CN110659627A - Intelligent video monitoring method based on video segmentation - Google Patents

Intelligent video monitoring method based on video segmentation Download PDF

Info

Publication number
CN110659627A
CN110659627A CN201910949694.6A CN201910949694A CN110659627A CN 110659627 A CN110659627 A CN 110659627A CN 201910949694 A CN201910949694 A CN 201910949694A CN 110659627 A CN110659627 A CN 110659627A
Authority
CN
China
Prior art keywords
video
method based
monitoring method
picture
segmentation
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
Application number
CN201910949694.6A
Other languages
Chinese (zh)
Inventor
汝佩哲
李锐
于治楼
Current Assignee (The listed assignees may be inaccurate. Google has not performed a legal analysis and makes no representation or warranty as to the accuracy of the list.)
Shandong Inspur Artificial Intelligence Research Institute Co Ltd
Original Assignee
Shandong Inspur Artificial Intelligence Research Institute Co Ltd
Priority date (The priority date 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 date listed.)
Filing date
Publication date
Application filed by Shandong Inspur Artificial Intelligence Research Institute Co Ltd filed Critical Shandong Inspur Artificial Intelligence Research Institute Co Ltd
Priority to CN201910949694.6A priority Critical patent/CN110659627A/en
Publication of CN110659627A publication Critical patent/CN110659627A/en
Pending legal-status Critical Current

Links

Images

Classifications

    • G—PHYSICS
    • G06—COMPUTING; CALCULATING OR COUNTING
    • G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
    • G06V20/00—Scenes; Scene-specific elements
    • G06V20/40—Scenes; Scene-specific elements in video content
    • 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
    • G06N3/084—Backpropagation, e.g. using gradient descent
    • 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
    • G06V10/26—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion
    • G06V10/267—Segmentation of patterns in the image field; Cutting or merging of image elements to establish the pattern region, e.g. clustering-based techniques; Detection of occlusion by performing operations on regions, e.g. growing, shrinking or watersheds
    • G—PHYSICS
    • G08—SIGNALLING
    • G08B—SIGNALLING OR CALLING SYSTEMS; ORDER TELEGRAPHS; ALARM SYSTEMS
    • G08B25/00—Alarm systems in which the location of the alarm condition is signalled to a central station, e.g. fire or police telegraphic systems
    • G08B25/01—Alarm systems in which the location of the alarm condition is signalled to a central station, e.g. fire or police telegraphic systems characterised by the transmission medium
    • G08B25/08—Alarm systems in which the location of the alarm condition is signalled to a central station, e.g. fire or police telegraphic systems characterised by the transmission medium using communication transmission lines
    • G—PHYSICS
    • G08—SIGNALLING
    • G08B—SIGNALLING OR CALLING SYSTEMS; ORDER TELEGRAPHS; ALARM SYSTEMS
    • G08B3/00—Audible signalling systems; Audible personal calling systems
    • G08B3/10—Audible signalling systems; Audible personal calling systems using electric transmission; using electromagnetic transmission
    • H—ELECTRICITY
    • H04—ELECTRIC COMMUNICATION TECHNIQUE
    • H04N—PICTORIAL COMMUNICATION, e.g. TELEVISION
    • H04N7/00—Television systems
    • H04N7/18—Closed-circuit television [CCTV] systems, i.e. systems in which the video signal is not broadcast

Landscapes

  • Engineering & Computer Science (AREA)
  • Physics & Mathematics (AREA)
  • Theoretical Computer Science (AREA)
  • General Physics & Mathematics (AREA)
  • Multimedia (AREA)
  • Biophysics (AREA)
  • Artificial Intelligence (AREA)
  • Computational Linguistics (AREA)
  • Data Mining & Analysis (AREA)
  • Evolutionary Computation (AREA)
  • General Health & Medical Sciences (AREA)
  • Molecular Biology (AREA)
  • Computing Systems (AREA)
  • General Engineering & Computer Science (AREA)
  • Biomedical Technology (AREA)
  • Mathematical Physics (AREA)
  • Software Systems (AREA)
  • Life Sciences & Earth Sciences (AREA)
  • Health & Medical Sciences (AREA)
  • Signal Processing (AREA)
  • Electromagnetism (AREA)
  • Business, Economics & Management (AREA)
  • Emergency Management (AREA)
  • Image Analysis (AREA)
  • Closed-Circuit Television Systems (AREA)

Abstract

The invention particularly relates to an intelligent video monitoring method based on video segmentation. The intelligent video monitoring method based on video segmentation is characterized in that a NetWarp structure is constructed on the basis of a convolutional neural network CNN at a computer end, continuous frames Frame in a monitoring video are analyzed, and objects in the video are segmented to achieve the purposes of intrusion detection and real-time monitoring of abnormal conditions; and sending out warning sound to the abnormal condition to inform by mail or short message. According to the intelligent video monitoring method based on video segmentation, the machine is used for completing the work, not only can security personnel be relieved from the complicated and boring task of observing the screen for a long time, the manpower input is reduced, but also key data to be found can be quickly searched in massive video data, and the probability of non-report, false report and false report is greatly reduced.

Description

Intelligent video monitoring method based on video segmentation
Technical Field
The invention relates to the technical field of deep learning, in particular to an intelligent video monitoring method based on video segmentation.
Background
In daily life, various illegal events are layered endlessly, and illegal means tend to be more complex along with the development of science and technology. In today's society, the prevention and exploration of illegal violations is largely through the analysis of video surveillance. However, this requires a lot of labor and time, and the prevention before the event occurs requires someone to stare at the video surveillance all the time, and the inspection after the event is issued requires analysis of a lot of video surveillance. Obviously, the traditional means of staring at and checking the video monitor manually cannot meet the requirements of the current social development. Therefore, it is an urgent problem to provide a new and more intelligent video monitoring method.
In order to relive the task of watching the screen for a long time of security personnel from being complicated and boring, the part of work is finished by a machine, and the labor input is reduced; meanwhile, in order to quickly search key data to be found in massive video data and greatly reduce the probability of non-report, false report and false report, the invention provides an intelligent video monitoring method based on video segmentation.
The intelligent video monitoring is to utilize computer vision technology to process, analyze and understand video signals, to locate, identify and track the change in the monitored scene through automatic analysis of sequence images without human intervention, and to analyze and judge the behavior of the target on the basis.
Video is composed of successive pictures from frame to frame. The intelligent video monitoring needs to utilize an image segmentation technology for video monitoring analysis. Image segmentation refers to the process of subdividing a digital image into a plurality of image sub-regions (sets of pixels), also referred to as superpixels. The purpose of image segmentation is to simplify or change the representation of the image so that the image is easier to understand and analyze. With the development of deep learning, image segmentation techniques have achieved good results. But the image segmentation technology cannot be simply utilized in video analysis because the pictures in the video have the characteristic of time-sequence correlation. For example, two frames of pictures in a video are taken, the same person in the two pictures is respectively indoors and outdoors, a technician can hardly judge whether the person enters the room from the inside or the outside by using a picture segmentation method, and the technician can easily judge by using the time sequence characteristic between the video frames.
Disclosure of Invention
In order to make up for the defects of the prior art, the invention provides a simple and efficient intelligent video monitoring method based on video segmentation.
The invention is realized by the following technical scheme:
an intelligent video monitoring method based on video segmentation is characterized by comprising the following steps:
firstly, shooting a monitoring video by using a video monitoring camera;
secondly, transmitting the monitoring video data to a computer end;
thirdly, constructing a NetWarp structure on the basis of a convolutional neural network CNN at a computer end, analyzing continuous frames Frame in the monitoring video, and segmenting objects in the video so as to achieve the purposes of intrusion detection and monitoring abnormal conditions in real time;
fourthly, sending out warning sound to the abnormal condition and carrying out mail or short message notification.
In the third step, the video is segmented, and the method comprises the following steps:
(1) flow Computation
Inputting two continuous frames of pictures ItAnd I(t-1)Calculating the relative offset of each pixel position in two continuous frames of pictures by using DIS-Flow algorithm to obtain a slave frame ItTo I(t-1)Optical flow data of (a);
(2) flow Transformation
FlowCNN pair Slave frame I through convolutional neural networktTo I(t-1)Optical flow data ofConverting to obtain converted optical flow data;
(3) distortion represents Warping reductions
Calculating current frame picture I by converted optical flow datatPixel position mapping to previous frame picture I(t-1)Warping the pixel position to obtain the previous frame of picture I(t-1)The warped convolution kernel of (1);
(4) representing binding of reactions
The previous frame picture I is calculated(t-1)The distortion convolution kernel and the current frame picture ItThe convolution kernels are linearly added, and the addition result is transmitted to the rest image convolution network layers;
(5) intrusion Detection (Intrusion Detection)
Performing difference processing on the segmented video frame images, and judging whether the segmented video frame images are abnormal invasion or not according to a preset threshold; and if the abnormal invasion is found, warning is given.
In the step (1), the offsets of the pixels in the horizontal and vertical directions are represented by a set of floating point numbers μ and ν, respectively, (x ', y') (x + μ, y + ν), where (x, y) represents the picture ItAt each pixel position, (x ', y') represents picture I(t-1)At each pixel location.
Since the optical flow data obtained in step (1) does not represent the propagation behavior between video frames well, it needs to be converted.
In the step (2), the convolutional neural network FlowCNN is connected with the original two channel flows, and the previous frame of picture I(t-1)With the current frame picture ItAnd the difference between the two frames forms an 11-channel tensor (tensor) as input to the convolutional neural network FlowCNN; the convolutional neural network FlowCNN itself consists of 4 convolutional layers using the ReLU nonlinear function, all convolutional layers are composed of 3 × 3 convolutional kernels (filters), and the output channels of the first three layers are 16, 32 and 2 respectively; connecting the output of the third layer with the optical flow data calculated in step (1) as input to the last layer of convolutional layers to obtain the final converted stream data.
All parameters in the convolutional neural network FlowCNN are learned by a standard back propagation algorithm (BackPropagation).
In the step (3), a Net-Warp module is realized on the kth layer of the image convolution neural network, and convolution kernels of two adjacent frames are respectively
Figure BDA0002225373510000031
And
Figure BDA0002225373510000032
for convenience of presentation, respectively using ZtAnd Z(t-1)To represent; z(t-1)By twisting and ZtAlignment:
Figure BDA0002225373510000033
wherein the content of the first and second substances,
Figure BDA0002225373510000034
for warped convolution kernels, FtRepresents optical flow information, and Λ (·) represents a FlowCNN network.
The above-mentioned
Figure BDA0002225373510000035
That is, the current frame picture I is calculated by the converted optical flow datatPixel position (x, y) is mapped to previous frame I(t-1)Warped representation of picture position (x ', y'), implementing Warp () as Z(t-1)Bilinear interpolation of points (x ', y'), using (x)1，y1),(x1，y2),(x2，y1) And (x)2，y2) Represents the corner points of the grid where (x ', y') is located:
Figure BDA0002225373510000036
wherein the content of the first and second substances,
Figure BDA0002225373510000037
in the step (4), the calculated previous frame picture I(t-1)The distortion convolution kernel and the current frame picture ItThe convolution kernel linear addition of (2):
wherein, w1And w2Is a weight vector, length and zkChannel number is the same, an represents scalar multiplication; w is a1And w2The parameters are learned by a standard Back Propagation algorithm (Back Propagation);
finally, the result is obtained
Figure BDA0002225373510000041
Passed to the remaining image convolutional network layers.
In the step (5), the difference processing is performed between the segmented video frame images, and the method comprises the following steps:
d(i,j,t)＝F(i,j,t)-F(i,j，t-1)
wherein F (i, j, t) represents the pixel position of t frame (i, j), and F (i, j, t-1) represents the pixel position of t-1 frame.
In the step (5), the threshold value threshold is 30, and if the difference value d (i, j, t) reaches the threshold value threshold, it is determined that an abnormal intrusion has occurred.
The invention has the beneficial effects that: according to the intelligent video monitoring method based on video segmentation, the machine is used for completing the work, not only can security personnel be relieved from the complicated and boring task of observing the screen for a long time, the manpower input is reduced, but also key data to be found can be quickly searched in massive video data, and the probability of non-report, false report and false report is greatly reduced.
Drawings
FIG. 1 is a schematic diagram of an intelligent video monitoring method based on video segmentation.
Detailed Description
In order to make those skilled in the art better understand the technical solution of the present invention, the technical solution in the embodiment of the present invention will be clearly and completely described below with reference to the embodiment of the present invention. It is to be understood that the described embodiments are merely exemplary of the invention, and not restrictive of the full scope of the invention. All other embodiments, which can be derived by a person skilled in the art from the embodiments given herein without making any creative effort, shall fall within the protection scope of the present invention.
The intelligent video monitoring method based on video segmentation comprises the following steps:
firstly, shooting a monitoring video by using a video monitoring camera;
secondly, transmitting the monitoring video data to a computer end;
thirdly, constructing a NetWarp structure on the basis of a convolutional neural network CNN at a computer end, analyzing continuous frames Frame in the monitoring video, and segmenting objects in the video so as to achieve the purposes of intrusion detection and monitoring abnormal conditions in real time;
fourthly, sending out warning sound to the abnormal condition and carrying out mail or short message notification.
In the third step, the video is segmented, and the method comprises the following steps:
(1) flow Computation
Inputting two continuous frames of pictures ItAnd I(t-1)Calculating the relative offset of each pixel position in two continuous frames of pictures by using DIS-Flow algorithm to obtain a slave frame ItTo I(t-1)Optical flow data of (a);
(2) flow Transformation
FlowCNN pair Slave frame I through convolutional neural networktTo I(t-1)Converting the optical flow data to obtain converted optical flow data;
(3) distortion represents Warping reductions
Calculating current frame picture I by converted optical flow datatPixel position mapping to previous frame picture I(t-1)Warping the pixel position to obtain the previous frame of picture I(t-1)The warped convolution kernel of (1);
(4) representing binding of reactions
The previous frame picture I is calculated(t-1)The distortion convolution kernel and the current frame picture ItThe convolution kernels are linearly added, and the addition result is transmitted to the rest image convolution network layers;
(5) intrusion Detection (Intrusion Detection)
Performing difference processing on the segmented video frame images, and judging whether the segmented video frame images are abnormal invasion or not according to a preset threshold; and if the abnormal invasion is found, warning is given.
In the step (1), the offsets of the pixels in the horizontal and vertical directions are represented by a set of floating point numbers μ and ν, respectively, (x ', y') (x + μ, y + ν), where (x, y) represents the picture ItAt each pixel position, (x ', y') represents picture I(t-1)At each pixel location.
Since the optical flow data obtained in step (1) does not represent the propagation behavior between video frames well, it needs to be converted.
In the step (2), the convolutional neural network FlowCNN is connected with the original two channel flows, and the previous frame of picture I(t-1)With the current frame picture ItAnd the difference between the two frames forms an 11-channel tensor (tensor) as input to the convolutional neural network FlowCNN; the convolutional neural network FlowCNN itself consists of 4 convolutional layers using the ReLU nonlinear function, all convolutional layers are composed of 3 × 3 convolutional kernels (filters), and the output channels of the first three layers are 16, 32 and 2 respectively; connecting the output of the third layer with the optical flow data calculated in step (1) as input to the last layer of convolutional layers to obtain the final converted stream data.
All parameters in the convolutional neural network FlowCNN are learned by a standard back propagation algorithm (BackPropagation).
In the step (3), a Net-Warp module is realized on the kth layer of the image convolution neural network, and convolution kernels of two adjacent frames are respectively
Figure BDA0002225373510000061
And
Figure BDA0002225373510000062
for convenience of presentation, respectively using ZtAnd Z(t-1)To represent; z(t-1)By twisting and ZtAlignment:
Figure BDA0002225373510000063
wherein the content of the first and second substances,for warped convolution kernels, FtRepresents optical flow information, and Λ (·) represents a FlowCNN network.
The above-mentionedThat is, the current frame picture I is calculated by the converted optical flow datatPixel position (x, y) is mapped to previous frame I(t-1)Warped representation of picture position (x ', y'), implementing Warp () as Z(t-1)Bilinear interpolation of points (x ', y'), using (x)1，y1),(x1，y2),(x2，y1) And (x)2，y2) Represents the corner points of the grid where (x ', y') is located:
Figure BDA0002225373510000066
wherein the content of the first and second substances,
Figure BDA0002225373510000067
in the step (4), the calculated previous frame picture I(t-1)The distortion convolution kernel and the current frame picture ItThe convolution kernel linear addition of (2):
Figure BDA0002225373510000068
wherein, w1And w2Is a weight vector, length and zkChannel number is the same,/A table scalar multiplication; w is a1And w2The parameters are learned by a standard Back Propagation algorithm (Back Propagation);
finally, the result is obtained
Figure BDA0002225373510000069
Passed to the remaining image convolutional network layers.
Because the quality of the monitoring video is greatly influenced by the video resolution, the background time, the weather and other reasons, the video frames may have larger difference and cannot be directly subjected to difference processing. After the video image is segmented, the influence caused by the change of the video background is greatly eliminated, and the change of people, vehicles and the like is focused.
In the step (5), the difference processing is performed between the segmented video frame images, and the method comprises the following steps:
d(i，j，t)＝F(i，j，t)-F(i，j，t-1)
wherein F (i, j, t) represents the pixel position of t frame (i, j), and F (i, j, t-1) represents the pixel position of t-1 frame.
In the step (5), the threshold value threshold is 30, and if the difference value d (i, j, t) reaches the threshold value threshold, it is determined that an abnormal intrusion has occurred.
The above describes an intelligent video monitoring method based on video segmentation in the embodiment of the present invention in detail. While the present invention has been described with reference to specific examples, which are provided to assist in understanding the core concepts of the present invention, it is intended that all other embodiments that can be obtained by those skilled in the art without departing from the spirit of the present invention shall fall within the scope of the present invention.

Claims (10)

1. An intelligent video monitoring method based on video segmentation is characterized by comprising the following steps:
firstly, shooting a monitoring video by using a video monitoring camera;
secondly, transmitting the monitoring video data to a computer end;
thirdly, constructing a NetWarp structure on the basis of a convolutional neural network CNN at a computer end, analyzing continuous frames Frame in the monitoring video, and segmenting objects in the video so as to achieve the purposes of intrusion detection and monitoring abnormal conditions in real time;
fourthly, sending out warning sound to the abnormal condition and carrying out mail or short message notification.
2. The intelligent video monitoring method based on video segmentation as claimed in claim 1, wherein: in the third step, the video is segmented, and the method comprises the following steps:
(1) flow Computation
Inputting two continuous frames of pictures ItAnd I(t-1)Calculating the relative offset of each pixel position in two continuous frames of pictures by using DIS-Flow algorithm to obtain a slave frame ItTo I(t-1)Optical flow data of (a);
(2) flow Transformation
FlowCNN pair Slave frame I through convolutional neural networktTo I(t-1)Converting the optical flow data to obtain converted optical flow data;
(3) distortion represents Warping reductions
Calculating current frame picture I by converted optical flow datatPixel position mapping to previous frame picture I(t-1)Warping the pixel position to obtain the previous frame of picture I(t-1)The warped convolution kernel of (1);
(4) representing binding of reactions
The previous frame picture I is calculated(t-1)The distortion convolution kernel and the current frame picture ItThe convolution kernels are linearly added, and the addition result is transmitted to the rest image convolution network layers;
(5) intrusion Detection
Performing difference processing on the segmented video frame images, and judging whether the segmented video frame images are abnormal invasion or not according to a preset threshold; and if the abnormal invasion is found, warning is given.
3. The intelligent video monitoring method based on video segmentation as claimed in claim 2, wherein: in the step (1), the offsets of the pixels in the horizontal and vertical directions are represented by a set of floating point numbers μ and ν, respectively, (x ', y') (x + μ, y + ν), where (x, y) represents the picture ItAt each pixel position, (x ', y') represents picture I(t-1)At each pixel location.
4. The intelligent video monitoring method based on video segmentation as claimed in claim 3, wherein: in the step (2), the convolutional neural network FlowCNN is connected with the original two channel flows, and the previous frame of picture I(t-1)With the current frame picture ItAnd the difference between the two frames forms an 11-channel tensor (tensor) as input to the convolutional neural network FlowCNN; the convolutional neural network FlowCNN itself consists of 4 convolutional layers using the ReLU nonlinear function, all convolutional layers are composed of 3 × 3 convolutional kernels (filters), and the output channels of the first three layers are 16, 32 and 2 respectively; connecting the output of the third layer with the optical flow data calculated in step (1) as input to the last layer of convolutional layers to obtain the final converted stream data.
5. The intelligent video monitoring method based on video segmentation as claimed in claim 4, wherein: all parameters in the convolutional neural network FlowCNN are learned by a standard back propagation algorithm.
6. The intelligent video monitoring method based on video segmentation as claimed in claim 5, wherein: in the step (3), a Net-Warp module is realized on a k layer of the image convolution neural network, and convolution kernels of two adjacent frames are respectively Zt kAnd
Figure FDA0002225373500000021
for convenience of presentation, respectively using ZtAnd Z(t-1)To represent; z(t-1)By twisting and ZtAlignment:
Figure FDA0002225373500000022
wherein the content of the first and second substances,
Figure FDA0002225373500000023
for warped convolution kernels, FtRepresents optical flow information, and Λ (·) represents a FlowCNN network.
7. The intelligent video monitoring method based on video segmentation as claimed in claim 6, wherein: the above-mentioned
Figure FDA0002225373500000024
That is, the current frame picture I is calculated by the converted optical flow datatPixel position (x, y) is mapped to previous frame I(t-1)Warped representation of picture position (x ', y'), implementing Warp () as Z(t-1)Bilinear interpolation of points (x ', y'), using (x)1，y1),(x1，y2),(x2，y1) And (x)2，y2) Represents the corner points of the grid where (x ', y') is located:
Figure FDA0002225373500000025
wherein the content of the first and second substances,
Figure FDA0002225373500000026
8. the intelligent video monitoring method based on video segmentation as claimed in claim 7, wherein: in the step (4), the calculated previous frame picture I(t-1)The distortion convolution kernel and the current frame picture ItThe convolution kernel linear addition of (2):
Figure FDA0002225373500000027
wherein, w1And w2Is a weight vector, length and zkChannel number is the same, an represents scalar multiplication; w is a1And w2The parameters are learned through a standard back propagation algorithm;
finally, the result is obtained
Figure FDA0002225373500000031
Passed to the remaining image convolutional network layers.
9. The intelligent video monitoring method based on video segmentation as claimed in claim 8, wherein: in the step (5), the difference processing is performed between the segmented video frame images, and the method comprises the following steps:
d(i，j，t)＝F(i，j，t)-F(i，j，t-1)
wherein F (i, j, t) represents the pixel position of t frame (i, j), and F (i, j, t-1) represents the pixel position of t-1 frame.
10. The intelligent video monitoring method based on video segmentation as claimed in claim 9, wherein: in the step (5), the threshold value threshold is 30, and if the difference value d (i, j, t) reaches the threshold value threshold, it is determined that an abnormal intrusion has occurred.
CN201910949694.6A 2019-10-08 2019-10-08 Intelligent video monitoring method based on video segmentation Pending CN110659627A (en)

Priority Applications (1)

Application Number Priority Date Filing Date Title
CN201910949694.6A CN110659627A (en) 2019-10-08 2019-10-08 Intelligent video monitoring method based on video segmentation

Applications Claiming Priority (1)

Application Number Priority Date Filing Date Title
CN201910949694.6A CN110659627A (en) 2019-10-08 2019-10-08 Intelligent video monitoring method based on video segmentation

Publications (1)

Publication Number Publication Date
CN110659627A true CN110659627A (en) 2020-01-07

Family

ID=69040050

Family Applications (1)

Application Number Title Priority Date Filing Date
CN201910949694.6A Pending CN110659627A (en) 2019-10-08 2019-10-08 Intelligent video monitoring method based on video segmentation

Country Status (1)

Country Link
CN (1) CN110659627A (en)

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113379046A (en) * 2020-03-09 2021-09-10 中国科学院深圳先进技术研究院 Method for accelerated computation of convolutional neural network, storage medium, and computer device
CN113487618A (en) * 2021-09-07 2021-10-08 北京世纪好未来教育科技有限公司 Portrait segmentation method, portrait segmentation device, electronic equipment and storage medium
WO2022012002A1 (en) * 2020-07-15 2022-01-20 Zhejiang Dahua Technology Co., Ltd. Systems and methods for video analysis

Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101266710A (en) * 2007-03-14 2008-09-17 中国科学院自动化研究所 An all-weather intelligent video analysis monitoring method based on a rule
US20140348390A1 (en) * 2013-05-21 2014-11-27 Peking University Founder Group Co., Ltd. Method and apparatus for detecting traffic monitoring video
CN107801000A (en) * 2017-10-17 2018-03-13 国网江苏省电力公司盐城供电公司 A kind of transmission line of electricity external force damage prevention intelligent video monitoring system
CN108009506A (en) * 2017-12-07 2018-05-08 平安科技（深圳）有限公司 Intrusion detection method, application server and computer-readable recording medium
CN108846335A (en) * 2018-05-31 2018-11-20 武汉市蓝领英才科技有限公司 Wisdom building site district management and intrusion detection method, system based on video image

Patent Citations (5)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN101266710A (en) * 2007-03-14 2008-09-17 中国科学院自动化研究所 An all-weather intelligent video analysis monitoring method based on a rule
US20140348390A1 (en) * 2013-05-21 2014-11-27 Peking University Founder Group Co., Ltd. Method and apparatus for detecting traffic monitoring video
CN107801000A (en) * 2017-10-17 2018-03-13 国网江苏省电力公司盐城供电公司 A kind of transmission line of electricity external force damage prevention intelligent video monitoring system
CN108009506A (en) * 2017-12-07 2018-05-08 平安科技（深圳）有限公司 Intrusion detection method, application server and computer-readable recording medium
CN108846335A (en) * 2018-05-31 2018-11-20 武汉市蓝领英才科技有限公司 Wisdom building site district management and intrusion detection method, system based on video image

Non-Patent Citations (1)

* Cited by examiner, † Cited by third party
Title
R.GADDE, ET AL.: "Semantic Video CNNs through RepresentationWarping", 《2017 IEEE INTERNATIONAL CONFERENCE ON COMPUTER VISION》 *

Cited By (3)

* Cited by examiner, † Cited by third party
Publication number Priority date Publication date Assignee Title
CN113379046A (en) * 2020-03-09 2021-09-10 中国科学院深圳先进技术研究院 Method for accelerated computation of convolutional neural network, storage medium, and computer device
WO2022012002A1 (en) * 2020-07-15 2022-01-20 Zhejiang Dahua Technology Co., Ltd. Systems and methods for video analysis
CN113487618A (en) * 2021-09-07 2021-10-08 北京世纪好未来教育科技有限公司 Portrait segmentation method, portrait segmentation device, electronic equipment and storage medium

Similar Documents

Publication Publication Date Title
CN104966304B (en) Multi-target detection tracking based on Kalman filtering and nonparametric background model
CN109711318B (en) Multi-face detection and tracking method based on video stream
CN110232380A (en) Fire night scenes restored method based on Mask R-CNN neural network
CN110659627A (en) Intelligent video monitoring method based on video segmentation
CN112396635B (en) Multi-target detection method based on multiple devices in complex environment
EP4090036A1 (en) Privacy shielding processing method and apparatus, electronic device, and monitoring system
CN110096945B (en) Indoor monitoring video key frame real-time extraction method based on machine learning
CN110751630A (en) Power transmission line foreign matter detection method and device based on deep learning and medium
CN112257643A (en) Smoking behavior and calling behavior identification method based on video streaming
JP3486229B2 (en) Image change detection device
CN111582074A (en) Monitoring video leaf occlusion detection method based on scene depth information perception
CN111369548A (en) No-reference video quality evaluation method and device based on generation countermeasure network
CN111369557B (en) Image processing method, device, computing equipment and storage medium
CN105930814A (en) Method for detecting personnel abnormal gathering behavior on the basis of video monitoring platform
CN116740110A (en) Image edge measuring system based on secondary determination
CN111147815A (en) Video monitoring system
CN112532999B (en) Digital video frame deletion tampering detection method based on deep neural network
CN111797761A (en) Three-stage smoke detection system, method and readable medium
CN112183310A (en) Method and system for filtering redundant monitoring pictures and screening invalid monitoring pictures
CN113052878A (en) Multi-path high-altitude parabolic detection method and system for edge equipment in security system
EP4136832A1 (en) Enhanced person detection using face recognition and reinforced, segmented field inferencing
CN112307936A (en) Passenger flow volume analysis method, system and device based on head and shoulder detection
CN113158963B (en) Method and device for detecting high-altitude parabolic objects
CN111191593A (en) Image target detection method and device, storage medium and sewage pipeline detection device
CN117011288B (en) Video quality diagnosis method and system

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
RJ01 Rejection of invention patent application after publication

Application publication date: 20200107