CN110766894A

CN110766894A - Community fence crossing early warning method, system, server and computer storage medium

Info

Publication number: CN110766894A
Application number: CN201911352404.6A
Authority: CN
Inventors: 蒋宇
Original assignee: Evergrande Intelligent Technology Co Ltd
Current assignee: Evergrande Intelligent Technology Co Ltd
Priority date: 2019-12-25
Filing date: 2019-12-25
Publication date: 2020-02-07

Abstract

The invention provides a community fence crossing early warning method, which comprises the following steps: arranging a camera and an infrared sensing device around a community enclosure; the server monitors abnormal signals output by the infrared sensing devices around the preset enclosing wall, assigns corresponding Boolean values to the preset first variable and enters a monitoring image analysis mode; the server acquires a monitoring video stream after entering the monitoring image analysis mode, wherein a certain mapping relation is formed between a selected interval of the monitoring video stream and the set position of the infrared sensing device, and the server assigns a Boolean value to the preset second variable according to an analysis result obtained in the monitoring image analysis mode; and the server performs AND gate operation on the first variable and the second variable to obtain a third variable, and if the Boolean value of the third variable is true, a preset early warning message is sent to the property management center. The invention has high early warning timeliness and accuracy.

Description

Community fence crossing early warning method, system, server and computer storage medium

Technical Field

The embodiment of the invention relates to the field of image processing, in particular to a community fence crossing early warning method, a community fence crossing early warning system, a community fence crossing early warning server and a computer storage medium.

Background

At present, a plurality of communities are surrounded by enclosing walls or guardrails, and the silence and the safety of the communities are ensured. However, people often climb over the enclosing wall in the cell to facilitate the drawing, or thieves climb over the enclosing wall to illegally enter the cell.

Some of the most primitive methods simply add sharp objects such as glass frit to the top of the enclosure to prevent an illegal enclosure from climbing over. With the popularization of video monitoring, a plurality of cameras are added to the wall, and when some events occur, images shot by the cameras are called to check, and the cameras are mostly used for paths for collecting evidences afterwards.

An infrared sensor is adopted in the following process, and the alarm is given out when the infrared sensor senses an infrared signal above the enclosing wall. This kind of mode misjudgment rate is higher, for example child has thrown a brick or other article in toward the enclosure, can arouse infrared sensor to produce the misjudgment when brick etc. passes through the top of enclosure top, and then triggers the alarm, for example the enclosure limit always has planted trees again, and the fallen leaves of trees drop and shelter from the light of infrared sensing equipment, also produces the misjudgment equally, and then triggers the alarm, and is visible, still can't avoid the various misjudgment phenomena that infrared sensor produced in the enclosure security protection detects among the prior art at present.

Disclosure of Invention

In order to solve the problems, the invention provides a community enclosure crossing early warning method with higher enclosure security detection precision, wherein cameras and infrared sensing devices are arranged around a community enclosure, and the community enclosure crossing early warning method comprises the following steps:

the method comprises the steps that a server monitors abnormal signals output by an infrared sensing device around a preset enclosing wall, the abnormal signals are triggered by shielding of light beams emitted by the infrared sensing device and are low-level signals, and the server endows a preset first variable with a corresponding Boolean value according to the fact whether the abnormal signals are received or not;

if the server obtains the abnormal signal uploaded by the infrared sensing device through monitoring, the server enters a monitoring image analysis mode; after the server enters the image analysis mode, acquiring a monitoring video stream, wherein the selection interval of the monitoring video stream is related to the placement position of the infrared sensing device; the server extracts a frame image from the monitoring video stream to process the frame image and obtain an analysis result generated according to the image analysis mode, and a corresponding Boolean value is given to the preset second variable according to the analysis result;

the server performs AND gate operation on the first variable and the second variable to obtain a third variable, and if the Boolean value of the third variable is true, a preset early warning message is sent to a property management center;

wherein, the image analysis mode comprises the following steps:

the server extracts frame images from the monitoring video stream;

the server takes the frame image as an original input image and inputs the frame image into a feature extraction structure formed by interweaving five convolution layers and three pooling layers to obtain a first output layer extracted from the original input image;

mapping each point of the first output layer back to the original input image to obtain coordinates of each point of the first output layer, wherein a coordinate system corresponding to the coordinates takes the original input image as reference;

inputting the first output layer into a preset target selection model to obtain a region judged to contain human body features and a candidate frame corresponding to the region containing the human body features;

the server captures the subsequent frame images of the original input image, performs the same processing on the subsequent frame images, and updates the candidate frame position according to the data obtained by the processing;

and when the position of the candidate frame is updated every time, judging whether the candidate frame touches a preset warning line position, and if so, assigning a true value to the second variable by the server.

Preferably, the step of performing an and gate operation on the first variable and the second variable by the server to obtain a third variable, and if a boolean value of the third variable is true, sending a preset warning message to the property management center further includes:

and if the Boolean value of the third variable is false, the server enters a flameout mode, numerical values of the first variable, the second variable and the third variable are refreshed, the server only keeps a communication module to supply power and eliminates a service process related to the image analysis mode in the flameout mode, and the communication module is used for monitoring abnormal signals uploaded by the infrared sensing device.

Preferably, the step of determining whether the candidate frame touches a preset warning line position includes:

and judging whether any frame of the candidate frame intersects with the warning line, if so, judging that the candidate frame touches the preset warning line, and if the analysis result is that a community enclosure crossing event is generated, assigning a true value to the second variable, and if no frame of the candidate frame intersects with the warning line, assigning a false value to the second variable.

Preferably, the step of determining whether any frame side of the candidate frame intersects with the warning line includes:

setting the end point of the warning line as

Andthe warning line is in a vector representation form as a vector

；

Let the candidate frame end point be

,

,

And

each side of the candidate frame is respectively a vector in a vector representation mode

Vector of motion

Vector of motion

And vector

;

Selecting a side long vector of the candidate frame

Creating a virtual vector as a first order judgment object

Vector of motion

Vector of motion

Vector of motion

；

Computing vectors

And

obtaining a first cross product by the cross product of the first step;

computing vectors

And

obtaining a second cross product by the cross product;

judging whether the first cross product and the second cross product are abnormal signs or not, if so, judging that a frame edge of the candidate frame is intersected with the warning line, and if so, touching the warning line by the candidate frame to assign a true value to the second variable;

if not, the same calculation steps are executed on the rest side length vectors, if all the side length vectors are not intersected with the warning line, the output result is that the candidate frame is not intersected with the warning line, and the second variable is assigned with a false value.

Preferably, the step of setting the warning line is:

receiving auxiliary positioning points input by community personnel, sequencing the positioning points according to the input sequence, and generating a plurality of auxiliary lines according to the principle that two adjacent points generate auxiliary lines, wherein the auxiliary lines are used for calculating and generating the warning lines;

extracting intersections between the auxiliary linesArray, obtaining new polygon vertex according to preset external polygon algorithm

Array of the polygon verticesThe array is used for representing the warning line, and the external polygon algorithm comprises:

let the intersection point of two adjacent auxiliary lines be

The vector of two adjacent auxiliary lines is

And

said point of intersectionThe side edge has a point

，

=

+(

)。

Preferably, in the step of inputting the first output layer to a preset target selection model to obtain a region determined to include human body features and a candidate frame corresponding to the region including human body features, the target selection model is composed of 2 layers of full-link layers and a loss function layer, where n is multiplied by 1.

Preferably, the loss function in the loss function layer is:

wherein,

is the sample distribution value of the real label,

is composed ofW is a weight coefficient,

is a preset trimming factor constant, j is the total weight number, n is an integer and n is more than or equal to 1.

Preferably, the step of arranging the cameras and the infrared sensing devices around the community fence includes:

and arranging an infrared sensing device on the ground outwards from the community enclosing wall.

Preferably, the step of arranging the camera and the infrared sensing device around the enclosure of the rebirth further comprises:

and arranging infrared sensing devices on the community enclosing walls outwards and the communities inwards on the ground.

Preferably, the step of associating the selected interval of the monitoring video stream with the placement position of the infrared sensing device includes:

△t=

+Q

wherein △ t is the selected interval of the monitoring video stream,

and v is the distance between the infrared sensing device and the enclosing wall, the preset human body moving speed and Q is a time delay factor.

The embodiment of the invention also provides a community fence crossing early warning system, which comprises:

the infrared module is used for monitoring abnormal signals output by the infrared sensing devices around the preset enclosing wall by the server, the abnormal signals are triggered by shielding of light beams emitted by the infrared sensing devices and are low-level signals, and the server gives corresponding Boolean values to the preset first variable according to the fact whether the abnormal signals are received or not;

the image analysis module is used for entering a monitoring image analysis mode if the server monitors and obtains the abnormal signals uploaded by the infrared sensing device; after the server enters the image analysis mode, acquiring a monitoring video stream, wherein the selection interval of the monitoring video stream is related to the placement position of the infrared sensing device; the server extracts a frame image from the monitoring video stream to process the frame image and obtain an analysis result generated according to the image analysis mode, and a corresponding Boolean value is given to the preset second variable according to the analysis result;

the output module is used for the server to carry out AND gate operation on the first variable and the second variable to obtain a third variable, and if the Boolean value of the third variable is true, a preset early warning message is sent to the property management center;

wherein the image analysis module is further to:

the server extracts frame images from the monitoring video stream;

and judging whether the candidate frame touches a preset warning line position or not when the position of the candidate frame is updated every time, and if so, assigning a true value to the second variable by the server.

Preferably, the output module further includes:

and the flameout unit is used for entering a flameout mode by the server and refreshing the numerical values of the first variable, the second variable and the third variable if the Boolean value of the third variable is false, the server only keeps the power supply of the communication module in the flameout mode and eliminates the service process related to the image analysis mode, and the communication module is used for monitoring the abnormal signals uploaded by the infrared sensing device.

Preferably, the image analysis module further comprises:

and the intersection judging unit is used for judging whether any frame edge of the candidate frame intersects with the warning line, if so, judging that the candidate frame touches the preset warning line, and if the analysis result is that a community enclosure crossing event is generated, assigning a true value to the second variable, and if no frame edge of the candidate frame intersects with the warning line, assigning a false value to the second variable.

Preferably, the intersection determination unit is further configured to:

setting the end point of the warning line as

And

the warning line is in a vector representation form as a vector

；

Let the candidate frame end point be

,

,

And

Vector of motion

Vector of motion

And vector

;

Selecting a side long vector of the candidate frame

Creating a virtual vector as a first order judgment object

Vector of motion

Vector of motion

Vector of motion

；

Computing vectors

And

obtaining a first cross product by the cross product of the first step;

computing vectors

And

obtaining a second cross product by the cross product;

Preferably, the image analysis module further comprises:

the warning line unit is used for receiving auxiliary positioning points input by community personnel, sequencing the positioning points according to the input sequence, and generating a plurality of auxiliary lines according to the principle that two adjacent points generate the auxiliary lines, wherein the auxiliary lines are used for calculating and generating the warning lines;

extracting intersections between the auxiliary lines

Array, obtaining new polygon vertex according to preset external polygon algorithm

Array of the polygon vertices

The array is used for representing the warning line, and the external polygon algorithm comprises:

let the intersection point of two adjacent auxiliary lines be

The vector of two adjacent auxiliary lines is

And

said point of intersection

The side edge has a point，

=+(

)。

Preferably, the image analysis module is further configured to input the first output layer to a preset target selection model, and in the step of obtaining a region determined to include human body features and a candidate frame corresponding to the region including human body features, the target selection model is composed of 2 full-connected layers obtained by multiplying n by 1 and a loss function layer.

Preferably, the loss function in the loss function layer of the image analysis module is:

wherein,

is the sample distribution value of the real label,

is composed of

W is a weight coefficient,

Preferably, the infrared module is further configured to:

Preferably, the infrared module further sets a function of the selected interval related to the placement position of the infrared sensing device as follows:

△t=

+Q

wherein △ t is the selected interval of the monitoring video stream,and the distance between the infrared sensing device and the enclosing wall is V, the preset human body moving speed is v, and Q is a delay factor.

The invention also provides a server, which comprises a memory, a processor and a computer program stored on the memory and capable of running on the processor, and is characterized in that the computer program is executed by the processor to implement the community fence crossing warning method.

The present invention also provides a computer storage medium having a computer program stored therein, where the computer program can be executed by at least one processor, so that the at least one processor executes the community fence crossing warning method as described above.

According to the community enclosure crossing early warning method, the community enclosure crossing early warning system, the computer equipment and the storage medium, whether an abnormal signal exists is detected through the infrared sensing device, if the abnormal signal exists, the monitoring image analysis mode is entered, whether the abnormal signal can generate an alarm message needing to be alarmed is judged after analysis, and the alarm message is sent to the property management center, so that the occurrence of misjudgment caused by only adopting infrared sensing detection is reduced through a multiple verification mode. In addition, when the monitoring video stream is obtained, a certain mapping relation is set between the time interval selection and the position of the infrared sensing device, so that the process of intercepting the whole event is better realized, a more complete data source is provided for the follow-up analysis result, the probability of misjudgment is greatly reduced, and the accuracy of early warning of crossing of the community enclosure is greatly improved.

Drawings

FIG. 1 is a flowchart illustrating steps of a community fence crossing warning method according to the present invention;

FIG. 2 is a schematic diagram of a fully connected layer structure in the prior art;

FIG. 3 is a schematic diagram of a fully-connected layer design according to the present invention;

FIG. 4 is a schematic diagram of program modules of the community fence crossing warning system according to the present invention;

FIG. 5 is a diagram of the hardware structure of the computer device of the present invention.

FIG. 6 is a schematic view of a position of an infrared sensor apparatus according to another embodiment of the present invention;

FIG. 7 is a schematic view of a position of an infrared sensing device according to another embodiment of the present invention;

FIG. 8 is a schematic view of a warning line according to the present invention;

FIG. 9 is a schematic diagram of a human candidate box according to the present invention;

FIG. 10 is a schematic diagram of a human body candidate frame touching a warning line according to the present invention;

FIG. 11 is a diagram of the weight components of the fully-connected layer according to the present invention.

Detailed Description

In order to make the objects, technical solutions and advantages of the present invention more apparent, the present invention is described in further detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative of the invention and are not intended to limit 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 terminology used in the embodiments of the invention is for the purpose of describing particular embodiments only and is not intended to be limiting of the invention. As used in the examples of the present invention and the appended claims, the singular forms "a," "an," and "the" are intended to include the plural forms as well, unless the context clearly indicates otherwise.

It should be understood that the term "and/or" as used herein is merely one type of association that describes an associated object, meaning that three relationships may exist, e.g., a and/or B may mean: a exists alone, A and B exist simultaneously, and B exists alone. In addition, the character "/" herein generally indicates that the former and latter related objects are in an "or" relationship.

It should be understood that although the terms first, second, etc. may be used to describe the designated key in embodiments of the present invention, the designated key should not be limited to these terms. These terms are only used to distinguish specified keywords from each other. For example, the first specified keyword may also be referred to as the second specified keyword, and similarly, the second specified keyword may also be referred to as the first specified keyword, without departing from the scope of embodiments of the present invention.

The word "if" as used herein may be interpreted as referring to "at … …" or "when … …" or "corresponding to a determination" or "in response to a detection", depending on the context. Similarly, the phrase "if determined" or "if detected (a stated condition or time)" may be interpreted as "when determined" or "in response to a determination" or "when detected (a stated condition or event)" or "in response to a detection (a stated condition or event)", depending on the context.

Referring to fig. 1, an embodiment of the present invention provides a method for early warning crossing of a community fence, where cameras and infrared sensing devices are arranged around the community fence, and the method for early warning crossing of the community fence includes:

step S100: the server monitors abnormal signals output by the infrared sensing devices around the preset enclosing wall, the abnormal signals are triggered by the fact that light beams emitted by the infrared sensing devices are shielded, the abnormal signals are low-level signals, and the server gives corresponding Boolean values to the preset first variable according to the fact that whether the abnormal signals are received or not.

Specifically, the infrared sensing device is composed of two infrared sensors, namely a first infrared sensor and a second infrared sensor, the first infrared sensor sends infrared rays, the second infrared sensor receives infrared rays, the second infrared sensor is communicated with the server, when no light beam shielding occurs, the second infrared sensor receives the infrared rays sent by the first infrared sensor, signals transmitted to the server are always in a high level, when the light beam shielding occurs, the second infrared sensor does not receive the infrared rays sent by the first infrared sensor, a ground state high level signal transmitted by the second infrared sensor is changed into a low level signal, the server identifies the level signal as an abnormal signal according to the requirement setting, a corresponding boolean value, namely 'true' or 'false', is given to a preset first variable, and then a subsequent monitoring image analysis mode is entered, and the first variable is used for representing the judgment result of the infrared sensing device.

Step S200: if the server obtains the abnormal signal uploaded by the infrared sensing device through monitoring, the server enters a monitoring image analysis mode; after the server enters the monitoring image analysis mode, a monitoring video stream is obtained, and a certain mapping relation is formed between the selected interval of the monitoring video stream and the set position of the infrared sensing device; and the server extracts a frame image from the monitoring video stream, processes the frame image to obtain an analysis result generated according to the image analysis mode, and assigns a corresponding Boolean value to the preset second variable according to the analysis result.

Specifically, a result generated by the image analysis mode is monitored and represented by a preset second variable, and if the result generated by the image analysis mode is that a human body and a human body crossing event are detected, a Boolean value 'true' is given to the second variable, otherwise, a Boolean value 'false' is given.

The value of the first variable is also used for triggering the server to enter the monitoring image analysis mode, namely, the server can enter the monitoring image analysis mode only when the second infrared sensor sends a low level signal, and various service processes required by each preset monitoring image analysis are started, otherwise, the server cannot start the various service processes, so that electric energy is saved, data storage and operation resources of the server are saved, and the situations of overheating of self components and the like caused by excessive operation and the like are prevented.

In addition, a mapping function is arranged between the selection interval of the monitoring video stream and the placement position of the infrared sensing device, so that the accuracy of selecting the monitoring video stream sample can be improved, and specific functions are explained later.

Step S300: and the server performs AND gate operation on the first variable and the second variable to obtain a third variable, and if the Boolean value of the third variable is true, a preset early warning message is sent to the property management center.

Specifically, a third variable is further provided for carrying a final determination result, where an exemplary scheme of whether an enclosure crossing event occurs or not to perform a final confirmation mode is as follows:

setting a 'signal snap low level transmitted by an infrared sensor' event as a variable

Setting the result bearing variable generated by the monitoring image analysis mode in the subsequent step as a variable

The final result carrying variable is a variable，

=

×

If true, the final confirmation result is that a crossover event has occurred;

=

×

if not, the final confirmation result is that the crossing event does not occur;

when the second infrared sensor does not receive the infrared rays emitted by the first infrared sensor, the second infrared sensor sends a low level signal to the server, the server receives the low level signal sent by the second sensor, recognizes the low level signal as an abnormal signal, assigns 'true' to the first variable, triggers the server to enter a monitoring image analysis mode, then when the judgment result of the monitoring image analysis mode is 'true', the judgment result of the monitoring image analysis mode represents that a human body is recognized in a monitoring video stream and the human body crosses an alarm line, assigns 'true' to the second variable, performs AND gate operation on the first variable and the second variable, namely 'x' operation, obtains a third variable value 'true', and when the third variable value is 'true', the final judgment output result of the whole system represents that a 'community wall-turning event occurs', the server sends a preset early warning message text to the property management center along a set network path, and sends the enclosure positions where the community wall turning event occurs are identified together, the property management center receives data transmitted by the server and displays the data on a screen, and property management personnel can go to the abnormal detection enclosure points according to the displayed enclosure positions after seeing the early warning message.

If the second variable value generated in the monitoring image mode is assigned as "false", the third variable value is also "false" after the and gate operation, and the system determines that the community crossing event does not occur, which is likely to be a phenomenon that other objects such as birds and leaves block the infrared rays emitted by the first infrared sensor, so that the second infrared sensor cannot receive the infrared rays emitted by the first infrared sensor, and then the server enters a "flameout mode", which is specifically described later, and gives an opportunity of appropriate heat reduction buffering to hardware components which generate heat due to excessive calculation amount.

Wherein, the step S200 of monitoring the image analysis mode includes the following steps:

step S201: the server extracts frame images from the monitoring video stream;

specifically, the extraction of the frame image can extract video stream data by using an opencv vector machine, most of monitoring systems in the existing market are IP network monitoring systems, a server is not required to be directly connected with a camera, the server only needs to send a required time interval point to an IP network monitoring system center, and the IP network monitoring system center can call monitoring images in a corresponding time interval in a storage in the network of the IP network monitoring system center to return.

Step S202: the server takes the frame image as an original input image and inputs the frame image into a feature extraction structure formed by interweaving five convolution layers and three pooling layers to obtain a first output layer extracted from the original input image;

specifically, the feature extraction structure designed by the invention is formed by interweaving five convolution layers and three pooling layers, and is an optimal combination ratio of early warning timeliness and calculated quantity loss aiming at the comprehensive performance related to the system, if the feature extraction structure is six convolution layers and three pooling layers, the accuracy of the output identification result is only improved by 0.5%, but the calculation time length is increased by about 10% and is increased by about 15ms, if the feature extraction structure is four convolution layers and three pooling layers, the accuracy of the output identification result is reduced by 5% -8%, and if the feature extraction structure is five convolution layers and two pooling layers, the identification accuracy is reduced by about 8%.

In addition, the feature extraction structure of the invention has lower depth than the convolution layers of face recognition and object recognition, has less calculation time, and the interweaving design structure of five convolution levels and three pooling layers can reduce the calculation time required from input to output in a monitoring image analysis mode, ensure the recognition precision and reduce the misjudgment rate.

Step S203: mapping each point of the first output layer back to the original input image to obtain coordinates of each point of the first output layer, wherein a coordinate system corresponding to the coordinates takes the original input image as reference;

step S204: inputting the first output layer into a preset target selection model to obtain a region judged to contain human body features and a candidate frame corresponding to the region containing the human body features;

step S205: the server captures the subsequent frame images of the original input image, performs the same processing on the subsequent frame images, and updates the candidate frame position according to the data obtained by the processing;

step S206: and judging whether the candidate frame touches a preset warning line position or not when the position of the candidate frame is updated every time, and if so, assigning a true value to the second variable by the server.

According to the community enclosure crossing early warning method, the community enclosure crossing early warning system, the computer equipment and the storage medium, whether an abnormal signal exists is detected through the infrared sensing device, if the abnormal signal exists, the monitoring image analysis mode is entered, whether the abnormal signal can generate an alarm message needing to be alarmed is judged after analysis, and the alarm message is sent to the property management center, so that the occurrence of misjudgment caused by only adopting infrared sensing detection is reduced through a multiple verification mode. In addition, when the monitoring video stream is obtained, the time interval is selected and the position of the infrared sensing device is provided with a mapping function, so that the whole event occurrence process is better intercepted, a more complete data source is provided for the follow-up analysis result, the misjudgment probability is greatly reduced, and the accuracy of early warning of crossing of the community enclosing wall is greatly improved.

Optionally, step S300: the server performs AND gate operation on the first variable and the second variable to obtain a third variable, and if the Boolean value of the third variable is true, the server also comprises a step of sending a preset early warning message to a property management center, wherein the step further comprises

Specifically, the "flameout mode" is one of unique design points of the present invention, and if a third variable value bearing a final determination result is false, the server eliminates various program processes related to the monitored image analysis mode, only maintains a service process for communication monitoring with the infrared sensor on software, only maintains power supply of the communication module on hardware, wakes up required items such as various processes of image analysis when the infrared sensor transmits an abnormal signal, and the flameout mode greatly reduces loss of hardware components of the device, also provides buffering time for reducing frequency and heat of a CPU in a motherboard, and closes other components of the server to reduce loss of electric energy and electronic components, thereby increasing stability and durability of the entire system.

Optionally, referring to fig. 8, 9, 10, the step of determining whether the candidate frame touches the preset alert line position in step 206 includes:

step S207 judges whether any frame of the candidate frame intersects with the warning line, if so, the candidate frame is judged to touch the preset warning line, the analysis result is that a community enclosure crossing event is generated, the second variable is assigned with true, and if no frame of the candidate frame intersects with the warning line, the second variable is assigned with false.

Optionally, the step of determining whether any frame side of the candidate frame intersects with the warning line in step S207 further includes:

s207-1, the end point of the warning line is set asAnd

the warning line is in a vector representation form as a vector

；

S207-2 sets the candidate frame end point to

,

,

And

Vector of motion

Vector of motion

And vector

;

S207-3 selecting a side length vector of the candidate frame

Creating a virtual vector as a first order judgment object

Vector of motion

Vector of motion

Vector of motion

；

S207-4 calculating a vectorAnd

obtaining a first cross product by the cross product of the first step;

s207-5 calculating a vector

And

obtaining a second cross product by the cross product;

s207-6, judging whether the first cross product and the second cross product are abnormal numbers, if so, judging that a frame edge of the candidate frame intersects with the warning line, and if so, giving a true value to the second variable by touching the warning line;

and if not, performing the same calculation step on the rest side length vectors, and if all the side length vectors are not intersected with the guard line, outputting a result that the candidate frame is not intersected with the guard line, and assigning a false value to the second variable.

Compared with the method that only one infrared sensor is used for judging, the accuracy is higher, the calculation amount generated when the candidate frame is judged to be intersected with the warning line is lower, and the required calculation time is only 5-10 ms. Therefore, data operation resources of the server are greatly saved.

Optionally, the step of setting a warning line in step 207 is:

S207-A receives auxiliary positioning points input by community personnel, sequences the positioning points according to the input sequence, and generates a plurality of auxiliary lines according to the principle that two adjacent points generate auxiliary lines, wherein the auxiliary lines are used for calculating and generating the warning lines;

S207-B extracts intersections between the auxiliary linesArray, obtaining new polygon vertex according to preset external polygon algorithm

Array of the polygon vertices

S207-C sets the intersection point of two adjacent auxiliary lines asThe vector of two adjacent auxiliary lines is

And

said point of intersection

The side edge has a point，

=

+(

)。

The scheme for generating the warning line can be adapted to all communities, has high adaptation degree, only needs community related personnel or technical personnel to select the auxiliary positioning points on the enclosing wall in the image, and then forms the auxiliary line according to the selection sequence of the auxiliary positioning points by the packaged program.

Optionally, in step S204, the first output layer is input to a preset target selection model, and a region determined to include human body features and a candidate frame corresponding to the region including human body features are obtained, where the target selection model is composed of 2 full-connection layers and a loss function layer, where n is multiplied by 1.

Referring to fig. 2, fig. 2 shows a conventional full-link layer design, in fig. 2, it can be seen that the full-link layer is a structure formed by interleaving a plurality of nodes a 1-a 6 with a plurality of nodes B1-B6, a small structure formed by arranging the nodes a 1-a 6 is a first layer, and a small structure formed by arranging the nodes B1-B6 is a second layer, so the conventional full-link layer structure is a two-layer interleaving structure of n x n. Referring to fig. 3, the present invention changes a fully-connected layer into a two-layer structure of n x 1, where n corresponds to a small structure formed by a plurality of node a series, which is the first layer, and the second layer has only 1B 1, i.e. a connection node, since it is not necessary to identify whether it is a brick or a fallen leaf when identifying the wall turning over of a human body, and only needs to identify whether it is a human body, by modifying the fully-connected layer, the amount of computation generated in the fully-connected layer can be greatly reduced, because in practical applications, if the turning over of a community is not only a pragmatic person, but also is a thief with a definite target, the turning over of the thief will go to a target floor or a specific position after turning over the enclosure of the small area, therefore, the computation time duration generated in the structure of the entire image detection is absolutely not long enough, and the fully-connected layer structure designed by the present invention, the method can greatly reduce the calculated amount generated on the full connection layer, and is greatly suitable for the actual scene which is easy to occur, thereby improving the timeliness of the early warning on the speed.

Optionally, the loss function in the loss function layer is:

wherein,

is the sample distribution value of the real label,

is composed of

W is a weight coefficient,

Illustratively, referring to FIG. 11, the input to the first layer of the fully-connected layer is

And correspondingly arranged corresponding to each connection

~

The weight value w can be obtained by training according to the sample, and the value of the second layer of the full connection layer is y, y =

;

Then, the value of y is taken as

And inputting the result into the loss function to obtain a result, wherein the smaller the result is, the more accurate the design value of the w weight is.

In addition, in order to prevent the system from being crashed due to the fact that the value of the weight value is too large in the self-training process and exceeds the maximum value allowed by the data type, the loss function is added

Therefore, the phenomenon that the weighted value is too large and further breakdown occurs is prevented, and the stability of the system is improved.

Optionally, the step of arranging the camera and the infrared sensing device around the community fence includes:

arranging infrared sensing device on the ground with the community enclosing wall facing outwards

Specifically, referring to fig. 6, fig. 6 is a schematic diagram of an infrared sensing device in an embodiment, where an outward direction is a road, a community enclosure is an inward side, the infrared sensing device is disposed on the community enclosure toward the outside of the road, general infrared sensing devices are disposed on brick surfaces of enclosure walls, and the infrared sensing device of the present invention is disposed on the ground of the community enclosure toward the outside of the road, and is specially used to cooperate with a subsequent image monitoring mode, once the infrared sensing device senses a sheltered object, a monitoring image analysis mode is started, and since the monitoring image analysis mode is optimized, a certain calculation time is required, therefore, a position boundary point of infrared detection of the present invention is prior to a design of "disposing the infrared sensing device on the enclosure wall top", and if the infrared sensing device is disposed on the enclosure wall top, after a suspicious molecule climbs to the enclosure wall top, the last time for crossing the fence is very short, and at the moment, the infrared sensing device uploads the abnormal signal to the server, the server carries out monitoring image analysis, and the suspicious molecule is deep into the community, so that the arrangement mode of the infrared sensing device of the embodiment of the invention can advance the monitoring time boundary of the whole system, and the timeliness of early warning is greatly improved.

The invention can be equal to or even higher than the prior infrared sensing arrangement in the timeliness of early warning, and is lossless in timeliness, and the judgment accuracy is greatly improved.

In this embodiment, the time limit for capturing the video stream is changed to the video stream data after the time stamp for uploading the abnormal signal by the infrared sensor device.

Optionally, the step of arranging camera and infrared sensing device around the community enclosure further includes:

Specifically, please refer to fig. 7, fig. 7 is a schematic diagram of an infrared sensing device in another embodiment, which is a road outward, a community enclosure is an inner side, the description mainly wants to show the relative position of the infrared sensing device, the infrared sensing device is set on the ground at a certain distance from the enclosure at the outer side of the road, and is also set on the ground at a certain distance from the inner side of the community enclosure, so that the intruder can monitor when entering the community, and can perform positioning analysis when escaping from the community, so that the manager can know the escaping direction of the intruder and turn over the wall to detect the triggering place at the first time.

In addition, in the embodiment, the time limit for intercepting the video stream is changed into the video stream data after the time stamp for uploading the abnormal signal by the infrared sensing device,

optionally, the function in the step of associating the selected interval of the monitoring video stream with the placement position of the infrared sensing device in step S200 includes:

△t=

+Q

wherein △ t is the selected interval of the monitoring video stream,and v is the distance between the infrared sensing device and the enclosing wall, the preset human body moving speed and Q is a time delay factor.

The server sends a data pulling request to the digital network monitoring system, and the monitoring system feeds back monitoring data of the related camera to the server. In the process, the embodiment calculates the time interval of the monitoring image selection by using the function, and pulls the monitoring image, wherein the pulled monitoring image is closer to the whole process of the event.

the infrared module 100 is configured to monitor an abnormal signal output by the infrared sensing device around a preset fence, where the abnormal signal is triggered by shielding a light beam emitted by the infrared sensing device and is a low-level signal, and the server assigns a corresponding boolean value to a preset first variable according to whether the abnormal signal is received;

the image analysis module 200 is used for entering a monitoring image analysis mode if the server monitors and obtains the abnormal signals uploaded by the infrared sensing device; after the server enters the image analysis mode, acquiring a monitoring video stream, wherein the selection interval of the monitoring video stream is related to the placement position of the infrared sensing device; the server extracts a frame image from the monitoring video stream to process the frame image and obtain an analysis result generated according to the image analysis mode, and a corresponding Boolean value is given to the preset second variable according to the analysis result;

the output module 300 is configured to perform an and gate operation on the first variable and the second variable by the server to obtain a third variable, and send a preset early warning message to the property management center if a boolean value of the third variable is true;

wherein the image analysis module 200 is further configured to:

the server extracts frame images from the monitoring video stream;

Optionally, the output module 300 further includes:

and a flameout unit 301, configured to, if the boolean value of the third variable is false, enter a flameout mode by the server, and refresh the numerical values of the first variable, the second variable, and the third variable, in the flameout mode, the server only keeps a communication module powered on, and eliminates a service process related to the image analysis mode, and the communication module is configured to monitor an abnormal signal uploaded by the infrared sensing device.

Optionally, the image analysis module 200 further includes:

an intersection determining unit 201, configured to determine whether any frame of the candidate frame intersects with the warning line, if so, determine that the candidate frame has touched the preset warning line, and if an analysis result is that a community fence crossing event has occurred, assign a true value to the second variable, and if no frame of the candidate frame intersects with the warning line, assign a false value to the second variable.

Optionally, the intersection determination unit 201 is further configured to:

setting the end point of the warning line asAnd

the warning line is in a vector representation form as a vector；

Let the candidate frame end point be,

,

And

Vector of motion

Vector of motion

And vector

;

Selecting a side long vector of the candidate frame

Creating a virtual vector as a first order judgment object

Vector of motion

Vector of motion

Vector of motion

；

Computing vectors

And

obtaining a first cross product by the cross product of the first step;

computing vectors

And

obtaining a second cross product by the cross product;

Optionally, the image analysis module 200 further includes:

a warning line unit 202, configured to receive auxiliary positioning points input by community staff, sort the positioning points according to the input order, and generate a plurality of auxiliary lines according to a principle that two adjacent points generate auxiliary lines, where the auxiliary lines are used for calculation generation of the warning line;

extracting intersections between the auxiliary lines

let the intersection point of two adjacent auxiliary lines be

The vector of two adjacent auxiliary lines is

And

said point of intersection

The side edge has a point，

=+()。

Optionally, the image analysis module 200 is further configured to input the first output layer to a preset target selection model, and in the step of obtaining a region determined to include human body features and a candidate frame corresponding to the region including human body features, the target selection model is composed of 2 full-connected layers obtained by multiplying n by 1 and a loss function layer.

Optionally, the loss function in the loss function layer of the image analysis module 200 is:

wherein,

is the sample distribution value of the real label,

is composed of

W is a weight coefficient,

Optionally, the infrared module 100 is further configured to:

Optionally, the infrared module 100 further sets a function of the selected interval related to the placement position of the infrared sensing device as:

△t=

+Q

wherein △ t is the selected interval of the monitoring video stream,

Specifically, when the infrared sensing device is placed on the outer side of the enclosing wall, the suspicious molecules touch the infrared rays, the infrared sensing device uploads abnormal signals, but the suspicious molecules are likely to observe or wander nearby the enclosing wall first and do not turn over the wall, the monitoring image acquisition and analysis are invalid at the moment, the setting of the delay factor Q better avoids the invalid analysis, the starting point of the monitoring image selection interval at the moment is a time point which takes the abnormal signals uploaded by the infrared sensing device as a boundary point and is spaced backward by △ t, and the later ending point can be set for 15 minutes or 30 minutes or set according to an actual scene.

When the infrared sensing devices are placed on the outer side and the inner side of the enclosing wall, the image intervals selected by monitoring triggered by the infrared sensing devices on the outer side are set according to the same content, Q is a positive value, and the image intervals selected by monitoring triggered by the infrared sensing devices on the inner side are negative values.

Processor 22 may be a Central Processing Unit (CPU), controller, microcontroller, microprocessor or other data processing chip in some embodiments. The processor 22 is typically used to control the overall operation of the computer device 2. In this embodiment, the processor 22 is configured to execute the program code stored in the memory 21 or process data, for example, execute the community fence crossing warning system 20, so as to implement the community fence crossing warning method.

The network interface 23 may comprise a wireless network interface or a limited network interface, and the network interface 23 is typically used for establishing a communication connection between the computer device 2 and other electronic apparatuses. For example, the network interface 23 is used to connect the computer device 2 with an external terminal necklace, establish a data transmission channel and a communication connection between the computer device 2 and an external interrupt, and the like via a network. The network may be a wireless or wired network such as an Intranet (Intranet), the Internet (Internet), a global system for mobile communications (GSM), Wideband Code Division Multiple Access (WCDMA), a 4G network, a 5G network, Bluetooth (Bluetooth), or Wi-Fi.

In this embodiment, the community fence crossing warning system 20 stored in the memory 21 may be further divided into one or more program modules, and the one or more program modules are stored in the memory 21 and executed by one or more processors (in this embodiment, the processor 22) to complete the present invention.

In addition, the present embodiment also provides a computer storage medium, such as a flash memory, a hard disk, a multimedia card, a card-type memory (e.g., SD or DX memory, etc.), a Random Access Memory (RAM), a Static Random Access Memory (SRAM), a read-only memory (ROM), an electrically erasable programmable read-only memory (EEPROM), a programmable read-only memory (PROM), a magnetic memory, a magnetic disk, an optical disk, a server, an App application store, etc., on which a computer program is stored, which when executed by a processor implements a corresponding function. The computer storage medium of the embodiment is used for the community fence crossing warning system 20, and when being executed by a processor, the computer storage medium implements the community fence crossing warning method of the invention.

Through the above description of the embodiments, those skilled in the art will clearly understand that the method of the above embodiments can be implemented by software plus a necessary general hardware platform, and certainly can also be implemented by hardware, but in many cases, the former is a better implementation manner.

The above description is only a preferred embodiment of the present invention, and not intended to limit the scope of the present invention, and all modifications of equivalent structures and equivalent processes, which are made by using the contents of the present specification and the accompanying drawings, or directly or indirectly applied to other related technical fields, are included in the scope of the present invention.

Claims

1. The community enclosure crossing early warning method is characterized in that a camera and an infrared sensing device are arranged around a community enclosure, and the community enclosure crossing early warning method comprises the following steps:

if the server obtains the abnormal signal uploaded by the infrared sensing device through monitoring, the server enters a monitoring image analysis mode; after the server enters the monitoring image analysis mode, a monitoring video stream is obtained, and a certain mapping relation is formed between the selected interval of the monitoring video stream and the set position of the infrared sensing device;

the server extracts frame images from the monitoring video stream;

judging whether the candidate frame touches a preset warning line position or not when the position of the candidate frame is updated every time, and if so, assigning a value to a second variable to be true by the server;

and the server performs AND gate operation on the first variable and the second variable to obtain a third variable, and if the Boolean value of the third variable is true, a preset early warning message is sent to the property management center.

2. The community fence crossing warning method of claim 1, wherein the server performs an and gate operation on the first variable and the second variable to obtain the third variable, and after the step of sending a preset warning message to the property management center if the boolean value of the third variable is true, the method further comprises:

and if the Boolean value of the third variable is false, the server enters a flameout mode, numerical values of the first variable, the second variable and the third variable are refreshed, the server only keeps power supply of an internal communication module and stops a service process related to the image analysis mode in the flameout mode, and the communication module is used for monitoring abnormal signals uploaded by the infrared sensing device.

3. The community fence crossing warning method of claim 1, wherein the step of determining whether the candidate box touches a preset fence position comprises:

4. The community fence crossing warning method of claim 3, wherein the step of determining whether any frame of the candidate box intersects the warning line comprises:

setting the end point of the warning line as

And

the warning line is in a vector representation form as a vector

；

Let the candidate frame end point be

,

,And

each side of the candidate frame is respectively a vector in a vector representation modeVector of motionVector of motion

And vector

;

Selecting a side long vector of the candidate frame

Creating a virtual vector as a first order judgment object

Vector of motion

Vector of motionVector of motion

；

Computing vectors

And

obtaining a first cross product by the cross product of the first step;

computing vectors

And

obtaining a second cross product by the cross product;

5. The community fence crossing warning method according to claim 3, wherein the warning line is set by the steps of:

extracting intersections between the auxiliary lines

Array, obtaining new polygon vertex according to preset external polygon algorithmArray of the polygon vertices

An array for representing the warning lineThe external polygon algorithm includes:

let the intersection point of two adjacent auxiliary lines beThe vector of two adjacent auxiliary lines is

And

said point of intersection

The side edge has a point

，=

+(

)。

6. The community fence crossing warning method according to claim 1, wherein in the step of inputting the first output layer to a preset target selection model to obtain a region including human body features and a candidate frame corresponding to the region including human body features, the target selection model is composed of 2 full-connection layers and a loss function layer, wherein n is multiplied by 1.

7. The community fence crossing warning method of claim 6, wherein the loss function in the loss function layer is:

wherein,

is the sample distribution value of the real label,

is composed of

W is a weight coefficient,

8. The community fence crossing early warning method according to claim 1, wherein the step of arranging the cameras and the infrared sensing devices around the community fence comprises:

and an infrared sensing device is arranged on the ground outside the community enclosing wall.

9. The community fence crossing early warning method according to claim 1, wherein the step of arranging the cameras and the infrared sensing devices around the community fence further comprises:

and infrared sensing devices are arranged on the ground outside and inside the community enclosing wall.

10. The community fence crossing warning method according to claim 1, wherein a mapping relationship between a selected interval of the monitoring video stream and the set position of the infrared sensing device satisfies the following function:

△t=

+Q

wherein △ t is the selected interval of the monitoring video stream,

11. The utility model provides a community's enclosure early warning system that climbs which characterized in that includes:

the image analysis module is used for entering a monitoring image analysis mode if the server monitors and obtains the abnormal signals uploaded by the infrared sensing device; after the server enters the image analysis mode, acquiring a monitoring video stream, wherein the selection interval of the monitoring video stream is related to the placement position of the infrared sensing device; the server extracts frame images from the monitoring video stream;

when the position of the candidate frame is updated every time, judging whether the candidate frame touches a preset warning line position, if so, assigning a second variable to be true by the server;

and the output module is used for carrying out AND gate operation on the first variable and the second variable by the server to obtain a third variable, and if the Boolean value of the third variable is true, sending a preset early warning message to the property management center.

12. The community fence crossing warning system of claim 11, wherein the output module further comprises:

13. The community fence crossing warning system of claim 11, wherein the image analysis module further comprises:

14. The community fence crossing warning system of claim 13, wherein the intersection determination unit is further configured to:

setting the end point of the warning line as

And

the warning line is in a vector representation form as a vector；

Let the candidate frame end point be

,

,

Andeach side of the candidate frame is respectively a vector in a vector representation modeVector of motion

Vector of motion

And vector;

Selecting a side long vector of the candidate frame

Creating a virtual vector as a first order judgment object

Vector of motion

Vector of motionVector of motion

；

Computing vectors

And

obtaining a first cross product by the cross product of the first step;

computing vectors

And

obtaining a second cross product by the cross product;

15. The community fence crossing warning system of claim 13, wherein the image analysis module further comprises:

extracting intersections between the auxiliary lines

let the intersection point of two adjacent auxiliary lines be

The vector of two adjacent auxiliary lines is

And

said point of intersection

The side edge has a point

，

=

+(

)。

16. The community fence crossing warning system of claim 11, wherein the image analysis module is further configured to input the first output layer to a preset target selection model, and in the step of obtaining the region including the human body feature and the candidate frame corresponding to the region including the human body feature, the target selection model is composed of 2 fully-connected layers where n is multiplied by 1 and a loss function layer.

17. The community fence crossing warning system of claim 11, wherein the loss function in the loss function layer of the image analysis module is:

wherein,is the sample distribution value of the real label,is composed of

W is a weight coefficient,

18. The community fence crossing warning system of claim 11, wherein the infrared module is further configured to:

19. The community fence crossing warning system of claim 11, wherein the infrared module is further configured to:

20. The community fence crossing early warning system according to claim 11, wherein the infrared module further sets a mapping relationship between a selection interval and the setting position of the infrared sensing device to satisfy the following function:

△t=

+Q

21. A server comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, wherein the computer program when executed by the processor implements the community fence crossing warning method of any of claims 1 to 10.

22. A computer-readable storage medium, having a computer program stored therein, the computer program being executable by at least one processor to cause the at least one processor to perform the community fence crossing warning method of any one of claims 1 to 10.