CN115063730A - Video track analysis-based method and system for early warning of intrusion of workers into borderline area - Google Patents
Video track analysis-based method and system for early warning of intrusion of workers into borderline area Download PDFInfo
- Publication number
- CN115063730A CN115063730A CN202210821827.3A CN202210821827A CN115063730A CN 115063730 A CN115063730 A CN 115063730A CN 202210821827 A CN202210821827 A CN 202210821827A CN 115063730 A CN115063730 A CN 115063730A
- Authority
- CN
- China
- Prior art keywords
- target
- track
- video
- area
- early warning
- 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 36
- 238000004458 analytical method Methods 0.000 title claims abstract description 27
- 238000013507 mapping Methods 0.000 claims abstract description 15
- 238000001514 detection method Methods 0.000 claims abstract description 14
- 238000004364 calculation method Methods 0.000 claims abstract description 5
- 239000013598 vector Substances 0.000 claims description 18
- 238000010586 diagram Methods 0.000 claims description 17
- 238000004590 computer program Methods 0.000 claims description 14
- 238000012545 processing Methods 0.000 claims description 8
- 230000006870 function Effects 0.000 claims description 7
- 238000010606 normalization Methods 0.000 claims description 6
- 238000007781 pre-processing Methods 0.000 claims description 5
- 238000004422 calculation algorithm Methods 0.000 claims description 4
- 238000009499 grossing Methods 0.000 claims description 4
- 239000011159 matrix material Substances 0.000 claims description 4
- 238000000354 decomposition reaction Methods 0.000 claims description 3
- 238000001914 filtration Methods 0.000 claims description 3
- 238000013528 artificial neural network Methods 0.000 claims description 2
- 230000008859 change Effects 0.000 claims description 2
- 238000000605 extraction Methods 0.000 claims description 2
- 230000009466 transformation Effects 0.000 claims description 2
- 238000010276 construction Methods 0.000 description 6
- 238000012549 training Methods 0.000 description 5
- 230000008901 benefit Effects 0.000 description 4
- 238000012360 testing method Methods 0.000 description 4
- 230000008569 process Effects 0.000 description 3
- 230000006399 behavior Effects 0.000 description 2
- 238000002372 labelling Methods 0.000 description 2
- 238000004519 manufacturing process Methods 0.000 description 2
- 230000004048 modification Effects 0.000 description 2
- 238000012986 modification Methods 0.000 description 2
- 210000002569 neuron Anatomy 0.000 description 2
- 230000003287 optical effect Effects 0.000 description 2
- 208000027418 Wounds and injury Diseases 0.000 description 1
- 230000009471 action Effects 0.000 description 1
- 230000003044 adaptive effect Effects 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000006378 damage Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 208000014674 injury Diseases 0.000 description 1
- 238000013178 mathematical model Methods 0.000 description 1
- 238000012544 monitoring process Methods 0.000 description 1
- 230000035484 reaction time Effects 0.000 description 1
- 230000003068 static effect Effects 0.000 description 1
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
- G06V20/41—Higher-level, semantic clustering, classification or understanding of video scenes, e.g. detection, labelling or Markovian modelling of sport events or news items
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/762—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using clustering, e.g. of similar faces in social networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
-
- 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
- G06V20/46—Extracting features or characteristics from the video content, e.g. video fingerprints, representative shots or key frames
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V2201/00—Indexing scheme relating to image or video recognition or understanding
- G06V2201/07—Target detection
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Multimedia (AREA)
- General Physics & Mathematics (AREA)
- Physics & Mathematics (AREA)
- Software Systems (AREA)
- Evolutionary Computation (AREA)
- Computing Systems (AREA)
- Medical Informatics (AREA)
- General Health & Medical Sciences (AREA)
- Databases & Information Systems (AREA)
- Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Artificial Intelligence (AREA)
- Computational Linguistics (AREA)
- Image Analysis (AREA)
Abstract
The invention belongs to the field of intelligent behavior identification, and provides a method and a system for early warning that a worker breaks into an edge area based on video track analysis, wherein the method comprises the steps of acquiring video data to be detected; obtaining the position of a target in a video according to video data to be detected and a target detection network, and tracking the target to obtain a target tracking track; adopting a social force motion field and 3D geometric position relation mapping calculation to obtain an energy field value of the target; based on the target tracking track and the energy field value of the target, the energy estimation probability density of the target is obtained through the probability model of the marginal area crossed by the track, and when the energy estimation probability density of the target exceeds the threshold value of the marginal area, an alarm is given out, so that whether a worker crosses the marginal area or not can be accurately calculated.
Description
Technical Field
The invention belongs to the field of intelligent behavior identification, and particularly relates to a method and a system for early warning that a worker breaks into an edge area based on video track analysis.
Background
The statements in this section merely provide background information related to the present disclosure and may not necessarily constitute prior art.
Construction workers' working environment is complicated, working strength is relatively large, and lucky psychology and violation behavior are easy to occur, thereby causing safe production accidents. Therefore, the safety management of the construction site is an important link of the safety management of the construction enterprise. Among all safety accidents, accidents such as high fall, mechanical injury and the like caused by the fact that workers mistakenly enter the adjacent area are common. With the development of intelligent video monitoring technology, the early warning of danger caused by the fact that workers break into the borderline area based on an artificial intelligence method is researched, and the method has great industrial requirements and important practical significance.
At present, the scheme that the border region adopted is gone into to the workman mistake to the current is: through a system consisting of a video frame acquisition module, an identification module and a crossing judgment module, after a real target area is acquired in a camera, a preset virtual monitoring area about the real target area and an acquired video frame are identified by the judgment module to a moving target, and then the virtual monitoring area is configured based on dynamic programming, so that the crossing of the target through a dangerous area can be judged.
However, the above solution has the following disadvantages: the timeliness of switching between a real area and a virtual area and solving a dynamic programming problem is poor, the task of a video detection worker intruding into an adjacent area cannot be completed in time, and the safety of the worker cannot be well guaranteed.
Disclosure of Invention
In order to solve at least one technical problem in the background art, the invention provides a method and a system for early warning of the intrusion of a worker into an edge area based on video track analysis.
In order to achieve the purpose, the invention adopts the following technical scheme:
the invention provides a method for early warning of intrusion of workers into a boundary area based on video track analysis, which comprises the following steps:
acquiring video data to be detected;
obtaining the position of a target in a video according to video data to be detected and a target detection network, and carrying out track tracking on the target to obtain a target tracking track;
adopting a social force motion field and 3D geometric position relation mapping calculation to obtain an energy field value of the target;
and based on the target tracking track and the energy field value of the target, acquiring the energy estimation probability density of the target by traversing the marginal area probability model through the track, and alarming when the energy estimation probability density of the target exceeds a marginal area threshold value.
A second aspect of the present invention provides a video trajectory analysis-based worker intrusion into an edge area early warning system, including:
the data acquisition module is used for acquiring video data to be detected;
the target tracking track module is used for obtaining the position of a target in a video according to the video data to be detected and a target detection network, and carrying out track tracking on the target to obtain a target tracking track;
the track analysis module is used for mapping and calculating the social force motion field and the 3D geometric position relationship to obtain an energy field value of the target;
and the marginal area early warning module is used for acquiring the energy estimation probability density of the target by passing through the marginal area probability model based on the target tracking track and the energy field value of the target, and giving an alarm when the energy estimation probability density of the target exceeds a marginal area threshold value.
A third aspect of the invention provides a computer-readable storage medium.
A computer-readable storage medium, on which a computer program is stored, which when executed by a processor implements the steps in the video-trajectory-analysis-based method for early warning of intrusion of a worker into an edge area as described above.
A fourth aspect of the invention provides a computer apparatus.
A computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, the processor when executing the program implementing the steps of the video trajectory analysis based method for early warning of intrusion of a worker into an edge area as described above.
Compared with the prior art, the invention has the beneficial effects that:
the method comprises the steps of adopting a social force motion field and 3D geometric position relation mapping calculation to obtain an energy field value of a target; based on the target tracking track and the energy field value of the target, the energy estimation probability density of the target is obtained through the probability model of the track crossing the marginal area, when the energy estimation probability density of the target exceeds the threshold value of the marginal area, an alarm is given, and whether a worker crosses the marginal area or not is accurately calculated.
The invention selects the marks of the adjacent edge areas in a manual calibration mode, and then establishes a DCN adjacent edge area mark identification model, so that the adjacent edge areas of the construction site can be automatically and accurately screened out by a computer.
When the target tracks are tracked, the similarity of the images of the workers in each frame of video is calculated by adopting the Simese network architecture, so that the tracks of the workers are tracked, and the track tracking accuracy is improved.
Advantages of additional aspects of the invention will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of the invention.
Drawings
The accompanying drawings, which are incorporated in and constitute a part of this specification, are included to provide a further understanding of the invention, and are incorporated in and constitute a part of this specification, illustrate exemplary embodiments of the invention and together with the description serve to explain the invention and not to limit the invention.
FIG. 1 is a diagram of a video trajectory analysis-based method and system for early warning of an intrusion of a constructor into an adjacent area;
FIG. 2 is a schematic diagram of target detection according to an embodiment of the present invention;
FIG. 3 is a schematic diagram of a Siamese network according to an embodiment of the present invention;
fig. 4 is a schematic diagram of a smoothed trajectory of a wavelet packet according to an embodiment of the present invention.
Detailed Description
The invention is further described with reference to the following figures and examples.
It is to be understood that the following detailed description is exemplary and is intended to provide further explanation of the invention as claimed. Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.
It is noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of exemplary embodiments according to the invention. As used herein, the singular forms "a", "an" and "the" are intended to include the plural forms as well, and it should be understood that when the terms "comprises" and/or "comprising" are used in this specification, they specify the presence of stated features, steps, operations, devices, components, and/or combinations thereof, unless the context clearly indicates otherwise.
Example one
As shown in fig. 1, the embodiment provides a method for early warning that a worker breaks into an edge area based on video trajectory analysis, which includes the following steps:
step 1: acquiring video data;
in one or more embodiments, the video data may be obtained from a camera at a construction site, the video is converted into an image sequence frame by frame according to a time sequence, and the image sequence is used as an input vector of a subsequent step for preprocessing and operation. Step 2: and (3) data preprocessing, namely labeling and selecting the adjacent edge area in the sample, and identifying the mark of the adjacent edge area through the DCN so as to obtain the label of the adjacent edge area.
As one or more embodiments, in step 1, the data preprocessing includes the following specific steps:
(1) denoising and normalizing the image;
dividing original data X into a training sample and a testing sample, and carrying out denoising and normalization operations on input training and testing images.
The image denoising module configured to: calculating different weights A (i, j) of noise pixels and non-noise pixels by using an adaptive mean filtering algorithm at pixel points at the (i, j) th position of an input image, and adaptively searching for the optimal window size, as shown in formula (1):
where L is the filter radius.
The image normalization module configured to: the filtered vector is subjected to normalization processing of image data on the original data set X by traversing each pixel in the image matrix using the Z-score algorithm, as shown in formula (2):
wherein,
is the mean of x (t), and σ (x (t)) is the standard deviation of x (t).
(2) Target detection model training
From the normalized data x p (t) taking out x q (t)、x k (t) and x v (t) the three column vectors are arranged in rows as Q, K and V, respectively, as the channel headers for different layers of the concealment. Introducing a bias matrix B having an element x q (t)、x k (t) and x v (t) as the relative positional deviation b from the center pixel point of the image at the head of the channel ij The similarity a (Q, K, V) between two vectors is obtained as the weight ω by calculating the multi-head attention of the three matrices.
The calculation formula is as follows:
wherein d is the vector x q (t)、x k (t) and x v The dimension of (t).
(3) DCN border region mark identification
And (3) training the mark recognition of the edge region by using a DCN network, inputting the input vector subjected to denoising and normalization, then passing through an embedding layer, and then entering a backbone network. The network has a total of l layers, using the output x of the l layer l (t) and the calculated weight ω l The output x of the l +1 th layer can be calculated l+1 ：
Wherein x is 0 (t) is the output of the embedding layer, b l Is the corresponding cross constant of the l-th layer.
And (3) splicing the crossed output vectors obtained after the treatment of the formula (4), and outputting a mark y (t) which can be identified by the edge region after l layers.
(3) Boundary region calibration
And (3) combining the crossed items of the adjacent edge area mark y (t) obtained by the DCN and the thick frame y '(t) marked by the expert, and searching the shielded boundary frame after the combination of the adjacent edge area mark y (t) and the thick frame y' (t) marked by the expert to obtain a final output vector th of the boundary of the adjacent edge area.
The method comprises the steps that training sample images are delivered to building industry experts, rough labeling is carried out on an adjacent area by the experts, the adjacent area in the images is drawn by a large-range selected frame, and a manually calibrated thick frame of the adjacent area is set as y' (t).
The scheme has the advantages that the marks of the adjacent edge areas are selected in a manual calibration mode, and then the DCN adjacent edge area mark identification model is established, so that the adjacent edge areas of the construction site can be automatically and accurately screened out by a computer;
and 3, step 3: inputting the processed test sample into a target detection backbone network to obtain the position of a target in a video, and then tracking the track of the result obtained by target detection.
As one or more embodiments, step 3 specifically includes:
(1) target detection
As shown in fig. 2, the test sample x after denoising and normalization is processed p And (t) inputting the image into a trained target detection model in the data preprocessing module, selecting a boundary frame of a constructor from the sample image to obtain a characteristic diagram of the constructor, and taking the characteristic diagram as a template for target tracking.
(2) Simese network computation similarity
Firstly, an image detected by a target is converted into a gray image and is input into a twin neural network, and a schematic diagram of the siemese network is shown in fig. 3.
Obtaining a multidimensional characteristic Z after the shot gray level image passes through a main characteristic extraction network 1 And Z 2 Tiling the feature vectors on one dimension in a flatten mode to obtain one-dimensional feature vectors X i And calculating the L1 norm of the feature vector interpolation:
the formula (5) is that the distance between every two one-dimensional vectors is obtained, the distance is fully connected twice, the distance is fully connected to a neuron for the second time, sigmoid is taken from the result of the neuron, the value of sigmoid is between 0 and 1, and the similarity degree of two input pictures is represented
(3) Target tracking trajectory
Calculating the similarity of the target in the k frame image
Similarity with the target in the k +1 frame image
The difference ε of (c):
if epsilon is larger, the two tracks are not the same target track; otherwise, the position of the target is recorded and the target track is updated. And traversing all input vectors to obtain complete tracks of different targets in the video.
The scheme has the advantage that the similarity of the images of the workers in each frame of video is calculated by using the Simese network architecture, so that the tracking of the whereabouts of the workers is realized.
And 4, step 4: and smoothing the target tracking track, calculating the energy field value of the target by adopting a social force motion field and 3D geometric position relation mapping, and finally combining the energy field value with a calibrated edge region model to realize the function of analyzing the position relation between the target track and the edge region.
As shown in fig. 4, as one or more embodiments, in step 3, the smoothing processing on the target tracking trajectory specifically includes:
regarding the trajectory T in the images as a continuous time sequence, and assuming that the depth of wavelet packet decomposition is L for the T + k frame image, the input sequence can be expressed as: x ═ T t+1 ,T t+2 ,...,T t+k ],k＝1,...,2 L . In the wavelet packet decomposition tree, each layer uses a threshold function eta (-) to perform zero treatment on the wavelet coefficients of the track segments exceeding the threshold, thereby achieving the effect of smoothing the track. The wavelet packet sequence after thresholding is represented as:
and reconstructing the sequence by using the smooth wavelet packet representation through wavelet packet inverse transformation to obtain a smooth target moving track
As one or more embodiments, in step 3, the calculating to obtain the energy field value of the target by using the social force motion field and the 3D geometric position relationship mapping specifically includes:
for the target i, the advancing direction of the target i can be judged according to the facing direction of the target i in the k frame image, and the self-driving force of the target i is calculated
The formula is as follows:
wherein m is i Is the mass of the pedestrian, v e (t) is the desired speed of the vehicle,
in the desired direction, τ i In order to achieve the reaction time,
is based onThe external environment and its own factors constantly regulate the speed.
Calculating the social force of the target i by combining the self-driving force, the repulsive force of the target with other targets and surrounding obstacles and the like and the random force of random change of the target itself
The formula is as follows:
wherein,
is the sum of the repulsive forces between target i and target j,
is the sum of the repulsive forces between the object i and the obstacle o,
is a random force that randomly varies with the target itself.
After the motion state of the target in each frame of image is obtained, according to the zero parallax state principle of human vision, establishing a 3D geometric position mapping relation corresponding to different motion states and the position of the same target, thereby deducing the three-dimensional state and the position of the target in the current space;
fusing social forces of target i
And calculating to obtain the energy field value of the target at the time t according to the 3D geometric position mapping relation.
And 5: calculating the track obtained by tracking the target and estimating the numerical value of the target energy field, obtaining the energy estimation probability density of the target by establishing a probability model of the track crossing the marginal area, judging whether the energy estimation probability density exceeds a threshold value, and performing early warning.
As one or more embodiments, in step 4, the probability model of the track crossing the borderline region specifically includes:
re-acquiring the state of the target on each frame of video image picture along each smooth action track
Wherein,
for the appearance feature of target i at the kth frame,
for the size of the target i at the kth frame,
is the motion state of the target i at the k-th frame.
Take k 1.. t, i.e. traverse all static states
And calculating the joint probability density, and taking the maximum value as a standard template of the target i.
A target energy field is estimated. Fusing the aforementioned target social force
And 3D geometric location mapping
And calculating to obtain an energy field value rho of the target corresponding to the time t.
Setting a threshold th of a marginal area in an early warning module according to the established mathematical model of the marginal area, obtaining the motion state of the target in the current frame image according to the energy rho of the target constructor, and further judging whether the target is in the current frame image or notCrossing the adjacent edge area. For target i, if the estimated energy probability density from time t to time t + k
Triggering an early warning device to prevent the target from crossing the adjacent area.
The method has the advantages that the target energy field is estimated through the social force motion field and the 3D geometric position relation mapping, the probability density of the worker crossing the edge region is obtained, and whether the worker crosses the edge region is accurately calculated.
Example two
This embodiment provides workman break-in limb area early warning system based on video track analysis, includes:
the data acquisition module is used for acquiring video data to be detected;
the target tracking track module is used for obtaining the position of a target in a video according to the video data to be detected and a target detection network, and carrying out track tracking on the target to obtain a target tracking track;
the track analysis module is used for mapping and calculating the social force motion field and the 3D geometric position relationship to obtain an energy field value of the target;
and the marginal area early warning module is used for acquiring the energy estimation probability density of the target by passing through the marginal area probability model based on the target tracking track and the energy field value of the target, and giving an alarm when the energy estimation probability density of the target exceeds a marginal area threshold value.
EXAMPLE III
The present embodiment provides a computer-readable storage medium having stored thereon a computer program which, when executed by a processor, implements the steps in the video-trajectory-analysis-based method for early warning of intrusion of a worker into an edge area as described above.
Example four
The embodiment provides a computer device, which comprises a memory, a processor and a computer program stored on the memory and capable of running on the processor, wherein the processor executes the program to realize the steps of the method for early warning of intrusion of a worker into an edge area based on video track analysis.
As will be appreciated by one skilled in the art, embodiments of the present invention may be provided as a method, system, or computer program product. Accordingly, the present invention may take the form of a hardware embodiment, a software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, optical storage, and the like) having computer-usable program code embodied therein.
The present invention is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each flow and/or block of the flow diagrams and/or block diagrams, and combinations of flows and/or blocks in the flow diagrams and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
It will be understood by those skilled in the art that all or part of the processes of the methods of the embodiments described above can be implemented by a computer program, which can be stored in a computer-readable storage medium, and when executed, can include the processes of the embodiments of the methods described above. The storage medium may be a magnetic disk, an optical disk, a Read-Only Memory (ROM), a Random Access Memory (RAM), or the like.
The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention, and various modifications and changes may be made by those skilled in the art. Any modification, equivalent replacement, or improvement made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (10)
1. The method for early warning the intrusion of workers into the borderline area based on video track analysis is characterized by comprising the following steps:
acquiring video data to be detected;
obtaining the position of a target in a video according to video data to be detected and a target detection network, and carrying out track tracking on the target to obtain a target tracking track;
adopting a social force motion field and 3D geometric position relation mapping calculation to obtain an energy field value of the target;
and based on the target tracking track and the energy field value of the target, acquiring the energy estimation probability density of the target by traversing the marginal area probability model through the track, and alarming when the energy estimation probability density of the target exceeds a marginal area threshold value.
2. The video track analysis-based method for early warning that a worker breaks into an edge area according to claim 1, wherein the method for obtaining the position of the target in the video according to the video data to be detected and a target detection network, and performing track tracking on the target to obtain a target tracking track comprises the following steps:
selecting a boundary frame of a constructor from the kth frame image to obtain a characteristic diagram of the constructor, and taking the characteristic diagram as a target tracking template;
inputting the kth frame image into a twin neural network, and obtaining multi-dimensional characteristics through a trunk characteristic extraction network;
tiling the multidimensional features on one dimension to obtain one-dimensional feature vectors, solving the distance between the one-dimensional feature vectors, performing full connection on the distance twice, and outputting the similarity degree between the image and the target tracking template;
and calculating the difference between the similarity of the target in the kth frame image and the similarity of the target in the (k + 1) th frame image, if the difference is greater than a set threshold value, the difference is not the track of the same target, otherwise, recording the position of the target and updating the track of the target.
3. The video trajectory analysis-based worker intrusion critical area early warning method according to claim 1, wherein the calculating of the energy field value of the target by using the social force motion field and the 3D geometric position relationship map comprises:
obtaining the self-driving force of the target i according to the target motion state of the target i in the k frame image;
combining the self-driving force of the target, the repulsive force of the target with other targets and surrounding obstacles and the random force of random change of the target per se to obtain the social force of the target i;
based on the social force of a target i in each frame of image, establishing a 3D geometric position mapping relation corresponding to different motion states and the position of the same target according to the zero parallax state principle of human vision;
and fusing the social force of the target i and the 3D geometric position mapping relation, and calculating to obtain the energy field value of the target at the time t.
4. The video trajectory analysis-based worker intrusion critical area early warning method according to claim 1, wherein the critical area is obtained by the following steps:
identifying based on the video image frame to be detected and the DCN border area mark identification model to obtain a border area mark;
and combining the boundary region mark with the cross term of the coarse frame of the boundary region marked by the expert, and searching the boundary frame shielded after the combination of the boundary region mark and the coarse frame of the boundary region to obtain the final boundary of the boundary region.
5. The video track analysis-based worker intrusion into the borderline area early warning method according to claim 4, wherein the identifying based on the video image frame to be detected and the DCN borderline area mark identification model to obtain the borderline area mark comprises:
calculating the output of the l +1 th layer by using the output of the l layer and the corresponding weight, splicing the obtained cross output vectors, and outputting a mark which can be identified by an adjacent area after passing through the l layer;
the corresponding weight is obtained in the following manner:
extracting x from video image frame to be detected q (t)、x k (t) and x v (t) the three column vectors are used as different channel headers of the hidden layer and are respectively arranged into Q, K and V according to rows;
introducing a bias matrix B having an element x q (t)、x k (t) and x v (t) relative positional deviation with respect to a central pixel point of the image when the channel header is taken;
the similarity between the two vectors is obtained as a weight by calculating the multi-headed attention of the three matrices Q, K and V.
6. The video trajectory analysis-based worker intrusion critical area early warning method according to claim 1, wherein after the target tracking trajectory is obtained, wavelet packet trajectory smoothing processing is performed, and the method comprises the following steps:
and (3) regarding the target tracking track as a continuous time sequence, performing zero treatment on wavelet coefficients of track segments exceeding a threshold value by using a threshold function in each layer in a wavelet packet decomposition tree on the basis of the continuous time sequence to obtain a wavelet packet sequence, and reconstructing the smooth wavelet packet sequence by using wavelet packet inverse transformation to obtain a smooth target moving track.
7. The video trajectory analysis-based worker intrusion critical area early warning method as claimed in claim 1, wherein the preprocessing of data in acquiring video data to be detected comprises:
calculating different weights of noise pixels and non-noise pixels by using a self-adaptive mean filtering algorithm for pixel points of an input image, and filtering after searching for the optimal window size in a self-adaptive manner;
and (4) carrying out normalization processing on the image data by traversing each pixel in the image matrix by using a Z-score algorithm.
8. Workman break-in limb area early warning system based on video track analysis, its characterized in that includes:
the data acquisition module is used for acquiring video data to be detected;
the target tracking track module is used for obtaining the position of a target in a video according to the video data to be detected and a target detection network, and carrying out track tracking on the target to obtain a target tracking track;
the track analysis module is used for mapping and calculating the social force motion field and the 3D geometric position relationship to obtain an energy field value of the target;
and the marginal area early warning module is used for acquiring the energy estimation probability density of the target by passing through the marginal area probability model based on the target tracking track and the energy field value of the target, and giving an alarm when the energy estimation probability density of the target exceeds a marginal area threshold value.
9. A computer-readable storage medium, on which a computer program is stored, which, when being executed by a processor, carries out the steps of the method for early warning of intrusion of a worker into an area bordering on a border, based on video trajectory analysis, as set forth in any one of claims 1 to 7.
10. A computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, wherein the processor when executing the program implements the steps of the video trajectory analysis based method for early warning of intrusion of a worker into an area adjacent to a border as claimed in any one of claims 1 to 7.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210821827.3A CN115063730A (en) | 2022-07-13 | 2022-07-13 | Video track analysis-based method and system for early warning of intrusion of workers into borderline area |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210821827.3A CN115063730A (en) | 2022-07-13 | 2022-07-13 | Video track analysis-based method and system for early warning of intrusion of workers into borderline area |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115063730A true CN115063730A (en) | 2022-09-16 |
Family
ID=83206750
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210821827.3A Pending CN115063730A (en) | 2022-07-13 | 2022-07-13 | Video track analysis-based method and system for early warning of intrusion of workers into borderline area |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115063730A (en) |
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115620228A (en) * | 2022-10-13 | 2023-01-17 | 南京信息工程大学 | Subway shield door passenger door-rushing early warning method based on video analysis |
| CN117218112A (en) * | 2023-11-03 | 2023-12-12 | 江苏森途信息技术有限公司 | System and method for identifying excessive dangerous waste warehouse based on video analysis |
| CN118230261A (en) * | 2024-05-27 | 2024-06-21 | 四川省建筑科学研究院有限公司 | Intelligent construction site construction safety early warning method and system based on image data |
| CN118298363A (en) * | 2024-06-03 | 2024-07-05 | 珠海安士佳电子有限公司 | Method and system for monitoring and early warning of personnel going to and from home based on track analysis |
-
2022
- 2022-07-13 CN CN202210821827.3A patent/CN115063730A/en active Pending
Cited By (6)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN115620228A (en) * | 2022-10-13 | 2023-01-17 | 南京信息工程大学 | Subway shield door passenger door-rushing early warning method based on video analysis |
| CN117218112A (en) * | 2023-11-03 | 2023-12-12 | 江苏森途信息技术有限公司 | System and method for identifying excessive dangerous waste warehouse based on video analysis |
| CN117218112B (en) * | 2023-11-03 | 2024-02-06 | 江苏森途信息技术有限公司 | System and method for identifying excessive dangerous waste warehouse based on video analysis |
| CN118230261A (en) * | 2024-05-27 | 2024-06-21 | 四川省建筑科学研究院有限公司 | Intelligent construction site construction safety early warning method and system based on image data |
| CN118230261B (en) * | 2024-05-27 | 2024-07-19 | 四川省建筑科学研究院有限公司 | Intelligent construction site construction safety early warning method and system based on image data |
| CN118298363A (en) * | 2024-06-03 | 2024-07-05 | 珠海安士佳电子有限公司 | Method and system for monitoring and early warning of personnel going to and from home based on track analysis |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN115063730A (en) | Video track analysis-based method and system for early warning of intrusion of workers into borderline area | |
| CN104680508B (en) | Convolutional neural networks and the target object detection method based on convolutional neural networks | |
| CN105894701B (en) | The identification alarm method of transmission line of electricity external force damage prevention Large Construction vehicle | |
| CN111598066A (en) | Helmet wearing identification method based on cascade prediction | |
| CN113989702B (en) | Target identification method and device | |
| Qu et al. | Moving vehicle detection with convolutional networks in UAV videos | |
| CN104268594A (en) | Method and device for detecting video abnormal events | |
| CN109241829A (en) | The Activity recognition method and device of convolutional neural networks is paid attention to based on space-time | |
| CN113065431B (en) | Human body violation prediction method based on hidden Markov model and recurrent neural network | |
| CN116259002A (en) | Human body dangerous behavior analysis method based on video | |
| Abdullah et al. | Objects detection and tracking using fast principle component purist and kalman filter. | |
| Gal | Automatic obstacle detection for USV’s navigation using vision sensors | |
| Ali et al. | Deep Learning Algorithms for Human Fighting Action Recognition. | |
| CN111862147A (en) | Method for tracking multiple vehicles and multiple human targets in video | |
| Chughtai et al. | Object detection and segmentation for scene understanding via random forest | |
| Kang et al. | Robust visual tracking framework in the presence of blurring by arbitrating appearance-and feature-based detection | |
| CN117392565A (en) | Automatic identification method for unmanned aerial vehicle power inspection defects | |
| CN110111358B (en) | Target tracking method based on multilayer time sequence filtering | |
| CN111951298A (en) | Target tracking method fusing time series information | |
| DE102020201939A1 (en) | Method and device for evaluating an image classifier | |
| Lyasheva et al. | Application of the weight model to detect the state registration number of the vehicle in the image | |
| CN113536934B (en) | Active hiding method for patrol robot during tracking task execution | |
| CN116798117A (en) | Video understanding-based method for identifying abnormal actions under mine | |
| CN106023168A (en) | Method and device for edge detection in video surveillance | |
| CN113449663B (en) | Collaborative intelligent security method and device based on polymorphic fitting |
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 |