Picture forwarding method, system and equipment capable of dynamically selecting route and medium

Abstract

The invention provides a picture forwarding method, system and device capable of dynamically selecting routes and a medium, and the method comprises the steps: monitoring the delay data of each forwarding route based on a pre-constructed bidirectional deep neural network in response to an initiated picture forwarding request; and obtaining route information of each forwarding route, the route information including the delay data, dynamically determining a target route with the minimum delay by comparing and analyzing the route information of each forwarding route, and controlling the target route to perform picture forwarding. The method is more suitable for scenes such as 1400 view library collection service with large request data volume, dense requests and multiple forwarding routes, the optimal route can be adaptively adjusted, normal communication can be ensured even if congestion or faults occur in the optimal route, and the concurrent processing performance of service forwarding pictures is improved to the greatest extent.

Description

Technical Field

[0001] The present invention relates to the technical field of view transmission application, and in particular to a method, system, device and medium for forwarding an image capable of dynamically selecting a route. Background Art

[0002] With the development of edge analysis technology of front-end cameras, the application of front-end cameras with intelligent capture is becoming more and more common. In the field of video surveillance, in many cases, the backend needs to forward the captured images of faces, people or vehicles pushed by the front-end devices to multiple superior or third-party platforms at the same time. The view library service and the front-end devices interact with each other based on the GAT1400 protocol. The push interface has the characteristics of large single request data (more than several hundred KB), few requests during idle time, and frequent requests during peak hours. At the same time, the image data processing is easily affected by the docking service and network environment.

[0003] Since each view library collection service needs to access multiple front-end devices, during holidays, rush hours and other occasions with dense crowds of people and vehicles, when the number of pictures captured by the front-end is very large, it brings great processing pressure to the service to forward pictures. The push of pictures may be delayed due to network delays and network congestion, affecting the business analysis and processing of superiors and third-party platforms.

[0004] How to choose the appropriate route to push pictures is important for ensuring the real-time transmission of pictures, avoiding the accumulation of data in the view library service when the front-end devices push pictures concurrently during peak hours, and improving the concurrent performance of the view library acquisition service, which is of great significance to improving the stability and reliability of the entire view library system.

[0005] Currently, the view library collection service generally uses the http service framework. Since the image forwarding route of a single link is very susceptible to factors such as network delay and network congestion, image forwarding delays are prone to occur during peak hours. This will not only cause delays in business analysis and processing on the upper-level platform (such as AI alarms), but will also cause the view library service to accumulate too many images due to link congestion, resulting in rapid memory growth and exhaustion or service fuses. Summary of the invention

[0006] The embodiment of the present invention provides a method for forwarding an image that can dynamically select a route to solve the problems existing in the related art. The technical solution is as follows:

[0007] In a first aspect, an embodiment of the present invention provides a method for forwarding an image capable of dynamically selecting a route, comprising:

[0008] In response to the initiated image forwarding request, the delay data of each forwarding route is monitored based on the pre-built bidirectional deep neural network;

[0009] Obtain the routing information of each forwarding route, which includes delay data. By comparing and analyzing the routing information of each forwarding route, dynamically determine the target route with the minimum delay, and control the target route to forward the image.

[0010] In one embodiment, the bidirectional deep neural network includes two independent LSTM neural networks, and the input sequence is input into the two LSTM neural networks in forward order and reverse order respectively for feature extraction, and then the output vectors of the two LSTM neural networks are spliced.

[0011] In one embodiment, the routing information includes a naming tag, a state, and a delay table; the state includes an initialization state, an operation state, and a fault disabled state; the delay table includes multiple historical delay data and delay data obtained by bidirectional deep neural network prediction.

[0012] In one implementation, the method for determining the target route with the minimum delay is:

[0013] Among the forwarding routes with the same naming label, the forwarding route with the smallest delay is selected as the target route for image forwarding.

[0014] In one embodiment, it further includes:

[0015] Determine whether there is a forwarding route in the initialization state. If so, randomly select one of all forwarding routes in the initialization state for data forwarding to complete the initialization of the forwarding route. After the initialization is completed, monitor the delay data of the forwarding route.

[0016] In one embodiment, it further includes:

[0017] Determine whether there are multiple forwarding routes in operation with the same time delay. If so, use the forwarding route with the highest ranking as the target route for image forwarding. The ranking of the routes with the same time delay is based on the delay sequence returned by the forwarding routes.

[0018] In one embodiment, it further includes:

[0019] Determine whether the historical delay data in the delay table of the forwarding route exceeds the preset timeout period, delete the time delay data that exceeds the timeout period, and adjust the arrangement order of the remaining time delay data in turn; and the state of the forwarding route changes to the initialization state, and after the next forwarding request is completed, new delay data is generated and added to the delay table, and delay data predicted from the bidirectional deep neural network is generated to re-determine the target route with the smallest delay.

[0020] In a second aspect, an embodiment of the present invention provides a picture forwarding system that executes the picture forwarding method capable of dynamically selecting a route as described above.

[0021] In a third aspect, an embodiment of the present invention provides an electronic device, the device comprising: a memory and a processor. The memory and the processor communicate with each other through an internal connection path, the memory is used to store instructions, the processor is used to execute the instructions stored in the memory, and when the processor executes the instructions stored in the memory, the processor executes the method in any one of the above-mentioned embodiments.

[0022] In a fourth aspect, an embodiment of the present invention provides a computer-readable storage medium, which stores a computer program. When the computer program runs on a computer, the method in any one of the above-mentioned embodiments is executed.

[0023] The advantages or beneficial effects of the above technical solution include at least:

[0024] Compared with the traditional forwarding strategy algorithm, the present invention adds delay prediction based on CNN-BiLSTM neural network in the key processing link, and dynamically adjusts the forwarding route according to the predicted delay data. The neural network can output predictions for all routes in parallel, and the prediction speed is fast. It is more suitable for scenarios such as 1400 view library acquisition services with large request data volume, intensive requests, and many forwarding routes than traditional forwarding strategies, and can adaptively adjust the best route to ensure normal communication even when congestion or failure occurs in the best optimal route, thereby maximizing the concurrent processing performance of the service forwarding images.

[0025] The above summary is for illustrative purposes only and is not intended to be limiting in any way. In addition to the illustrative aspects, embodiments and features described above, further aspects, embodiments and features of the present invention will be readily apparent by reference to the accompanying drawings and the following detailed description. BRIEF DESCRIPTION OF THE DRAWINGS

[0026] In the accompanying drawings, unless otherwise specified, the same reference numerals throughout the multiple drawings represent the same or similar parts or elements. These drawings are not necessarily drawn to scale. It should be understood that these drawings only depict some embodiments disclosed in the present invention and should not be regarded as limiting the scope of the present invention.

[0027] Figure 1 A schematic diagram of the process flow of the method for forwarding pictures capable of dynamically selecting a route according to the present invention;

[0028] Figure 2 Schematic diagram of the architecture of the CNN-BiLSTM bidirectional deep neural network of the present invention;

[0029] Figure 3 This is a schematic diagram of the LSTM hidden layer structure principle of the present invention;

[0030] Figure 4 It is a structural block diagram of the electronic device of the present invention. DETAILED DESCRIPTION

[0031] In the following, only some exemplary embodiments are briefly described. As those skilled in the art will appreciate, the described embodiments may be modified in various ways without departing from the spirit or scope of the present invention. Therefore, the drawings and descriptions are considered to be exemplary and non-restrictive in nature.

[0032] Embodiment 1

[0033] This embodiment provides a method for forwarding images that can dynamically select routes. For the scenario where the amount of image forwarding data is large during the peak period of the GAT1400 view library acquisition service, the optimal forwarding route is selected through the delay predicted by the neural network trained by CNN-BiLSTM, and dynamic adaptive adjustment of the forwarding route is achieved to improve the concurrent forwarding performance and real-time performance of peak period images.

[0034] like Figure 1 As shown, the image forwarding method capable of dynamically selecting a route includes the following steps:

[0035] In response to the initiated image forwarding request, the delay data of each forwarding route is monitored based on the pre-built bidirectional deep neural network;

[0036] Obtain the routing information of each forwarding route, which includes delay data. By comparing and analyzing the routing information of each forwarding route, dynamically determine the target route with the minimum delay, and control the target route to forward the image.

[0037] The image forwarding request may be a request initiated when image concurrency occurs, or a request initiated during idle time. This embodiment is mainly to solve the image forwarding problem in the image concurrency scenario. Therefore, the image forwarding request in this embodiment mainly refers to a request initiated when image concurrency occurs, and the following image forwarding method is also a data processing process in the image concurrency scenario.

[0038] like Figure 2 As shown in the figure, the pre-built bidirectional deep neural network is a CNN-BiLSTM model; the CNN-BiLSTM model consists of an input layer, a CNN layer, a Bi-LSTM layer, a fully connected layer and an output layer; wherein the CNN layer is composed of a stack of convolutional layers and maximum pooling layers, and the BiLSTM layer is composed of a stack of BiLSTM layers. Dropout layers are added at the end of the CNN layer and the BiLSTM layer to randomly discard nodes to prevent overfitting.

[0039] Among them, the neural network structure model of the Bi-LSTM layer is mainly divided into two independent LSTM neural networks. The input sequence is input into the two LSTM neural networks in forward order and reverse order for feature extraction, and the word vector formed by concatenating the two output vectors (i.e., the extracted feature vectors) is used as the final feature expression of the word.

[0040] The Bi-LSTM neural network extracts features through forward and reverse order, so that the feature data obtained at time t can contain information between the past and the future at the same time; experiments have shown that the efficiency and performance of this neural network structure model for text feature extraction are better than those of a single LSTM structure model.

[0041] Among them, the full name of LSTM is Long Short-Term Memory, which is a type of RNN (Recurrent Neural Network). The LSTM neural network model is similar to the RNN model in terms of its main structure, mainly adding three gate structures in the hidden layer h, namely the forget gate, input gate, and output gate; at the same time, a hidden state (cell state) is added.

[0042] The LSTM hidden layer structure principle is as follows Figure 3 As shown in the figure, f(t), i(t), and o(t) represent the values ​​of the forget gate, input gate, and output gate at time t, respectively, and a(t) represents the preliminary feature extraction of h(t–1) and x(t) at time t.

[0043] The specific calculation process is:

[0044] f(t)=σ(Wfh t-1 +Ufx t +bf);

[0045] i(t)=σ(W i h t-1 +U i x t +b i );

[0046] a(t)=tanh(W a h t-1 +U a x t +b a );

[0047] o(t)=σ(W o h t-1 +U o x t +b o );

[0048] Among them, x t represents the input at time t;

[0049] h t-1 represents the hidden layer state value at time t–1;

[0050] W f , W i , W o and W a They represent the forget gate, input gate, output gate and h in the feature extraction process respectively. t-1 The weight coefficient of

[0051] U f , U i , U o and U a Respectively represent the weight coefficients of xt in the forget gate, input gate, output gate and feature extraction process;

[0052] b f 、b i 、b o and b a Respectively represent the bias values ​​in the forget gate, input gate, output gate and feature extraction process;

[0053] tanh represents the tangent hyperbolic function, σ represents the activation function Sigmoid,

[0054]

[0055]

[0056] The results calculated by the forget gate and the input gate act on c(t-1) to form the cell state c(t) at time t, which can be expressed as:

[0057] c(t)=c(t-1)ef(t)+i(t)ea(t);

[0058] Where e is the Hadamard product.

[0059] Finally, the hidden layer state h(t) at time t is calculated by the output gate o(t) and the cell state c(t) at the current moment: h(t) = o(t) e tanh(c(t)).

[0060] The LSTM model solves the long-term dependency problem of the recurrent neural network RNN. Its unique "gate" structure can avoid gradient explosion and gradient disappearance, and has the advantage of strong long-term memory. BiLSTM not only has the advantages of LSTM, but also considers the forward and backward bidirectional time series information in the time dimension, making the prediction more comprehensive and accurate. CNN is suitable for extracting local spatial features, and BiLSTM takes into account the combination of bidirectional time series information, which can more comprehensively analyze the routing delay data and make the prediction results more fit.

[0061] The CNN-BiLSTM bidirectional deep neural network is pre-trained with a large amount of sample data. Real image forwarding request data is selected as sample data so that the delay characteristics of the actual network system can be truly restored. With reference to the size of general network delay, 200 different delay samples are obtained, of which 50% are training samples and 50% are test samples.

[0062] During the neural network training process, the parameters of Bi-LSTM are updated through the time-based back-propagation algorithm, which calculates the gradient of the loss function for each parameter and then uses the stochastic gradient descent algorithm to update the parameters. The time series-based back-propagation algorithm (or BPTT) is the application of the back-propagation training algorithm in recursive neural networks. The specific supplement of the time series back-propagation algorithm has been disclosed in the prior art, and the principle steps are simply described here:

[0063] 1. Present the network with a sequence of input and output pairs of time steps;

[0064] 2. Expand the network, then calculate and accumulate the error at each time step;

[0065] 3. Roll up the network and update the weights;

[0066] 4. Repeat.

[0067] The method for predicting the delay data of each forwarding route based on the trained CNN-BiLSTM bidirectional deep neural network is:

[0068] The input data of the neural network is constructed by sliding time window, that is, the delay values ​​of the previous few time points are input into the neural network to predict the delay value of the next time point, thereby improving the accuracy of the network. In this embodiment, a multi-layer Bi-LSTM network is constructed to predict the delay data of the time window feature; the experimental results show that the proposed method has higher prediction accuracy than the traditional prediction method.

[0069] After predicting the delay data of each forwarding route through the CNN-BiLSTM bidirectional deep neural network, the forwarding route can be dynamically selected.

[0070] This embodiment adds the routing information Stat of each forwarding route to the forwarding table (FIB table). The routing information Stat includes a naming tag, a state, and a delay table. A character string is used to represent the naming tag; a simple number is used to represent the state. For each forwarding route interface, there is only one state, 0 represents the initialization state, 1 represents the running state, and 2 represents the fault disabled state. The delay table includes multiple historical delay data and delay data obtained by prediction by a bidirectional deep neural network. The historical delay data can be five historical delay data D1-D5. When forwarding, the forwarding route to be taken can be determined by simply comparing the data in the routing information Stat of each route.

[0071] In order to adapt to the scenario with large concurrent image volume, Stat interface is added to improve the bandwidth of data transmission; due to the addition of Stat interface, under the premise of forwarding routes, it is assumed that several forwarding routes have named labels pointing to data packets.

[0072] If there is only one forwarding route, the corresponding data is forwarded in this direction. If there are multiple forwarding routes, the data in Stat is judged and the following forwarding strategy is implemented:

[0073] (1) The data selects the forwarding route with the smallest delay among the forwarding routes with the same naming label to forward the image.

[0074] (2) If there is a forwarding route in the initialization state, one of the forwarding routes in the initialization state will be randomly selected to forward the data to the forwarding route to complete the initialization of the forwarding route. If there are interfaces with the same delay in the forwarding routes in the running state, the interface with the highest ranking will be forwarded. Among them, the ranking can be set according to the preset rules. In this embodiment, after forwarding the picture, the Stats of each interface and its respective returned delay are first stored, and an increment is added after the delay, and the number will not be repeated. Then sort by delay from small to large, and sort by increment for the same time delay.

[0075] (3) When the data transmission is completed, the delay data of D1 in the historical delay data is deleted, the time delay data is moved to the left in sequence, the new actual time delay is stored in D5, and the sixth delay is predicted. If the forwarding interface changes, the historical delay data in the delay table is changed to the delay data of the new route.

[0076] In actual use, it is necessary to pre-set the image receiving route, initialize each forwarding route, and mark the forwarding route with a named label. After the route initialization is completed, it participates in the competition based on the network delay predicted by the neural network, and dynamically selects the optimal route for image forwarding.

[0077] The routing information of each forwarding route needs to be updated in real time. The following situations exist:

[0078] (1) When a new forwarding route is added, the forwarding route is in the initialization state 0, and the delay table data is empty. At this time, the server will preferentially assign requests to the new forwarding route. After the delay table data is filled, the state of the forwarding route changes from 0 to 1, and the delay data predicted by the neural network is generated to participate in the data forwarding competition.

[0079] (2) The historical delay data in the delay table of each forwarding route needs to be compared with the preset timeout period T. If any of the historical delay data exceeds the timeout period T, the data will be deleted and the time delay data will move to the left in sequence. At this time, the interface state will change from 1 to 0 and re-enter the initialization state. After the next forwarding request is completed, new delay data will be generated and added to the delay table, and delay data predicted by the neural network will be generated to re-participate in the data forwarding competition.

[0080] (3) After forwarding, first store the routing information Stats of each forwarding route and its respective return delay, and add an increment after the delay, and the number will not be repeated. Then sort by delay from small to large, and sort by increment for the same time delay.

[0081] (4) If the request fails or the predicted delay data exceeds T1 (T1 is m1 times the timeout period T, where m1 is determined based on the specific conditions of the network), the routing state is set to 2 and the route enters the fault disabled state. The route enters the sleep state and maintains the sleep time of T2 (the length of T2 is determined by the error level and the delay period T1). After the sleep period ends, the route re-enters the initialization state.

[0082] (5) When the forwarding route changes, the original route information Stat is deleted, the new forwarding route becomes initialized, and participates in data forwarding competition in the same way as the newly added route, so as to avoid the influence of the previous route delay data and reduce the storage pressure of the local time delay table.

[0083] Since the network itself has highly dynamic nodes, it will bring topological uncertainty and even routing interruption. In order to ensure normal communication even when congestion or failure occurs in the best optimal route, a dynamic adaptive routing mechanism and fault rerouting mechanism are proposed, that is, the status judgment of the route and link is based on the timer's timeout retransmission count. Each routing node sets a timeout retransmission counter for the corresponding content. When the timeout retransmission occurs multiple times in a short period of time, the link is considered to be faulty. Each routing node can quickly detect the fault and take corresponding measures according to the specific severity to ensure the quality of subsequent communication. The adjustment process is divided into the following three aspects:

[0084] (1) When a timeout retransmission occurs, if the frequency is not too high, the optimal route is considered to be congested or faulty, and the suboptimal route is used for communication first. The route with the second smallest delay is selected through the routing information Stat to perform communication. At this time, the route state is set to 2, entering the fault disabled state, and this route will sleep for T2 (a short time), and will re-enter the initialization state after the sleep ends.

[0085] (2) If timeout retransmissions occur frequently within a short period of time, the previous route is considered to be completely interrupted. At this time, the route status is set to 2 and enters the fault disabled state. This route will sleep for T2 (a longer time) and will re-enter the initialization state after the sleep ends.

[0086] (3) If all routes under a named label fail, it is necessary to re-establish the routes based on the retransmitted data packets and update the FIB forwarding table (ForwardingInformation) data based on the pre-built neural network prediction delay data to obtain the complete routing information of the newly added routes to participate in the subsequent data forwarding competition.

[0087] (4) When the sleep time of the faulty node ends, it will re-initialize. After the initialization is completed, it will re-participate in the data forwarding competition based on the delay data predicted by the neural network to ensure that the traffic converges to the current optimal route as quickly as possible.

[0088] Compared with the traditional forwarding strategy algorithm, this embodiment adds a delay prediction based on the CNN-BiLSTM neural network in the key processing link, and dynamically adjusts the forwarding route according to the predicted delay data. The neural network can output predictions for all routes in parallel, and the prediction speed is fast. It is more suitable for scenarios such as the 1400 view library acquisition service with large request data volume, intensive requests, and many forwarding routes than traditional forwarding strategies, and can adaptively adjust the optimal route, ensuring normal communication even when congestion or failure occurs in the best optimal route, thereby maximizing the concurrent processing performance of the service forwarding images.

[0089] Embodiment 2

[0090] This embodiment provides a picture forwarding system with dynamic routing selection, which executes the picture forwarding method with dynamic routing selection as described in the first embodiment. The system specifically includes:

[0091] Model building module, used to build CNN-BiLSTM neural network and train the neural network to predict network delay;

[0092] The delay prediction module is used to monitor the delay data of each forwarding route based on the initiated image forwarding request based on a pre-built bidirectional deep neural network;

[0093] The routing selection module is used to obtain the routing information of each forwarding route, the routing information includes delay data, and dynamically determine the target route with the minimum delay by comparing the routing information of each forwarding route;

[0094] The image forwarding module is used to control the target route to forward images.

[0095] The functions of each module in the system of the embodiment of the present invention can be found in the corresponding description of the above method, which will not be repeated here.

[0096] Embodiment 3

[0097] Figure 4 FIG. 2 shows a structural block diagram of an electronic device according to an embodiment of the present invention. Figure 4 As shown, the electronic device includes: a memory 100 and a processor 200, wherein the memory 100 stores a computer program that can be run on the processor 200. When the processor 200 executes the computer program, the method for forwarding a picture with dynamic routing selection in the above embodiment is implemented. The number of the memory 100 and the processor 200 can be one or more.

[0098] The electronic device also includes:

[0099] The communication interface 300 is used to communicate with external devices and perform data exchange transmission.

[0100] If the memory 100, the processor 200 and the communication interface 300 are implemented independently, the memory 100, the processor 200 and the communication interface 300 can be connected to each other through a bus and communicate with each other. The bus can be an Industry Standard Architecture (ISA) bus, a Peripheral Component Interconnect (PCI) bus or an Extended Industry Standard Architecture (EISA) bus. The bus can be divided into an address bus, a data bus, a control bus, etc. For ease of representation, Figure 4 Only one thick line is used in the diagram, but this does not mean that there is only one bus or only one type of bus.

[0101] Optionally, in a specific implementation, if the memory 100, the processor 200 and the communication interface 300 are integrated on a chip, the memory 100, the processor 200 and the communication interface 300 can communicate with each other through an internal interface.

[0102] An embodiment of the present invention provides a computer-readable storage medium storing a computer program, which implements the method provided in the embodiment of the present invention when executed by a processor.

[0103] An embodiment of the present invention further provides a chip, which includes a processor for calling and executing instructions stored in the memory from the memory, so that a communication device equipped with the chip executes the method provided by the embodiment of the present invention.

[0104] An embodiment of the present invention also provides a chip, including: an input interface, an output interface, a processor and a memory, wherein the input interface, the output interface, the processor and the memory are connected via an internal connection path, and the processor is used to execute the code in the memory. When the code is executed, the processor is used to execute the method provided by the embodiment of the invention.

[0105] It should be understood that the processor may be a central processing unit (CPU), or other general-purpose processors, digital signal processors (DSP), application specific integrated circuits (ASIC), field programmable gate arrays (FPGA) or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. The general-purpose processor may be a microprocessor or any conventional processor, etc. It is worth noting that the processor may be a processor supporting the advanced RISC machines (ARM) architecture.

[0106] Further, optionally, the above-mentioned memory may include a read-only memory and a random access memory, and may also include a non-volatile random access memory. The memory may be a volatile memory or a non-volatile memory, or may include both volatile and non-volatile memories. Among them, the non-volatile memory may include a read-only memory (ROM), a programmable read-only memory (PROM), an erasable programmable read-only memory (EPROM), an electrically erasable programmable read-only memory (EEPROM), or a flash memory. The volatile memory may include a random access memory (RAM), which is used as an external cache. By way of example but not limitation, many forms of RAM are available. For example, static RAM (SRAM), dynamic random access memory (DRAM), synchronous DRAM (SDRAM), double data rate synchronous dynamic random access memory (DDR SDRAM), enhanced synchronous dynamic random access memory (ESDRAM), synchronous link dynamic random access memory (SLDRAM) and direct rambus RAM (DR RAM).

[0107] In the above embodiments, it can be implemented in whole or in part by software, hardware, firmware or any combination thereof. When implemented using software, it can be implemented in whole or in part in the form of a computer program product. The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, the process or function according to the present invention is generated in whole or in part. The computer can be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions can be stored in a computer-readable storage medium, or transmitted from one computer-readable storage medium to another computer-readable storage medium.

[0108] In the description of this specification, the description with reference to the terms "one embodiment", "some embodiments", "example", "specific example", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner. In addition, those skilled in the art may combine and combine different embodiments or examples described in this specification and the features of different embodiments or examples, unless they are contradictory.

[0109] In addition, the terms "first" and "second" are used for descriptive purposes only and should not be understood as indicating or implying relative importance or implicitly indicating the number of the indicated technical features. Therefore, the features defined as "first" and "second" may explicitly or implicitly include at least one of the features. In the description of the present invention, the meaning of "plurality" is two or more, unless otherwise clearly and specifically defined.

[0110] Any process or method description in the flow chart or otherwise described herein can be understood to represent a module, segment or portion of a code including one or more executable instructions for implementing the steps of a specific logical function or process. And the scope of the preferred embodiment of the present invention includes other implementations, in which the functions may not be performed in the order shown or discussed, including in a substantially simultaneous manner or in a reverse order according to the functions involved.

[0111] The logic and / or steps represented in the flowchart or otherwise described herein, for example, can be considered as an ordered list of executable instructions for implementing logical functions, which can be embodied in any computer-readable medium for use by an instruction execution system, apparatus or device (such as a computer-based system, a system including a processor or other system that can fetch instructions from an instruction execution system, apparatus or device and execute instructions), or used in combination with these instruction execution systems, apparatuses or devices.

[0112] It should be understood that the various parts of the present invention can be implemented by hardware, software, firmware or a combination thereof. In the above embodiments, multiple steps or methods can be implemented by software or firmware stored in a memory and executed by a suitable instruction execution system. All or part of the steps of the above embodiment method can be completed by instructing the relevant hardware through a program, which can be stored in a computer-readable storage medium, and when the program is executed, it includes one of the steps of the method embodiment or a combination thereof.

[0113] In addition, each functional unit in each embodiment of the present invention may be integrated into a processing module, or each unit may exist physically separately, or two or more units may be integrated into one module. The above-mentioned integrated module may be implemented in the form of hardware or in the form of a software functional module. If the above-mentioned integrated module is implemented in the form of a software functional module and sold or used as an independent product, it may also be stored in a computer-readable storage medium. The storage medium may be a read-only memory, a disk or an optical disk, etc.

[0114] The above is only a specific embodiment of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art can easily think of various changes or substitutions within the technical scope disclosed by the present invention, which should be included in the protection scope of the present invention. Therefore, the protection scope of the present invention should be based on the protection scope of the claims.

Claims

1. A method for forwarding an image with dynamic routing selection, characterized in that: include: In response to the initiated image forwarding request, the delay data of each forwarding route is monitored based on the pre-built bidirectional deep neural network; The routing information of each forwarding route is obtained, the routing information includes the delay data, and a target route with the smallest delay is dynamically determined by comparing and analyzing the routing information of each forwarding route, and the target route is controlled to forward the image.

2. The method for forwarding images with dynamic routing selection according to claim 1, characterized in that: The bidirectional deep neural network includes two independent LSTM neural networks. The input sequence is input into the two LSTM neural networks in forward order and reverse order respectively for feature extraction, and then the output vectors of the two LSTM neural networks are spliced.

3. The method for forwarding pictures with dynamic routing selection according to claim 1, characterized in that: The routing information includes a naming tag, a state and a delay table; the state includes an initialization state, an operation state and a fault disabled state; the delay table includes a plurality of historical delay data and the delay data obtained by the bidirectional deep neural network prediction.

4. The method for forwarding images with dynamic routing selection according to claim 1, characterized in that: The method for determining the target route with the minimum delay is: The forwarding route with the smallest delay is selected from the forwarding routes with the same naming tag as the target route for image forwarding.

5. The method for forwarding pictures with dynamic routing selection according to claim 1, characterized in that: Also includes: Determine whether there is the forwarding route in the initialization state. If so, randomly select one of all the forwarding routes in the initialization state for data forwarding to complete the initialization of the forwarding route. After the initialization is completed, monitor the delay data of the forwarding route.

6. The method for forwarding pictures with dynamic routing selection according to claim 1, characterized in that: Also includes: Determine whether there are multiple forwarding routes in operation and with the same time delay. If so, use the forwarding route with the highest ranking as the target route for image forwarding; wherein the ranking of the routes with the same time delay is sorted according to the delay order returned by the forwarding routes.

7. The method for forwarding pictures with dynamic routing selection according to claim 1, characterized in that: Also includes: Determine whether the historical delay data in the delay table of the forwarding route exceeds the preset timeout time, delete the time delay data that exceeds the timeout time, and adjust the arrangement order of the remaining time delay data in turn; and the state of the forwarding route changes to the initialization state, and after the next forwarding request is completed, new delay data is generated and added to the delay table, and the delay data predicted by the bidirectional deep neural network is generated, and the target route with the smallest delay is re-determined.

8. A picture forwarding system, characterized in that: Execute the image forwarding method capable of dynamically selecting a route as described in any one of claims 1 to 7.

9. An electronic device, characterized in that: include: A processor and a memory, wherein the memory stores instructions, and the instructions are loaded and executed by the processor to implement the image forwarding method capable of dynamically selecting a route as claimed in any one of claims 1 to 7.

10. A computer-readable storage medium, wherein a computer program is stored in the computer-readable storage medium, and when the computer program is executed by a processor, the method for forwarding an image with dynamic routing selection according to any one of claims 1 to 7 is implemented.

Images