CN105760847B - A kind of visible detection method of pair of helmet of motorcycle driver wear condition - Google Patents
A kind of visible detection method of pair of helmet of motorcycle driver wear condition Download PDFInfo
- Publication number
- CN105760847B CN105760847B CN201610116365.XA CN201610116365A CN105760847B CN 105760847 B CN105760847 B CN 105760847B CN 201610116365 A CN201610116365 A CN 201610116365A CN 105760847 B CN105760847 B CN 105760847B
- Authority
- CN
- China
- Prior art keywords
- image
- helmet
- motorcycle
- feature
- classification
- 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.)
- Active
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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/20—Image preprocessing
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Multimedia (AREA)
- Theoretical Computer Science (AREA)
- Image Analysis (AREA)
- Traffic Control Systems (AREA)
Abstract
The present invention relates to the visible detection method of a kind of pair of helmet of motorcycle driver wear condition, mentioned method is divided into two stages: detection whether motorvehicle segmentation and classification (for judging whether target of interest is motorcycle) and the helmet are worn.Classify for motorvehicle, using common Haar feature as descriptor and SVM classifier；The helmet is detected, we extract head zone with Circle Hough Transform (circular hough transform, that is, CHT), and then utilization orientation histogram of gradients (HOG) descriptor extracts characteristics of image, and target is classified with multilayer neural network (MLP) classifier；The accuracy of this method detection is high, and real-time is good, has engineering practical value.
Description
Technical field
The invention belongs to technical field of image processing, are related to image segmentation and image information acquisition, and in particular to a kind of right
The visible detection method of helmet of motorcycle driver wear condition is mainly used in traffic safety monitoring and management.
Background technique
Traffic safety problem has become global significant problem, influence of the safety of motorcycle to human life's property
It is self-evident.In many countries, with the increase of motorcycle usage quantity, travel speed is also accordingly accelerated, and traffic is transported in addition
It is defeated increasingly busy, as do not wear the helmet and caused by motorcycle accident increase, and a large amount of casualties and property is brought to damage
It loses, which, which has become one, cannot be neglected social concern.Currently, relying only on the Motor-cyclist of traffic police's artificial judgment not
The supervision of the wearing helmet has been increasingly difficult to such a separated to meet growing motorcycle ownership and not wear the helmet
Judicial act.Therefore, how utilize extensive road monitoring, using visual sensing due to have contain much information, low-cost spy
Point solves the problems, such as that not wearing the helmet detects this, has display meaning.
Helmet detection technique, which refers to, searches and determines to the driver in image using image sensing means, goes out it with decision
Whether the process of the helmet is worn.The technology and methods of this aspect also compare shortage at present.In view of this, the invention proposes one
The helmet of the Motor-cyclist travelled on public way detects computer vision system.
The contents of the present invention are described for convenience, it is necessary first to which some concepts are illustrated.
Area-of-interest: in field of image processing, area-of-interest refers to the area, topography selected from image
Domain, this region are image analysis emphasis of interest.The region is determined to be further processed.Use area-of-interest
The processing time can be often reduced, precision is increased.
Summary of the invention
The invention proposes the visible detection method of a kind of pair of helmet of motorcycle driver wear condition, mentioned method is divided into
Two stages: whether motorvehicle segmentation and classification (for judging whether target of interest is motorcycle) and the helmet are worn
Detection.Classify for motorvehicle, using common Haar feature as descriptor and SVM classifier；The helmet is detected, I
Extract head zone with Circle Hough Transform (circular hough transform, that is, CHT), and then utilization orientation gradient is straight
Side's figure (HOG) descriptor extracts characteristics of image, and target is classified with multilayer neural network (MLP) classifier.The party
The accuracy of method detection is high, and real-time is good, has engineering practical value.
The visible detection method of a kind of pair of helmet of motorcycle driver wear condition, includes the following steps:
The acquisition of step 1) background image: image collecting device is installed at roadside, road operation conditions is adopted
Collection, the road real image extracted from video flowing are set as A, are mentioned in video streaming with adaptive GMM algorithm
The background image without moving object is taken to be set as B；
The segmentation of step 2) moving target: moving target is split from the image of shooting: firstly, by step 1)
Resulting image A and image B subtract each other, and obtain image C；Then, to image C binaryzation: with otsu threshold segmentation method by image C into
Row binary conversion treatment obtains image D；Finally, to moving meshes: carrying out edge detection to image D, and be closed with morphology
Operator is to image except dry, removal picture noise, the subgraph E for the representative moving object being partitioned into；
Step 3) target classification: subgraph E obtained in step 2) is divided into two kinds: motorcycle and non-motorcycle；It is first
Detected object is first mapped as a high dimensional feature vector with Haar feature, then judges image object category with SVM classifier
In which kind of；If being judged as YES motorcycle, enter next step；
Step 4) determines area-of-interest and head child window: firstly, will be deemed as is upper the 1/ of the E image of motorcycle
6~1/4 parts are defined as area-of-interest, are denoted as image G；Then, the determination of head child window: with Circle Hough Transform come
The circle in image G is calculated, circumscribed square corresponding to the subgraph in image G with best circularity is denoted as image I；
Step 5) feature extraction: carry out feature extraction to the image I in step 4) with HOG descriptor: wherein HOG describes quilt
Nine pieces are separated into, every piece is divided into nine junior unit lattice, then produces a feature vector being made of 81 subcharacters
H；
The classification of step 6) child window: after the image I in step 4) is carried out feature extraction by step 5), one can all be obtained
This series of feature vector H is input to antithetical phrase in multilayer neural network MLP classifier by the feature vector H in a step 5)
Window is classified, so that the head zone image classification of driver is to have the helmet and without helmet two major classes, to finally realize
Detection whether helmet of motorcycle driver is worn.
Further, image collector is set to CCD camera in the step 1).
Further, in the step 4) the E image of motorcycle upper 1/5 be area-of-interest.
Detailed description of the invention
Attached drawing 1 is embodiment of the present invention flow chart；
The utility model has the advantages that
1. this method is in two stages: the detection that motorvehicle is divided and classifies and the helmet uses.For motorvehicle
Classification, we are using Haar feature as descriptor and SVM model as classifier；The helmet is detected, we use circle Hough
It converts (CHT) and carries out head zone extraction, extract characteristics of image using gradient orientation histogram (HOG) descriptor, and with more
Layer neural network (MLP) classifier classifies target, realizes finally to realize to rub by the classification method
Detection whether motorcycle helmet of driver is worn.
2. twin-stage inspection policies whether use by vehicles segmentation and classification and the helmet, realize the whole of helmet detection
Process.
3. this method utilize extensive road monitoring, using visual sensing due to have contain much information, low-cost spy
Point solves the problems, such as that not wearing the helmet detects this.
Specific embodiment
The visible detection method of a kind of pair of helmet of motorcycle driver wear condition, includes the following steps:
The acquisition of step 1) background image: image collecting device is installed at roadside, road operation conditions is adopted
Collection, the road real image extracted from video flowing are set as A, are mentioned in video streaming with adaptive GMM algorithm
The background image without moving object is taken to be set as B；
The segmentation of step 2) moving target: moving target is split from the image of shooting: firstly, by step 1)
Resulting image A and image B subtract each other, and obtain image C；Then, to image C binaryzation: with otsu threshold segmentation method by image C into
Row binary conversion treatment obtains image D；Finally, to moving meshes: carrying out edge detection to image D, and be closed with morphology
Operator is to image except dry, removal picture noise, the subgraph E for the representative moving object being partitioned into；
Step 3) target classification: subgraph E obtained in step 2) is divided into two kinds: motorcycle and non-motorcycle；It is first
Detected object is first mapped as a high dimensional feature vector with Haar feature, then judges image object category with SVM classifier
In which kind of；If being judged as YES motorcycle, enter next step；
Step 4) determines area-of-interest and head child window: firstly, will be deemed as is upper the 1/ of the E image of motorcycle
6~1/4 parts are defined as area-of-interest, are denoted as image G；Then, the determination of head child window: with Circle Hough Transform come
The circle in image G is calculated, circumscribed square corresponding to the subgraph in image G with best circularity is denoted as image I；
Step 5) feature extraction: carry out feature extraction to the image I in step 4) with HOG descriptor: wherein HOG describes quilt
Nine pieces are separated into, every piece is divided into nine junior unit lattice, then produces a feature vector being made of 81 subcharacters
H；
The classification of step 6) child window: after the image I in step 4) is carried out feature extraction by step 5), one can all be obtained
This series of feature vector H is input to antithetical phrase in multilayer neural network MLP classifier by the feature vector H in a step 5)
Window is classified, so that the head zone image classification of driver is to have the helmet and without helmet two major classes, to finally realize
Detection whether helmet of motorcycle driver is worn.
Wherein, image collector is set to CCD camera in the step 1).
Upper the 1/5 of the E image of motorcycle is area-of-interest in the step 4).
Specific embodiment
Step 1: background image obtains: image collecting device CCD camera being installed at roadside, image collector is adjusted
The position set and posture, to obtain the image (video) of high quality；If the road real image being extracted from video frame
For A；It is extracted in video streaming with adaptive GMM algorithm (algorithm is algorithms most in use, is not added and repeats) without fortune
The background image of animal body is B；
Step 2: the segmentation of moving target: moving target being split from the image of shooting, following steps can be passed through
It realizes；
Step 2.1: Background difference: gained image A and B in step 1 are subtracted each other, image C is obtained:
Step 2.2: image binaryzation.With Otsu thresholding method (algorithm be algorithms most in use, be not added and repeat) by step
Image C in 2.1 carries out binary conversion treatment, and obtain image D: selected threshold value is calculated automatically from by Otsu；Using the threshold
Value carries out binaryzation to image C；Higher than the pixel set of the threshold value；By rest of pixels reset；
Step 2.3: moving meshes: edge detection being carried out to the image D in step 2.2, and is closed and is calculated with morphology
Son is to scene image partition, the subgraph E for the representative moving object being partitioned into；
Step 3: target classification: under traffic scene, the object (i.e. subgraph E) that step 2 is partitioned into can be divided into
Two kinds: motorcycle and non-motorcycle；Here using the method for classical Haar feature combination SVM classifier: special with Haar first
Detected object is mapped as a high dimensional feature vector by sign；Then judge which kind of image object belongs to SVM classifier: if
It is judged as YES motorcycle, then enters step 4；
Step 4: determining RoI and its child window；
Step 4.1: area-of-interest (RoI) extracts: will be deemed as being that upper 1/5 part of the E image of motorcycle is defined
For area-of-interest, be denoted as image G: this area-of-interest is count by the image for obtaining the vehicles segmentation stage
The empirical value arrived, head zone are usually located at 1/5 image-region；
Step 4.2: the determination of head child window.Possible circle in image G is calculated with Circle Hough Transform CHT；By image
Circumscribed square corresponding to subgraph in G with best circularity is denoted as image I, which is considered as Motor-cyclist
Head zone；
Step 5: feature extraction feature extraction: being carried out to the image I in step 4.2 with HOG descriptor.Wherein HOG is described
Symbol is divided into nine pieces, and every piece is divided into nine junior unit lattice, then produces a feature being made of 81 subcharacters
Vector H；
Step 6: the classification of child window: after feature extraction in step 5, the child window of each generation can obtain a spy
Vector is levied, this series of feature vector H is input to child windows in multilayer neural network MLP classifier and is classified, it will
The head zone image classification of driver is to there is the helmet and without helmet two major classes, to finally realize helmet of motorcycle driver
Detection whether wearing；
Step 1 to step 6 is twin-stage inspection policies whether use by vehicles segmentation and classification and the helmet, realizes head
The all processes of helmet detection.
Using method of the invention, helmet of motorcycle driver inspection software is write using C Plus Plus first；Then it will take the photograph
Camera is mounted on roadside suitable position, and is acquired in vehicle travel process to vehicle image；Then, taking
Original image is input in helmet inspection software and is handled；The resolution ratio of the video be 1280*720 pixel and 30 frames/second,
The total time of video is 150 minutes, and the accuracy rate of vehicle classification result has reached 97.78%.Helmet detection algorithm then reaches
91.37% accuracy rate.Running environment is WinXP, CPU 2.4GHz.
In conclusion the present invention is detected as an integral framework with what vehicles segmentation and classification and the helmet used, sufficiently
High-precision that Haar feature descriptor and SVM model are obtained as classifier in the vehicle classification stage is utilized and with circle
Accurate extraction of the Hough transform (CHT) to head zone, and then image is extracted using gradient orientation histogram (HOG) descriptor
Target be sorted in the high evaluation index of helmet detection-phase acquisition by feature with multilayer neural network (MLP) classifier
Feature, to realize the method for accurately detecting whether driver wears the helmet from provided input source images.
The embodiment is a preferred embodiment of the present invention, but present invention is not limited to the embodiments described above, not
In the case where substantive content of the invention, any conspicuous improvement that those skilled in the art can make, replacement
Or modification all belongs to the scope of protection of the present invention.
Claims (3)
1. the visible detection method of a kind of pair of helmet of motorcycle driver wear condition, which comprises the steps of:
The acquisition of step 1) background image: image collecting device being installed at roadside and is acquired to road operation conditions, from
The road real image extracted in video flowing is set as A, is extracted in video streaming not with adaptive GMM algorithm
Background image containing moving object is set as B；
The segmentation of step 2) moving target: moving target is split from the image of shooting: firstly, by gained in step 1)
Image A and image B subtract each other, obtain image C；Then, to image C binaryzation: image C being carried out two with otsu threshold segmentation method
Value processing, obtains image D；Finally, to moving meshes: carrying out edge detection to image D, and be closed operator with morphology
To scene image partition, picture noise, the subgraph E for the representative moving object being partitioned into are removed；
Step 3) target classification: subgraph E obtained in step 2) is divided into two kinds: motorcycle and non-motorcycle；It uses first
Detected object is mapped as a high dimensional feature vector by Haar feature, then judges which image object belongs to SVM classifier
It is a kind of；If being judged as YES motorcycle, enter next step；
Step 4) determines area-of-interest and head child window: firstly, will be deemed as be the E image of motorcycle upper 1/6~
1/4 part is defined as area-of-interest, is denoted as image G；Then, it the determination of head child window: is calculated with Circle Hough Transform
Circumscribed square corresponding to subgraph in image G with best circularity is denoted as image I by the circle in image G；
Step 5) feature extraction: carry out feature extraction to the image I in step 4) with HOG descriptor: wherein HOG description is separated
At nine pieces, every piece is divided into nine junior unit lattice, then produces a feature vector H being made of 81 subcharacters；
The classification of step 6) child window: after the image I in step 4) is carried out feature extraction by step 5), a step can all be obtained
It is rapid 5) in feature vector H, this series of feature vector H is input to child windows in multilayer neural network MLP classifier
Classify, so that the head zone image classification of driver is to have the helmet and without helmet two major classes, rub to finally realize
Detection whether motorcycle helmet of driver is worn.
2. the visible detection method of a kind of pair of helmet of motorcycle driver wear condition according to claim 1, feature
It is, image collector is set to CCD camera in the step 1).
3. the visible detection method of a kind of pair of helmet of motorcycle driver wear condition according to claim 1, feature
It is, upper the 1/5 of the E image of motorcycle is area-of-interest in the step 4).
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610116365.XA CN105760847B (en) | 2016-03-01 | 2016-03-01 | A kind of visible detection method of pair of helmet of motorcycle driver wear condition |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201610116365.XA CN105760847B (en) | 2016-03-01 | 2016-03-01 | A kind of visible detection method of pair of helmet of motorcycle driver wear condition |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN105760847A CN105760847A (en) | 2016-07-13 |
| CN105760847B true CN105760847B (en) | 2019-04-02 |
Family
ID=56332294
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201610116365.XA Active CN105760847B (en) | 2016-03-01 | 2016-03-01 | A kind of visible detection method of pair of helmet of motorcycle driver wear condition |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN105760847B (en) |
Families Citing this family (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106372662B (en) * | 2016-08-30 | 2020-04-28 | 腾讯科技（深圳）有限公司 | Detection method and device for wearing of safety helmet, camera and server |
| CN110516518A (en) * | 2018-05-22 | 2019-11-29 | 杭州海康威视数字技术股份有限公司 | A kind of illegal manned detection method of non-motor vehicle, device and electronic equipment |
| CN111251994B (en) * | 2018-11-30 | 2021-08-24 | 华创车电技术中心股份有限公司 | Method and system for detecting objects around vehicle |
| CN109697430A (en) * | 2018-12-28 | 2019-04-30 | 成都思晗科技股份有限公司 | The detection method that working region safety cap based on image recognition is worn |
| CN109919182B (en) * | 2019-01-24 | 2021-10-22 | 国网浙江省电力有限公司电力科学研究院 | Terminal side electric power safety operation image identification method |
| EP3941811A4 (en) * | 2019-03-18 | 2022-11-30 | JJC Imports, LLC | Deployable safety device apparatus |
| CN110866479A (en) * | 2019-11-07 | 2020-03-06 | 北京文安智能技术股份有限公司 | Method, device and system for detecting that motorcycle driver does not wear helmet |
| CN111708098A (en) * | 2020-06-18 | 2020-09-25 | 山西省交通科技研发有限公司 | Traffic safety device and safety detection method |
| CN112036360B (en) * | 2020-09-10 | 2023-11-28 | 杭州云栖智慧视通科技有限公司 | Riding helmet attribute identification method |
| CN113537183B (en) * | 2021-09-17 | 2021-12-10 | 江苏巨亨智能科技有限公司 | Electric vehicle helmet identification method based on artificial intelligence |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101017573A (en) * | 2007-02-09 | 2007-08-15 | 南京大学 | Method for detecting and identifying moving target based on video monitoring |
| CN102044151A (en) * | 2010-10-14 | 2011-05-04 | 吉林大学 | Night vehicle video detection method based on illumination visibility identification |
| CN104200668A (en) * | 2014-07-28 | 2014-12-10 | 四川大学 | Image-analysis-based detection method for helmet-free motorcycle driving violation event |
| CN104700068A (en) * | 2014-12-17 | 2015-06-10 | 安徽清新互联信息科技有限公司 | SVM based detection method of safety belt of driver |
-
2016
- 2016-03-01 CN CN201610116365.XA patent/CN105760847B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101017573A (en) * | 2007-02-09 | 2007-08-15 | 南京大学 | Method for detecting and identifying moving target based on video monitoring |
| CN102044151A (en) * | 2010-10-14 | 2011-05-04 | 吉林大学 | Night vehicle video detection method based on illumination visibility identification |
| CN104200668A (en) * | 2014-07-28 | 2014-12-10 | 四川大学 | Image-analysis-based detection method for helmet-free motorcycle driving violation event |
| CN104700068A (en) * | 2014-12-17 | 2015-06-10 | 安徽清新互联信息科技有限公司 | SVM based detection method of safety belt of driver |
Non-Patent Citations (1)
| Title |
|---|
| 基于方向可变Haar特征和双曲线模型的车道线检测方法;王海，蔡英凤，林国余，张为公;《交通运输工程学报》;20141015;全文 |
Also Published As
| Publication number | Publication date |
|---|---|
| CN105760847A (en) | 2016-07-13 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN105760847B (en) | A kind of visible detection method of pair of helmet of motorcycle driver wear condition | |
| Hadi et al. | Vehicle detection and tracking techniques: a concise review | |
| CN108549864B (en) | Vehicle-mounted thermal imaging pedestrian detection-oriented region-of-interest filtering method and device | |
| Chen et al. | Vehicle detection, tracking and classification in urban traffic | |
| WO2019196130A1 (en) | Classifier training method and device for vehicle-mounted thermal imaging pedestrian detection | |
| CN109918971B (en) | Method and device for detecting number of people in monitoring video | |
| CN105404857A (en) | Infrared-based night intelligent vehicle front pedestrian detection method | |
| CN107622258A (en) | A kind of rapid pedestrian detection method of combination static state low-level image feature and movable information | |
| CN111832461B (en) | Method for detecting wearing of non-motor vehicle riding personnel helmet based on video stream | |
| CN104951784A (en) | Method of detecting absence and coverage of license plate in real time | |
| CN103971524B (en) | A kind of traffic flow detection method based on machine vision | |
| CN109697430A (en) | The detection method that working region safety cap based on image recognition is worn | |
| CN103400113B (en) | Freeway tunnel pedestrian detection method based on image procossing | |
| WO2019119515A1 (en) | Face analysis and filtering method, device, embedded apparatus, dielectric and integrated circuit | |
| CN106548131A (en) | A kind of workmen's safety helmet real-time detection method based on pedestrian detection | |
| CN109313699A (en) | Apparatus and method for object recognition of input image of vehicle | |
| Su et al. | A new local-main-gradient-orientation HOG and contour differences based algorithm for object classification | |
| CN106570440A (en) | People counting method and people counting device based on image analysis | |
| CN104899559B (en) | A kind of rapid pedestrian detection method based on video monitoring | |
| AG et al. | Helmet detection using single shot detector (ssd) | |
| Agorku et al. | Real-time helmet violation detection using yolov5 and ensemble learning | |
| KR101711328B1 (en) | Method for classifying children and adult by using head and body height in images obtained from camera such as CCTV | |
| Arshad et al. | Detection of Two-Wheeler Traffic Rule Violation Using Deep Learning | |
| Xia et al. | An automobile detection algorithm development for automated emergency braking system | |
| Alcantarilla et al. | Automatic daytime road traffic control and monitoring system |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| C06 | Publication | ||
| PB01 | Publication | ||
| C10 | Entry into substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |