CN113642362A - Crowd density estimation method for intelligent escape in dense place - Google Patents
Crowd density estimation method for intelligent escape in dense place Download PDFInfo
- Publication number
- CN113642362A CN113642362A CN202010391976.1A CN202010391976A CN113642362A CN 113642362 A CN113642362 A CN 113642362A CN 202010391976 A CN202010391976 A CN 202010391976A CN 113642362 A CN113642362 A CN 113642362A
- Authority
- CN
- China
- Prior art keywords
- picture
- image
- crowd density
- estimation method
- person
- 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
- 238000000034 method Methods 0.000 title claims abstract description 32
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 10
- 238000012549 training Methods 0.000 claims description 3
- 238000013528 artificial neural network Methods 0.000 claims description 2
- 238000012163 sequencing technique Methods 0.000 claims 1
- 238000012217 deletion Methods 0.000 abstract description 2
- 230000037430 deletion Effects 0.000 abstract description 2
- 238000004364 calculation method Methods 0.000 description 2
- 230000010485 coping Effects 0.000 description 2
- 238000002474 experimental method Methods 0.000 description 2
- 241000283070 Equus zebra Species 0.000 description 1
- 230000003044 adaptive effect Effects 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000005259 measurement Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Data Mining & Analysis (AREA)
- General Health & Medical Sciences (AREA)
- Biomedical Technology (AREA)
- Biophysics (AREA)
- Computational Linguistics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Evolutionary Computation (AREA)
- Artificial Intelligence (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Health & Medical Sciences (AREA)
- Image Analysis (AREA)
Abstract
The invention relates to the technical field of escape in crowded places, in particular to a crowd density estimation method for intelligent escape in crowded places; the method comprises the steps of obtaining a real-time picture; deleting the background of the picture; graying the picture; importing the picture into a model based on a multi-column convolutional neural network, an OpenCV model or a scale algorithm; and outputting a counting result. Background deletion is carried out on the picture by adopting the prior art, the method is simple, quick and effective, and the picture is converted into a gray-scale picture, so that color information is reduced, and the operation requirement is reduced; the number of people is estimated by adopting various means, so that a large counting error caused by crowding, overlapping and the like is avoided.
Description
Technical Field
The invention relates to the technical field of escape in crowded places, in particular to a crowd density estimation method for intelligent escape in crowded places.
Background
In the occasion with dense crowds, for example, on a zebra crossing at a peak time of work or in a subway channel, people rush, and when an accident occurs, a decision maker needs to sense the real-time crowd density, particularly the specific number of people in a certain picture, so that the adaptive coping mode can be conveniently selected.
Most of the existing visual counting methods are to learn through human body characteristics, then to identify human body through images, and finally to calculate the specific number of people. However, the above scheme has the following disadvantages: in crowded occasions where people surmount, most characteristics of human bodies are overlapped, complete human body recognition cannot be completed actually, and secondly, the whole calculation process has huge data and needs to consume huge calculation power.
Disclosure of Invention
Aiming at the defects of the prior art, the invention provides the crowd density estimation method which does not extract the specific characteristics of the human body and is used for intelligent escape in the dense place with quick operation.
The technical scheme of the invention is as follows:
the crowd density estimation method for intelligent escape in the dense place is characterized by comprising the following steps of: it comprises the following steps:
step one, acquiring a real-time picture;
step two, deleting the background of the picture;
graying the picture;
step four, importing the pictures into a model based on a multi-column convolutional neural network, an OpenCV model or a scale algorithm;
and step five, outputting a counting result.
Specifically, the method for deleting the picture background in the second step is as follows: and (3) importing the picture into Word, and quickly removing the areas except the human body in the picture by using a background deleting function in Word software.
In one embodiment, the fourth step adopts a multi-column convolutional neural network, and the specific method is as follows:
training with CNN to estimate the crowd density map in the input image, first describe how to convert the image of the tagged person's head into a map of crowd density, which we will represent as delta function Δ (x-xi) if there is a head at pixel xi, so the image tagged with N heads can be represented as a function: h (x) Δ (x-xi);
convolving the function with a gaussian kernel σ to convert it into a continuous density function such that the density is f (x) h (x) σ (x), determining an extension parameter σ based on the size of each person's head in the image, adaptively determining each person's extension parameter based on the average distance of each person to it;
for each column, using a different size filter to model a density map corresponding to a different scale of the head; specifically, a mode of optimizing Euclidean loss is adopted to enable a density graph output by the network to return to a standard density graph.
In one embodiment, the OpenCV model is used in the fourth step, and the method includes:
s1 binarizing the picture to obtain a clear black and white contour;
s2 determining the head position by searching for a connected domain;
s3 identifying the overlapping person 'S head and the count of overlapping persons' heads; identifying and counting connected domains with the number of edges being more than 4, subtracting 4 from the total number of edges of the connected domains with the number of edges being more than 4 and dividing the subtracted result by 2 (obviously, decimals such as 0.5 can be obtained, but the method is used for estimation, so the decisionmaker's decision is not influenced by the decimal), obtaining the number which is the total number of the heads represented by the connected domains, identifying and counting the connected domains with the number of edges being less than or equal to 4, and counting the connected domains with the number of edges being less than or equal to 4 as one person;
s4, calculating the total number of people and outputting the estimated number of people.
In one embodiment, the step four adopts a scale algorithm model, and the method is as follows:
converting the gray level image obtained in the previous step into a binary image, and multiplying the total area of the image by the ratio of black/total pixels to obtain the area of the image; and uniformly selecting a plurality of heads according to the orientation of the cameras to calculate the occupied area of the heads, then calculating the average value of the occupied area, and finally dividing the total area by the average value to obtain the number of people.
In some embodiments, the fourth step adopts a combination of three models based on a multi-column convolutional neural network, an OpenCV model or a scale algorithm model, and sorts the calculated people number by using the largest value as the estimation result. Most countermeasures require more than actual demand for coping with sufficient strength as necessary for the group event. For example, a football pitch of one hundred thousand people, it is generally based on this consideration that security requirements only allow eight thousand people to approach the field.
The invention has the beneficial effects that: background deletion is carried out on the picture by adopting the prior art, the method is simple, quick and effective, and the picture is converted into a gray-scale picture, so that color information is reduced, and the operation requirement is reduced; the number of people is estimated by adopting various means, so that a large counting error caused by crowding, overlapping and the like is avoided.
Description of the drawings:
fig. 1 and 2 are graphs for experiments.
Detailed Description
The following is further described in conjunction with the detailed description:
example 1
The crowd density estimation method for intelligent escape in the dense place is characterized by comprising the following steps of: it comprises the following steps:
step one, acquiring a real-time picture;
step two, deleting the background of the picture;
graying the picture;
step four, importing the pictures into a model based on a multi-column convolutional neural network, an OpenCV model or a scale algorithm;
and step five, outputting a counting result.
Specifically, the method for deleting the picture background in the second step is as follows: and (3) importing the picture into Word, and quickly removing the areas except the human body in the picture by using a background deleting function in Word software.
The fourth step adopts a neural network based on multi-column convolution, and the specific method is as follows:
training with CNN to estimate the crowd density map in the input image, first describe how to convert the image of the tagged person's head into a map of crowd density, which we will represent as delta function Δ (x-xi) if there is a head at pixel xi, so the image tagged with N heads can be represented as a function: h (x) Δ (x-xi);
convolving the function with a gaussian kernel σ to convert it into a continuous density function such that the density is f (x) h (x) σ (x), determining an extension parameter σ based on the size of each person's head in the image, adaptively determining each person's extension parameter based on the average distance of each person to it;
for each column, using a different size filter to model a density map corresponding to a different scale of the head; specifically, a mode of optimizing Euclidean loss is adopted to enable a density graph output by the network to return to a standard density graph.
Example 2
The fourth step adopts an OpenCV model, and the method comprises the following steps:
s1 binarizing the picture to obtain a clear black and white contour;
s2 determining the head position by searching for a connected domain;
s3 identifying the overlapping person 'S head and the count of overlapping persons' heads; identifying and counting connected domains with the number of edges being more than 4, subtracting 4 from the total number of edges of the connected domains with the number of edges being more than 4 and dividing the subtracted result by 2 (obviously, decimals such as 0.5 can be obtained, but the method is used for estimation, so the decisionmaker's decision is not influenced by the decimal), obtaining the number which is the total number of the heads represented by the connected domains, identifying and counting the connected domains with the number of edges being less than or equal to 4, and counting the connected domains with the number of edges being less than or equal to 4 as one person;
s4, calculating the total number of people and outputting the estimated number of people.
Example 3
The fourth step adopts a scale algorithm model, and the method comprises the following steps:
converting the gray level image obtained in the previous step into a binary image, and multiplying the total area of the image by the ratio of black/total pixels to obtain the area of the image; and uniformly selecting a plurality of heads according to the orientation of the cameras to calculate the occupied area of the heads, then calculating the average value of the occupied area, and finally dividing the total area by the average value to obtain the number of people.
Example 4
In actual measurement experiments, pictures 1 and 2 are introduced into examples 1, 2 and 3, respectively, and the counting results are shown in the following table:
item | Example 1 | Example 2 | Example 3 |
FIG. 1 shows a schematic view of a | 91 | 103 | 87 |
FIG. 2 | 1320 | 1399 | 1577 |
The foregoing embodiments and description have been presented only to illustrate the principles and preferred embodiments of the invention, and various changes and modifications may be made therein without departing from the spirit and scope of the invention as hereinafter claimed.
Claims (6)
1. The crowd density estimation method for intelligent escape in the dense place is characterized by comprising the following steps of: it comprises the following steps:
step one, acquiring a real-time picture;
step two, deleting the background of the picture;
graying the picture;
step four, importing the pictures into a model based on a multi-column convolutional neural network, an OpenCV model or a scale algorithm;
and step five, outputting a counting result.
2. The crowd density estimation method for intelligent escape in dense places according to claim 1, wherein: the method for deleting the picture background in the second step comprises the following steps: and (3) importing the picture into Word, and quickly removing the areas except the human body in the picture by using a background deleting function in Word software.
3. The crowd density estimation method for intelligent escape in dense places according to claim 2, wherein: the fourth step adopts a neural network based on multi-column convolution, and the specific method is as follows:
training with CNN to estimate the crowd density map in the input image, first describe how to convert the image of the tagged person's head into a map of crowd density, which we will represent as delta function Δ (x-xi) if there is a head at pixel xi, so the image tagged with N heads can be represented as a function: h (x) Δ (x-xi);
convolving the function with a gaussian kernel σ to convert it into a continuous density function such that the density is f (x) h (x) σ (x), determining an extension parameter σ based on the size of each person's head in the image, adaptively determining each person's extension parameter based on the average distance of each person to it;
for each column, using a different size filter to model a density map corresponding to a different scale of the head; specifically, a mode of optimizing Euclidean loss is adopted to enable a density graph output by the network to return to a standard density graph.
4. The crowd density estimation method for intelligent escape in dense places according to claim 2, wherein: the fourth step adopts an OpenCV model, and the method comprises the following steps:
s1 binarizing the picture to obtain a clear black and white contour;
s2 determining the head position by searching for a connected domain;
s3 identifying the overlapping person 'S head and the count of overlapping persons' heads; identifying and counting the connected domains with the number of edges being more than 4, subtracting 4 from the total number of edges and dividing by 2 for each connected domain with the number of edges being more than 4 to obtain the number which is the total number of the heads represented by the connected domain, identifying and counting the connected domains with the number of edges being less than or equal to 4, and counting the connected domains with the number of edges being less than or equal to 4 as one person;
s4, calculating the total number of people and outputting the estimated number of people.
5. The crowd density estimation method for intelligent escape in dense places according to claim 2, wherein: the fourth step adopts a scale algorithm model, and the method comprises the following steps:
converting the gray level image obtained in the previous step into a binary image, and multiplying the total area of the image by the ratio of black/total pixels to obtain the area of the image; and uniformly selecting a plurality of heads according to the orientation of the cameras to calculate the occupied area of the heads, then calculating the average value of the occupied area, and finally dividing the total area by the average value to obtain the number of people.
6. The crowd density estimation method for intelligent escape in dense places according to claim 2, wherein: and fourthly, combining three models based on a multi-column convolutional neural network, an OpenCV model or a scale algorithm model, sequencing the calculated number of people, and taking the maximum numerical value as an estimation result.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010391976.1A CN113642362A (en) | 2020-05-11 | 2020-05-11 | Crowd density estimation method for intelligent escape in dense place |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010391976.1A CN113642362A (en) | 2020-05-11 | 2020-05-11 | Crowd density estimation method for intelligent escape in dense place |
Publications (1)
Publication Number | Publication Date |
---|---|
CN113642362A true CN113642362A (en) | 2021-11-12 |
Family
ID=78415372
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202010391976.1A Pending CN113642362A (en) | 2020-05-11 | 2020-05-11 | Crowd density estimation method for intelligent escape in dense place |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN113642362A (en) |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
TWI796033B (en) * | 2021-12-07 | 2023-03-11 | 巨鷗科技股份有限公司 | People flow analysis and identification system |
Citations (14)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101751668A (en) * | 2009-11-13 | 2010-06-23 | 北京智安邦科技有限公司 | Method and device for detecting crowd density |
CN104573697A (en) * | 2014-12-31 | 2015-04-29 | 西安丰树电子科技发展有限公司 | Construction hoist lift car people counting method based on multi-information fusion |
CN104794790A (en) * | 2015-04-23 | 2015-07-22 | 南京信息工程大学 | Scenic spot tourist counting and evacuating method |
CN105893962A (en) * | 2016-03-31 | 2016-08-24 | 成都信息工程大学 | Method for counting passenger flow at airport security check counter |
CN106548158A (en) * | 2016-11-07 | 2017-03-29 | 华南理工大学 | Crowd density intelligent monitor system and method based on machine vision |
CN107122714A (en) * | 2017-03-28 | 2017-09-01 | 天棣网络科技(上海)有限公司 | A kind of real-time pedestrian detection method based on edge constraint |
CN108717549A (en) * | 2018-04-26 | 2018-10-30 | 东华大学 | Crowd density analysis method based on unmanned plane vision and support vector cassification |
CN108876774A (en) * | 2018-06-07 | 2018-11-23 | 浙江大学 | A kind of people counting method based on convolutional neural networks |
CN109635763A (en) * | 2018-12-19 | 2019-04-16 | 燕山大学 | A kind of crowd density estimation method |
US20190333241A1 (en) * | 2017-01-13 | 2019-10-31 | Canon Kabushiki Kaisha | People flow analysis apparatus, people flow analysis system, people flow analysis method, and non-transitory computer readable medium |
CN110427894A (en) * | 2019-08-06 | 2019-11-08 | 齐鲁工业大学 | A kind of crowd's abnormal motion monitoring method, system, equipment and medium |
CN110503017A (en) * | 2019-08-12 | 2019-11-26 | 北京交通大学 | Wisdom energy conservation occupancy detection system and method based on image procossing |
CN110705408A (en) * | 2019-09-23 | 2020-01-17 | 东南大学 | Indoor people counting method and system based on mixed Gaussian people distribution learning |
CN110969561A (en) * | 2019-12-23 | 2020-04-07 | 河北大学 | Emergency evacuation method for dense crowd |
-
2020
- 2020-05-11 CN CN202010391976.1A patent/CN113642362A/en active Pending
Patent Citations (14)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN101751668A (en) * | 2009-11-13 | 2010-06-23 | 北京智安邦科技有限公司 | Method and device for detecting crowd density |
CN104573697A (en) * | 2014-12-31 | 2015-04-29 | 西安丰树电子科技发展有限公司 | Construction hoist lift car people counting method based on multi-information fusion |
CN104794790A (en) * | 2015-04-23 | 2015-07-22 | 南京信息工程大学 | Scenic spot tourist counting and evacuating method |
CN105893962A (en) * | 2016-03-31 | 2016-08-24 | 成都信息工程大学 | Method for counting passenger flow at airport security check counter |
CN106548158A (en) * | 2016-11-07 | 2017-03-29 | 华南理工大学 | Crowd density intelligent monitor system and method based on machine vision |
US20190333241A1 (en) * | 2017-01-13 | 2019-10-31 | Canon Kabushiki Kaisha | People flow analysis apparatus, people flow analysis system, people flow analysis method, and non-transitory computer readable medium |
CN107122714A (en) * | 2017-03-28 | 2017-09-01 | 天棣网络科技(上海)有限公司 | A kind of real-time pedestrian detection method based on edge constraint |
CN108717549A (en) * | 2018-04-26 | 2018-10-30 | 东华大学 | Crowd density analysis method based on unmanned plane vision and support vector cassification |
CN108876774A (en) * | 2018-06-07 | 2018-11-23 | 浙江大学 | A kind of people counting method based on convolutional neural networks |
CN109635763A (en) * | 2018-12-19 | 2019-04-16 | 燕山大学 | A kind of crowd density estimation method |
CN110427894A (en) * | 2019-08-06 | 2019-11-08 | 齐鲁工业大学 | A kind of crowd's abnormal motion monitoring method, system, equipment and medium |
CN110503017A (en) * | 2019-08-12 | 2019-11-26 | 北京交通大学 | Wisdom energy conservation occupancy detection system and method based on image procossing |
CN110705408A (en) * | 2019-09-23 | 2020-01-17 | 东南大学 | Indoor people counting method and system based on mixed Gaussian people distribution learning |
CN110969561A (en) * | 2019-12-23 | 2020-04-07 | 河北大学 | Emergency evacuation method for dense crowd |
Non-Patent Citations (3)
Title |
---|
LUYANG WANG等: "Removing background interference for crowd counting viade-background detail convolutional network", 《NEUROCOMPUTING》, vol. 332, 30 December 2018 (2018-12-30), pages 360 - 371, XP085582498, DOI: 10.1016/j.neucom.2018.12.047 * |
杨林等: "像素特征与粘连人体分割结合的人数统计方法", 《计算机工程与设计》, vol. 40, no. 2, 28 February 2019 (2019-02-28), pages 455 - 461 * |
陈思秦: "基于全卷积神经网络的人群计数", 《电子设计工程》, vol. 26, no. 2, 31 January 2018 (2018-01-31), pages 75 - 79 * |
Cited By (1)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
TWI796033B (en) * | 2021-12-07 | 2023-03-11 | 巨鷗科技股份有限公司 | People flow analysis and identification system |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN110555368B (en) | Fall-down behavior identification method based on three-dimensional convolutional neural network | |
CN110348376B (en) | Pedestrian real-time detection method based on neural network | |
US20210224609A1 (en) | Method, system and device for multi-label object detection based on an object detection network | |
CN108388896A (en) | A kind of licence plate recognition method based on dynamic time sequence convolutional neural networks | |
CN109800817B (en) | Image classification method based on fusion semantic neural network | |
CN111709300B (en) | Crowd counting method based on video image | |
CN106960176B (en) | Pedestrian gender identification method based on transfinite learning machine and color feature fusion | |
CN109829924B (en) | Image quality evaluation method based on principal feature analysis | |
CN104424457A (en) | Method for identifying two-dimensional code under the condition of nonlinear distortion | |
CN111652213A (en) | Ship water gauge reading identification method based on deep learning | |
Tereikovskyi et al. | The method of semantic image segmentation using neural networks | |
CN113486894A (en) | Semantic segmentation method for satellite image feature component | |
CN107516083A (en) | A kind of remote facial image Enhancement Method towards identification | |
CN114022872A (en) | Multi-crop leaf disease identification method based on dynamic neural network | |
CN109934835B (en) | Contour detection method based on deep strengthening network adjacent connection | |
CN113642362A (en) | Crowd density estimation method for intelligent escape in dense place | |
CN1858773A (en) | Image identifying method based on Gabor phase mode | |
CN111931630B (en) | Dynamic expression recognition method based on facial feature point data enhancement | |
CN110349119B (en) | Pavement disease detection method and device based on edge detection neural network | |
CN109886320B (en) | Human femoral X-ray intelligent recognition method and system | |
CN116543338A (en) | Student classroom behavior detection method based on gaze target estimation | |
CN110633705A (en) | Low-illumination imaging license plate recognition method and device | |
CN112084874B (en) | Object detection method and device and terminal equipment | |
CN114742760A (en) | Automatic and rapid diagnosis method for distal radius fracture by X-ray film | |
Chaudhary et al. | An effective and robust technique for the binarization of degraded document images |
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 |