CN109492575A - A kind of staircase safety monitoring method based on YOLOv3 - Google Patents
A kind of staircase safety monitoring method based on YOLOv3 Download PDFInfo
- Publication number
- CN109492575A CN109492575A CN201811315462.7A CN201811315462A CN109492575A CN 109492575 A CN109492575 A CN 109492575A CN 201811315462 A CN201811315462 A CN 201811315462A CN 109492575 A CN109492575 A CN 109492575A
- Authority
- CN
- China
- Prior art keywords
- staircase
- target area
- picture
- data set
- network model
- 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
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V40/00—Recognition of biometric, human-related or animal-related patterns in image or video data
- G06V40/20—Movements or behaviour, e.g. gesture recognition
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/29—Graphical models, e.g. Bayesian networks
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Evolutionary Computation (AREA)
- Bioinformatics & Computational Biology (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Artificial Intelligence (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Biology (AREA)
- General Health & Medical Sciences (AREA)
- Psychiatry (AREA)
- Social Psychology (AREA)
- Human Computer Interaction (AREA)
- Multimedia (AREA)
- Alarm Systems (AREA)
Abstract
The present invention provides a kind of staircase safety monitoring method based on YOLOv3, include the following steps: step 1: having the picture of passenger's boarding in interception history staircase monitor video, marks out personage part in picture and be used as target area and make data set one according to PASCAL VOC data set format；Classify according to the posture of the personage in target area and makes data set two according to PASCAL VOC data set format；Step 2: using the training YOLOv3 network model one of data set one；Use the training YOLOv3 network model two of data set two；Step 3: Real-time security monitoring being carried out to staircase, target area position coordinates are exported by YOLOv3 network model one；The input that target area picture is YOLOv3 network model two identifies the posture of passenger on staircase.Technical solution of the present invention solves in the method for tional identification staircase uplink human action that model is excessively complicated, is not able to satisfy the problem of real time monitoring requires.
Description
Technical field
The present invention relates to technical field of machine vision, specifically, more particularly to a kind of staircase safety based on YOLOv3
Monitoring method.
Background technique
Traditional staircase safety detection method, which depends on, detects single one physical amount signal, normal for staircase
The information such as the passenger behavior occurred when operation can not identify that accuracy and efficiency is all relatively low.It is traditional for action recognition
Method complexity is relatively high, is lacking in real-time.And with the rise of machine vision, for action recognition research more
Deepen into the action identification method based on deep learning frame is suggested, and for action recognition, accuracy rate and speed are most heavy
The two indices wanted, and accuracy rate height then means that real-time is poor, to comprehensively consider two indices in practical applications.
So far, the recognition methods of the hazardous act of pedestrian on staircase can be substantially divided into: based on conventional physical amount,
Based on traditional algorithm and it is based on deep learning frame.Passenger when conventional physical amount relies primarily on external sensor to staircase work
Various semaphores carry out real-time monitoring；Traditional algorithm is then mainly handled to extract passenger's feature, in turn monitoring picture
It is tracked, is analyzed by motion profile, Optical-flow Feature or other features to passenger to identify passenger's dangerous play；And
It mainly extracts pedestrian's feature based on deep learning frame to be input to neural network model and classify, or based on building bone
Mould is realized.But all there is respective defects for these methods, such as the method based on deep learning frame, if net
Network model is excessively complicated, although improving a lot in accuracy, calculation amount is too big, can hardly meet wanting for real-time
It asks.
Summary of the invention
It is excessively complicated according to model in the method for tional identification staircase uplink human action set forth above, it is not able to satisfy real-time prison
The technical issues of control requires, and a kind of staircase safety monitoring method based on YOLOv3 is provided.The present invention mainly utilizes YOLOv3
Network model one determines the posture that the position coordinates of passenger in video recycle YOLOv3 network model two to identify passenger, to rise
The effect of the safety of passenger on to real time monitoring staircase.
The technological means that the present invention uses is as follows:
A kind of staircase safety monitoring method based on YOLOv3, includes the following steps:
Step 1: having the picture of passenger's boarding in interception history staircase monitor video, data enhancing, mark are carried out to picture
Personage part is used as target area and makes data set one according to PASCAL VOC data set format in picture out；
Classify according to the posture of the personage in target area and makes data according to PASCAL VOC data set format
Collection two, wherein the posture of personage includes standing, fall, squat down and climbing handrail；
Step 2: using the training YOLOv3 network model one of data set one；Use the training YOLOv3 network model of data set two
Two；
For YOLOv3 network model one for determining target area position coordinates, the input of model is staircase video monitoring figure
Piece exports as target area position coordinates；
YOLOv3 network model two is used to carry out feature extraction to target area picture and classify to the posture of personage,
The input of model is target area picture, is exported as personage's gesture recognition result；
Step 3: Real-time security monitoring being carried out to staircase, the frame picture in staircase monitor video is intercepted, passes through each second
YOLOv3 network model one exports target area position coordinates, is regarded according to target area position coordinates using opencv interception staircase
Target area picture in frequency monitoring picture；By target area picture be YOLOv3 network model two input to passenger on staircase
Posture identified；It in the posture of personage, stands as security posture, falls, squats down, climbing handrail as dangerous boarding posture.
Compared with the prior art, the invention has the following advantages that
1, the staircase safety monitoring method provided by the invention based on YOLOv3, is sampled, phase by the time of every 1s
Than the safety detection method for needing to carry out target tracing detection in tradition, data volume and calculation amount are smaller, and detection efficiency is higher.
2, the staircase safety monitoring method provided by the invention based on YOLOv3, the YOLOv3 of use are a kind of novel targets
Detection algorithm has used for reference the thought of ResNet network, has deepened network depth, improves detection accuracy, and relatively before version
The detection effect for Small object is improved, the requirement of discrimination and real-time under scene condition is met.
3, the staircase safety monitoring method provided by the invention based on YOLOv3, by danger occur when timing node and
The preservation of dangerous pose presentation is conducive to monitoring check.
To sum up, it applies the technical scheme of the present invention and is sampled by the time of every 1s, compared to needing to target in the past
The method for carrying out tracing detection, the data volume and calculation amount of this method are smaller, and detection efficiency is higher.Therefore, technology of the invention
Scheme solves in the method for tional identification staircase uplink human action in the prior art that model is excessively complicated, is not able to satisfy in real time
The problem of monitoring requirement.
The present invention can be widely popularized in fields such as public staircase passenger safety monitoring based on the above reasons.
Detailed description of the invention
In order to more clearly explain the embodiment of the invention or the technical proposal in the existing technology, to embodiment or will show below
There is attached drawing needed in technical description to do simply to introduce, it should be apparent that, the accompanying drawings in the following description is this hair
Bright some embodiments for those of ordinary skill in the art without any creative labor, can be with
It obtains other drawings based on these drawings.
Fig. 1 is staircase safety monitoring method course of work flow chart of the present invention.
Fig. 2 is the schematic network structure of staircase safety monitoring method of the present invention.
Fig. 3 is the one testing result schematic diagram of YOLOv3 network model.
Fig. 4 is the two testing result schematic diagram of YOLOV3 network model.
Fig. 5 is the two testing result schematic diagram of YOLOV3 network model.
Specific embodiment
In order to enable those skilled in the art to better understand the solution of the present invention, below in conjunction in the embodiment of the present invention
Attached drawing, technical scheme in the embodiment of the invention is clearly and completely described, it is clear that described embodiment is only
The embodiment of a part of the invention, instead of all the embodiments.Based on the embodiments of the present invention, ordinary skill people
The model that the present invention protects all should belong in member's every other embodiment obtained without making creative work
It encloses.
It should be noted that description and claims of this specification and term " first " in above-mentioned attached drawing, "
Two " etc. be to be used to distinguish similar objects, without being used to describe a particular order or precedence order.It should be understood that using in this way
Data be interchangeable under appropriate circumstances, so as to the embodiment of the present invention described herein can in addition to illustrating herein or
Sequence other than those of description is implemented.In addition, term " includes " and " having " and their any deformation, it is intended that cover
Cover it is non-exclusive include, for example, the process, method, system, product or equipment for containing a series of steps or units are not necessarily limited to
Step or unit those of is clearly listed, but may include be not clearly listed or for these process, methods, product
Or other step or units that equipment is intrinsic.
Embodiment 1
As shown in Figs. 1-2, the present invention provides a kind of staircase safety monitoring method based on YOLOv3, including walk as follows
It is rapid:
Step 1: having the picture of passenger's boarding in interception history staircase monitor video, data enhancing is carried out to picture and expands number
According to amount, marks out personage part in picture and be used as target area and make data set one according to PASCAL VOC data set format；
Wherein, the mode of data enhancing includes rotation, flip horizontal, shearing, changes size and increase picture noise etc.；
Preferably, the present invention goes out personage part in picture using Labelimg software
Classify according to the posture of the personage in target area and makes data according to PASCAL VOC data set format
Collection two, wherein the posture of personage includes standing, fall, squat down and climbing handrail；
Step 2: using the training YOLOv3 network model one of data set one, until the loss function of YOLOv3 network model one
≤ 0.0001, obtain trained network model；Using the training YOLOv3 network model two of data set two, until YOLOv3 network
Loss function≤0.0001 of model one obtains trained network model；
The loss function uses binary crossentropy loss function, and formula is as follows:
For YOLOv3 network model one for determining target area position coordinates, the input of model is staircase video monitoring figure
Piece exports as target area position coordinates；
YOLOv3 network model two is used to carry out feature extraction to target area picture and classify to the posture of personage,
The input of model is target area picture, is exported as personage's gesture recognition result；
The YOLOv3 neural network that the application uses is a kind of target detection neural network with 53 layers of convolutional layer, is added
Residual error network, can under three kinds of different scales to pre- come performance degradation caused by solving the problems, such as to deepen with network depth
It surveys frame to be predicted, has good performance for wisp, intensive and the case where blocking；
The training of YOLOv3 network model one and YOLOv3 network model two uses multiple dimensioned training, method particularly includes: it is logical
It crosses python script file and data set one or data set two is generated into picture path list, training list and proof listing, it is corresponding
Data set one is respectively modified data/.names, cfg/.data, cfg/.cfg file, operation order line load cfg/.data,
Cfg/.cfg and initial weight file are trained YOLOv3 network model one or network model two；
Step 3: as in Figure 3-5, Real-time security monitoring being carried out to staircase, intercepts one in staircase monitor video each second
Frame picture exports target area position coordinates by YOLOv3 network model one, is used according to target area position coordinates
Opencv intercepts the target area picture in staircase video monitoring picture；It is YOLOv3 network model two by target area picture
Input identifies the posture of passenger on staircase；In the posture of personage, stands as security posture, fall, squat down, climb handrail
For dangerous boarding posture；
When monitoring has passenger dangerous boarding posture occur on staircase, dangerous posture picture is saved, and record corresponding danger
For the timing node that dangerous posture occurs to monitor check, staff can take voice reminder or tight according to monitoring result in time
Emergency stop stops the safety measures such as staircase work and guarantees passenger safety, realizes the real-time monitoring for helping passenger on staircase boarding posture.
Finally, it should be noted that the above embodiments are merely illustrative of the technical solutions of the present invention, rather than its limitations；Although
Present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that: it still may be used
To modify to technical solution documented by previous embodiment, or some or all of the technical features are equal
Replacement；And these are modified or replaceed, the model for technical solution of the embodiment of the present invention that it does not separate the essence of the corresponding technical solution
It encloses.
Claims (1)
1. a kind of staircase safety monitoring method based on YOLOv3, which comprises the steps of:
Step 1: having the picture of passenger's boarding in interception history staircase monitor video, data enhancing is carried out to picture, marks out figure
Personage part is used as target area and makes data set one according to PASCAL VOC data set format in piece；
Classify according to the posture of the personage in target area and make data set two according to PASCAL VOC data set format,
Wherein, the posture of personage includes standing, fall, squat down and climbing handrail；
Step 2: using the training YOLOv3 network model one of data set one；Use the training YOLOv3 network model two of data set two；
For YOLOv3 network model one for determining target area position coordinates, the input of model is staircase video monitoring picture, defeated
It is out target area position coordinates；
YOLOv3 network model two is used to carry out feature extraction to target area picture and classify to the posture of personage, model
Input be target area picture, export as personage's gesture recognition result；
Step 3: Real-time security monitoring being carried out to staircase, the frame picture in staircase monitor video is intercepted, passes through each second
YOLOv3 network model one exports target area position coordinates, is regarded according to target area position coordinates using opencv interception staircase
Target area picture in frequency monitoring picture；By target area picture be YOLOv3 network model two input to passenger on staircase
Posture identified；It in the posture of personage, stands as security posture, falls, squats down, climbing handrail as dangerous boarding posture.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811315462.7A CN109492575A (en) | 2018-11-06 | 2018-11-06 | A kind of staircase safety monitoring method based on YOLOv3 |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811315462.7A CN109492575A (en) | 2018-11-06 | 2018-11-06 | A kind of staircase safety monitoring method based on YOLOv3 |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109492575A true CN109492575A (en) | 2019-03-19 |
Family
ID=65693967
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811315462.7A Pending CN109492575A (en) | 2018-11-06 | 2018-11-06 | A kind of staircase safety monitoring method based on YOLOv3 |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109492575A (en) |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109993122A (en) * | 2019-04-02 | 2019-07-09 | 中国石油大学(华东) | A kind of pedestrian based on depth convolutional neural networks multiplies staircase anomaly detection method |
| CN110211173A (en) * | 2019-04-03 | 2019-09-06 | 中国地质调查局发展研究中心 | A kind of paleontological fossil positioning and recognition methods based on deep learning |
| CN110598633A (en) * | 2019-09-12 | 2019-12-20 | 杭州品茗安控信息技术股份有限公司 | Tumble behavior identification method, device and system |
| CN112800856A (en) * | 2021-01-06 | 2021-05-14 | 南京通盛弘数据有限公司 | Livestock position and posture recognition method and device based on YOLOv3 |
| CN112836667A (en) * | 2021-02-20 | 2021-05-25 | 上海吉盛网络技术有限公司 | Method for judging falling and retrograde of passenger on ascending escalator |
| CN113011290A (en) * | 2021-03-03 | 2021-06-22 | 上海商汤智能科技有限公司 | Event detection method and device, electronic equipment and storage medium |
| CN113657165A (en) * | 2020-08-10 | 2021-11-16 | 广东电网有限责任公司 | Dangerous climbing behavior recognition algorithm in electric power field operation |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN106503632A (en) * | 2016-10-10 | 2017-03-15 | 南京理工大学 | A kind of escalator intelligent and safe monitoring method based on video analysis |
| CN107358223A (en) * | 2017-08-16 | 2017-11-17 | 上海荷福人工智能科技（集团）有限公司 | A kind of Face datection and face alignment method based on yolo |
| US20180039840A1 (en) * | 2015-03-04 | 2018-02-08 | Hitachi Systems, Ltd. | Situation ascertainment system using camera picture data, control device, and situation ascertainment method using camera picture data |
| CN108062526A (en) * | 2017-12-15 | 2018-05-22 | 厦门美图之家科技有限公司 | A kind of estimation method of human posture and mobile terminal |
| CN108639921A (en) * | 2018-07-05 | 2018-10-12 | 江苏瑞奇海力科技有限公司 | A kind of staircase passenger safety prior-warning device and method |
-
2018
- 2018-11-06 CN CN201811315462.7A patent/CN109492575A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20180039840A1 (en) * | 2015-03-04 | 2018-02-08 | Hitachi Systems, Ltd. | Situation ascertainment system using camera picture data, control device, and situation ascertainment method using camera picture data |
| CN106503632A (en) * | 2016-10-10 | 2017-03-15 | 南京理工大学 | A kind of escalator intelligent and safe monitoring method based on video analysis |
| CN107358223A (en) * | 2017-08-16 | 2017-11-17 | 上海荷福人工智能科技（集团）有限公司 | A kind of Face datection and face alignment method based on yolo |
| CN108062526A (en) * | 2017-12-15 | 2018-05-22 | 厦门美图之家科技有限公司 | A kind of estimation method of human posture and mobile terminal |
| CN108639921A (en) * | 2018-07-05 | 2018-10-12 | 江苏瑞奇海力科技有限公司 | A kind of staircase passenger safety prior-warning device and method |
Non-Patent Citations (1)
| Title |
|---|
| 黎德源: "基于机器视觉的扶梯安全检测方法的研究", 《中国优秀硕士学位论文全文数据库 信息科技辑》 * |
Cited By (10)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN109993122A (en) * | 2019-04-02 | 2019-07-09 | 中国石油大学(华东) | A kind of pedestrian based on depth convolutional neural networks multiplies staircase anomaly detection method |
| CN110211173A (en) * | 2019-04-03 | 2019-09-06 | 中国地质调查局发展研究中心 | A kind of paleontological fossil positioning and recognition methods based on deep learning |
| CN110598633A (en) * | 2019-09-12 | 2019-12-20 | 杭州品茗安控信息技术股份有限公司 | Tumble behavior identification method, device and system |
| CN110598633B (en) * | 2019-09-12 | 2023-04-07 | 品茗科技股份有限公司 | Tumble behavior identification method, device and system |
| CN113657165A (en) * | 2020-08-10 | 2021-11-16 | 广东电网有限责任公司 | Dangerous climbing behavior recognition algorithm in electric power field operation |
| CN112800856A (en) * | 2021-01-06 | 2021-05-14 | 南京通盛弘数据有限公司 | Livestock position and posture recognition method and device based on YOLOv3 |
| CN112836667A (en) * | 2021-02-20 | 2021-05-25 | 上海吉盛网络技术有限公司 | Method for judging falling and retrograde of passenger on ascending escalator |
| CN112836667B (en) * | 2021-02-20 | 2022-11-15 | 上海吉盛网络技术有限公司 | Method for judging falling and reverse running of passengers going upstairs escalator |
| CN113011290A (en) * | 2021-03-03 | 2021-06-22 | 上海商汤智能科技有限公司 | Event detection method and device, electronic equipment and storage medium |
| WO2022183661A1 (en) * | 2021-03-03 | 2022-09-09 | 上海商汤智能科技有限公司 | Event detection method and apparatus, electronic device, storage medium, and program product |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109492575A (en) | A kind of staircase safety monitoring method based on YOLOv3 | |
| CN109615019B (en) | Abnormal behavior detection method based on space-time automatic encoder | |
| Huang et al. | Detection algorithm of safety helmet wearing based on deep learning | |
| CN109117827B (en) | Video-based method for automatically identifying wearing state of work clothes and work cap and alarm system | |
| CN110147743B (en) | Real-time online pedestrian analysis and counting system and method under complex scene | |
| CN102663452B (en) | Suspicious act detecting method based on video analysis | |
| CN107679471B (en) | Indoor personnel air post detection method based on video monitoring platform | |
| CN103020992B (en) | A kind of video image conspicuousness detection method based on motion color-associations | |
| CN103279737B (en) | A kind of behavioral value method of fighting based on space-time interest points | |
| CN103020596B (en) | A kind of based on Human bodys' response method abnormal in the power generation of block models | |
| CN110414400B (en) | Automatic detection method and system for wearing of safety helmet on construction site | |
| CN108197589A (en) | Semantic understanding method, apparatus, equipment and the storage medium of dynamic human body posture | |
| CN102938058A (en) | Method and system for video driving intelligent perception and facing safe city | |
| CN102324025A (en) | Human face detection and tracking method based on Gaussian skin color model and feature analysis | |
| CN107301376B (en) | Pedestrian detection method based on deep learning multi-layer stimulation | |
| CN110298297A (en) | Flame identification method and device | |
| CN103020985A (en) | Video image saliency detection method based on field quantity analysis | |
| US10282601B2 (en) | Electronic device and gesture recognition method applied therein | |
| CN104834916A (en) | Multi-face detecting and tracking method | |
| CN105654505B (en) | A kind of collaboration track algorithm and system based on super-pixel | |
| CN107844742A (en) | Facial image glasses minimizing technology, device and storage medium | |
| Sawant | Sign language recognition system to aid deaf-dumb people using PCA | |
| CN107886110A (en) | Method for detecting human face, device and electronic equipment | |
| Alksasbeh et al. | Smart hand gestures recognition using K-NN based algorithm for video annotation purposes | |
| Wu et al. | Automated visual helmet identification based on deep convolutional neural networks |
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: 20190319 |
|
| RJ01 | Rejection of invention patent application after publication |