Training method of state echo network, and monitoring track determination method and device

Abstract

The application provides a state echo network training method, a monitoring track determination method and device. The state echo network comprises an input ESN, a chaotic ESN and a feedback classifier. The training method comprises the following steps: acquiring attractor sample data, wherein the attractor sample data comprises position information of monitoring tracks arranged in a plurality of regions of interest in a monitorable region of a monitoring camera; pre-training a connection matrix of the input ESN by using the attractor sample data; inputting the attractor sample data into the pre-trained input ESN to obtain states of nodes of the pre-trained input ESN; training a connection matrix of the chaotic ESN by using the states of the nodes output by the pre-trained input ESN to obtain a trained chaotic ESN; and training a connection matrix of the feedback classifier by using the states of the nodes output by the pre-trained input ESN to obtain a trained feedback classifier.

Description

Technical Field

[0001] This invention relates to the field of security monitoring technology, and in particular to a training method, a monitoring trajectory determination method and device for a state echo network. Background Technology

[0002] In the existing security monitoring field, surveillance cameras can be set up to focus on areas of interest (also known as key areas) as needed. However, the trajectories of these cameras as they sweep across these areas of interest are usually fixed, and the transition trajectories between areas of interest are typically monotonous. Under this monitoring method, the surveillance camera trajectories are easily predicted by criminals, leading to loss of life and property. Summary of the Invention

[0003] This invention provides a training method, a monitoring trajectory determination method, and an apparatus for state echo networks, which address the problem that the monitoring trajectories of surveillance cameras are usually fixed and easily predictable by criminals in existing security monitoring processes.

[0004] To solve the above-mentioned technical problems, the present invention is implemented as follows:

[0005] In a first aspect, embodiments of the present invention provide a training method for a State Echo Network, the State Echo Network comprising: an input ESN, a chaotic ESN, and a feedback classifier, the method comprising:

[0006] Acquire attractor sample data, which includes location information of monitoring trajectories set in multiple regions of interest within the monitorable area of ​​the surveillance camera;

[0007] The connection matrix of the input ESN is pre-trained using the attractor sample data to obtain the pre-trained input ESN.

[0008] The attractor sample data is input into the pre-trained input ESN to obtain the state of the nodes of the pre-trained input ESN;

[0009] The connection matrix of the chaotic ESN is trained using the state of the nodes output by the pre-trained input ESN to obtain the trained chaotic ESN.

[0010] The connection matrix of the feedback classifier is trained using the state of the nodes output by the pre-trained input ESN to obtain the trained feedback classifier, which is used to realize the autonomous transformation between the regions of interest.

[0011] Optionally, obtaining attractor sample data includes:

[0012] The surveillance camera is controlled to collect images of the monitored area using a traversal method.

[0013] The images of all the monitored areas collected are stitched together to obtain the overall monitored area image;

[0014] Identify the region of interest in the overall monitorable area image;

[0015] Obtain the monitoring trajectory set within the area of ​​interest;

[0016] The location information of the monitoring trajectory set within the region of interest is used as the attractor sample data.

[0017] Optionally, the step of pre-training the connection matrix of the input ESN using the attractor sample data to obtain the pre-trained input ESN includes:

[0018] The attractor sample data is input into the input ESN to obtain the state of the node of the input ESN output by the input ESN;

[0019] The first loss function is determined based on the state of the node output by the input ESN and the pre-set target state of the node of the input ESN;

[0020] The connection matrix of the input ESN is optimized using the first loss function.

[0021] Optionally, the number of nodes participating in the training of the input ESN is half of the total number of nodes in the input ESN.

[0022] Optionally, the input ESN is represented as:

[0023]

[0024] Where t is time, τ is the proportionality coefficient, and x in For the state of the nodes in the input ESN, g in J is the scaling parameter. in Let u be the connection matrix of the input ESN. in (s) represents the attractor sample data.

[0025] Optionally, the first loss function is:

[0026]

[0027] Among them, C 1-in Let x be the first loss function, t be time, and x be the loss function. s (t) represents the state of the node that inputs the ESN at time t. Let L be the target state of the node that inputs the ESN at time t. innate Let be the maximum value at time t, s be the monitoring trajectory set within the region of interest, and S be the set of monitoring trajectories set within the region of interest.

[0028] Optionally, the connection matrix of the chaotic ESN is trained using the states of the nodes output by the pre-trained input ESN to obtain the trained chaotic ESN, including:

[0029] The state of the node output by the pre-trained input ESN is input into the chaotic ESN to obtain the monitoring position of the surveillance camera output by the chaotic ESN.

[0030] The second loss function is determined based on the monitoring location of the surveillance camera output by the chaotic ESN and the specified monitoring location;

[0031] The connection matrix of the chaotic ESN is optimized using the second loss function.

[0032] Optionally, the chaotic ESN is represented as:

[0033]

[0034] Where t is time, τ is the proportionality coefficient, and x ch For the state of a node in the chaotic ESN, g ch J is the scaling parameter. ch Let J be the connection matrix of the chaotic ESN. ic Let x be the connection matrix between the input ESN and the chaotic ESN. in The state of the nodes in the input ESN.

[0035] Optionally, the second loss function is:

[0036]

[0037] Among them, C 1-out Let x be the second loss function, t be time, and x be the value of x. s (t) includes the state of the node that inputs the ESN and the state of the node that outputs the ESN at time t, x s (t)=[x in ;x ch ], f s (t) represents the specified monitoring location at time t, L out w is the maximum value at time t. out Let be the connection matrix between the chaotic ESN and the output layer, s be the monitoring trajectory set within the region of interest, and S be the set of monitoring trajectories set within the region of interest.

[0038] Optionally, the connection matrix of the feedback classifier is trained using the states of the nodes output by the pre-trained input ESN to obtain the trained feedback classifier, including:

[0039] The state of the node output by the pre-trained input ESN is input into the feedback classifier to obtain the monitoring position of the surveillance camera at the next moment, as output by the feedback classifier.

[0040] The third loss function is determined based on the monitoring position of the surveillance camera at the next moment output by the feedback classifier and the preset monitoring position of the surveillance camera at the next moment.

[0041] The connection matrix of the feedback classifier is optimized based on the third loss function.

[0042] Optionally, the connection matrix of the feedback classifier is trained using the states of the nodes output by the pre-trained input ESN to obtain the trained feedback classifier, including:

[0043] The state of the nodes output by the pre-trained input ESN and the transition probability between the pre-set regions of interest are input into the feedback classifier to obtain the monitoring position of the surveillance camera at the next moment, as output by the feedback classifier.

[0044] The fourth loss function is determined based on the monitoring position of the surveillance camera at the next moment output by the feedback classifier and the preset monitoring position of the surveillance camera at the next moment.

[0045] The connection matrix of the feedback classifier is optimized based on the fourth loss function.

[0046] Optionally, the feedback classifier is represented as:

[0047]

[0048] Among them, f max Let x(t) be the feedback classifier, t be time, and x(t) include the state of the node input to the ESN and the state of the node outputting the ESN at time t, x(t) = [x in ;x ch ], w is the connection matrix of the feedback classifier, s is the monitoring trajectory set in the region of interest, and S is the set of monitoring trajectories set in the region of interest.

[0049] Optionally, the third loss function is:

[0050]

[0051] Where C2 is the third loss function, t is time, and s is the time interval. sper (t+Δt) represents the monitoring position of the surveillance camera at the next moment, as output by the feedback classifier; x(t) represents the state of the node of the input ESN at time t; w is the connection matrix of the feedback classifier; s is the monitoring trajectory set within the region of interest; S is the set of monitoring trajectories set within the region of interest; T rec This represents the maximum value for time.

[0052] Optionally, the fourth loss function is:

[0053]

[0054] Where C3 is the fourth loss function, t is time, and s is the time interval. sto (t+Δt) represents the monitoring position of the surveillance camera at the next moment, output by the feedback classifier based on the pre-set transition probabilities between regions of interest; x(t) represents the state of the node of the input ESN at time t; w is the connection matrix of the feedback classifier; s is the monitoring trajectory set within the region of interest; S is the set of monitoring trajectories set within the region of interest; T rec The maximum value for time is used, and the conversion probability is determined based on the importance of the region of interest.

[0055] Secondly, embodiments of the present invention provide a method for determining a monitoring trajectory, including:

[0056] Obtain the current monitoring location of the surveillance camera;

[0057] The current monitoring location is input into the state echo network to obtain the monitoring location of the surveillance camera at the next moment, which is output by the state echo network. The state echo network includes: an input ESN, a chaotic ESN, and a feedback classifier; the feedback classifier is used to realize the autonomous conversion between regions of interest of the surveillance camera.

[0058] Control the rotation of the surveillance camera to obtain an image of the monitored location at the next moment.

[0059] Optionally, the input ESN is represented as:

[0060]

[0061] Where t is time, τ is the proportionality coefficient, and x in For the state of the nodes in the input ESN, g in J is the scaling parameter. in Let u be the connection matrix of the input ESN. in (s(t)) represents the monitoring location at time t.

[0062] Optionally, the chaotic ESN is represented as:

[0063]

[0064] Where t is time, τ is the proportionality coefficient, and x ch For the state of a node in the chaotic ESN, g ch J is the scaling parameter. ch Let J be the connection matrix of the chaotic ESN. ic Let x be the connection matrix between the input ESN and the chaotic ESN. in The state of the nodes in the input ESN.

[0065] Optionally, the feedback classifier is represented as:

[0066]

[0067] Among them, f max Let x(t) be the feedback classifier, t be time, and x(t) include the state of the node input to the ESN and the state of the node outputting the ESN at time t, x(t) = [x in ;x ch ], w is the connection matrix of the feedback classifier, s is the monitoring trajectory set in the region of interest, and S is the set of monitoring trajectories set in the region of interest.

[0068] Optionally, inputting the current monitoring location into the status echo network includes:

[0069] The conversion probability between the current monitoring location and the region of interest of the monitoring camera is input into the state echo network.

[0070] Thirdly, embodiments of the present invention provide a training apparatus for a state echo network, the state echo network comprising: an input ESN, a chaotic ESN, and a feedback classifier, the apparatus comprising:

[0071] The first acquisition module is used to acquire attractor sample data, which includes the location information of the monitoring trajectory set in multiple regions of interest within the monitorable area of ​​the surveillance camera;

[0072] The first training module is used to pre-train the connection matrix of the input ESN using the attractor sample data to obtain the pre-trained input ESN.

[0073] The second acquisition module is used to input the attractor sample data into the pre-trained input ESN to obtain the state of the nodes of the pre-trained input ESN;

[0074] The second training module is used to train the connection matrix of the chaotic ESN using the state of the nodes output by the pre-trained input ESN, so as to obtain the trained chaotic ESN.

[0075] The third training module is used to train the connection matrix of the feedback classifier using the state of the nodes output by the pre-trained input ESN, so as to obtain the trained feedback classifier, which is used to realize the autonomous transformation between the regions of interest.

[0076] Fourthly, embodiments of the present invention provide a monitoring trajectory determination device, comprising:

[0077] The acquisition module is used to acquire the current monitoring location of the surveillance camera;

[0078] The trajectory determination module is used to input the current monitoring position into the state echo network to obtain the monitoring position of the monitoring camera at the next moment, which is output by the state echo network. The state echo network includes: an input ESN, a chaotic ESN, and a feedback classifier; the feedback classifier is used to realize the autonomous conversion between regions of interest of the monitoring camera head.

[0079] The control module is used to control the rotation of the surveillance camera to obtain an image of the monitoring position at the next moment.

[0080] Fifthly, embodiments of the present invention provide an electronic device, including: a processor, a memory, and a program stored in the memory and executable on the processor. When the program is executed by the processor, it implements the steps of the training method for the state echo network as described in the first aspect above, or, when the program is executed by the processor, it implements the steps of the monitoring trajectory determination method as described in the second aspect above.

[0081] In a sixth aspect, embodiments of the present invention provide a computer-readable storage medium, characterized in that the computer-readable storage medium stores a computer program, which, when executed by a processor, implements the steps of the training method for the state echo network as described in the first aspect above; or, when executed by a processor, the computer program implements the steps of the monitoring trajectory determination method as described in the second aspect above.

[0082] In this embodiment of the invention, the location information of the monitoring trajectory set in the region of interest within the monitorable area is used as attractor sample data to train a state echo network based on chaotic travel. Because chaos is highly sensitive to initial conditions, the monitoring trajectory planned by the trained state echo network has random and unpredictable characteristics. Furthermore, due to the presence of attractors, the probability of the monitoring trajectory planned by the trained state echo network appearing near the attractors is high. Applying this state echo network to the field of security monitoring maintains the unplanned, random, and unpredictable characteristics of chaos while enabling a higher scanning frequency of the region of interest, making it impossible for criminals to effectively predict the monitoring trajectory and improving the security of the security monitoring system. Attached Figure Description

[0083] Various other advantages and benefits will become apparent to those skilled in the art upon reading the following detailed description of preferred embodiments. The accompanying drawings are for illustrative purposes only and are not intended to limit the invention. Furthermore, the same reference numerals denote the same parts throughout the drawings. In the drawings:

[0084] Figure 1 This is a flowchart illustrating the training method of the state echo network according to an embodiment of the present invention.

[0085] Figure 2 This is a schematic diagram of the state echo network according to an embodiment of the present invention;

[0086] Figure 3 This is a schematic diagram of a method for acquiring an image of the entire monitorable area according to an embodiment of the present invention;

[0087] Figure 4 This is a schematic diagram of a Markov matrix according to an embodiment of the present invention;

[0088] Figure 5 This is a flowchart illustrating the monitoring trajectory determination method according to an embodiment of the present invention;

[0089] Figure 6 This is a schematic diagram of the structure of a training device for a state echo network according to an embodiment of the present invention;

[0090] Figure 7 This is a schematic diagram of the monitoring trajectory determination device according to an embodiment of the present invention;

[0091] Figure 8 This is a schematic diagram of the structure of an electronic device according to an embodiment of the present invention. Detailed Implementation

[0092] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0093] The following is a brief introduction to the technical terms related to this invention.

[0094] Chaos refers to the random and irregular motion that exists within a deterministic system. Its behavior is characterized by uncertainty, non-repeatability, and unpredictability; this is the phenomenon of chaos. Chaos is an inherent characteristic of nonlinear dynamic systems and a ubiquitous phenomenon in nonlinear systems.

[0095] Chaotic tours are a frequently observed nonlinear phenomenon in high-dimensional dynamical systems, characterized by chaotic tour transitions between locally contracting domains (i.e., quasi-attractors).

[0096] Chaotic attractors: They have complex stretched and twisted structures. Attractors are the product of the combined effects of overall stability and local instability in chaotic systems. They have self-similarity and fractal structure.

[0097] State Echo Network (ESN): A type of recurrent neural network. An ESN typically consists of an input layer, a storage layer (also called a storage layer, hidden layer, or other hidden layers), and an output layer. The input layer's input and connection matrix w... in The product of the input layer's connection matrix and the storage layer's connection matrix is ​​input into the storage layer. The storage layer contains N nodes (also called neurons), each corresponding to a state, for example, denoted by x(t). The storage layer processes the input from the input layer to obtain its state vector. The output layer combines the state vector output from the storage layer with the connection matrix w. out Multiply the (connection matrix from storage layer to output layer) and output the product.

[0098] Please refer to Figure 1 This invention provides a training method for a state echo network. Please refer to [link / reference]. Figure 2 The state echo network includes an input ESN, a chaotic ESN, and a feedback classifier (not shown in the figure). The input ESN and the chaotic ESN form the storage layer of the state echo network, and the input ESN serves as the interface between the input layer and the chaotic ESN. Both the input ESN and the chaotic ESN contain multiple nodes.

[0099] The training method for the state echo network includes:

[0100] Step 11: Obtain attractor sample data, which includes the location information of the monitoring trajectory set in multiple regions of interest (ROI) within the monitorable area of ​​the surveillance camera;

[0101] Step 12: Use the attractor sample data to pre-train the connection matrix of the input ESN to obtain the pre-trained input ESN;

[0102] Step 13: Input the attractor sample data into the pre-trained input ESN to obtain the state of the nodes of the pre-trained input ESN;

[0103] Step 14: Use the states of the nodes output by the pre-trained input ESN to train the connection matrix of the chaotic ESN to obtain the trained chaotic ESN.

[0104] Step 15: Use the state of the nodes output by the pre-trained input ESN to train the connection matrix of the feedback classifier to obtain the trained feedback classifier, which is used to realize the autonomous transformation between the regions of interest.

[0105] The pre-trained input ESN, the trained chaotic ESN, and the trained feedback classifier constitute the trained state echo network.

[0106] In this embodiment of the invention, the location information of the monitoring trajectory set in the region of interest within the monitorable area is used as attractor sample data to train a state echo network based on chaotic travel. Because chaos is highly sensitive to initial conditions, the monitoring trajectory planned by the trained state echo network has random and unpredictable characteristics. Furthermore, due to the presence of attractors, the probability of the monitoring trajectory planned by the trained state echo network appearing near the attractors is high. Applying this state echo network to the field of security monitoring maintains the unplanned, random, and unpredictable characteristics of chaos while enabling a higher scanning frequency of the region of interest, making it impossible for criminals to effectively predict the monitoring trajectory and improving the security of the security monitoring system.

[0107] Optionally, in this embodiment of the invention, obtaining attractor sample data includes:

[0108] Step 111: Control the surveillance camera to collect images of the monitorable area using a traversal method;

[0109] Optionally, the monitorable area of ​​the surveillance camera is determined based on the movable range of the surveillance camera, and images of the monitorable area are collected sequentially using a traversal method.

[0110] Step 112: Stitch together all the collected images of the monitorable area to obtain the overall monitorable area image; please refer to... Figure 3 .

[0111] Step 113: Determine the region of interest in the overall monitorable area image;

[0112] In this embodiment of the invention, the region of interest can be selected manually.

[0113] Step 114: Obtain the monitoring trajectory set within the area of ​​interest;

[0114] In this embodiment of the invention, the monitoring trajectory within the area of ​​interest can be set manually or automatically and randomly.

[0115] In this embodiment of the invention, different types of monitoring trajectories can be set according to the size of the region of interest.

[0116] The monitoring trajectory can be circular, square, or an irregular line (such as the Lissajous curve), etc.

[0117] Step 115: Use the location information of the monitoring trajectory set in the region of interest as the attractor sample data.

[0118] In this embodiment of the invention, optionally, the step of pre-training the connection matrix of the input ESN using the attractor sample data to obtain the pre-trained input ESN includes:

[0119] Step 121: Input the attractor sample data into the input ESN to obtain the state of the node of the input ESN output by the input ESN;

[0120] Step 122: Determine the first loss function based on the state of the node output by the input ESN and the pre-set target state of the node of the input ESN;

[0121] Step 123: Optimize the connection matrix of the input ESN using the first loss function.

[0122] In this embodiment of the invention, optionally, the nodes participating in the training of the input ESN are half of all the nodes of the input ESN, so as to reduce the redundancy of the connection matrix of the input ESN. Of course, in some other embodiments of the invention, the nodes participating in the training of the input ESN may also be all the nodes of the input ESN.

[0123] In this embodiment of the invention, optionally, the input ESN (having N) in (a number of nodes) is represented as:

[0124]

[0125] Where t is time, τ is a scaling factor that affects the update rate, and x in For the state of the nodes in the input ESN, g in For scaling parameters (i.e., for the connection matrix J) in Scaling (i.e., normalization) is performed, J in J is the connection matrix of the input ESN. in The elements of u follow a normal distribution. in (s) represents the attractor sample data. tanh() is the hyperbolic tangent function, an activation function.

[0126] In this embodiment of the invention, optionally, the first loss function is:

[0127]

[0128] Among them, C 1-in Let x be the first loss function, t be time, and x be the loss function. s (t) represents the state of the node that inputs the ESN at time t. Let L be the target state of the node that inputs the ESN at time t. innate Let be the maximum value at time t, s be the monitoring trajectory set within the region of interest, and S be the set of monitoring trajectories set within the region of interest.

[0129] Of course, in other embodiments of the present invention, the first loss function may also be other loss functions, and the present invention is not limited thereto.

[0130] In this embodiment of the invention, optionally, the connection matrix of the chaotic ESN is trained using the states of the nodes output by the pre-trained input ESN to obtain the trained chaotic ESN, including:

[0131] Step 141: Input the state of the node output by the pre-trained input ESN into the chaotic ESN to obtain the monitoring position of the monitoring camera output by the chaotic ESN.

[0132] Step 142: Determine the second loss function based on the monitoring position of the surveillance camera output by the chaotic ESN and the specified monitoring position;

[0133] Step 143: Optimize the connection matrix of the chaotic ESN using the second loss function.

[0134] In this embodiment of the invention, optionally, the chaotic ESN (having N) ch (a number of nodes) is represented as:

[0135]

[0136] Where t is time, τ is the proportionality coefficient, and x ch For the state of a node in the chaotic ESN, g ch For scaling parameters (i.e., for the connection matrix J) ch Scaling (i.e., normalization) is performed, J ch Let J be the connection matrix of the chaotic ESN. ic Let x be the connection matrix between the input ESN and the chaotic ESN. in The state of the nodes in the input ESN.

[0137] In some embodiments of the present invention, τ = 10.0, g can be used. in =0.9g ch =1.5 makes the input ESN non-chaotic, and the chaotic ESN chaotic.

[0138] In this embodiment of the invention, optionally, the second loss function is:

[0139]

[0140] Among them, C 1-out Let x be the second loss function, t be time, and x be the value of x. s (t) includes the state of the node that inputs the ESN and the state of the node that outputs the ESN at time t, x s (t)=[x in ;x ch ], f s (t) represents the specified monitoring location at time t, L out w is the maximum value at time t. out Let be the connection matrix between the chaotic ESN and the output layer, s be the monitoring trajectory set within the region of interest, and S be the set of monitoring trajectories set within the region of interest.

[0141] Of course, in other embodiments of the present invention, the second loss function may also be other loss functions, and the present invention is not limited thereto.

[0142] In this embodiment of the invention, the training process of the input ESN and / or output ESN can be performed using the first-order gradient error method (First-order–reduced and controlled-error learning). This training method does not involve matrix inversion and a large number of convolution calculations, thus optimizing the computation method and improving the computation speed.

[0143] In this embodiment of the invention, to prevent the chaotic ESN from becoming non-chaotic, the link between the input ESN and the chaotic ESN is trained to converge to a transiently stable state of 0 during the above training process.

[0144] To enable the itinerant transitions between attractors, a feedback loop is set between the storage layer and the input layer. This involves adding a feedback classifier to the trained state echo network (which feeds back the actual output around the set attractor trajectory to the input loop to achieve state transitions; for example, the output of the storage layer is from state A to state B, then to state C, and the feedback loop allows the A, B, C sequences to be executed) to autonomously generate specific symbol dynamics.

[0145] In some embodiments of the present invention, optionally, the connection matrix of the feedback classifier is trained using the states of the nodes output by the pre-trained input ESN to obtain the trained feedback classifier, including:

[0146] Step 151: Input the state of the node output by the pre-trained input ESN into the feedback classifier to obtain the monitoring position of the surveillance camera at the next moment output by the feedback classifier.

[0147] Step 152: Determine the third loss function based on the monitoring position of the surveillance camera at the next moment output by the feedback classifier and the preset monitoring position of the surveillance camera at the next moment;

[0148] Step 153: Optimize the connection matrix of the feedback classifier according to the third loss function.

[0149] In this embodiment of the invention, optionally, the feedback classifier is represented as:

[0150]

[0151] Among them, f max Let x(t) be the feedback classifier, t be time, and x(t) include the state of the node input to the ESN and the state of the node outputting the ESN at time t, x(t) = [x in ;x ch ], w is the connection matrix of the feedback classifier, s is the monitoring trajectory set in the region of interest, and S is the set of monitoring trajectories set in the region of interest.

[0152] In this embodiment of the invention, optionally, the third loss function is:

[0153]

[0154] Where C2 is the third loss function, t is time, and s is the time interval. sper(t+Δt) represents the monitoring position of the surveillance camera at the next moment, as output by the feedback classifier; x(t) represents the state of the node of the input ESN at time t; w is the connection matrix of the feedback classifier; s is the monitoring trajectory set within the region of interest; S is the set of monitoring trajectories set within the region of interest; T rec This represents the maximum value for time.

[0155] In the formula, 1{s sper The meaning of (t+Δt)=s} is: when s sper (t+Δt) = s, which is 1 in this case and 0 otherwise.

[0156] In some other embodiments of the present invention, optionally, the connection matrix of the feedback classifier is trained using the states of the nodes output by the pre-trained input ESN to obtain the trained feedback classifier, including:

[0157] Step 151': Input the state of the node output by the pre-trained input ESN and the pre-set transition probability between regions of interest into the feedback classifier to obtain the monitoring position of the surveillance camera at the next moment output by the feedback classifier.

[0158] Step 152': Determine the fourth loss function based on the monitoring position of the surveillance camera at the next moment output by the feedback classifier and the preset monitoring position of the surveillance camera at the next moment;

[0159] Step 153': Optimize the connection matrix of the feedback classifier according to the fourth loss function.

[0160] In this embodiment of the invention, optionally, the feedback classifier is represented as:

[0161]

[0162] Among them, f max Let x(t) be the feedback classifier, t be time, and x(t) include the state of the node input to the ESN and the state of the node outputting the ESN at time t, x(t) = [x in ;x ch ], w is the connection matrix of the feedback classifier, s is the monitoring trajectory set in the region of interest, and S is the set of monitoring trajectories set in the region of interest.

[0163] In this embodiment of the invention, optionally, the fourth loss function is:

[0164]

[0165] Where C3 is the fourth loss function, t is time, and s is the time interval.sto (t+Δt) represents the monitoring position of the surveillance camera at the next moment, output by the feedback classifier based on the pre-set transition probabilities between regions of interest; x(t) represents the state of the node of the input ESN at time t; w is the connection matrix of the feedback classifier; s is the monitoring trajectory set within the region of interest; S is the set of monitoring trajectories set within the region of interest; T rec The maximum value for time is used, and the conversion probability is determined based on the importance of the region of interest.

[0166] In the formula for the fourth loss function mentioned above, 1{s sto The meaning of (t+Δt)=s} is: when s sto (t+Δt) = s, which is 1; otherwise, it is 0.

[0167] In this embodiment of the invention, the conversion probability between regions of interest includes: the conversion probability between different regions of interest and the conversion probability between regions of interest themselves.

[0168] In this embodiment of the invention, the conversion probability between regions of interest can be determined based on the importance of the regions of interest. Please refer to... Figure 4 Assuming the importance of ROI1 is 20, the importance of ROI2 is 50, and the importance of ROI3 is 30, the conversion probabilities between the various regions of interest are as follows: Figure 4 As shown, these transition probabilities form a Markov matrix. In this embodiment of the invention, the importance of a region of interest can be determined according to its danger level. For example, if a region of interest has a large number of people, then the attention given to that region needs to be increased, the danger level is higher, and the probability of other regions of interest switching to that region will increase.

[0169] Please refer to Figure 5 This invention also provides a method for determining a monitoring trajectory, comprising:

[0170] Step 51: Obtain the current monitoring location of the surveillance camera;

[0171] Step 52: Input the current monitoring location into the state echo network to obtain the monitoring location of the monitoring camera at the next moment, which is output by the state echo network. The state echo network includes: input ESN, chaotic ESN and feedback classifier; the feedback classifier is used to realize the autonomous conversion between regions of interest of the monitoring camera.

[0172] Step 53: Control the rotation of the surveillance camera to obtain an image of the monitoring position at the next moment.

[0173] In this embodiment of the invention, optionally, the input ESN is represented as:

[0174]

[0175] Where t is time, τ is the proportionality coefficient, and x in For the state of the nodes in the input ESN, g in J is the scaling parameter. in Let u be the connection matrix of the input ESN. in (s(t)) represents the monitoring location at time t.

[0176] In this embodiment of the invention, optionally, the chaotic ESN is represented as:

[0177]

[0178] Where t is time, τ is the proportionality coefficient, and x ch For the state of a node in the chaotic ESN, g ch J is the scaling parameter. ch Let J be the connection matrix of the chaotic ESN. ic Let x be the connection matrix between the input ESN and the chaotic ESN. in The state of the nodes in the input ESN.

[0179] In this embodiment of the invention, optionally, the feedback classifier is represented as:

[0180]

[0181] Among them, f max Let x(t) be the feedback classifier, t be time, and x(t) include the state of the node input to the ESN and the state of the node outputting the ESN at time t, x(t) = [x in ;x ch ], w is the connection matrix of the feedback classifier, s is the monitoring trajectory set in the region of interest, and S is the set of monitoring trajectories set in the region of interest.

[0182] Optionally, in this embodiment of the invention, inputting the current monitoring location into the state echo network includes: inputting the conversion probability between the current monitoring location and the region of interest of the monitoring camera into the state echo network.

[0183] Please refer to Figure 6 This invention also provides a training device 60 for a state echo network, the state echo network including: an input ESN, a chaotic ESN, and a feedback classifier, the device 60 including:

[0184] The first acquisition module 61 is used to acquire attractor sample data, which includes the location information of the monitoring trajectory set in multiple regions of interest within the monitorable area of ​​the monitoring camera.

[0185] The first training module 62 is used to pre-train the connection matrix of the input ESN using the attractor sample data to obtain the pre-trained input ESN.

[0186] The second acquisition module 63 is used to input the attractor sample data into the pre-trained input ESN to obtain the state of the nodes of the pre-trained input ESN;

[0187] The second training module 64 is used to train the connection matrix of the chaotic ESN using the state of the nodes output by the pre-trained input ESN, so as to obtain the trained chaotic ESN.

[0188] The third training module 65 is used to train the connection matrix of the feedback classifier using the state of the nodes output by the pre-trained input ESN, so as to obtain the trained feedback classifier, which is used to realize the autonomous transformation between the regions of interest.

[0189] In this embodiment of the invention, optionally, the first acquisition module 61 is configured to control the monitoring camera to acquire images of the monitorable area using a traversal method; stitch together all the acquired images of the monitorable area to obtain an overall monitorable area image; determine the region of interest in the overall monitorable area image; acquire the monitoring trajectory set within the region of interest; and use the location information of the monitoring trajectory set within the region of interest as the attractor sample data.

[0190] In this embodiment of the invention, optionally, the first training module 62 is used to input the attractor sample data into the input ESN to obtain the state of the nodes of the input ESN output by the input ESN; determine a first loss function based on the state of the nodes output by the input ESN and the preset target state of the nodes of the input ESN; and optimize the connection matrix of the input ESN using the first loss function.

[0191] In this embodiment of the invention, optionally, the number of nodes participating in the training of the input ESN is half of the total number of nodes in the input ESN.

[0192] In this embodiment of the invention, optionally, the input ESN is represented as:

[0193]

[0194] Where t is time, τ is the proportionality coefficient, and x inFor the state of the nodes in the input ESN, g in J is the scaling parameter. in Let u be the connection matrix of the input ESN. in (s) represents the attractor sample data.

[0195] In this embodiment of the invention, optionally, the first loss function is:

[0196]

[0197] Among them, C 1-in Let x be the first loss function, t be time, and x be the loss function. s (t) represents the state of the node that inputs the ESN at time t. Let L be the target state of the node that inputs the ESN at time t. innate Let be the maximum value at time t, s be the monitoring trajectory set within the region of interest, and S be the set of monitoring trajectories set within the region of interest.

[0198] In this embodiment of the invention, optionally, the second training module 64 is used to input the state of the nodes output by the pre-trained input ESN into the chaotic ESN to obtain the monitoring position of the monitoring camera output by the chaotic ESN; determine a second loss function based on the monitoring position of the monitoring camera output by the chaotic ESN and the specified monitoring position; and optimize the connection matrix of the chaotic ESN using the second loss function.

[0199] In this embodiment of the invention, optionally, the chaotic ESN is represented as:

[0200]

[0201] Where t is time, τ is the proportionality coefficient, and x ch For the state of a node in the chaotic ESN, g ch J is the scaling parameter. ch Let J be the connection matrix of the chaotic ESN. in Let x be the connection matrix between the input ESN and the chaotic ESN. in The state of the nodes in the input ESN.

[0202] In this embodiment of the invention, optionally, the second loss function is:

[0203]

[0204] Among them, C 1-out Let x be the second loss function, t be time, and x be the value of x. s(t) includes the state of the node that inputs the ESN and the state of the node that outputs the ESN at time t, x s (t)=[x in ;x ch ], f s (t) represents the specified monitoring location at time t, L out w is the maximum value at time t. out Let be the connection matrix between the chaotic ESN and the output layer, s be the monitoring trajectory set within the region of interest, and S be the set of monitoring trajectories set within the region of interest.

[0205] In this embodiment of the invention, optionally, the third training module 65 is used to input the state of the nodes output by the pre-trained input ESN into the feedback classifier to obtain the monitoring position of the surveillance camera at the next moment output by the feedback classifier; determine a third loss function based on the monitoring position of the surveillance camera at the next moment output by the feedback classifier and a preset monitoring position of the surveillance camera at the next moment; and optimize the connection matrix of the feedback classifier based on the third loss function.

[0206] In this embodiment of the invention, optionally, the third training module 65 is used to input the state of the nodes output by the pre-trained input ESN and the pre-set transition probabilities between regions of interest into the feedback classifier to obtain the monitoring position of the surveillance camera at the next moment output by the feedback classifier; determine a fourth loss function based on the monitoring position of the surveillance camera at the next moment output by the feedback classifier and the preset monitoring position of the surveillance camera at the next moment; and optimize the connection matrix of the feedback classifier based on the fourth loss function.

[0207] In this embodiment of the invention, optionally, the feedback classifier is represented as:

[0208]

[0209] Among them, f max Let x(t) be the feedback classifier, t be time, and x(t) include the state of the node input to the ESN and the state of the node outputting the ESN at time t, x(t) = [x in ;x ch ], w is the connection matrix of the feedback classifier, s is the monitoring trajectory set in the region of interest, and S is the set of monitoring trajectories set in the region of interest.

[0210] In this embodiment of the invention, optionally, the third loss function is:

[0211]

[0212] Where C2 is the third loss function, t is time, and s is the time interval. sper (t+Δt) represents the monitoring position of the surveillance camera at the next moment, as output by the feedback classifier; x(t) represents the state of the node of the input ESN at time t; w is the connection matrix of the feedback classifier; s is the monitoring trajectory set within the region of interest; S is the set of monitoring trajectories set within the region of interest; T rec This represents the maximum value for time.

[0213] In the formula for the third loss function mentioned above, 1{s sper The meaning of (t+Δt)=s} is: when s sper (t+Δt) = s, which is 1 in this case and 0 otherwise.

[0214] In this embodiment of the invention, optionally, the fourth loss function is:

[0215]

[0216] Where C3 is the fourth loss function, t is time, and s is the time interval. sto (t+Δt) represents the monitoring position of the surveillance camera at the next moment, output by the feedback classifier based on the pre-set transition probabilities between regions of interest; x(t) represents the state of the node of the input ESN at time t; w is the connection matrix of the feedback classifier; s is the monitoring trajectory set within the region of interest; S is the set of monitoring trajectories set within the region of interest; T rec The maximum value for time is used, and the conversion probability is determined based on the importance of the region of interest.

[0217] In the formula for the fourth loss function mentioned above, 1{s sto The meaning of (t+Δt)=s} is: when s sto (t+Δt) = s, which is 1; otherwise, it is 0.

[0218] Please refer to Figure 7 This invention also provides a monitoring trajectory determination device 70, comprising:

[0219] The acquisition module 71 is used to acquire the current monitoring location of the surveillance camera;

[0220] The trajectory determination module 72 is used to input the current monitoring position into the state echo network to obtain the monitoring position of the monitoring camera at the next moment output by the state echo network. The state echo network includes: an input ESN, a chaotic ESN, and a feedback classifier; the feedback classifier is used to realize the autonomous conversion between the regions of interest of the monitoring camera head.

[0221] The control module 73 is used to control the rotation of the monitoring camera to obtain an image of the monitoring position at the next moment.

[0222] Optionally, the input ESN is represented as:

[0223]

[0224] Where t is time, τ is the proportionality coefficient, and x in For the state of the nodes in the input ESN, g in J is the scaling parameter. in Let u be the connection matrix of the input ESN. in (s(t)) represents the monitoring location at time t.

[0225] Optionally, the chaotic ESN is represented as:

[0226]

[0227] Where t is time, τ is the proportionality coefficient, and x ch For the state of a node in the chaotic ESN, g ch J is the scaling parameter. ch Let J be the connection matrix of the chaotic ESN. ic Let x be the connection matrix between the input ESN and the chaotic ESN. in The state of the nodes in the input ESN.

[0228] Optionally, the feedback classifier is represented as:

[0229]

[0230] Among them, f max Let x(t) be the feedback classifier, t be time, and x(t) include the state of the node input to the ESN and the state of the node outputting the ESN at time t, x(t) = [x in ;x ch ], w is the connection matrix of the feedback classifier, s is the monitoring trajectory set in the region of interest, and S is the set of monitoring trajectories set in the region of interest.

[0231] Optionally, the trajectory determination module 72 is used to input the conversion probability between the current monitoring location and the region of interest of the monitoring camera into the state echo network.

[0232] Please refer to Figure 8The present invention also provides an electronic device 80, including a processor 81, a memory 82, and a computer program stored in the memory 82 and executable on the processor 81. When the computer program is executed by the processor 81, it implements the various processes of the above-described training method or monitoring trajectory determination method of the state echo network and achieves the same technical effect. To avoid repetition, it will not be described again here.

[0233] This invention also provides a computer-readable storage medium storing a computer program. When executed by a processor, the computer program implements the various processes of the above-described training method or monitoring trajectory determination method for state echo networks, achieving the same technical effects. To avoid repetition, these details are not repeated here. The computer-readable storage medium may be a read-only memory (ROM), a random access memory (RAM), a magnetic disk, or an optical disk.

[0234] It should be noted that, in this document, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Unless otherwise specified, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or apparatus that includes that element.

[0235] Through the above description of the embodiments, those skilled in the art can clearly understand that the methods of the above embodiments can be implemented by means of software plus necessary general-purpose hardware platforms. Of course, they can also be implemented by hardware, but in many cases the former is a better implementation method. Based on this understanding, the technical solution of the present invention, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a storage medium (such as ROM / RAM, magnetic disk, optical disk), and includes several instructions to cause a terminal (which may be a mobile phone, computer, server, air conditioner, or network device, etc.) to execute the methods described in the various embodiments of the present invention.

[0236] The embodiments of the present invention have been described above with reference to the accompanying drawings. However, the present invention is not limited to the specific embodiments described above. The specific embodiments described above are merely illustrative and not restrictive. Those skilled in the art can make many other forms under the guidance of the present invention without departing from the spirit and scope of the claims, and all of these forms are within the protection scope of the present invention.

Claims

1. A training method for a state echo network, characterized in that, The state echo network includes: an input ESN, a chaotic ESN, and a feedback classifier. The input ESN and the chaotic ESN constitute the storage layer of the state echo network. The input ESN is the interface between the input layer of the state echo network and the chaotic ESN. The input ESN and the chaotic ESN each include a connection matrix and multiple nodes. The method includes: Acquire attractor sample data, which includes location information of monitoring trajectories set in multiple regions of interest within the monitorable area of ​​the surveillance camera; The connection matrix of the input ESN is pre-trained using the attractor sample data to obtain the pre-trained input ESN. The attractor sample data is input into the pre-trained input ESN to obtain the state of the nodes of the pre-trained input ESN; The connection matrix of the chaotic ESN is trained using the state of the nodes output by the pre-trained input ESN to obtain the trained chaotic ESN. The connection matrix of the feedback classifier is trained using the state of the nodes output by the pre-trained input ESN to obtain the trained feedback classifier, which is used to realize the autonomous transformation between the regions of interest.

2. The method according to claim 1, characterized in that, The acquisition of attractor sample data includes: The surveillance camera is controlled to collect images of the monitored area using a traversal method. The images of all the monitored areas collected are stitched together to obtain the overall monitored area image; Identify the region of interest in the overall monitorable area image; Obtain the monitoring trajectory set within the area of ​​interest; The location information of the monitoring trajectory set within the region of interest is used as the attractor sample data.

3. The method according to claim 1, characterized in that, The step of pre-training the connection matrix of the input ESN using the attractor sample data to obtain the pre-trained input ESN includes: The attractor sample data is input into the input ESN to obtain the state of the node of the input ESN output by the input ESN; The first loss function is determined based on the state of the node output by the input ESN and the pre-set target state of the node of the input ESN; The connection matrix of the input ESN is optimized using the first loss function.

4. The method according to claim 1 or 3, characterized in that, The number of nodes participating in the training of the input ESN is half of the total number of nodes in the input ESN.

5. The method according to claim 1, characterized in that, The input ESN is represented as: ； Where t is time, This is the proportionality coefficient. The state of the nodes in the input ESN. For scaling parameters, The connection matrix of the input ESN. The attractor sample data.

6. The method according to claim 3, characterized in that, The first loss function is: ； in, Let be the first loss function, and t be time. Let t be the state of the node that inputs the ESN. Let t be the target state of the node that inputs the ESN. Let be the maximum value at time t, s be the monitoring trajectory set within the region of interest, and S be the set of monitoring trajectories set within the region of interest.

7. The method according to claim 1, characterized in that, The connection matrix of the chaotic ESN is trained using the states of the nodes output from the pre-trained input ESN to obtain the trained chaotic ESN, including: The state of the node output by the pre-trained input ESN is input into the chaotic ESN to obtain the monitoring position of the surveillance camera output by the chaotic ESN. The second loss function is determined based on the monitoring location of the surveillance camera output by the chaotic ESN and the specified monitoring location; The connection matrix of the chaotic ESN is optimized using the second loss function.

8. The method according to claim 1 or 7, characterized in that, The chaotic ESN is represented as: ； Where t is time, This is the proportionality coefficient. The state of a node in the chaotic ESN. For scaling parameters, Let be the connection matrix of the chaotic ESN. The connection matrix between the input ESN and the chaotic ESN. The state of the nodes in the input ESN.

9. The method according to claim 7, characterized in that, The second loss function is: ； in, Let t be the second loss function, and t be time. This includes the state of the node that inputs the ESN and the state of the node that outputs the ESN at time t. =[ ; ], To specify the monitoring location at time t, The maximum value at time t. Let be the connection matrix between the chaotic ESN and the output layer, s be the monitoring trajectory set within the region of interest, and S be the set of monitoring trajectories set within the region of interest.

10. The method according to claim 1, characterized in that, The connection matrix of the feedback classifier is trained using the states of the nodes output by the pre-trained input ESN to obtain the trained feedback classifier, which includes: The state of the node output by the pre-trained input ESN is input into the feedback classifier to obtain the monitoring position of the surveillance camera at the next moment, as output by the feedback classifier. The third loss function is determined based on the monitoring position of the surveillance camera at the next moment output by the feedback classifier and the preset monitoring position of the surveillance camera at the next moment. The connection matrix of the feedback classifier is optimized based on the third loss function.

11. The method according to claim 1, characterized in that, The connection matrix of the feedback classifier is trained using the states of the nodes output by the pre-trained input ESN to obtain the trained feedback classifier, which includes: The state of the nodes output by the pre-trained input ESN and the transition probability between the pre-set regions of interest are input into the feedback classifier to obtain the monitoring position of the surveillance camera at the next moment, as output by the feedback classifier. The fourth loss function is determined based on the monitoring position of the surveillance camera at the next moment output by the feedback classifier and the preset monitoring position of the surveillance camera at the next moment. The connection matrix of the feedback classifier is optimized based on the fourth loss function.

12. The method according to any one of claims 1, 10, or 11, characterized in that, The feedback classifier is represented as: ； in, Let t be the feedback classifier, and t be time. This includes the state of the node that inputs the ESN and the state of the node that outputs the ESN at time t. =[ ; ], w is the connection matrix of the feedback classifier, s is the monitoring trajectory set in the region of interest, and S is the set of monitoring trajectories set in the region of interest.

13. The method according to claim 10, characterized in that, The third loss function is: ； in, Let t be the third loss function, and t be time. The next moment's monitoring position of the surveillance camera is output by the feedback classifier. Let be the state of the node input to the ESN at time t, w be the connection matrix of the feedback classifier, s be the monitoring trajectory set within the region of interest, and S be the set of monitoring trajectories set within the region of interest. This represents the maximum value for time.

14. The method according to claim 11, characterized in that, The fourth loss function is: ； in, Let t be the fourth loss function, and t be time. The monitoring position of the surveillance camera at the next moment is output by the feedback classifier based on the pre-set transition probabilities between regions of interest. Let be the state of the node input to the ESN at time t, w be the connection matrix of the feedback classifier, s be the monitoring trajectory set within the region of interest, and S be the set of monitoring trajectories set within the region of interest. The maximum value for time is used, and the conversion probability is determined based on the importance of the region of interest.

15. A method for determining a monitoring trajectory, characterized in that, include: Obtain the current monitoring location of the surveillance camera; The current monitoring location is input into the state echo network to obtain the next monitoring location of the surveillance camera, which is output by the state echo network. The state echo network includes an input ESN, a chaotic ESN, and a feedback classifier. The input ESN and the chaotic ESN form the storage layer of the state echo network. The input ESN is the interface between the input layer of the state echo network and the chaotic ESN. The input ESN and the chaotic ESN each include a connection matrix and multiple nodes. The feedback classifier is used to realize the autonomous conversion between regions of interest of the surveillance camera. Control the rotation of the surveillance camera to obtain an image of the monitored location at the next moment.

16. The method according to claim 15, characterized in that, The input ESN is represented as: ； Where t is time, This is the proportionality coefficient. The state of the nodes in the input ESN. For scaling parameters, The connection matrix of the input ESN. Let t be the monitoring location at time t.

17. The method according to claim 15, characterized in that, The chaotic ESN is represented as: ； Where t is time, This is the proportionality coefficient. The state of a node in the chaotic ESN. For scaling parameters, Let be the connection matrix of the chaotic ESN. The connection matrix between the input ESN and the chaotic ESN. The state of the nodes in the input ESN.

18. The method according to claim 15, characterized in that, The feedback classifier is represented as: ； in, Let t be the feedback classifier, and t be time. This includes the state of the node that inputs the ESN and the state of the node that outputs the ESN at time t. =[ ; ], w is the connection matrix of the feedback classifier, s is the monitoring trajectory set in the region of interest, and S is the set of monitoring trajectories set in the region of interest.

19. The method according to claim 15, characterized in that, Inputting the current monitoring location into the status echo network includes: The conversion probability between the current monitoring location and the region of interest of the monitoring camera is input into the state echo network.

20. A training device for a state echo network, characterized in that, The state echo network includes: an input ESN, a chaotic ESN, and a feedback classifier. The input ESN and the chaotic ESN constitute the storage layer of the state echo network. The input ESN is the interface between the input layer of the state echo network and the chaotic ESN. The input ESN and the chaotic ESN each include a connection matrix and multiple nodes. The device includes: The first acquisition module is used to acquire attractor sample data, which includes the location information of the monitoring trajectory set in multiple regions of interest within the monitorable area of ​​the surveillance camera; The first training module is used to pre-train the connection matrix of the input ESN using the attractor sample data to obtain the pre-trained input ESN. The second acquisition module is used to input the attractor sample data into the pre-trained input ESN to obtain the state of the nodes of the pre-trained input ESN; The second training module is used to train the connection matrix of the chaotic ESN using the state of the nodes output by the pre-trained input ESN, so as to obtain the trained chaotic ESN. The third training module is used to train the connection matrix of the feedback classifier using the state of the nodes output by the pre-trained input ESN, so as to obtain the trained feedback classifier, which is used to realize the autonomous transformation between the regions of interest.

21. A monitoring trajectory determination device, characterized in that, include: The acquisition module is used to acquire the current monitoring location of the surveillance camera; The trajectory determination module is used to input the current monitoring location into the state echo network to obtain the monitoring location of the monitoring camera at the next moment, output by the state echo network. The state echo network includes: an input ESN, a chaotic ESN, and a feedback classifier; the input ESN and the chaotic ESN form the storage layer of the state echo network, and the input ESN is the interface between the input layer of the state echo network and the chaotic ESN. The input ESN and the chaotic ESN each include a connection matrix and multiple nodes; the feedback classifier is used to realize the autonomous conversion between the regions of interest of the monitoring camera. The control module is used to control the rotation of the surveillance camera to obtain an image of the monitoring position at the next moment.

22. An electronic device, characterized in that, include: A processor, a memory, and a program stored in the memory and executable on the processor, wherein the program, when executed by the processor, implements the steps of the training method for a state echo network as described in any one of claims 1 to 14, or, when executed by the processor, the program implements the steps of the monitoring trajectory determination method as described in any one of claims 15 to 19.

23. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores a computer program that, when executed by a processor, implements the steps of the training method for a state echo network as described in any one of claims 1 to 14; or, when executed by a processor, the computer program implements the steps of the monitoring trajectory determination method as described in any one of claims 15 to 19.

Images