction to light emitting source and drive
The leakage of the person's of sailing eyes.
3, the automobile of normally travel never opens any lamp in opposed vehicle, pre-adjusts, opens suddenly greatly in opposed vehicle
When lamp, because there is pre-adjusting, the opposite vehicle headlight light for injecting this vehicle driver's eyes is made to be reduced to acceptable degree.
4, in the case where that can not identify opposed vehicle, unexpected strong light guide causes driver dazzling, system call interception is opposite
Automobile front lamp is to the light transmittance of this vehicle driver's eyes light to driver to acceptable degree.
5, existing intense light source is shielded or is substituted by the present invention, and predicts that is be likely to occur causes dazzling intense light source feelings
Condition, and according to prediction adjust corresponding device in advance, it would be possible to light intensity light source enter driver's eye light light transmittance
Reducing to the light when light intensity light source issues strong light and reaching driver's eyes is not in dazzling degree, can also be covered possible
Light intensity light source enters the light of driver's eye and separately shows that image substitutes in corresponding position.
6, of the invention to predict that possible intense light source position, the position of intense light source may is that
The object and specific location of strong light may currently be issued；
Next time point may issue the motion profile of strong light object.
7, system of the present invention, all will not shadow no matter the eyes of driver are seen whither when blocking opposite intense light source
The sight of driver is rung, the eyes of people automatic can be seen to opposite intense light source in the dark, after system of the present invention, be driven
The sight for the person of sailing will not be attracted by opposite strong light, not will cause caused by being attracted because of sight without paying attention to direction of advance object
It is dangerous.
Detailed description of the invention
Fig. 1 is the anti-glare method flow schematic diagram of automobile intelligent of the embodiment of the present invention.
Fig. 2 is the anti-glare system comprising modules schematic diagram of automobile intelligent of the embodiment of the present invention.
Specific embodiment
Illustrate specific working mode of the invention below by embodiment.
Anti-glare 1 process of the method and system embodiment such as attached drawing 1 of automobile intelligent of the present invention, 1 system group of the embodiment of the present invention
At module such as attached drawing 2, the embodiment of the present invention 1 includes 201 information collection components point, 202 control sections, 203 light difference control units
Point, 204 driver head's state detection portions and 205 communications portion, specific each section it is as follows:
1,201 information collection components point shine light intensity on shield glass and picture and measure for detecting
The distance of object, information collection component are divided into two cameras and computing device NVIDIA GTX 1060, have liquid crystal screening on camera
Blocking means, can adjust the light transmittance of camera every bit under control, and two cameras are mounted on automobile two sides A column and are clapped
It takes the photograph, functional module is:
1) object extraction module, the purpose of Objective extraction are that the vehicle that will be needed is extracted from the picture that camera is shot
Come, extraction process is to propose scene image sequence inputting, the adaptive background extraction based on mixed Gaussian background modeling, frame difference method
It takes vehicle target, morphology closed operation, boundary rectangle operation, size modes puppet target discrimination, extract vehicle target；To vehicle mesh
Mark carries out vehicle cab recognition, special using the picture extraction ORB of the preceding ratio different with all types of automobiles, each angle, covering in system
Levy training SVM classifier；Knowledge is otherwise:
Feature is extracted, the ORB feature of vehicle is extracted；
It is identified using vehicle characteristics with SVM, identifies vehicle model information；
The case where car light can only be taken for darkness, can not identify vehicle, directly use car light as target, on picture
All light sources to remove remaining calculate of target that object extraction module has identified be car light target.
2) it identifies intensity of light source module, identifies that intense light source, intense light source are and surrounding on the picture taken from camera
Part of the environment compared to brightness beyond setting threshold values, knowing otherwise is the method identification region for using Morphological Gradient, has been handled
Photo afterwards is separated into multiple regions with edge, calculates the brightness value of each regional center point, calculates the brightness value of picture, such as
The average brightness of the brightness conversation structure in fruit region is greater than setting threshold values, then the liquid crystal radical occlusion device on camera is by this region
Light transmission capacity reduce, until the absolute value of the difference of the average brightness of the brightness and picture in certain region is less than setting threshold values, record is saturating
Light quantity decreasing value, according to light transmission capacity decreasing value judge specific light source whether be vehicle headlight.
3) target range module is detected, is calculated according to the difference (parallax) of the same time shooting picture of two cameras
The distance of extracted target out, distance can be obtained with Z=fB/d, and wherein f is camera focus, and B is two image centers away from d
The parallax that target is detected on image is shot for two cameras same time；For automobile, calculate opposite automobile left-front corner away from
From predicting leftward position, distance using vehicle, right positions, range data if automobile left-front corner is blocked；For not having
There is the light emitting source for identifying vehicle directly to calculate the distance of light emitting source central point；Target for distance beyond reservation threshold is direct
It deletes.
2,202 control section, using full will T7 chip, for controlling the other parts in this system, functional module has light
Identifing source module, distance calculates and position prediction module, eye position computing module, light difference control module, specific as follows:
1) light source identification module, the information point being collected into according to the information collection component are identified and are had been sent from and may
Intense light source position, possible position is calculated to the vehicle model information, the location information process that identify in the light source for issuing strong light,
Possible position is the position of this headlight when front truck blocks one Side light of rear car；The feelings of car light can only be taken for darkness
Condition can not identify vehicle, and car light position is directly used to carry out the position of two Side lights of prediction as the position of a side lamp, this prediction
It is to obtain the possible position range of two Side lights according to all model datas of analysis, because vehicle is unknown, this position is basis
The range that the empirical value counted in advance obtains.
2) distance calculating and position prediction module, shine according to the possibility of the speed of this vehicle, angle of turn, vehicle front
Object information, and the opposite possible speed of object (being up to the road speed limit in navigation information), obtained according to vehicle tire
Maximum dipping and heaving amplitude, turning rate using this vehicle as coordinate origin it is calculated under to the object that may issue strong light
The possible position at one time point, distance, and predict object may issue strong light part next time point it is possible
Position, distance, object may issue the part size of strong light；The possible position of next time point same object has
One range；For identifying the opposite target of vehicle, predicted according to opposite target current location when the position of headlight, with
The speed limit of road and next detection time point prediction go out the position of next time point current goal headlight, according to the wheel of vehicle
Tyre data predicts possible fluctuating range, and predicting next time point current goal turning according to automobile turning ability causes
The variation of headlight position records the position range of all possible headlight；The case where car light can only be taken for darkness, knows
Do not go out vehicle, directly predicted with vehicle lamp light source position, the possibility for using vehicle lamp light source to carry out two sides as a side lights is strong
Light source position prediction and next time point position prediction, vehicle tire data can use a setting value when prediction；Prediction makes
With the statistical data of well in advance, the positional relationship of the good various known vehicle car lights of statistics in advance, according to a side lights when prediction,
It is predicted according to the car light positional relationship of all known vehicles the position of the other side.
3) eye position computing module, with the calculated driver's eye of the data of driver head's state detection portion
Two eyes of driver of driver head's state detection portion shooting are found using the Haar classifier trained in the position of eyeball
The central point of eyes in image is carried out the distance that eyes are calculated apart from detection with two cameras, according to eye by the image of eyeball
Position of the distance, eyes of eyeball on picture, calculates the position of eyes.
4) light difference control module calculates according to the eye position and adjusts the light transmission capacity needed in control area,
Divide the position of dual camera, distance calculating and position pre- according to the position of eyes, the position of light difference control, information collection component
The position for measuring out light source prediction, calculates the region that light difference control section needs to control light transmittance, and this region is become
Secretly.
It 3,203 light difference control section can be according to control section to be attached to the liquid crystal film on front windshield
Control instruction reduces some region of light transmittance.
4,204 driver head's state detection portion is used for for the binocular camera being mounted in front of driver's cabin driver
Detect the position of driver's eyes.
5,205 communications portion is CAN interface, for obtaining speed, the angle of turn information of vehicle.
1 workflow of the embodiment of the present invention is as follows:
The outer picture of 1.101 collecting vehicles, opens the dual camera on automobile A column, shoots required picture, shooting picture is right
Picture is identified, identifies the light source of the appearance such as automobile, battery truck, tricycle, and obtains the intensity of light source；
2.102, according to target information, calculate target range；
3.103 identification vehicles, the light source that can not be identified are calculated according to car light；
4.104 obtain current light source position, obtain current possible intense light source according to vehicle model information and distance, location information
Position；The light source that can not be identified is set according to the lamp position that one side lights of automobile calculate possible other two sides；
5.105 prediction light source positions in future 1, according to this current vehicle speed, opposite vehicle distance, lane speed limit, opposite vehicle vehicle
The information predictions such as type go out the position of next time point opposite direction vehicle intense light source；
6.106 prediction light source positions in future 2, according to this current vehicle speed, the light source that can not be identified according to automobile side
The lamp location information that car light calculates possible other two sides obtains the location information at next time point of three points；
7.107 obtain current eye locations information；
8.108 calculate occlusion area position, calculate the position that light difference control section needs to reduce light transmittance；
9.109 light difference control sections reduce the light transmittance for needing part according to above-mentioned calculating data.
2 process of the embodiment of the present invention such as attached drawing 1,2 system comprising modules of the embodiment of the present invention such as attached drawing 2, the present invention are implemented
Example 2 includes 201 information collection components point, 202 control sections, 203 light difference control sections, 204 driver head's state-detections
Partially, 205 communications portion, specific each section are as follows:
1,201 information collection components point shine light intensity on shield glass and picture and measure for detecting
The distance of object, information collection component are divided into a camera and an Ouster OS-1-16 line laser radar and computing device
NVIDIA GTX 1060 has liquid crystal radical occlusion device on camera, can adjust the light transmittance of camera every bit under control,
Camera and laser radar are mounted on outside the front windshield of inside-automobile rear mirror and are shot and detected, and functional module is:
1) Objective extraction, the purpose of Objective extraction are that the vehicle that will be needed is extracted from the picture that camera is shot,
It extracts the trained deep learning algorithm of the image of preceding laser radar with depth, identifies that laser radar is sent with deep learning when extraction
The image with depth information, obtain the information such as automobile, battery truck, tricycle.Vehicle cab recognition is carried out to vehicle target, in system
Using preceding with all types of automobiles, each angle, the picture extraction ORB feature training SVM classifier for covering different ratios；Identification
Mode is:
Feature is extracted, the ORB feature of target is extracted；
It is identified using target signature with SVM, identifies vehicle model information；
For can not identifying vehicle, directly using car light as target, removing Objective extraction mould with all light sources on picture
The target that block identifies is calculated as car light target.
2) it identifies the intensity of light source, identifies intense light source on the picture taken from camera, intense light source is and ambient enviroment
Part compared to brightness beyond setting threshold values uses the border circular areas on Hough transformation identification frame out；Calculate each region
The brightness value of central point calculates the brightness value of picture, if the average brightness of the brightness conversation structure in certain region is greater than setting threshold values,
Then the light transmission capacity in this region is reduced, until the absolute value of the difference of the average brightness of the brightness and picture in certain region is less than setting valve
Value, record light transmission capacity decreasing value, according to light transmission capacity decreasing value judge specific light source whether be vehicle headlight.
3) detect target range, according to the vehicle target identified, obtained from the data of laser radar target away from
From data.For automobile, the distance for taking out opposite automobile left-front corner uses vehicle, right lateral position if automobile left-front corner is blocked
It sets, range data predicts leftward position, distance；For not identifying that the light emitting source of vehicle directly takes out light emitting source central point
Distance.
2,202 control section, using full will T7 chip, for controlling the other parts in this system, the function of completion is
Light source identification, position prediction, eye position calculates, light difference controls, specific as follows:
1) light source identifies, according to the information that the information collection component point is collected into, identifies and has been sent from and may issue
Vehicle model information that the light source of strong light will identify that, position obtain the possible position of intense light source；For can not identify the light source of vehicle
Position is directly predicted with car light position.
2) position prediction, according to the object information that the possibility of the speed of this vehicle, angle of turn, vehicle front shines, with this
Vehicle is coordinate origin, calculates it in the possible position at next time point, distance to the object that may issue strong light, and predict
Object may issue the position at part next time point of strong light out, distance, object may issue the part size of strong light；In advance
The speed of opposite object, the maximum dipping and heaving amplitude obtained according to vehicle tire, turning rate are considered when survey, when next
Between the position put have it is multiple；For identifying the target of vehicle, predicted according to current location when the position of headlight, with road
Speed limit and next detection time point prediction go out the position of next time point current goal headlight, according to the tire number of vehicle
It is predicted that possible fluctuating out, predicting next time point current goal turning according to automobile turning ability leads to headlight position
Variation, the position of all possible headlight is recorded；For can not identify the light source of vehicle, directly with vehicle lamp light source position into
Possibility intense light source position prediction and next time point position of the vehicle lamp light source as side lights progress two sides are used in row prediction
Prediction；
3) eye position calculates, with the calculated driver's eyes of data of driver head's state detection portion
Two eyes of driver of driver head's state detection portion shooting are found using the Haar classifier trained in position
The central point of image is carried out the distance that eyes are calculated apart from detection with two cameras by image, according to the distance of eyes,
Position of the eyes on picture, calculates the position of eyes.
4) light difference controls, and is calculated according to the eye position and adjusts the light transmission capacity needed in control area, according to
The position of eyes, the position of light difference control, the position of information collection component shunt excitation optical radar, distance calculates and position prediction obtains
The position of light source prediction out, calculates the region that light difference control section needs to control light transmittance, and this region is dimmed.
It 3,203 light difference control section can be according to control section to be attached to the liquid crystal film on front windshield
Control instruction reduces the light transmittance of certain a part.
4,204 driver head's state detection portion is used for for the binocular camera being mounted in front of driver's cabin driver
Detect the position of driver's eyes.
5,205 communications portion is used for and vehicle communication for CAN interface, obtains speed, the angle of turn information of vehicle.
2 workflow of the embodiment of the present invention is as follows:
The outer picture of 1.101 collecting vehicles, opens laser radar and camera, shooting picture, obtain picture and every bit away from
From information, picture, range information are identified, identify the light source of the appearance such as automobile, battery truck, tricycle, and obtains light
The intensity in source；
2.102, according to target information, calculate target range；
3.103 identification vehicles, the light source that can not be identified are calculated according to car light；
4.104 obtain current light source position, obtain current possible intense light source according to vehicle model information and distance, location information
Position；The light source that can not be identified is set according to the lamp position that one side lights of automobile calculate possible other two sides；
5.105 prediction light source positions in future 1, according to this current vehicle speed, opposite vehicle distance, lane speed limit, opposite vehicle vehicle
The information predictions such as type go out the position of next time point opposite direction vehicle intense light source；
6.106 prediction light source positions in future 2, according to this current vehicle speed, the light source that can not be identified according to automobile side
The lamp location information that car light calculates possible other two sides obtains the location information at next time point of three points；
7.107 obtain current eye locations information；
8.108 calculate occlusion area position, calculate the position that light difference control section needs to reduce light transmittance；
9.109 light difference control sections reduce the light transmittance for needing part according to above-mentioned calculating data.
3 process of the embodiment of the present invention such as attached drawing 1,3 system comprising modules of the embodiment of the present invention such as attached drawing 2, the present invention are implemented
Example 3 includes 201 information collection components point, 202 control sections, 203 light difference control sections, 204 driver head's state-detections
Partially, 205 communications portion, in addition to 203 light difference control sections, other parts are the same as the embodiment of the present invention 1,203 light
Realize as follows with function in difference control section:
Light difference control section is the double-layer films being tightly attached on front part of vehicle windshield, and internal layer membrane is self-luminous
OLED film, outer layer is liquid crystal film, is respectively intended to block certain zonal ray under control performed by the control section and certain region is aobvious
Diagram picture, function are when there is opposite strong illumination to come, and external Qiang Guangyu eyes line is outside under control performed by the control section
Intersection point on layer film becomes opaque, and intersection point of the external Qiang Guangyu eyes line on internal layer membrane shows what camera photographed
The picture of corresponding position.
4 process of the embodiment of the present invention such as attached drawing 1,4 system comprising modules of the embodiment of the present invention such as attached drawing 2, the present invention are implemented
Example 4 includes 201 information collection components point, 202 control sections, 203 light difference control sections, 204 driver head's state-detections
Partially, 205 communications portion, exceptionally except 201 information collection components, other parts are collected with the embodiment of the present invention 1,201 information
Partial target extraction algorithm is as follows:
1, CNN network structure is constructed；
Two convolutional layers+pond layer finally connect two full articulamentums.
First layer convolution uses the convolution kernel of 32 7x7x1, step-length 1, and BORDER PROCESSING mode is " SAME ", activation primitive
For ReLU, it is followed by the pond layer of a 2x2, mode is Max Pooling；
Second layer convolution uses the convolution kernel of 50 3x3x32, step-length 1, and BORDER PROCESSING mode is " SAME ", activates letter
Number is ReLU, is followed by the pond layer of a 2x2, and mode is Max Pooling；
The full articulamentum of first layer: use 1024 neurons, activation primitive be ReLU；
The full articulamentum of the second layer: 10 neurons are used, using softmax classifier, for exporting result；
2, using log logarithm loss function, the optimizer of Adam algorithm is configured, the calculation expression of accuracy is established；
3, it identifies.
CNN network is trained with data before identification.
5 process of the embodiment of the present invention such as attached drawing 1,5 system comprising modules of the embodiment of the present invention such as attached drawing 2, the present invention are implemented
Example 5 includes 201 information collection components point, 202 control sections, 203 light difference control sections, 204 driver head's state-detections
Partially, 205 communications portion, exceptionally except 201 information collection components, other parts are the same as 2,201 information collection component of the embodiment of the present invention point
Object extraction algorithm uses Fast YOLO, and Fast YOLO structure is as follows:
Construct Fast YOLO network structure；
6 convolutional layers+pond layer finally connect 3 Dense hidden layers；
First layer convolution uses the convolution kernel of 16 3x3x1, step-length 1, and BORDER PROCESSING mode is " SAME ", activation primitive
For (LeakyReLU, is followed by the pond layer of a 2x2, and mode is Max Pooling；
Second layer convolution uses the convolution kernel of 32 3x3x1, step-length 1, and BORDER PROCESSING mode is " SAME ", activation primitive
For LeakyReLU, it is followed by the pond layer of a 2x2, mode is Max Pooling；
Third layer convolution uses the convolution kernel of 64 3x3x1, step-length 1, and BORDER PROCESSING mode is " SAME ", activation primitive
For LeakyReLU, it is followed by the pond layer of a 2x2, mode is Max Pooling；
4th layer of convolution uses the convolution kernel of 128 3x3x1, step-length 1, and BORDER PROCESSING mode is " SAME ", activates letter
Number is LeakyReLU, is followed by the pond layer of a 2x2, and mode is Max Pooling；
Layer 5 convolution uses the convolution kernel of 256 3x3x1, step-length 1, and BORDER PROCESSING mode is " SAME ", activates letter
Number is LeakyReLU, is followed by the pond layer of a 2x2, and mode is Max Pooling；
Layer 6 convolution uses the convolution kernel of 512 3x3x1, step-length 1, and BORDER PROCESSING mode is " SAME ", activates letter
Number is LeakyReLU, is followed by the pond layer of a 2x2, and mode is Max Pooling；
Multi-layer data is become into one-dimensional data
Dense layers of first layer: using parameter 256；
Dense layers of the second layer: use parameter 4096, activation primitive be ReLU；
Dense layers of third layer: using parameter 1470.
6 process of the embodiment of the present invention such as attached drawing 1,6 system comprising modules of the embodiment of the present invention such as attached drawing 2, the present invention are implemented
Example 6 includes 201 information collection components point, 202 control sections, 203 light difference control sections, 204 driver head's state-detections
Partially, 205 communications portion, exceptionally except 201 information collection components, other parts are the same as 1,201 information collection component of the embodiment of the present invention point
Vehicle cab recognition is as follows:
Vehicle cab recognition is carried out using Baidu's AI cloud identification interface.
Although the present invention has been described by way of example and in terms of the preferred embodiments, embodiment and attached drawing are not for limiting the present invention
's.Without departing from the spirit and scope of the invention, any equivalent change or retouch done, also belongs to the protection of the present invention
Range.Therefore protection scope of the present invention should be based on the content defined in the claims of this application.
Claims (10)
1. a kind of anti-glare system of automobile intelligent, it is characterised in that: it includes information collection component point, control section, light difference
Control section, driver head's state detection portion, each section are specific as follows:
The information collection component point may issue the object of strong light for obtaining, in front of detecting vehicle within the scope of pilot's line of vision,
And the position of object, range information are obtained, also obtain the speed, angle of turn information of this vehicle；
The control section, for controlling the other parts in this system, the function of completion is light source identification, light source distance calculating
Position prediction, eye position calculate, light difference controls；
Light difference control section, for control enter eyes light each region light transmittance or cover certain area
Image is shown in this region while the entering light of domain；
Driver head's state detection portion, the state of driver head obtains two positions for identification.
2. the anti-glare system of automobile intelligent according to claim 1, it is characterised in that:
The information collection component divides acquirement vehicle external light source information and the vehicle model information identified, and the control section, which is identified, works as
The intense light source of preceding outside is calculated by the driver eye positions' information obtained from driver head's state detection portion
Light difference control section needs the position of light light transmittance reduction controlling light difference control section for phase out
Partial light permeability rate is answered to reduce, the strong illumination for issuing intense light source is controllable into the light intensity of driver's eyes；
The information collection component divides the light source information of acquirement vehicle external environment and the vehicle model information identified, and the control section is pre-
The location information for measuring the intense light source that next time point is likely to occur, the intense light source position being likely to occur according to next time point
Confidence breath and the speed of this vehicle vehicle, angle of turn, this front side traffic information predict next time point intense light source may
The position of appearance calculates institute by the driver eye positions' information obtained from driver head's state detection portion
Light difference control section is stated to need the position of light light transmittance reduction controlling light difference control section for corresponding portion
Point light transmittance reduces, and the strong illumination for issuing intense light source in the position that next time point is likely to occur is into driver's eyes
Light intensity is controllable.
3. the anti-glare system of automobile intelligent according to claim 1, it is characterised in that:
The information collection component point, collection mode is:
Data based on camera, radar data, identify front possible intense light source information, information include position, away from
From, the intensity of light source；
From CAN, acceleration transducer, gyroscope, GPS, this vehicle speed of Beidou one or more of which, angle of turn letter
Breath；
The function of the control section, completion is as follows:
Light source identification identifies according to the information that the information collection component point is collected into and has been sent from or may issue strong light
Light source, to the vehicle model information identified, location information is by being calculated intense light source position, possible position；For identification
The light source for not going out vehicle, directly uses car light position to carry out the position of two Side lights of prediction as the position of a side lamp；
Distance calculate and position prediction, to the object that may issue strong light calculate its next time point possible position,
Distance, and predict that object may issue the position at part next time point of strong light, distance, object may issue strong light
Part size；The maximum dipping and heaving amplitude that is obtained when prediction using the speed of opposite object, according to vehicle tire, with speed phase
The turning rate of pass is predicted；
Eye position calculates, according to the calculated driver's eyes of data of driver head's state detection portion offer
Position and the distance calculate and position prediction in the size of object luminous component, distance, position, next time point away from
Size from, the position at next time point, next time point object luminous component calculates the light difference control unit
The point for needing to control on point；
The control of light difference is adjusted the light transmission capacity in the region for needing to control according to the eye position calculated result.
4. according to claim 1 to one of 3 the anti-glare system of automobile intelligent, it is characterised in that: light difference control when use
Dilation operation handles the region for needing to reduce light transmission capacity, and operation result will be corresponding for light difference control section
Partial light permeability rate reduces.
5. according to claim 1 to one of 3 the anti-glare system of automobile intelligent, it is characterised in that: the information collection component point is
Two cameras and computing device have liquid crystal radical occlusion device on camera, adjust the light transmittance of camera every bit under control,
The function of completion is also:
Objective extraction extracts vehicle target from the picture that camera is shot；
It identifies the intensity of light source, identifies that intense light source, intense light source are bright compared with ambient enviroment on the picture taken from camera
Degree beyond setting threshold values part, judge specific light source whether be vehicle headlight；
Target range is detected, the target extracted is calculated according to the difference of the same time shooting picture of two cameras
Distance, distance are obtained with Z=fB/d, and wherein f is camera focus, and B is two image centers away from d is two camera same times
Shoot the parallax that target is detected on image；
It light difference control section can be according to the control of control section to be attached to the liquid crystal film on front windshield
Instruction reduces the light transmittance of certain a part；
Driver head's state detection portion, for the binocular camera being mounted in front of driver's cabin driver, for detecting
The position of driver's eyes.
6. according to claim 1 to one of 3 the anti-glare system of automobile intelligent, it is characterised in that: the information collection component point is
One camera A and a camera laser radar two-in-one system have liquid crystal radical occlusion device on camera A, adjust under control
The light transmittance of every bit on camera A, the function of completion is:
Objective extraction, extracts the trained deep learning algorithm of the image of preceding laser radar with depth, and when extraction is known with deep learning
The image with depth information that other laser radar is sent obtains target information；
Target range is detected, according to the vehicle target identified, the range data of target is obtained from the data of laser radar；
Light difference control section, to be attached to the liquid crystal film on front windshield, according to the control instruction of control section by certain
The light transmittance of a part reduces；
Driver head's state detection portion drives for the binocular camera being mounted in front of driver's cabin driver for detecting
The position of member's eyes.
7. a kind of anti-glare method of automobile intelligent, it is characterised in that:
Step 1: the outer picture of collecting vehicle, opens the dual camera on automobile A column, or open laser radar and camera, shooting
Required picture, shooting picture identify picture, identify the light source of the appearance such as automobile, battery truck, tricycle, and obtain
To the intensity of light source；
Step 2: calculating target range according to target information；
Step 3: identification vehicle, the light source that can not be identified are calculated according to car light；
Step 4: obtaining current light source position, current possible intense light source position is obtained according to vehicle model information and distance, location information
It sets；The light source that can not be identified is set according to the lamp position that one side lights of automobile calculate possible other two sides；
Step 5: prediction light source position in future 1, according to this current vehicle speed, opposite vehicle distance, lane speed limit, opposite vehicle vehicle
Equal information predictions go out the position of next time point opposite direction vehicle intense light source；
Step 6: prediction light source position in future 2, according to this current vehicle speed, the light source that can not be identified according to automobile side vehicle
The lamp location information that lamp calculates possible other two sides obtains the location information at next time point of three points；
Step 7: obtaining current eye locations information；
Step 8: calculating occlusion area position, the position that light difference control section needs to reduce light transmittance is calculated；
Step 9: light difference control section reduces the light transmittance for needing part according to above-mentioned calculating data.
8. the anti-glare method of automobile intelligent according to claim 7, it is characterised in that: in step 1, the outer picture of vehicle mentions after collecting
The process taken be by scene image sequence inputting, adaptive background extract, frame difference method extract vehicle target, morphology closed operation,
Boundary rectangle operation, extracts vehicle target at size modes puppet target discrimination；The case where car light can only be taken for darkness, knows
Do not go out vehicle, directly uses car light as target.
9. the anti-glare method of automobile intelligent according to claim 7, it is characterised in that: in step 2, calculate target range, use
The Haar classifier trained finds the image of two eyes of driver of driver head's state detection portion shooting, will scheme
The central point of picture carries out the distance that eyes are calculated apart from detection with two cameras, according to the distance of eyes, eyes in picture
On position, calculate the position of eyes.
10. according to the anti-glare method of the automobile intelligent of one of claim 7 to 9, it is characterised in that: in step 3,
Identify that vehicle uses SVM classifier, training classifier before identifying is used preceding each using multiple SVM classifiers in system
A kind of automobile of SVM classifier, each angle, the picture extraction feature training SVM classifier for covering different ratios；Identification process
It is as follows:
Feature is extracted, the ORB feature of target is extracted；
It is identified using target signature with SVM；
The light source that can not be identified is to obtain the possible position of two Side lights according to all model datas of analysis, is a range.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811600490.3A CN109624666A (en) | 2018-12-26 | 2018-12-26 | A kind of anti-glare method and system of automobile intelligent |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811600490.3A CN109624666A (en) | 2018-12-26 | 2018-12-26 | A kind of anti-glare method and system of automobile intelligent |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109624666A true CN109624666A (en) | 2019-04-16 |
Family
ID=66077700
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811600490.3A Pending CN109624666A (en) | 2018-12-26 | 2018-12-26 | A kind of anti-glare method and system of automobile intelligent |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109624666A (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110466432A (en) * | 2019-09-18 | 2019-11-19 | 宝能汽车有限公司 | The working method of Antiglaring rear mirror, automobile and Antiglaring rear mirror |
| CN110569971A (en) * | 2019-09-09 | 2019-12-13 | 吉林大学 | convolutional neural network single-target identification method based on LeakyRelu activation function |
| CN112198965A (en) * | 2020-12-04 | 2021-01-08 | 宁波圻亿科技有限公司 | AR (augmented reality) glasses eye protection automatic control method and device |
| CN114520880A (en) * | 2020-11-18 | 2022-05-20 | 华为技术有限公司 | Exposure parameter adjusting method and device |
Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2005035384A (en) * | 2003-07-14 | 2005-02-10 | Nissan Motor Co Ltd | Automatic shielding device for vehicle |
| EP2583847A1 (en) * | 2011-10-18 | 2013-04-24 | Signum Bildtechnik GmbH | Device and method for reducing glare |
| CN103273826A (en) * | 2013-06-13 | 2013-09-04 | 北京汽车股份有限公司 | Automatic anti-dazzling system of front windshield of car |
| CN103985334A (en) * | 2014-05-04 | 2014-08-13 | 京东方科技集团股份有限公司 | Transparent display system and device with transparent display system |
| CN106143072A (en) * | 2015-03-11 | 2016-11-23 | 刘韵凯 | A kind of vehicle front-viewing intelligence anti-dazzle system based on electrochomeric glass |
| CN106994884A (en) * | 2017-05-05 | 2017-08-01 | 上海小糸车灯有限公司 | The anti-glare method and device of active based on OLED Display Techniques |
| CN108583237A (en) * | 2018-03-23 | 2018-09-28 | 广州市微龙电子科技有限公司 | Automobile intelligent electronic anti-dazzle system and control method |
-
2018
- 2018-12-26 CN CN201811600490.3A patent/CN109624666A/en active Pending
Patent Citations (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| JP2005035384A (en) * | 2003-07-14 | 2005-02-10 | Nissan Motor Co Ltd | Automatic shielding device for vehicle |
| EP2583847A1 (en) * | 2011-10-18 | 2013-04-24 | Signum Bildtechnik GmbH | Device and method for reducing glare |
| CN103273826A (en) * | 2013-06-13 | 2013-09-04 | 北京汽车股份有限公司 | Automatic anti-dazzling system of front windshield of car |
| CN103985334A (en) * | 2014-05-04 | 2014-08-13 | 京东方科技集团股份有限公司 | Transparent display system and device with transparent display system |
| CN106143072A (en) * | 2015-03-11 | 2016-11-23 | 刘韵凯 | A kind of vehicle front-viewing intelligence anti-dazzle system based on electrochomeric glass |
| CN106994884A (en) * | 2017-05-05 | 2017-08-01 | 上海小糸车灯有限公司 | The anti-glare method and device of active based on OLED Display Techniques |
| CN108583237A (en) * | 2018-03-23 | 2018-09-28 | 广州市微龙电子科技有限公司 | Automobile intelligent electronic anti-dazzle system and control method |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110569971A (en) * | 2019-09-09 | 2019-12-13 | 吉林大学 | convolutional neural network single-target identification method based on LeakyRelu activation function |
| CN110569971B (en) * | 2019-09-09 | 2022-02-08 | 吉林大学 | Convolutional neural network single-target identification method based on LeakyRelu activation function |
| CN110466432A (en) * | 2019-09-18 | 2019-11-19 | 宝能汽车有限公司 | The working method of Antiglaring rear mirror, automobile and Antiglaring rear mirror |
| CN114520880A (en) * | 2020-11-18 | 2022-05-20 | 华为技术有限公司 | Exposure parameter adjusting method and device |
| WO2022105381A1 (en) * | 2020-11-18 | 2022-05-27 | 华为技术有限公司 | Exposure parameter adjustment method and apparatus |
| CN114520880B (en) * | 2020-11-18 | 2023-04-18 | 华为技术有限公司 | Exposure parameter adjusting method and device |
| CN112198965A (en) * | 2020-12-04 | 2021-01-08 | 宁波圻亿科技有限公司 | AR (augmented reality) glasses eye protection automatic control method and device |
| CN112198965B (en) * | 2020-12-04 | 2021-03-09 | 宁波圻亿科技有限公司 | AR (augmented reality) glasses eye protection automatic control method and device |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109624666A (en) | A kind of anti-glare method and system of automobile intelligent | |
| US11341667B2 (en) | Camera systems using filters and exposure times to detect flickering illuminated objects | |
| CN105799574B (en) | Night intelligence DAS (Driver Assistant System) | |
| Wu et al. | Applying a functional neurofuzzy network to real-time lane detection and front-vehicle distance measurement | |
| WO2019192418A1 (en) | Automobile head-up display system and obstacle prompting method thereof | |
| CN111619324A (en) | Intelligent anti-dazzling method and system for sight tracking automobile | |
| CN106934808B (en) | Method for identifying and tracking tail lamp of automobile headlight under visual perception | |
| CN106965675B (en) | A kind of lorry swarm intelligence safety work system | |
| US11881039B2 (en) | License plate reading system with enhancements | |
| CN109624832A (en) | A kind of method and system of the anti-glare headlight of automobile intelligent | |
| CN105739705A (en) | Human-eye control method and apparatus for vehicle-mounted system | |
| CN102712282A (en) | Method for controlling a headlight arrangement for a vehicle and such a headlight arrangement | |
| CN102712283A (en) | Method for controlling a headlight arrangement for a vehicle and such a headlight arrangement | |
| CN104097565B (en) | A kind of automobile dimming-distance light lamp control method and device | |
| CN110450706A (en) | A kind of adaptive distance light lamp control system and image processing algorithm | |
| US10715738B1 (en) | Asymmetrical license plate reading (ALPR) camera system | |
| CN109624667A (en) | A kind of method and system that the anti-traffic photographic device of automobile intelligent is dazzling | |
| CN109584176A (en) | Motor vehicle driving vision enhancement system | |
| US11682219B2 (en) | Asymmetrical license plate reading (ALPR) camera system | |
| CN115116035A (en) | Road traffic light identification system and method based on neural network | |
| CN112330558A (en) | Road image recovery early warning system and method based on foggy weather environment perception | |
| Yu et al. | Vehicle and pedestrian target detection in auto driving scene | |
| CN108171747A (en) | A kind of efficient overnight sight vision positioning method | |
| CN116883978A (en) | Driver distraction fatigue detection method, system and storage medium based on infrared camera | |
| CN115447476A (en) | Method and system for controlling headlights of vehicle |
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 | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20190416 |
|
| RJ01 | Rejection of invention patent application after publication |