CN114359024B - Method for shunting and scheduling picture to edge, edge picture recognition box and system - Google Patents
Method for shunting and scheduling picture to edge, edge picture recognition box and system Download PDFInfo
- Publication number
- CN114359024B CN114359024B CN202210019902.4A CN202210019902A CN114359024B CN 114359024 B CN114359024 B CN 114359024B CN 202210019902 A CN202210019902 A CN 202210019902A CN 114359024 B CN114359024 B CN 114359024B
- Authority
- CN
- China
- Prior art keywords
- edge
- box
- picture
- edge image
- cooperation
- 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.)
- Active
Links
Images
Landscapes
- Image Analysis (AREA)
- Image Processing (AREA)
Abstract
The invention discloses a method, electronic equipment and a medium for shunting and scheduling pictures to edges, wherein the method comprises the following steps: s1, receiving pictures to be identified sent by an X-ray machine of a security check point in real time, storing the received pictures to be identified into task queues TQ and S2, and judging the number X of the pictures to be identified in the task queues TQ 0 And a predetermined threshold u 0 If x is the size of 0 ＞u 0 Turning to the step S3; s3, sending a cooperation request to other edge image recognition boxes; s4, receiving a returned cooperation response; s5, if the cooperation response carries information for receiving the cooperation request, state information and channel information are obtained; and S6, obtaining the minimum scheduling overhead time based on the state information and the channel information, and scheduling the picture to be identified with the preset adjustment amount alpha to a cooperative edge image identifying box corresponding to the minimum scheduling overhead time. The invention solves the problem of improving the efficiency of identifying the picture amount in the peak period on the premise of not increasing the cost by a method of shunting and scheduling the picture to the edge.
Description
Technical Field
The invention relates to the field of security inspection intelligent image judgment, in particular to a method for shunting and dispatching images to edges, an edge image recognition box and a system.
Background
In a subway scene, each security check point corresponds to a security check machine, one security check machine is connected with an edge recognition box, the security check machine is used for imaging passenger packages by X-ray machines, the edge recognition box is responsible for intelligently recognizing package pictures transmitted by the security check machine connected with the edge recognition box, and contraband articles in the package pictures are marked.
In the processing process of the edge map recognizing box, the intelligent map recognizing process needs a great amount of calculation power, so that a relatively high time delay is generated, especially in the peak period of passenger flow and under the condition of excessive passenger capacity, the edge map recognizing box possibly reaches the bottleneck of the capability of processing packages, and in the process, how to improve the efficiency of recognizing the picture amount in the peak period on the premise of not increasing the cost as much as possible becomes a difficult point and a subject.
In a subway security inspection scene, the passenger flow conditions of each security inspection point at the same time are different, so that the image recognition tasks required to be completed by the edge image recognition boxes corresponding to each security inspection point are also different greatly. Therefore, how to distribute the picture detection task of the security check point with high passenger flow to the security check station with low passenger flow becomes a feasible option, namely the edge map identifying box carries out map identifying task scheduling to other edge map identifying boxes.
Disclosure of Invention
The invention aims to provide a method for shunting and scheduling pictures to edges, an edge picture recognition box and a system, which are used for solving the problem of improving the efficiency of recognizing the picture amount in the peak period on the premise of not increasing the cost in the prior art by detecting the picture recognition capability of the current edge picture recognition box and the number of the pictures to be recognized in real time and initiating an assistance request to other edge picture recognition boxes when the picture recognition capability cannot meet the number of the pictures to be recognized.
A method for shunting and dispatching pictures to edges is applied to edge computing nodes formed by a plurality of edge map-identifying boxes, wherein each edge map-identifying box is connected with each other, each edge map-identifying box is connected with an X-ray machine of a security check point where the edge map-identifying box is located, and the following steps are executed:
s1, receiving a picture to be identified sent by an X-ray machine of a security check point in real time, storing the received picture to be identified into a task queue TQ,
s2, judging the number x of the pictures to be identified in the task queue TQ 0 And a predetermined threshold u 0 The size of u 0 For recognizing the current edgeThe maximum value of the picture to be identified is processed by the box every second, if x 0 ＞u 0 Turning to step S3;
s3, sending a cooperation request to other edge image recognizing boxes, wherein the cooperation request carries a preset quantity alpha, so that the other edge image recognizing boxes judge whether to accept the cooperation request according to the preset quantity alpha and the quantity of the images to be recognized of the other edge image recognizing boxes;
s4, receiving the cooperation response returned by the other edge image recognizing boxes;
s5, if the cooperation response carries information for receiving the cooperation request, acquiring state information of a cooperation edge image recognition box, and channel information of the cooperation edge image recognition box and a current edge image recognition box, wherein the cooperation edge image recognition box is an edge image recognition box corresponding to the cooperation response carrying the information for receiving the cooperation request;
and S6, obtaining the minimum scheduling overhead time based on the state information and the channel information, and scheduling the picture to be identified with the preset adjustment amount alpha to a cooperative edge image identifying box corresponding to the minimum scheduling overhead time.
Further, the state information includes: coordinated edge recognition box i processing delay y for each picture i I =0, 1.. And n, n is the total number of all cooperating edge map boxes.
Further, the channel information includes: the current edge image recognition box dispatches the pre-regulated amount of the image to be recognized to the transmission time delay t of the cooperative edge image recognition box i 0i And the transmission time delay t of the cooperative edge image recognition box i for returning the image recognition result according to the preset quantity of the image to be recognized i0 I =0, 1.. And n, n is the total number of all cooperating edge map boxes.
Further, the step S6 specifically includes:
s001, based on the state information and the channel information, calculating the scheduling overhead time Y of the front edge image recognition box for scheduling the pre-adjusted quantity of the images to be recognized to the cooperative edge image recognition box i i I =0, 1.. N, n is the total number of edge map boxes for all collaborations, and the scheduling overhead time Y is i ＝α×y i +t 0i +t i0 ；
S002, scheduling the picture to be identified with the preset adjustment amount alpha to the minimum scheduling overhead time Y min And (5) correspondingly recognizing the image box at the cooperative edge, and turning to the step S2.
Further, the
X is said i Number of pictures to be recognized, u, processed per second for a collaborative edge recognition box i i Processing the maximum value of the picture to be recognized, y, per second for the cooperating edge recognition box i i The delay of processing each picture for the cooperating edge aware box i.
Further, the
And generating a cooperative edge image recognition box i by counting historical image processing data to be recognized according to least square polynomial curve fitting.
Further, the pre-adjustment amount α satisfies
X is said l Is a minimum scheduling overhead time Y min Number of pictures to be recognized, u, processed per second of the corresponding cooperating edge recognition box 1 l Is a minimum scheduling overhead time Y min The corresponding cooperating edge map box 1 processes the maximum value of the picture to be recognized every second.
Further, the
Listen to the above i (b North) number of bytes, v, of pictures to be identified of a pre-adjusted quantity α scheduled to a cooperating edge-aware map box i i (bit/s) is the uplink rate of the channel,
the above-mentioned i (b North)' is a cooperative edge map box i rootThe image recognition result, v, returned according to the byte number of the picture to be recognized of the preset quantity alpha i (bit/s)' is the downlink rate of the channel.
An edge recognition carton comprising:
one or more processors;
a storage unit, configured to store one or more programs, which when executed by the one or more processors, enable the one or more processors to implement the method for picture splitting scheduling to an edge, where the one or more programs include:
the picture receiving module is used for receiving the picture to be identified sent by the X-ray machine of the security check point in real time and storing the received picture to be identified into the task queue TQ,
a size judging module for judging the number x of the pictures to be identified in the task queue TQ 0 And a predetermined threshold u 0 The size of u 0 Processing the maximum value of the picture to be identified per second for the current edge image identifying box, if x 0 ＞u 0 If yes, executing the cooperation request module;
the cooperation request module is used for sending a cooperation request to other edge image recognizing boxes, wherein the cooperation request carries a preset quantity alpha, so that the other edge image recognizing boxes judge whether to accept the cooperation request according to the preset quantity alpha and the quantity of the images to be recognized of the other edge image recognizing boxes;
the receiving response module is used for receiving the cooperation response returned by the other edge map recognizing box;
an information obtaining module, configured to obtain, if the cooperation response carries information for receiving the cooperation request, state information of a cooperative edge image recognizing box, and channel information of the cooperative edge image recognizing box and a current edge image recognizing box, where the cooperative edge image recognizing box is an edge image recognizing box corresponding to the cooperation response carrying the information for receiving the cooperation request;
and the scheduling module is used for obtaining the minimum scheduling overhead time based on the state information and the channel information and scheduling the picture to be identified with the preset quantity alpha to the cooperative edge image identifying box corresponding to the minimum scheduling overhead time.
A system for picture forking scheduling to an edge, comprising:
an edge computing node formed by a plurality of edge image recognizing boxes and an X-ray machine of a plurality of security check points; wherein, each edge map-identifying box is connected with each other, and each edge map-identifying box is connected with the X-ray machine of the security check point where the edge map-identifying box is located;
each edge map box is used for: receiving the picture to be identified sent by the X-ray machine of the security check point in real time, storing the received picture to be identified into a task queue TQ,
judging the number x of the pictures to be identified in the task queue TQ 0 And a predetermined threshold u 0 Of the size u 0 Processing the maximum value of the picture to be recognized per second for the current edge recognition box, if x 0 ＞u 0 Sending a cooperation request to other edge image recognizing boxes, wherein the cooperation request carries a preset quantity alpha, so that the other edge image recognizing boxes judge whether to accept the cooperation request according to the preset quantity alpha of the image to be recognized and the quantity of the image to be recognized;
receiving a cooperation response returned by the other edge image recognizing box;
if the cooperation response carries information for receiving the cooperation request, acquiring state information of a cooperation edge image recognition box, channel information of the cooperation edge image recognition box and a current edge image recognition box, wherein the cooperation edge image recognition box is an edge image recognition box corresponding to the cooperation response carrying the information for receiving the cooperation request;
obtaining the minimum scheduling overhead time based on the state information and the channel information, and scheduling the picture to be identified with the preset quantity alpha to a cooperative edge image identifying box corresponding to the minimum scheduling overhead time;
the edge recognition box is further used for: receiving a cooperation request sent by other edge image recognition boxes, wherein the cooperation request carries a preset quantity alpha, and judging whether to accept the cooperation request or not according to the preset quantity alpha of the image to be recognized and the quantity of the image to be recognized: if the cooperation request is received, returning a cooperation response carrying information for receiving the cooperation request to a corresponding edge map recognition box so that the corresponding edge map recognition box acquires state information of the edge map recognition box, channel information of the edge map recognition box and the current edge map recognition box, wherein the edge map recognition box is the edge map recognition box corresponding to the cooperation response carrying the information for receiving the cooperation request, and based on the state information and the channel information, obtaining the minimum scheduling overhead time, scheduling the picture to be identified with the preset quantity alpha to the edge map recognition box corresponding to the minimum scheduling overhead time;
the X-ray machine of the security check point is used for scanning the passenger's packages, obtaining X-ray scanning pictures and sending the X-ray scanning pictures to the edge image identifying box of the security check point.
According to the invention, through test data analysis, curve fitting is carried out according to a least square method polynomial, mathematical modeling is carried out on the performance of the edge map box to obtain an edge map box performance curve, and minimum scheduling overhead time Y is obtained according to the performance curve and corresponding uplink and downlink rates of a channel during each scheduling min Scheduling the picture to be identified with the preset adjustment quantity alpha to the minimum scheduling overhead time Y min Corresponding assisted edge recognition boxes.
The invention has the following beneficial effects:
1. by detecting the image recognition capability of the current edge image recognition box and the number of the images to be recognized in real time, if the image recognition capability cannot meet the number of the images to be recognized, an assistance request is sent to other edge image recognition boxes, so that part of the images to be recognized are dispatched to the assisted edge image recognition boxes, the pressure of security check point image recognition in the peak period of passenger flow is reduced, and a network can provide more efficient image recognition service;
2. in each scheduling, the time delay of processing each picture by the edge image recognition box is obtained by counting historical picture processing data to be recognized according to least square polynomial curve fitting, and the minimum scheduling overhead time Y is obtained according to the uplink and downlink rates of a channel min Will lead toThe picture to be identified with the scheduling amount alpha is always scheduled to the minimum scheduling overhead time Y min Corresponding assisted edge recognition boxes.
Drawings
FIG. 1 is a schematic diagram of a method for scheduling image shunting to an edge according to the present invention;
FIG. 2 is a schematic diagram of a system for scheduling image offloading to an edge according to the present invention;
FIG. 3 is a schematic diagram of an edge map box performance curve of the present invention;
FIG. 4 is a schematic structural diagram of an edge map recognition system according to the present invention;
FIG. 5 is a schematic diagram illustrating a process of dispatching picture streams to edges according to the present invention;
Detailed Description
The present invention will be described in further detail with reference to examples and drawings, but the present invention is not limited to these examples.
In the description of the present invention, it should be noted that the terms "center", "upper", "lower", "left", "right", "vertical", "longitudinal", "lateral", "horizontal", "inner", "outer", "front", "rear", "top", "bottom", and the like indicate orientations or positional relationships that are based on the orientations or positional relationships shown in the drawings, or that are conventionally placed when the product of the present invention is used, and are used only for convenience in describing and simplifying the description, but do not indicate or imply that the device or element referred to must have a particular orientation, be constructed in a particular orientation, and be operated, and thus should not be construed as limiting the invention.
In the description of the present invention, it should also be noted that, unless otherwise explicitly specified or limited, the terms "disposed," "open," "mounted," "connected," and "connected" are to be construed broadly, e.g., as meaning either a fixed connection, a removable connection, or an integral connection; can be mechanically or electrically connected; they may be connected directly or indirectly through intervening media, or they may be interconnected between two elements. The specific meanings of the above terms in the present invention can be understood in specific cases to those skilled in the art.
Example 1
The implementation aims to provide a method for shunting and scheduling pictures to edges, which is a feasible option for distributing a package processing task of a security check point at a peak time of passenger flow to a security check station with low passenger flow, namely an edge map recognition box carries out map recognition task scheduling to other edge map recognition boxes.
Firstly, establishing a map recognizing box mathematical model:
specifically, in the details of the edge image recognition box, for a specific security check point, under the condition that the performance of the edge image recognition box connected with the security check point is processed, the performance test for recognizing contraband in the image shot by the X-ray machine is carried out, a series of test values are obtained, a curve as shown in fig. 3 is obtained, the horizontal axis is the image quantity sent to the image recognition box every second, the unit is one, the vertical axis is the time delay for processing each image, and the unit is second:
under the condition that the quantity of pictures per second does not reach the sending quantity of 10 pictures, the curves are relatively gentle, and under the condition that the sending quantity exceeds 10 pictures per second, the processing delay of each picture is increased suddenly, and the suddenly increased abrupt change point can be regarded as the upper limit of the capacity of a certain edge map box, namely the rated performance, and the rated performance of each edge map box of different models is different.
According to the analysis of test data, curve fitting is carried out according to a least square method polynomial, and the following mathematical modeling is carried out on the performance of the edge recognition graph box:
g x-u ＝y
wherein X is the number of pictures sent by the edge image recognition box X-ray machine per second;
y is the time delay of processing each picture by the edge image recognition box;
u is the maximum value (expressed as a catastrophe point which can not cause serious delay in a curve) of the pictures processed by the edge map identifying box per second, and is the capacity range of the edge map identifying box, namely the rated parameter of the edge map identifying box of a certain model;
g is an inherent parameter, is related to the parallel operation performance of the edge map-identifying box, and is also a rated parameter of the edge map-identifying box of a specific model;
in one specific example, the u-value (maximum number of pictures processed per second within the capability of the edge map box) for the edge map box is 10.
In a specific example, the u value is configured depending on the hardware and software of the edge map box, and the u value of the edge map box of the same model is a fixed value.
1. Scheduling principle of scheduling between two groups:
two edge image recognition boxes A and B, wherein A and B belong to the same type of edge image recognition box, u values are the same, and x is monitored if the current time point is monitored according to real-time measured data a If the number of pictures to be identified is more than u, the quantity of the pictures to be identified which are processed by the A edge image box exceeds the rated performance.
The A edge image recognition box sends an assistance request to the B edge image recognition box, if the current time point x b If u is less than u, the B edge image recognizing box sends a cooperation response carrying the information for receiving the cooperation request to the A edge image recognizing box.
At this moment, the edge a image recognition box is a assisted box, the edge B image recognition box is an assisting box, the number of the images to be recognized, which should be received and scheduled by the assisting box, between the assisted box and the assisting box should be within the capability range of the assisting box, that is, after the images to be recognized of the edge a image recognition box are received and scheduled, the number of the images additionally processed by the edge B image recognition box cannot be greater than u, that is, the number of the images processed by the edge B image recognition box cannot exceed the rated performance of the edge B image recognition box.
The number of pictures scheduled may slowly increase until x a Reduced to u, or x, equal to A boxes b Increasing to u equal to the B box. Said x a Graphic tasks to be completed for the A-edge graphic box, x b And (4) a graph recognition task required to be completed by the graph recognition box for the edge B.
The above is an ideal situation without considering network delay, and in practical application, the picture is sent from the A edge image recognition box to the B edge image recognition box for image recognition processing, and the image is recognized from the B edgeThe graph box returns the processed picture to the edge A, the graph box has certain network transmission delay, and the network delay from A to B is t ab The network delay from B to A is t ba For the edge recognition box a, the processing time of each picture to be recognized scheduled to the edge recognition box B is as follows:
y b +t ab +t ba
wherein: y is b ＝g x-u
The number x of pictures processed by the B edge image identifying box per second actually comprises the pictures x to be identified of the security check point where the B edge image identifying box is located b And the picture x to be recognized is dispatched to the picture box to be recognized from the edge recognition box A to the edge recognition box B every second ab ：
x＝x b +x ab
Combining the above equations, we obtain an equation as follows:
that is, for the edge-a box, the average time length of each picture processing of the edge-B box depends on the average number of pictures per second processed by the edge-B box, and the network time overhead of the pictures.
The required picture processing delay of the security check point where the A edge image recognition box is located is not larger than the rated performance of the model of the A edge image recognition box, so that under the condition that the A box and the B box belong to the same model:
the method is used for ensuring that the time scheduled for processing to other edge map boxes is better than the time for processing the edge map boxes.
The formula expression that the picture to be identified processed by the A edge image identifying box can not be larger than the rated maximum picture is as follows:
x a -x ab -u＜0
wherein g and u are nominal parameters of the edge map box, are known, and x a 、x b The X-ray machines A and B of the security check point A and the security check point B respectively send the number of pictures, t, to the X-ray machines A and B every second ab +t ba Is the network transmission delay.
When x appears ab In the case of no solution, the number of pictures to be recognized per second exceeds the rated performance of the box A and the box B, and the problem needs to be solved by adding boxes in practical use.
And (3) scheduling recovery between two groups:
in the above example, when the image box at the edge a is in the peak of image processing, the image box at the edge a dispatches part of the image to be identified to the image box at the edge B, and if the peak condition disappears after the image box at the edge a processes the image normally, we should reduce dispatching to the image box at the edge B.
In the process, the real-time measurement and calculation x are also needed a -x ab And the size of u, if x a -x ab If u, the A edge map recognition box is always in a saturated operation state, pictures exceeding u are dispatched to B, and if x is a -x ab If u, the A edge image recognition box is always in an undersaturation operation state for recovery, and the number of recovery images can be slowly increased, namely, the recovery images are dispatched to the x of the B edge image recognition box ab The picture to be recognized is slowly reduced until x a -x ab = u, or x ab Equal to 0.
2. Cluster scheduling:
when the edge map-identifying boxes of the whole subway network work, all the edge map-identifying boxes form a virtual cloud cluster for intelligent map identification.
In the cluster, assuming that there are n +1 edge map boxes, and they all belong to the same model, that is, the performance parameters g and u are the same, and the numbers are 0,1, 2, 3, 4, 5.
A box which consumes less than its rated performance, i.e. x < u, is being used;
a box is being used that consumes performance equal to its nominal performance, i.e. x = u;
a box that consumes more than its rated performance is being used, i.e., x > u;
assume that the edge box No. 0 is a box that is being used and consumes more than its nominal capacity, i.e., x 0 -u = p > 0, p being the number of pictures per second that need to be scheduled to other edge map boxes.
All the boxes which are in use and consume less than the rated performance are obtained by sending assistance requests to other edge boxes, and the boxes which are in use and consume less than the rated performance are assumed from No. 1 edge box to No. m edge box, namely x 1 -u＜0，x 2 -u＜0，x 3 -u＜0......x m -u＜0。
And simultaneously, obtaining uplink network delay and downlink network delay between the m edge boxes and the No. 0 edge box: t is t 01 、t 10 ，t 02 、t 20 ，t 03 、t 30 ，...t 0m 、t m0 。
For the edge box 0, selecting a picture to be identified for scheduling, and calculating the network scheduling time overhead from the edge box 0 to each edge box i as follows:
…
and selecting a minimum scheduling overhead time from the scheduling time overhead, scheduling one picture to be identified to the corresponding edge image recognition box, setting the minimum scheduling overhead time as the k number edge image recognition box, scheduling one picture to the k number edge image recognition box by the 0 number edge image recognition box every second, and processing one more picture in the k number box every second.
After the primary scheduling is finished, sending an assistance request to other edge boxes again to obtain all the boxes which are using and consuming the performance less than the rated performance, if the edge box from No. 1 to No. m is the box which is using and consuming the performance less than the rated performance, that is, x 1 -u＜0，x 2 -u＜0，x 3 -u＜0......x m -u＜0。
The time spent in the process of refreshing the edge map box No. 1 to the edge map box No. m relative to the scheduling cost of the edge map box No. 0 is changed into the following steps:
…
…
selecting a minimum scheduling overhead time from the scheduling time overhead, and selecting a picture to be identified
And dispatching to the corresponding edge image recognition box, and repeating the steps in a circulating way until the No. 0 box finishes dispatching all the p images to be recognized out of the rated performance of the No. 0 box, or no box which consumes less than the rated performance of the No. 0 box is used.
In this process, when some edge map boxes change from being used with a consumed performance less than their nominal performance to being used with an edge map box with a consumed performance equal to their nominal performance, these edge map boxes need to be deleted from the set of m edge map boxes.
Similarly, other edge map boxes, except edge map box 0, which are using the consumed performance greater than the rated performance, are scheduled as above until the whole cluster reaches a dynamic balance.
In some specific real-time modes, the number of schedules required for each cycle, that is, the preset amount α is adjustable, and may be one or more, may be a sequence of sequentially increasing numbers, or may be a value adaptively selected according to the actual situation of the security inspection point.
In the process of scheduling and recovering, the above manner is also used, the difference is that in each cycle, the edge image recognition box corresponding to the maximum scheduling overhead time is selected for scheduling, and the number of the scheduled images in each cycle is sequentially decreased, which is not described herein again.
Example 2
An object of this implementation is to provide an edge recognition box, it includes:
one or more processors;
a storage unit configured to store one or more programs that, when executed by the one or more processors, enable the one or more processors to implement a method for picture splitting scheduling to an edge, the one or more programs comprising:
the picture receiving module is used for receiving the picture to be identified sent by the X-ray machine of the security check point in real time and storing the received picture to be identified into the task queue TQ,
a size judging module for judging the number x of the pictures to be identified in the task queue TQ 0 And a predetermined threshold u 0 Of the size u 0 Processing the maximum value of the picture to be recognized per second for the current edge recognition box, if x 0 ＞u 0 If yes, executing the cooperation request module;
the cooperation request module is used for sending cooperation requests to other edge image recognizing boxes, wherein the cooperation requests carry a preset quantity alpha, so that the other edge image recognizing boxes judge whether to accept the cooperation requests according to the preset quantity alpha of the images to be recognized and the quantity of the images to be recognized;
the receiving response module is used for receiving the cooperation response returned by the other edge map recognizing box;
an information obtaining module, configured to obtain, if the collaboration response carries information for accepting the collaboration request, state information of a collaborative edge image recognition box, and channel information of the collaborative edge image recognition box and a current edge image recognition box, where the collaborative edge image recognition box is an edge image recognition box corresponding to the collaboration response that carries information for receiving the collaboration request;
and the scheduling module is used for obtaining the minimum scheduling overhead time based on the state information and the channel information and scheduling the picture to be identified with the preset quantity alpha to the cooperative edge image identifying box corresponding to the minimum scheduling overhead time.
Example 3
An object of this implementation is to provide a system for shunting and scheduling a picture to an edge, including:
an edge computing node formed by a plurality of edge image recognizing boxes and an X-ray machine of a plurality of security check points; wherein, each edge map-identifying box is connected with each other, and each edge map-identifying box is connected with the X-ray machine of the security check point where the edge map-identifying box is located;
each edge map box is used for: receiving the picture to be identified sent by the X-ray machine of the security check point in real time, storing the received picture to be identified into a task queue TQ,
judging the number x of the pictures to be identified in the task queue TQ 0 And a predetermined threshold u 0 The size of u 0 Processing the maximum value of the picture to be identified per second for the current edge image identifying box, if x 0 ＞u 0 Sending a cooperation request to other edge image recognizing boxes, wherein the cooperation request carries a preset quantity alpha, so that the other edge image recognizing boxes judge whether to accept the cooperation request according to the preset quantity alpha of the image to be recognized and the quantity of the image to be recognized;
receiving a cooperation response returned by the other edge map recognizing box;
if the cooperation response carries information for receiving the cooperation request, acquiring state information of a cooperative edge image recognition box, and channel information of the cooperative edge image recognition box and a current edge image recognition box, wherein the cooperative edge image recognition box is an edge image recognition box corresponding to the cooperation response carrying the information for receiving the cooperation request;
obtaining the minimum scheduling overhead time based on the state information and the channel information, and scheduling the picture to be identified with the preset quantity alpha to a cooperative edge image identifying box corresponding to the minimum scheduling overhead time;
the edge recognition box is further used for: receiving a cooperation request sent by other edge image recognition boxes, wherein the cooperation request carries a preset quantity alpha, and judging whether to accept the cooperation request or not according to the preset quantity alpha of the image to be recognized and the quantity of the image to be recognized: if the cooperation request is received, returning a cooperation response carrying information for receiving the cooperation request to a corresponding edge map identifying box so that the corresponding edge map identifying box can acquire state information of the edge map identifying box in cooperation and channel information of the edge map identifying box in cooperation and the current edge map identifying box in cooperation, wherein the edge map identifying box in cooperation is the edge map identifying box corresponding to the cooperation response carrying information for receiving the cooperation request, and the minimum scheduling overhead time is obtained based on the state information and the channel information so as to schedule the picture to be identified with the pre-scheduling amount alpha to the edge map identifying box in cooperation corresponding to the minimum scheduling overhead time;
the X-ray machine of the security check point is used for scanning the passenger's packages, obtaining X-ray scanning pictures and sending the X-ray scanning pictures to the edge image identifying box of the security check point.
The foregoing is only a preferred embodiment of the present invention, and the present invention is not limited thereto in any way, and any simple modification, equivalent replacement and improvement made to the above embodiment within the spirit and principle of the present invention still fall within the protection scope of the present invention.
Claims (7)
1. A method for shunting and dispatching pictures to edges is characterized by being applied to edge computing nodes formed by a plurality of edge map-identifying boxes, wherein each edge map-identifying box is connected with each other, each edge map-identifying box is connected with an X-ray machine of a security check point where the edge map-identifying box is located, and the following steps are executed:
s1, receiving a picture to be identified sent by an X-ray machine of a security check point in real time, storing the received picture to be identified into a task queue TQ,
s2, judging the number x of the pictures to be identified in the task queue TQ 0 And a predetermined threshold u 0 Of the size u 0 Processing the maximum value of the picture to be identified per second for the current edge image identifying box, if x 0 ＞u 0 Turning to the step S3;
s3, sending a cooperation request to other edge image recognizing boxes, wherein the cooperation request carries a preset quantity alpha, so that the other edge image recognizing boxes judge whether to accept the cooperation request according to the preset quantity alpha and the quantity of the images to be recognized of the other edge image recognizing boxes;
s4, receiving a cooperation response returned by the other edge image recognizing box;
s5, if the cooperation response carries information for receiving the cooperation request, acquiring state information of a cooperation edge image recognition box, and channel information of the cooperation edge image recognition box and a current edge image recognition box, wherein the cooperation edge image recognition box is an edge image recognition box corresponding to the cooperation response carrying the information for receiving the cooperation request;
s6, obtaining the minimum scheduling overhead time based on the state information and the channel information, and scheduling the picture to be identified with the preset adjustment amount alpha to a cooperative edge image identifying box corresponding to the minimum scheduling overhead time;
the state information includes: coordinated edge recognition box i processing delay y for each picture i I =0, 1.. N, n is the total number of all cooperating edge map boxes;
the channel information includes: the current edge image recognition box dispatches the picture to be recognized with the preset quantity to the transmission time delay t of the cooperative edge image recognition box i 0i And the transmission time delay t of the cooperative edge image recognition box i for returning the image recognition result according to the preset amount of the image to be recognized i0 I =0, 1.. N, n is the total number of all cooperating edge map boxes;
step S6 specifically includes:
s001, based on the state information and the channel information, calculating scheduling overhead time Y of the front edge image recognition box for scheduling the pre-adjusted quantity of the images to be recognized to a cooperative edge image recognition box i i I =0, 1.. N, n is the total number of edge map boxes for all collaborations, and the scheduling overhead time Y is i ＝α×y i +t 0i +t i0 ；
S002, scheduling the picture to be identified with the preset adjustment amount alpha to the minimum scheduling overhead time Y min And (3) identifying the image box by corresponding cooperative edges, and turning to the step S2.
2. The method of claim 1, wherein the picture shunting is scheduled to an edge
Said x i Number of pictures to be recognized per second, u, processed for a cooperating edge recognition box i i Processing the maximum value of the picture to be recognized, y, for a cooperative edge recognition box i per second i The delay of each picture is processed for the cooperating edge recognition box i.
3. The method according to claim 2, wherein the picture shunting is scheduled to the edge
And generating a cooperative edge image recognition box i by counting historical to-be-recognized image processing data according to least square polynomial curve fitting, wherein g is an inherent parameter.
4. The method of claim 1, wherein the pre-adjustment amount α satisfies the requirement
Said x l Is a minimum scheduling overhead time Y min Number of pictures to be identified, u, processed per second of the corresponding cooperating edge graphic boxes l l Is a minimum scheduling overhead time Y min The corresponding cooperating edge recognition box/processes the maximum value of the picture to be recognized per second.
5. The method according to claim 4, wherein the method for shunting and scheduling the pictures to the edge is characterized in that
The above-mentioned i (bit) number of bytes of picture to be recognized, v, of pre-modulation amount alpha scheduled to cooperative edge recognition box i i (bit/s) is the uplink rate of the channel,
l is i (bit)' is edge recognition for collaborationThe picture box i returns the picture recognition result v according to the byte number of the picture to be recognized of the preset quantity alpha i (bit/s)' is the downlink rate of the channel.
6. An edge recognition box, comprising:
one or more processors;
a storage unit configured to store one or more programs, which when executed by the one or more processors, enable the one or more processors to implement a method for shunting and scheduling a picture to an edge according to any one of claims 1 to 5, the one or more programs including:
the picture receiving module is used for receiving the picture to be identified sent by the X-ray machine of the security check point in real time and storing the received picture to be identified into the task queue TQ,
a size judging module for judging the number x of the pictures to be identified in the task queue TQ 0 And a predetermined threshold u 0 The size of u 0 Processing the maximum value of the picture to be identified per second for the current edge image identifying box, if x 0 ＞u 0 If yes, executing the cooperation request module;
the cooperation request module is used for sending a cooperation request to other edge image recognizing boxes, wherein the cooperation request carries a preset quantity alpha, so that the other edge image recognizing boxes judge whether to accept the cooperation request according to the preset quantity alpha and the quantity of the images to be recognized of the other edge image recognizing boxes;
the receiving response module is used for receiving the cooperation response returned by the other edge image recognizing boxes;
an information obtaining module, configured to obtain, if the collaboration response carries information for accepting the collaboration request, state information of a collaborative edge image recognition box, and channel information of the collaborative edge image recognition box and a current edge image recognition box, where the collaborative edge image recognition box is an edge image recognition box corresponding to the collaboration response that carries information for receiving the collaboration request;
and the scheduling module is used for obtaining the minimum scheduling overhead time based on the state information and the channel information and scheduling the picture to be identified with the pre-modulation amount alpha to the cooperative edge image identifying box corresponding to the minimum scheduling overhead time.
7. A system for shunting and scheduling picture to edge, comprising:
an edge computing node formed by a plurality of edge image recognizing boxes and an X-ray machine of a plurality of security check points; wherein, each edge map-identifying box is connected with each other, and each edge map-identifying box is connected with the X-ray machine of the security check point where the edge map-identifying box is positioned;
the edge map box executes one or more programs, including:
the picture receiving module is used for receiving the picture to be identified sent by the X-ray machine of the security check point in real time and storing the received picture to be identified into the task queue TQ,
a size judging module for judging the number x of the pictures to be identified in the task queue TQ 0 And a predetermined threshold u 0 Of the size u 0 Processing the maximum value of the picture to be identified per second for the current edge image identifying box, if x 0 ＞u 0 If so, executing the cooperation request module;
the cooperation request module is used for sending a cooperation request to other edge image recognizing boxes, wherein the cooperation request carries a preset quantity alpha, so that the other edge image recognizing boxes judge whether to accept the cooperation request according to the preset quantity alpha and the quantity of the images to be recognized of the other edge image recognizing boxes;
the receiving response module is used for receiving the cooperation response returned by the other edge image recognizing boxes;
an information obtaining module, configured to obtain, if the cooperation response carries information for receiving the cooperation request, state information of a cooperative edge image recognizing box, and channel information of the cooperative edge image recognizing box and a current edge image recognizing box, where the cooperative edge image recognizing box is an edge image recognizing box corresponding to the cooperation response carrying the information for receiving the cooperation request;
the scheduling module is used for obtaining the minimum scheduling overhead time based on the state information and the channel information and scheduling the picture to be identified with the preset quantity alpha to a coordinated edge image identifying box corresponding to the minimum scheduling overhead time;
the state information includes: coordinated edge recognition box i processing delay y for each picture i I =0, 1.. N, n is the total number of all cooperating edge map boxes;
the channel information includes: the current edge image recognition box dispatches the picture to be recognized with the preset quantity to the transmission time delay t of the cooperative edge image recognition box i 0i And the transmission time delay t of the cooperative edge image recognition box i for returning the image recognition result according to the preset quantity of the image to be recognized i0 I =0, 1.. N, n is the total number of all cooperating edge map boxes;
the scheduling module specifically includes:
a scheduling overhead time submodule for calculating the scheduling overhead time Y of the front edge image recognition box for scheduling the pre-adjusted quantity of the images to be recognized to the cooperative edge image recognition box i based on the state information and the channel information i I =0, 1.. N, n is the total number of edge map boxes for all collaborations, and the scheduling overhead time Y is i ＝α×y i +t 0i +t i0 ；
A minimum scheduling submodule for scheduling the picture to be identified with the preset amount alpha to a minimum scheduling overhead time Y min The corresponding edge recognition box in cooperation executes a size judgment module;
the edge recognition box is further used for: receiving other edge image recognition boxes to send cooperation requests, wherein the cooperation requests carry a preset quantity alpha, and judging whether to accept the cooperation requests or not according to the preset quantity alpha of the images to be recognized and the quantity of the images to be recognized: if the cooperation request is received, returning a cooperation response carrying information for receiving the cooperation request to a corresponding edge map recognition box so that the corresponding edge map recognition box acquires state information of the edge map recognition box, channel information of the edge map recognition box and the current edge map recognition box, wherein the edge map recognition box is the edge map recognition box corresponding to the cooperation response carrying the information for receiving the cooperation request, and based on the state information and the channel information, obtaining the minimum scheduling overhead time, scheduling the picture to be identified with the preset quantity alpha to the edge map recognition box corresponding to the minimum scheduling overhead time;
the X-ray machine of the security check point is used for scanning the packages of passengers, obtaining X-ray scanning pictures and sending the X-ray scanning pictures to the edge image identifying box of the security check point.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210019902.4A CN114359024B (en) | 2022-01-10 | 2022-01-10 | Method for shunting and scheduling picture to edge, edge picture recognition box and system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202210019902.4A CN114359024B (en) | 2022-01-10 | 2022-01-10 | Method for shunting and scheduling picture to edge, edge picture recognition box and system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114359024A CN114359024A (en) | 2022-04-15 |
| CN114359024B true CN114359024B (en) | 2023-01-10 |
Family
ID=81108055
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202210019902.4A Active CN114359024B (en) | 2022-01-10 | 2022-01-10 | Method for shunting and scheduling picture to edge, edge picture recognition box and system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114359024B (en) |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5754709A (en) * | 1994-11-10 | 1998-05-19 | Matsushita Electric Industrial Co., Ltd. | Method and apparatus for gradation correction and image edge extraction |
| CN113300854A (en) * | 2021-05-21 | 2021-08-24 | 重庆紫光华山智安科技有限公司 | Edge node capability expansion method, system and expansion box |
| CN113687947A (en) * | 2021-08-25 | 2021-11-23 | 京东方科技集团股份有限公司 | Edge box optimization method and device, storage medium and electronic equipment |
-
2022
- 2022-01-10 CN CN202210019902.4A patent/CN114359024B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US5754709A (en) * | 1994-11-10 | 1998-05-19 | Matsushita Electric Industrial Co., Ltd. | Method and apparatus for gradation correction and image edge extraction |
| CN113300854A (en) * | 2021-05-21 | 2021-08-24 | 重庆紫光华山智安科技有限公司 | Edge node capability expansion method, system and expansion box |
| CN113687947A (en) * | 2021-08-25 | 2021-11-23 | 京东方科技集团股份有限公司 | Edge box optimization method and device, storage medium and electronic equipment |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114359024A (en) | 2022-04-15 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112541426B (en) | Communication bandwidth self-adaptive data processing method based on unmanned aerial vehicle cluster cooperative sensing | |
| CN104935622B (en) | A kind of method and device for message distribution, consumption, the system for Message Processing | |
| CN106171032B (en) | Data transmission method and equipment | |
| CN105159782A (en) | Cloud host based method and apparatus for allocating resources to orders | |
| CN104168318A (en) | Resource service system and resource distribution method thereof | |
| WO2018077031A1 (en) | Method and related apparatus for resource allocation | |
| CN109815019A (en) | Method for scheduling task, device, electronic equipment and readable storage medium storing program for executing | |
| CN106793133A (en) | The dispatching method of multi-service QoS is ensured in a kind of electric power wireless communication system | |
| CN105049485B (en) | A kind of Load-aware cloud computing system towards real time video processing | |
| CN112966608A (en) | Target detection method, system and storage medium based on edge-side cooperation | |
| CN103699433A (en) | Method and system for performing dynamic adjustment on number of tasks in Hadoop platform | |
| CN113141321B (en) | Data transmission method based on edge calculation and electronic equipment | |
| CN114359024B (en) | Method for shunting and scheduling picture to edge, edge picture recognition box and system | |
| CN109179101A (en) | Elevator control method, device, computer equipment and computer readable storage medium | |
| CN107065616A (en) | Inter-linked controlling method and device, computer-readable recording medium | |
| CN105991588B (en) | A kind of method and device for defending message attack | |
| CN108762936A (en) | Distributed computing system based on artificial intelligence image recognition and method | |
| CN107704322A (en) | Ask distribution method and device | |
| JP2018182430A (en) | Radio communication terminal | |
| CN107204930A (en) | Token adding method and device | |
| CN103795653B (en) | A kind of data cache method, device and optical network unit | |
| CN105991750B (en) | A kind of business scheduling method of business cluster, equipment and business cluster | |
| CN109214400A (en) | Classifier training method, apparatus, equipment and computer readable storage medium | |
| CN114359025B (en) | Method for shunting and dispatching pictures to center, edge picture-identifying box and system | |
| CN107817944A (en) | A kind of data processing method and storage device |
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 | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |