CN106650567A - Seatbelt detection method and seatbelt detection device - Google Patents
Seatbelt detection method and seatbelt detection device Download PDFInfo
- Publication number
- CN106650567A CN106650567A CN201610792780.7A CN201610792780A CN106650567A CN 106650567 A CN106650567 A CN 106650567A CN 201610792780 A CN201610792780 A CN 201610792780A CN 106650567 A CN106650567 A CN 106650567A
- Authority
- CN
- China
- Prior art keywords
- safety belt
- region
- detection zone
- straightway
- belt detection
- 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
- G06V20/59—Context or environment of the image inside of a vehicle, e.g. relating to seat occupancy, driver state or inner lighting conditions
- G06V20/597—Recognising the driver's state or behaviour, e.g. attention or drowsiness
Abstract
The embodiments of the invention disclose a seatbelt detection method and a seatbelt detection device. The method comprises the following steps: acquiring a shot image of the vehicle running condition on a road; performing vehicle detection on the image; segmenting out a vehicle image area; detecting a window area from the vehicle image area, and detecting a face area and/or a steering wheel area from the window area; determining a seatbelt detection area based on the face area and/or the steering wheel area; and extracting a line segment from the seatbelt detection area, and determining whether there is a seatbelt contained in the seatbelt detection area based on the line segment extracted. An intelligent, automatic and efficient seatbelt detection process is realized.
Description
Technical field
The present invention relates to information detection technology field, more particularly, it relates to a kind of Safe belt detection method and device.
Background technology
At present, most of traffics are all judged when safety belt detection is carried out using artificial naked eyes, and artificial judgment is not
But accuracy and it is ageing vary with each individual, and the traffic monitoring data of magnanimity cause manual detection expend human cost it is quite huge
Greatly.Therefore, how that safety belt detection process is intelligent, automation, high efficiency becomes the active demand of traffic.
The content of the invention
It is an object of the invention to provide a kind of Safe belt detection method and device, to realize safety belt detection process intelligence
Change, automate and efficient.
For achieving the above object, the invention provides following technical scheme：
A kind of Safe belt detection method, including：
Acquisition carries out shooting the image for obtaining to road vehicle travel situations；
Vehicle detection is carried out to described image, vehicle image region is partitioned into；
Vehicle window is detected in the vehicle image region；
Human face region and/or steering wheel region are detected in vehicle window region；
Safety belt detection zone is determined based on the human face region and/or steering wheel region；
Straightway is extracted in the safety belt detection zone, the safety belt detection is determined based on the straightway for being extracted
Whether safety belt is included in region.
Said method, it is preferred that described to detect that human face region and/or steering wheel region include in vehicle window region：
Target area is determined in the vehicle window region, the target area is that driver detects candidate region, or, it is
Passenger detects candidate region；
Human face region and/or steering wheel region are detected in the target area.
Said method, it is preferred that described to extract straightway in the safety belt detection zone, based on the straight line for being extracted
Section determines whether include comprising safety belt in the safety belt detection zone：
Angle of inclination is detected in the safety belt detection zone in the range of predetermined angle, and length is more than preset length
The straightway of threshold value；
If angle of inclination is not detected by the safety belt detection zone in the preset range, and length is more than default
The straightway of length threshold, determines and do not include in the safety belt detection zone safety belt；
If two Line Segments are there are in the safety belt detection zone, and between two Line Segments
Distance determines and include in the safety belt detection zone safety belt in preset distance range；
If only existing straight line section in the safety belt detection zone, along the gradient phase with the straightway central point
With or rightabout scan for, if there is the contrary marginal point of gradient in the distance range of the straightway,
With the marginal point as border, a new straightway parallel with the straightway is generated, in the straightway and the new straight line
Number of the luminance difference less than the pixel of preset difference value threshold value is determined in region between section, if determined by described in number accounts for
The total percentage of the pixel in region is more than preset percentage threshold value, determines in the safety belt detection zone comprising peace
Full band.
Said method, it is preferred that described to detect angle of inclination in predetermined angle scope in the safety belt detection zone
It is interior, and length includes more than the straightway of pre-set length threshold：
Edge feature is extracted in the safety belt detection zone based on neighbor gray scale logarithmic difference；
Edge feature to extracting carries out Probabilistic Hough Transform, obtains some straightways, and the start-stop of each bar straightway
Position；
Angle of inclination is selected from some straightways in the range of predetermined angle, and length is more than preset length threshold
The straightway of value.
Said method, it is preferred that also include：
If being not detected by safety belt in the safety belt detection zone, the direction ladder of the safety belt detection zone is extracted
Degree histogram feature；
Based on the histograms of oriented gradients feature, using safety belt detection zone described in the detection of classifier for having trained
Whether there is safety belt in domain.
A kind of safety belt detection means, including：
Road vehicle travel situations are carried out shooting the image for obtaining by acquisition module for obtaining；
First detection module, for carrying out vehicle detection to described image, is partitioned into vehicle image region；
Second detection module, for detecting vehicle window in the vehicle image region；
3rd detection module, in vehicle window region human face region and/or steering wheel region；
First determining module, for determining safety belt detection zone based on the human face region and/or steering wheel region；
Second determining module, for extracting straightway in the safety belt detection zone, the straight line based on the extraction
Section determines in the safety belt detection zone whether include safety belt.
Said apparatus, it is preferred that the 3rd detection module includes：
First determining unit, for determining target area in the vehicle window region, the target area is driver's inspection
Astronomical observation favored area, or, it is passenger detection candidate region；
First detector unit, for detecting human face region and/or steering wheel region in the target area.
Said apparatus, it is preferred that second determining module, including：
Second detector unit, for the detection angle of inclination in the safety belt detection zone in the range of predetermined angle,
And length is more than the straightway of pre-set length threshold；
Second determining unit, if for being not detected by angle of inclination in the safety belt detection zone in preset range,
And length is more than the straightway of pre-set length threshold, determines and do not include in the safety belt detection zone safety belt；
3rd determining unit, if for there are two Line Segments, and described two in the safety belt detection zone
The distance between bar Line Segment determines and include in the safety belt detection zone safety belt in preset distance range；
4th determining unit, if for only existing straight line section in the safety belt detection zone, along straight with this
The gradient of line segment central point is identical or rightabout is scanned for, if there is ladder in the distance range of the straightway
The contrary marginal point of degree, then with the marginal point as border, generate a new straightway parallel with the straightway, in the straight line
Number of the luminance difference less than the pixel of preset difference value threshold value is determined in region between section and the new straightway, if institute
It is determined that the total percentage of pixel that accounts in the region of number more than preset percentage threshold value, determine the safety belt
Safety belt is included in detection zone.
Said apparatus, it is preferred that second detector unit includes：
Subelement is extracted, it is special for edge to be extracted in the safety belt detection zone based on neighbor gray scale logarithmic difference
Levy；
Conversion subelement, for carrying out Probabilistic Hough Transform to the edge feature for extracting, obtains some straightways, and
The start-stop position of each bar straightway；
Subelement is selected, it is for angle of inclination to be selected from some straightways in the range of predetermined angle and long
Straightway of the degree more than pre-set length threshold.
Said apparatus, it is preferred that also include：
Extraction module, if not detecting safety belt in safety belt detection zone for second determining module, carries
Take the histograms of oriented gradients feature of the safety belt detection zone；
Classification and Detection module, for based on the histograms of oriented gradients feature, using the grader inspection for having trained
Survey in the safety belt detection zone and whether there is safety belt.
A kind of Safe belt detection method provided by above scheme, the application and device, are obtained on road
Vehicle travel situations carry out shooting the image for obtaining；Vehicle detection is carried out to described image, vehicle image region is partitioned into；Institute
State and detect in vehicle image region vehicle window region, and human face region and/or steering wheel region are detected in vehicle window region；Based on institute
State human face region and/or steering wheel region determines safety belt detection zone；Straightway is extracted in the safety belt detection zone,
Whether determined in the safety belt detection zone comprising safety belt based on the straightway of the extraction.Realize safety belt detection process
Intelligent, automation and high efficiency.
Description of the drawings
In order to be illustrated more clearly that the embodiment of the present application or technical scheme of the prior art, below will be to embodiment or existing
The accompanying drawing to be used needed for having technology description is briefly described, it should be apparent that, drawings in the following description are only this
Some embodiments of invention, for those of ordinary skill in the art, on the premise of not paying creative work, can be with
Other accompanying drawings are obtained according to these accompanying drawings.
A kind of flowchart of the Safe belt detection method that Fig. 1 is provided for the embodiment of the present application；
One kind that human face region and/or steering wheel region are detected in vehicle window region that Fig. 2 is provided for the embodiment of the present application
Flowchart；
Fig. 3 detects angle of inclination in predetermined angle scope for what the embodiment of the present application was provided in safety belt detection zone
It is interior, and length is more than a kind of flowchart of the straightway of pre-set length threshold；
A kind of structural representation of the safety belt detection means that Fig. 4 is provided for the embodiment of the present application；
A kind of structural representation of the 3rd detection module that Fig. 5 is provided for the embodiment of the present application；
A kind of structural representation of the second determining module that Fig. 6 is provided for the embodiment of the present application；
Another kind of structural representation of the safety belt detection means that Fig. 7 is provided for the embodiment of the present application.
Term " first ", " second ", " the 3rd " " 4th " in specification and claims and above-mentioned accompanying drawing etc. (if
Exist) it is part for distinguishing similar, without for describing specific order or precedence.It should be appreciated that so using
Data can exchange in the appropriate case, so that embodiments herein described herein can be with except illustrating here
Order in addition is implemented.
Specific embodiment
Below in conjunction with the accompanying drawing in the embodiment of the present application, the technical scheme in the embodiment of the present application is carried out clear, complete
Site preparation is described, it is clear that described embodiment is only a part of embodiment of the invention, rather than the embodiment of whole.It is based on
Embodiment in the present invention, it is every other that those of ordinary skill in the art are obtained under the premise of creative work is not paid
Embodiment, belongs to the scope of protection of the invention.
Refer to Fig. 1, a kind of flowchart of the Safe belt detection method that Fig. 1 is provided for the embodiment of the present application can be with
Including：
Step S11：Acquisition carries out shooting the image for obtaining to road vehicle travel situations；
Due to being that safety belt is detected, and the optimal visibility scope of safety belt is the front windshield vehicle window region of vehicle,
Therefore in the embodiment of the present application, when shooting to road vehicle travel situations, the front of vehicle should be shot,
So as to photograph the front windshield vehicle window region of vehicle, and then safety belt is detected.Specifically, the shooting of capture apparatus
The direction of head should be with the travel direction of vehicle on road conversely, so as to vehicle is when the shooting area of capture apparatus, can clap
Take the photograph the direct picture of vehicle.
Step S12：Image to obtaining carries out vehicle detection, is partitioned into vehicle image region；
The gradient distribution of acquired image can be calculated, the shade candidate regions of underbody are determined according to the gradient distributed intelligence
Domain, according to the underbody shade candidate region vehicle image region is extracted.The method robustness is stronger, segmentation result by shade,
The impact of the aspects such as complex illumination is less.
Step S13：Vehicle window is detected in vehicle image region, vehicle window region is partitioned into；
Optionally, vehicle window region can be partitioned into vehicle image region based on AdaBoost methods.AdaBoost side
Method is realized based on Haar-like features.The Haar-like features of standard have 15 kinds, as shown in figure 3, including four classes：Edge
Feature, line feature, point feature (central feature) and diagonal feature.
In the embodiment of the present invention, Haar-like features are extended, except including 15 kinds of features shown in Fig. 3 also,
Corner features are also increased newly, as shown in figure 4, the 4 kinds of Haar-like features extended for the embodiment of the present invention, add the 15 of standard
Haar-like features are planted, totally 19 kinds of Haar-like features.
The specific implementation that vehicle window region is partitioned into vehicle image region based on AdaBoost methods can be：
AdaBoost cascade classifiers are built by Sample Storehouse；Positive sample wherein in Sample Storehouse for vehicle vehicle window (i.e.
Front windshield wind) position image (in the image have complete vehicle window), negative sample is to be not more than with the registration of window locations
30% vehicle other area images；
Composition Weak Classifier is trained to vehicle image by above-mentioned 19 kinds of Haar-like features, is then passed through
Weak Classifier is superimposed cascade of strong classifiers in series by Adaboost
Using the AdaBoost cascade classifiers for training, vehicle image region is detected, complete vehicle window and determine
Position.
Found by Experimental comparison, after extension Haar-like features, the vehicle window localization method based on AdaBoost methods
Robustness it is higher, locating accuracy is higher, hence it is evident that improve the vehicle window locating effect under complex environment.
Step S14:Human face region and/or steering wheel region are detected in vehicle window region；
For the position (i.e. main driving position) of driver, human face region can be only detected, or, can only detect direction
Disk area, or, human face region had both been detected, steering wheel region is detected again.
For the position of copilot, due to without steering wheel, then can only detect human face region.
Step S15：Safety belt detection zone is determined based on human face region and/or steering wheel region；
The detailed process of safety belt detection zone is illustrated in the present embodiment as a example by based on domestic driving rule.
If only detecting human face region, determine that the implementation of safety belt detection zone can be according to human face region：
For main driving region, safety belt detection zone determination mode is：
xs=xf-α×widthf,
widths=xw+widthw-xs,
heights=yw+heightw-ys,
For copilot region, safety belt detection zone determination mode is：
xs=xw
widths=xf+α×widthf,
heights=yw+heightw-ys,
Wherein, (xw,yw) for vehicle window region left upper apex coordinate；widthwFor the width in vehicle window region；heightwFor
The height in vehicle window region；(xf, yf) for human face region left upper apex coordinate；widthfFor the width of human face region；heightf
For the height of human face region；(xs, ys) for safety belt detection zone left upper apex coordinate；widthsFor safety belt detection zone
The width in domain；heightsFor the height of safety belt detection zone；α is pre-set factory, can be set according to service condition, optional
, the value of α can be 1.
If only detecting steering wheel region, can determine that the implementation of safety belt detection zone can be with according to steering wheel region
For：
xs=x0,
ys=yw,
widths=xw+widthw-xs,
heights=y0-ys,
Wherein, (x0,y0) be the steering wheel for fitting centre coordinate；(xs, ys) for safety belt detection zone upper left push up
The coordinate of point；(xw,yw) for vehicle window region left upper apex coordinate；widthwFor the width in vehicle window region；widthsFor safety belt
The width of detection zone；heightsFor the height of safety belt detection zone.
If both having detected human face region, steering wheel region is detected again, then safety is determined according to human face region and steering wheel region
Implementation with detection zone can be：
xs=min (x0,xf-α×widthf),
widths=xw+widthw-xs,
heights=y0-ys,
(xs, ys) for safety belt detection zone left upper apex coordinate；(x0,y0) be the steering wheel for fitting center
Coordinate；(xf, yf) for human face region left upper apex coordinate；xwFor the coordinate of the x-axis of vehicle window region left upper apex；widthsFor
The width of safety belt detection zone；heightsFor the height of safety belt detection zone；widthfFor the width of human face region；
heightfFor the height of human face region；widthwFor the width in vehicle window region.
Safety belt detection zone is determined according to human face region and/or steering wheel region, is pacified in the detection of safety belt detection zone
Full band so that safety belt testing result is more accurate.Reduce the probability of flase drop.
Step S16：Straightway is extracted in safety belt detection zone, determines that safety belt is detected based on the straightway for being extracted
Whether safety belt is included in region.
Because safety belt is usually vertical bar shaped sheet, therefore, pass through extracted straightway and can determine that safety belt is detected
Whether safety belt is included in region.
The Safe belt detection method that the embodiment of the present application is provided, acquisition carries out shooting to road vehicle travel situations
The image for arriving；Vehicle detection is carried out to described image, vehicle image region is partitioned into；Car is detected in the vehicle image region
Window region, and human face region and/or steering wheel region are detected in vehicle window region；Based on the human face region and/or steering wheel
Region determines safety belt detection zone；Straightway is extracted in the safety belt detection zone, the straightway based on the extraction
Whether determine in the safety belt detection zone comprising safety belt.Realize that safety belt detection process is intelligent, automate and efficient
Change.
Optionally, what the embodiment of the present application was provided detects the one of human face region and/or steering wheel region in vehicle window region
Flowchart is planted as shown in Fig. 2 can include：
Step S21：Determine target area in vehicle window region, the target area is that driver detects candidate region, or,
Candidate region is detected for copilot passenger；
For the vehicle of driving custom (keeping to the right) for meeting China, if whether detection driver fastens the safety belt, will
The right side area in vehicle window region detects candidate regions as driver.If whether detection copilot passenger fastens the safety belt, will
The left field in vehicle window region detects candidate region as copilot passenger.
The size of target area can be determined by experiment, to ensure that target area includes complete driver's image or complete
Whole copilot passenger image.
Step S22：Human face region and/or steering wheel region are detected in the target area.
Noise in reduce image, can detect in the target area before human face region and/or steering wheel region, first
Smothing filtering is carried out to target area using Gaussian filter, to improve the detection of human face region and/or steering wheel region
Precision.Then human face region and/or steering wheel region are detected in the target area after smothing filtering.
Optionally, due to steering wheel edge approximately with two circular arcs, it is therefore possible to use the side of random Hough transformation
Method carries out in the target area circular arc (circle, ellipse etc.) detection.By the method for random Hough transformation, can be fitted and obtain ellipse
Round central coordinate of circle (x0,y0) and oval radius, central coordinate of circle (x0,y0) be aforementioned steering wheel centre coordinate.
Optionally, what the embodiment of the present application was provided extracts straightway in safety belt detection zone, straight based on what is extracted
Line segment determines whether a kind of implementation comprising safety belt can be in safety belt detection zone：
Angle of inclination is detected in safety belt detection zone in the range of predetermined angle, and length is more than pre-set length threshold
Straightway；
In safety belt detection zone, angle of inclination may be can't detect in the range of predetermined angle, and length is more than pre-
If the preset straightway of length；It is likely to only detect an angle of inclination in the range of predetermined angle, and length is more than default
The preset straightway of length；Or, two or more angles of inclination are detected in the range of predetermined angle, and length is more than pre-
If the preset straightway of length.
If angle of inclination is not detected by safety belt detection zone in preset range, and length is more than pre-set length threshold
Straightway, to determine in safety belt detection zone and do not include safety belt；
If two Line Segments are there are in safety belt detection zone, and the distance between two Line Segments exist
In preset distance range, determine and include in safety belt detection zone safety belt；
If that is, detecting at least two straightways in safety belt detection zone, judging this at least two directly
In line segment, if there are two straightways being parallel to each other, and the distance between two straightways being parallel to each other are preset
Distance range in, if exist meeting two straightways of above-mentioned condition, determine to include safety belt in safety belt detection zone.
If only exist in safety belt detection zone straight line section (for convenience of describe, be designated as first straight line section), then along
It is identical with the gradient of first straight line section central point or rightabout is scanned for, if in the above-mentioned distance range apart from the straightway
Inside there is the contrary marginal point of gradient, then with the marginal point as border, generate a new straightway parallel with the straightway
(for convenience of describing, being designated as second straight line section), in the region between first straight line section and second straight line section luminance difference is determined
Less than the number of the pixel of preset difference value threshold value, if determined by the area that accounts between first straight line section and second straight line section of number
The total percentage of the pixel in domain is more than preset percentage threshold value, determines and include in safety belt detection zone safety belt.
Optionally, it is provided in an embodiment of the present invention to detect angle of inclination in predetermined angle scope in safety belt detection zone
It is interior, and length is more than a kind of flowchart of the straightway of pre-set length threshold as shown in figure 3, can include：
Step S31：Edge feature is extracted in safety belt detection zone based on neighbor gray scale logarithmic difference；
Specifically, edge feature detection process can be：
For any one pixel a(i,j), calculate the four pixel (as adjacent with the pixel(i-1,j), a(i+1,j),
a(i,j+1), a(i,j-1)) logarithm value, calculate pixel a(i,j)Two pixels in left and right logarithm value difference absolute value,
And pixel a(i,j)The difference of the logarithm value of two pixels up and down absolute value；If calculated two absolute values are equal
More than default edge threshold, it is determined that pixel a(i,j)For marginal point, i.e. pixel a(i,j)For an Edge Feature Points.
Step S32：Edge feature to extracting carries out Probabilistic Hough Transform (PPHT), obtains some straightways, and respectively
The start-stop position of bar straightway；
The length of straightway, and the angle of inclination of straightway can be calculated by the start-stop position of straightway.
Step S33：Angle of inclination is selected from above-mentioned some straightways in the range of predetermined angle, and length is more than pre-
If the straightway of length threshold.
Optionally, Safe belt detection method provided in an embodiment of the present invention can also include：
If not detecting that detection zone detects safety belt in safety belt by said method, can be to safety belt detection zone
Domain carries out secondary detection, to reduce loss.The specific implementation of secondary detection is carried out to safety belt detection zone can be：
Extract the histograms of oriented gradients feature (HOG features) of safety belt detection zone；
Based on the histograms of oriented gradients feature extracted, using grader (such as SVM classifier) detection for having trained
Whether there is safety belt in safety belt detection zone.
Corresponding with embodiment of the method, the embodiment of the present application also provides a kind of safety belt detection means.The embodiment of the present application
A kind of structural representation of the safety belt detection means of offer is as shown in figure 4, can include：
Acquisition module 41, first detection module 42, the second detection module 43, the 3rd detection module 44, the first determining module
45 and second determining module 46；Wherein,
Acquisition module 41 is used to obtain carries out shooting the image for obtaining to road vehicle travel situations；
Due to being that safety belt is detected, and the optimal visibility scope of safety belt is the front windshield vehicle window region of vehicle,
Therefore in the embodiment of the present application, when shooting to road vehicle travel situations, the front of vehicle should be shot,
So as to photograph the front windshield vehicle window region of vehicle, and then safety belt is detected.Specifically, the shooting of capture apparatus
The direction of head should be with the travel direction of vehicle on road conversely, so as to vehicle is when the shooting area of capture apparatus, can clap
Take the photograph the direct picture of vehicle.
First detection module 42 is used to carry out vehicle detection to described image, is partitioned into vehicle image region；
The gradient distribution of acquired image can be calculated, the shade candidate regions of underbody are determined according to the gradient distributed intelligence
Domain, according to the underbody shade candidate region vehicle image region is extracted.The method robustness is stronger, segmentation result by shade,
The impact of the aspects such as complex illumination is less.
Second detection module 43 is partitioned into vehicle window region for detecting vehicle window in the vehicle image region；
Optionally, vehicle window region can be partitioned into vehicle image region based on AdaBoost methods.AdaBoost side
Method is realized based on Haar-like features.The Haar-like features of standard have 15 kinds, as shown in figure 3, including four classes：Edge
Feature, line feature, point feature (central feature) and diagonal feature.
In the embodiment of the present invention, Haar-like features are extended, except including 15 kinds of features shown in Fig. 3 also,
Corner features are also increased newly, as shown in figure 4, the 4 kinds of Haar-like features extended for the embodiment of the present invention, add the 15 of standard
Haar-like features are planted, totally 19 kinds of Haar-like features.
The specific implementation that vehicle window region is partitioned into vehicle image region based on AdaBoost methods can be：
AdaBoost cascade classifiers are built by Sample Storehouse；Positive sample wherein in Sample Storehouse is the window locations of vehicle
Picture, negative sample is to be not more than 30% vehicle other area images with the registration of window locations；
Composition Weak Classifier is trained to vehicle image by above-mentioned 19 kinds of Haar-like features, is then passed through
Weak Classifier is superimposed cascade of strong classifiers in series by Adaboost.
Using the AdaBoost cascade classifiers for training, vehicle image region is detected, complete vehicle window and determine
Position.
Found by Experimental comparison, after extension Haar-like features, the vehicle window localization method based on AdaBoost methods
Robustness it is higher, locating accuracy is higher, hence it is evident that improve the vehicle window locating effect under complex environment.
3rd detection module 44, for detecting human face region and/or steering wheel region in vehicle window region；
For the position (i.e. main driving position) of driver, human face region can be only detected, or, can only detect direction
Disk area, or, human face region had both been detected, steering wheel region is detected again.
For the position of copilot, due to without steering wheel, then can only detect human face region.
First determining module 45 is used to determine safety belt detection zone based on the human face region and/or steering wheel region；
The detailed process of safety belt detection zone is illustrated in the present embodiment as a example by based on domestic driving rule.
If only detecting human face region, determine that the implementation of safety belt detection zone can be according to human face region：
For main driving region, safety belt detection zone determination mode is：
xs=xf-α×widthf,
widths=xw+widthw-xs,
heights=yw+heightw-ys,
For copilot region, safety belt detection zone determination mode is：
xs=xw
widths=xf+α×widthf,
heights=yw+heightw-ys,
Wherein, (xw,yw) for vehicle window region left upper apex coordinate；widthwFor the width in vehicle window region；heightwFor
The height in vehicle window region；(xf, yf) for human face region left upper apex coordinate；widthfFor the width of human face region；heightf
For the height of human face region；(xs, ys) for safety belt detection zone left upper apex coordinate；widthsFor safety belt detection zone
The width in domain；heightsFor the height of safety belt detection zone；α is pre-set factory, and optionally, the value of α can be 1.
If only detecting steering wheel region, can determine that the implementation of safety belt detection zone can be with according to steering wheel region
For：
xs=x0,
ys=yw,
widths=xw+widthw-xs,
heights=y0-ys,
Wherein, (x0,y0) be the steering wheel for fitting centre coordinate；(xs, ys) for safety belt detection zone upper left push up
The coordinate of point；(xw,yw) for vehicle window region left upper apex coordinate；widthwFor the width in vehicle window region；widthsFor safety belt
The width of detection zone；heightsFor the height of safety belt detection zone.
If both having detected human face region, steering wheel region is detected again, then safety is determined according to human face region and steering wheel region
Implementation with detection zone can be：
xs=min (x0,xf-α×widthf),
widths=xw+widthw-xs,
heights=y0-ys,
(xs, ys) for safety belt detection zone left upper apex coordinate；(x0,y0) it is the steering wheel for fitting
Heart coordinate；(xf, yf) for human face region left upper apex coordinate；xwFor the coordinate of the x-axis of vehicle window region left upper apex；widths
For the width of safety belt detection zone；heightsFor the height of safety belt detection zone；widthfFor the width of human face region；
heightfFor the height of human face region；widthwFor the width in vehicle window region.
Safety belt detection zone is determined according to human face region and/or steering wheel region, is pacified in the detection of safety belt detection zone
Full band so that safety belt testing result is more accurate.Reduce the probability of flase drop.
Second determining module 46 is used to extract straightway in the safety belt detection zone, the straight line based on the extraction
Section determines in the safety belt detection zone whether include safety belt.
Because safety belt is usually vertical bar shaped sheet, therefore, pass through extracted straightway and can determine that safety belt is detected
Whether safety belt is included in region.
The safety belt detection means that the embodiment of the present application is provided, take carries out shooting and obtains to road vehicle travel situations
Image；Vehicle detection is carried out to described image, vehicle image region is partitioned into；Vehicle window is detected in the vehicle image region
Region, and human face region and/or steering wheel region are detected in vehicle window region；Based on the human face region and/or direction panel
Domain determines safety belt detection zone；Straightway is extracted in the safety belt detection zone, the straightway based on the extraction is true
Whether safety belt is included in the fixed safety belt detection zone.Realize intelligent safety belt detection process, automation and high efficiency.
Optionally, a kind of structural representation of the 3rd detection module 44 that the embodiment of the present application is provided is as shown in figure 5, can be with
Including：
First determining unit 51 and the first detector unit 52；Wherein,
First determining unit 51 is used to determine target area in the vehicle window region that the target area to be driver's inspection
Astronomical observation favored area, or, it is passenger detection candidate region；
For the vehicle of driving custom (keeping to the right) for meeting China, if whether detection driver fastens the safety belt, will
The right side area in vehicle window region detects candidate regions as driver.If whether detection copilot passenger fastens the safety belt, will
The left field in vehicle window region detects candidate region as copilot passenger.
The size of target area can be determined by experiment, to ensure that target area includes complete driver's image or complete
Whole copilot passenger image.
First detector unit 52 is used to detect human face region and/or steering wheel region in the target area.
Noise in reduce image, can detect in the target area before human face region and/or steering wheel region, first
Smothing filtering is carried out to target area using Gaussian filter, to improve the detection of human face region and/or steering wheel region
Precision.Then human face region and/or steering wheel region are detected in the target area after smothing filtering.
Optionally, due to steering wheel edge approximately with two circular arcs, it is therefore possible to use the side of random Hough transformation
Method carries out in the target area circular-arc detection.
Optionally, a kind of structural representation of the second determining module 46 that the embodiment of the present application is provided is as shown in fig. 6, can be with
Including：
Second detector unit 61, the second determining unit 62, the 3rd determining unit 63 and the 4th determining unit 64；Wherein,
Second detector unit 61 is used to be detected in the safety belt detection zone angle of inclination in the range of predetermined angle,
And length is more than the straightway of pre-set length threshold；
In safety belt detection zone, angle of inclination may be can't detect in the range of predetermined angle, and length is more than pre-
If the preset straightway of length；It is likely to only inspection 43 and measures an angle of inclination in the range of predetermined angle, and length is more than pre-
If the preset straightway of length；Or, two or more angles of inclination are detected in the range of predetermined angle, and length is more than
The preset straightway of preset length.
If the second determining unit 62 is used to being not detected by angle of inclination in the safety belt detection zone in preset range,
And length is more than the straightway of pre-set length threshold, determines and do not include in the safety belt detection zone safety belt；
If the 3rd determining unit 63 is used to there are two articles of Line Segments, and described two in the safety belt detection zone
The distance between bar Line Segment determines and include in the safety belt detection zone safety belt in preset distance range；
If that is, detecting at least two straightways in safety belt detection zone, judging this at least two directly
In line segment, if there are two straightways being parallel to each other, and the distance between two straightways being parallel to each other are preset
Distance range in, if exist meeting two straightways of above-mentioned condition, determine to include safety belt in safety belt detection zone.
If the 4th determining unit 64 (for convenience of describing, is designated as only existing straight line section in safety belt detection zone
First straight line section), then along identical with the gradient of first straight line section central point or rightabout is scanned for, if straight apart from this
There is the contrary marginal point of gradient in the above-mentioned distance range of line segment, then with the marginal point as border, generate one with the straight line
The new straightway (for convenience of describing, being designated as second straight line section) of Duan Pinghang, between first straight line section and second straight line section
In region determine luminance difference less than preset difference value threshold value pixel number, if determined by number account for first straight line section with
The total percentage of the pixel in region between second straight line section is more than preset percentage threshold value, determines that safety belt is detected
Safety belt is included in region.
Optionally, the second detector unit 61 that the embodiment of the present application is provided can include：
Subelement is extracted, it is special for edge to be extracted in the safety belt detection zone based on neighbor gray scale logarithmic difference
Levy；
Specifically, edge feature detection process can be：
For any one pixel a(i,j), calculate the four pixel (as adjacent with the pixel(i-1,j), a(i+1,j),
a(i,j+1), a(i,j-1)) logarithm value, calculate pixel a(i,j)Two pixels in left and right logarithm value difference absolute value,
And pixel a(i,j)The difference of the logarithm value of two pixels up and down absolute value；If calculated two absolute values are equal
More than default edge threshold, it is determined that pixel a(i,j)For marginal point, i.e. pixel a(i,j)For an Edge Feature Points.
Conversion subelement, for carrying out Probabilistic Hough Transform to the edge feature for extracting, obtains some straightways, and
The start-stop position of each bar straightway；
The length of straightway, and the angle of inclination of straightway can be calculated by the start-stop position of straightway.
Subelement is selected, it is for angle of inclination to be selected from some straightways in the range of predetermined angle and long
Straightway of the degree more than pre-set length threshold.
Optionally, on the basis of embodiment illustrated in fig. 4, the embodiment of the present application provide safety belt detection means it is another
Structural representation is planted as shown in fig. 7, can also include：
Extraction module 71, if not detecting safety belt in safety belt detection zone for second determining module 46,
Then extract the histograms of oriented gradients feature (HOG features) of the safety belt detection zone；
Classification and Detection module 72 is used to be based on the histograms of oriented gradients feature, using the grader for having trained
(such as SVM classifier) is detected and whether there is in the safety belt detection zone safety belt.
In the present embodiment, if not detecting safety belt in safety belt detection zone, safety belt detection zone is carried out
Secondary detection, to reduce loss.
The step of method or algorithm with reference to described by the disclosure of invention, can be realized in the way of hardware, also may be used
By be by computing device software instruction in the way of realizing.Software instruction can be made up of corresponding software module, software mould
Block can be stored on RAM memory, flash memory, ROM memory, eprom memory, eeprom memory, register, hard disk,
In the storage medium of portable hard drive, CD-ROM or any other form well known in the art.A kind of exemplary storage medium
Coupled to processor, so as to enable a processor to from the read information, and can be to the storage medium write information.When
So, storage medium can also be the part of processor.Processor and storage medium may be located in ASIC.In addition, should
ASIC may be located in user equipment.Certainly, processor and storage medium can also be present in user equipment as discrete assembly
In.
Those skilled in the art it will be appreciated that in said one or multiple examples, work(described in the invention
Be able to can be realized with hardware, software, firmware or their any combination.When implemented in software, can be by these functions
It is stored in computer-readable medium or is transmitted as one or more instructions on computer-readable medium or code.
Computer-readable medium includes computer-readable storage medium and communication media, and wherein communication media includes being easy to from a place to another
Any medium of one place transmission computer program.Storage medium can be universal or special computer can access it is any
Usable medium.
The foregoing description of the disclosed embodiments, enables professional and technical personnel in the field to realize or using the present invention.
Various modifications to these embodiments will be apparent for those skilled in the art, as defined herein
General Principle can be realized in other embodiments without departing from the spirit or scope of the present invention.Therefore, the present invention
The embodiments shown herein is not intended to be limited to, and is to fit to and principles disclosed herein and features of novelty phase one
The most wide scope for causing.
Claims (10)
1. a kind of Safe belt detection method, it is characterised in that include：
Acquisition carries out shooting the image for obtaining to road vehicle travel situations；
Vehicle detection is carried out to described image, vehicle image region is partitioned into；
Vehicle window is detected in the vehicle image region；
Human face region and/or steering wheel region are detected in vehicle window region；
Safety belt detection zone is determined based on the human face region and/or steering wheel region；
Straightway is extracted in the safety belt detection zone, the safety belt detection zone is determined based on the straightway for being extracted
In whether include safety belt.
2. method according to claim 1, it is characterised in that described that human face region and/or side are detected in vehicle window region
Include to disk area：
Target area is determined in the vehicle window region, the target area is that driver detects candidate region, or, it is to ride
Personnel detect candidate region；
Human face region and/or steering wheel region are detected in the target area.
3. method according to claim 1, it is characterised in that described to extract straight line in the safety belt detection zone
Section, determines whether include comprising safety belt in the safety belt detection zone based on the straightway for being extracted：
Angle of inclination is detected in the safety belt detection zone in the range of predetermined angle, and length is more than pre-set length threshold
Straightway；
If angle of inclination is not detected by the safety belt detection zone in the preset range, and length is more than preset length
The straightway of threshold value, determines and do not include in the safety belt detection zone safety belt；
The distance between if two Line Segments are there are in the safety belt detection zone, and two Line Segments
In preset distance range, determine and include in the safety belt detection zone safety belt；
If only existing straight line section in the safety belt detection zone, along identical with the gradient of the straightway central point or
Rightabout is scanned for, if there is the contrary marginal point of gradient in the distance range of the straightway, with this
Marginal point is border, generate a new straightway parallel with the straightway, the straightway and the new straightway it
Between region in determine luminance difference less than preset difference value threshold value pixel number, if determined by number account for the region
The total percentage of interior pixel is more than preset percentage threshold value, determines in the safety belt detection zone comprising safety
Band.
4. method according to claim 3, it is characterised in that described to detect inclination angle in the safety belt detection zone
Degree is in the range of predetermined angle, and length includes more than the straightway of pre-set length threshold：
Edge feature is extracted in the safety belt detection zone based on neighbor gray scale logarithmic difference；
Edge feature to extracting carries out Probabilistic Hough Transform, obtains some straightways, and the start stop bit of each bar straightway
Put；
Angle of inclination is selected from some straightways in the range of predetermined angle, and length is more than pre-set length threshold
Straightway.
5. the method according to claim 1-4 any one, it is characterised in that also include：
If being not detected by safety belt in the safety belt detection zone, the direction gradient for extracting the safety belt detection zone is straight
Square figure feature；
Based on the histograms of oriented gradients feature, using in safety belt detection zone described in the detection of classifier for having trained
With the presence or absence of safety belt.
6. a kind of safety belt detection means, it is characterised in that include：
Road vehicle travel situations are carried out shooting the image for obtaining by acquisition module for obtaining；
First detection module, for carrying out vehicle detection to described image, is partitioned into vehicle image region；
Second detection module, for detecting vehicle window in the vehicle image region；
3rd detection module, in vehicle window region detection human face region and/or steering wheel region；
First determining module, for determining safety belt detection zone based on the human face region and/or steering wheel region；
Second determining module, for extracting straightway in the safety belt detection zone, the straightway based on the extraction is true
Whether safety belt is included in the fixed safety belt detection zone.
7. device according to claim 6, it is characterised in that the 3rd detection module includes：
First determining unit, for determining target area in the vehicle window region, the target area is that driver's detection is waited
Favored area, or, it is passenger detection candidate region；
First detector unit, for detecting human face region and/or steering wheel region in the target area.
8. device according to claim 6, it is characterised in that second determining module, including：
Second detector unit for detecting angle of inclination in the range of predetermined angle and long in the safety belt detection zone
Straightway of the degree more than pre-set length threshold；
Second determining unit, if for being not detected by angle of inclination in preset range in the safety belt detection zone and long
Degree determines and do not include in the safety belt detection zone safety belt more than the straightway of pre-set length threshold；
3rd determining unit, if for there are two Line Segments in the safety belt detection zone, and described two are put down
The distance between row straightway determines and include in the safety belt detection zone safety belt in preset distance range；
4th determining unit, if for only existing straight line section in the safety belt detection zone, along with the straightway
The gradient of central point is identical or rightabout is scanned for, if there is gradient phase in the distance range of the straightway
Anti- marginal point, then with the marginal point as border, generate a new straightway parallel with the straightway, in the straightway and
Number of the luminance difference less than the pixel of preset difference value threshold value is determined in region between the new straightway, if being determined
The total percentage of pixel that accounts in the region of number more than preset percentage threshold value, determine the safety belt detection
Safety belt is included in region.
9. device according to claim 8, it is characterised in that second detector unit includes：
Subelement is extracted, for extracting edge feature in the safety belt detection zone based on neighbor gray scale logarithmic difference；
Conversion subelement, for carrying out Probabilistic Hough Transform to the edge feature for extracting, obtains some straightways, and each bar
The start-stop position of straightway；
Subelement is selected, for angle of inclination to be selected from some straightways in the range of predetermined angle, and length is big
In the straightway of pre-set length threshold.
10. the device according to claim 6-9 any one, it is characterised in that also include：
Extraction module, if not detecting safety belt in safety belt detection zone for second determining module, extracts institute
State the histograms of oriented gradients feature of safety belt detection zone；
Classification and Detection module, for based on the histograms of oriented gradients feature, using the detection of classifier institute for having trained
State and whether there is in safety belt detection zone safety belt.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610792780.7A CN106650567B (en) | 2016-08-31 | 2016-08-31 | Safety belt detection method and device |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610792780.7A CN106650567B (en) | 2016-08-31 | 2016-08-31 | Safety belt detection method and device |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN106650567A true CN106650567A (en) | 2017-05-10 |
| CN106650567B CN106650567B (en) | 2019-12-13 |
Family
ID=58853178
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610792780.7A Active CN106650567B (en) | 2016-08-31 | 2016-08-31 | Safety belt detection method and device |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN106650567B (en) |
Cited By (8)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108470144A (en) * | 2018-01-31 | 2018-08-31 | 浙江工业大学 | A kind of vehicle steering wheel detection method based on curve segmentation |
| CN108470145A (en) * | 2018-01-31 | 2018-08-31 | 浙江工业大学 | A kind of vehicle steering wheel detection method based on slope of curve variation |
| CN109709541A (en) * | 2018-12-26 | 2019-05-03 | 杭州奥腾电子股份有限公司 | A kind of vehicle environment perception emerging system target erroneous detection processing method |
| CN110517261A (en) * | 2019-08-30 | 2019-11-29 | 上海眼控科技股份有限公司 | Seat belt status detection method, device, computer equipment and storage medium |
| CN111382617A (en) * | 2018-12-28 | 2020-07-07 | 北京嘀嘀无限科技发展有限公司 | Driver identification method and device |
| CN111950348A (en) * | 2020-06-29 | 2020-11-17 | 北京百度网讯科技有限公司 | Method and device for identifying wearing state of safety belt, electronic equipment and storage medium |
| CN113689447A (en) * | 2020-05-19 | 2021-11-23 | 华晨宝马汽车有限公司 | Method, device, apparatus and system for identifying strip-shaped fixing parts |
| WO2022135589A1 (en) * | 2020-12-26 | 2022-06-30 | 虹软科技股份有限公司 | Method and apparatus for detecting wearing of safety belt, and storage medium and processor |
Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20070195990A1 (en) * | 2006-02-16 | 2007-08-23 | Uri Levy | Vision-Based Seat Belt Detection System |
| CN103150556A (en) * | 2013-02-20 | 2013-06-12 | 西安理工大学 | Safety belt automatic detection method for monitoring road traffic |
| CN104123557A (en) * | 2014-06-29 | 2014-10-29 | 中南大学 | Method for detecting car safety belt fastening state based on road monitoring device |
| CN104657752A (en) * | 2015-03-17 | 2015-05-27 | 银江股份有限公司 | Deep learning-based safety belt wearing identification method |
| CN105373779A (en) * | 2015-11-10 | 2016-03-02 | 北京汉王智通科技有限公司 | Intelligent vehicle safety belt detection method and intelligent detection system |
| US20160078306A1 (en) * | 2014-09-15 | 2016-03-17 | Xerox Corporation | System and method for detecting seat belt violations from front view vehicle images |
-
2016
- 2016-08-31 CN CN201610792780.7A patent/CN106650567B/en active Active
Patent Citations (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20070195990A1 (en) * | 2006-02-16 | 2007-08-23 | Uri Levy | Vision-Based Seat Belt Detection System |
| CN103150556A (en) * | 2013-02-20 | 2013-06-12 | 西安理工大学 | Safety belt automatic detection method for monitoring road traffic |
| CN104123557A (en) * | 2014-06-29 | 2014-10-29 | 中南大学 | Method for detecting car safety belt fastening state based on road monitoring device |
| US20160078306A1 (en) * | 2014-09-15 | 2016-03-17 | Xerox Corporation | System and method for detecting seat belt violations from front view vehicle images |
| CN104657752A (en) * | 2015-03-17 | 2015-05-27 | 银江股份有限公司 | Deep learning-based safety belt wearing identification method |
| CN105373779A (en) * | 2015-11-10 | 2016-03-02 | 北京汉王智通科技有限公司 | Intelligent vehicle safety belt detection method and intelligent detection system |
Non-Patent Citations (1)
| Title |
|---|
| 黎华东: "智能交通中的违章识别算法研究", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108470144A (en) * | 2018-01-31 | 2018-08-31 | 浙江工业大学 | A kind of vehicle steering wheel detection method based on curve segmentation |
| CN108470145A (en) * | 2018-01-31 | 2018-08-31 | 浙江工业大学 | A kind of vehicle steering wheel detection method based on slope of curve variation |
| CN108470145B (en) * | 2018-01-31 | 2021-03-16 | 浙江工业大学 | Automobile steering wheel detection method based on curve slope change |
| CN108470144B (en) * | 2018-01-31 | 2021-05-04 | 浙江工业大学 | Automobile steering wheel detection method based on curve segmentation |
| CN109709541A (en) * | 2018-12-26 | 2019-05-03 | 杭州奥腾电子股份有限公司 | A kind of vehicle environment perception emerging system target erroneous detection processing method |
| CN111382617A (en) * | 2018-12-28 | 2020-07-07 | 北京嘀嘀无限科技发展有限公司 | Driver identification method and device |
| CN111382617B (en) * | 2018-12-28 | 2024-01-09 | 北京嘀嘀无限科技发展有限公司 | Driver identification method and device |
| CN110517261A (en) * | 2019-08-30 | 2019-11-29 | 上海眼控科技股份有限公司 | Seat belt status detection method, device, computer equipment and storage medium |
| CN113689447A (en) * | 2020-05-19 | 2021-11-23 | 华晨宝马汽车有限公司 | Method, device, apparatus and system for identifying strip-shaped fixing parts |
| CN111950348A (en) * | 2020-06-29 | 2020-11-17 | 北京百度网讯科技有限公司 | Method and device for identifying wearing state of safety belt, electronic equipment and storage medium |
| JP2021152966A (en) * | 2020-06-29 | 2021-09-30 | ベイジン バイドゥ ネットコム サイエンス アンド テクノロジー カンパニー リミテッド | Seat belt wearing state recognition method, recognition apparatus, electronic device, and recording medium |
| WO2022135589A1 (en) * | 2020-12-26 | 2022-06-30 | 虹软科技股份有限公司 | Method and apparatus for detecting wearing of safety belt, and storage medium and processor |
Also Published As
| Publication number | Publication date |
|---|---|
| CN106650567B (en) | 2019-12-13 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106650567A (en) | Seatbelt detection method and seatbelt detection device | |
| US9489586B2 (en) | Traffic sign recognizing apparatus and operating method thereof | |
| Aly | Real time detection of lane markers in urban streets | |
| US8194927B2 (en) | Road-lane marker detection using light-based sensing technology | |
| CN103235938B (en) | The method and system of car plate detection and indentification | |
| US8099213B2 (en) | Road-edge detection | |
| CN101350063B (en) | Method and apparatus for locating human face characteristic point | |
| CN109948552B (en) | Method for detecting lane line in complex traffic environment | |
| JP6220327B2 (en) | Traveling lane marking recognition device, traveling lane marking recognition program | |
| US8385601B2 (en) | In-vehicle white line recognition apparatus | |
| CN109800752B (en) | Automobile license plate character segmentation and recognition algorithm based on machine vision | |
| JP6723328B2 (en) | Vehicle detection method, night-time vehicle detection method and system based on dynamic light intensity | |
| US9367750B2 (en) | Lane departure warning system and method of controlling the same | |
| CN108009522B (en) | Road detection method, device and terminal | |
| CN103324930A (en) | License plate character segmentation method based on grey level histogram binaryzation | |
| CN107194393B (en) | Method and device for detecting temporary license plate | |
| WO2008020544A1 (en) | Vehicle detection device, vehicle detection method, and vehicle detection program | |
| EP2741234B1 (en) | Object localization using vertical symmetry | |
| CN112381870B (en) | Binocular vision-based ship identification and navigational speed measurement system and method | |
| JP5011268B2 (en) | Image detection apparatus and image detection method | |
| CN105303157A (en) | Algorithm to extend detecting range for AVM stop line detection | |
| KR101236223B1 (en) | Method for detecting traffic lane | |
| Huang et al. | Lane marking detection based on adaptive threshold segmentation and road classification | |
| CN105574869B (en) | A kind of cable architecture striation center line extraction method based on improvement Laplacian edge detections | |
| CN107977608B (en) | Method for extracting road area of highway video image |
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 |