Real-time generation method of UAV reconnaissance target status based on PHD filter
Through the grid map model based on PHD filter and random finite set modeling, the problem of data gap in the UAV reconnaissance system is solved, and real-time and accurate target state generation is achieved, providing data support for subsequent decision-making behaviors.
Patent Information
- Application Number
- CN202410740205.7
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-06-07
- Publication Date
- 2025-09-09
- Estimated Expiration
- 2044-06-07
AI Technical Summary
Existing technologies make it difficult to achieve real-time dynamic interaction between real drone reconnaissance and drone reconnaissance simulation systems, resulting in data gaps and information islands, and unable to effectively support the simulation and deduction of subsequent decision-making behaviors.
A method based on PHD filter is adopted. By establishing the association relationship between the grid map model and the UAV reconnaissance target, the target entity state is modeled and estimated using random finite sets, and the sensor measurement data is combined for prediction and update to generate the real-time UAV reconnaissance target state.
It improves the authenticity and accuracy of the generation of UAV reconnaissance target status, adapts to dynamic changes and clutter interference, and provides strong data support for the simulation and deduction of subsequent decision-making behaviors such as path planning.
Smart Images
Figure CN118761258B_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of UAV reconnaissance technology, and in particular to a method for real-time generation of UAV reconnaissance target status based on a PHD filter. Background Art
[0002] With the rapid development and widespread application of measurement technologies such as sensor networks, it has become easier to obtain real UAV reconnaissance measurement data, laying a good foundation for the virtual-real integration of real UAV reconnaissance and UAV reconnaissance simulation systems. Real-time dynamic interaction between the virtual and the real is a key means to eliminate the data gaps and information islands between real UAV reconnaissance and UAV reconnaissance simulation systems. It can solve the problems of data exchange difficulties and isolation between the two, and the inability to play the fundamental role of data. Based on the real-time generated current reconnaissance scene situation information, it can provide data support for the simulation and deduction of subsequent decision-making behaviors such as intention recognition and path planning, and support the formation of the optimal action plan to be provided to the command center in the real battlefield to assist in decision-making. This is of great significance for improving the accuracy of UAV reconnaissance simulation results and improving the efficiency of UAV reconnaissance. Summary of the Invention
[0003] Based on this, it is necessary to provide a real-time generation method for UAV reconnaissance target status based on PHD filter, which can generate target status in real time to address the above technical problems.
[0004] A method for real-time generation of UAV reconnaissance target status based on a PHD filter, the method comprising:
[0005] Obtain a drone reconnaissance area; model the drone reconnaissance area to obtain a grid map model; establish an association relationship between a drone reconnaissance target entity and one or more grids in the grid map model, where each grid corresponds to a drone reconnaissance target entity state point, and the state set of multiple drone reconnaissance target entity state points jointly describes the state of the drone reconnaissance target entity in the grid map model;
[0006] The UAV reconnaissance target entity state and sensor measurement data are modeled using random finite sets to obtain the UAV reconnaissance target entity state model and sensor measurement model;
[0007] The state of the entity is estimated in the grid map model based on the PHD filter. The persistent UAV reconnaissance target entity and the newly emerged UAV reconnaissance target entity are predicted using the UAV reconnaissance target entity state model and the sensor measurement model. The prediction results of the persistent UAV reconnaissance target entity state and the newly emerged UAV reconnaissance target entity state are obtained for each grid cell.
[0008] The pre-derived likelihood function is used to correct and update the prediction results of the UAV reconnaissance target entity state of each grid unit to obtain the updated UAV reconnaissance target entity state of each grid unit;
[0009] The updated UAV reconnaissance target entity states of all grid cells are summed to obtain the UAV reconnaissance target state generation model; the UAV reconnaissance target state generation model is solved according to the Gaussian sum approximation algorithm to obtain the UAV reconnaissance target state.
[0010] In one embodiment, the state and measurement data of the drone reconnaissance target entity are modeled according to a random finite set to obtain a drone reconnaissance target entity state model and a measurement model, including:
[0011] The UAV reconnaissance target entity state is modeled according to the random finite set, and the UAV reconnaissance target entity state model is obtained as follows:
[0012] x k+1 =f k+1|k (x k+1 |x k )x k +ξ k
[0013] Among them, f k+1|k (x k+1 |x k ) is the state transfer function, x k represents the target entity state at step k, ξ k is the zero-mean process noise of the Gaussian distribution.
[0014] In one embodiment, the sensor measurement data is modeled according to a random finite set, and the sensor measurement model is obtained as follows:
[0015]
[0016] Among them, ε represents the probability of unexpected conditions in the measurement, which include false alarms, missed detections, and clutter, and p(z k+1 ) represents the sensor measurement data z k+1 spatial distribution.
[0017] In one embodiment, the state of an entity is estimated in a grid map model based on a PHD filter, and a persistent UAV reconnaissance target entity and a newly emerged UAV reconnaissance target entity are predicted using a UAV reconnaissance target entity state model and a sensor measurement model, thereby obtaining a persistent UAV reconnaissance target entity state prediction result and a newly emerged UAV reconnaissance target entity state prediction result for each grid cell, including:
[0018] The state of the entity is estimated in the grid map model according to the PHD filter. The persistent UAV reconnaissance target entity state model and the sensor measurement model are used to predict the persistent UAV reconnaissance target entity. The prediction result of the persistent UAV reconnaissance target entity state of each grid unit is obtained as follows:
[0019]
[0020] Among them, c represents the grid unit number, k represents the time step number, X k+1 Indicates the state of the drone reconnaissance target entity at the k+1th step, x k+1 Represents the entity state model of the drone reconnaissance target, To predict the probability of the continued existence of the target entity in the drone reconnaissance, The probability density function for predicting the state of the persistent drone reconnaissance target entity.
[0021] In one embodiment, the state of the entity is estimated in the grid map model according to the PHD filter, and the newly generated drone reconnaissance target entity is predicted using the drone reconnaissance target entity state model and the sensor measurement model, and the newly generated drone reconnaissance target entity state prediction result for each grid unit is obtained as follows:
[0022]
[0023] in, It represents the probability of predicting the new generation of the target entity during the k+1-step drone reconnaissance.
[0024] In one embodiment, a likelihood function is used to describe the likelihood of measurement data corresponding to a target entity detected by a UAV. The derivation process of the likelihood function includes:
[0025] Define that there is a measurement in a certain grid cell c, and the probability of generating measurement data when there is a UAV reconnaissance target entity is: That is, the probability that the sensor measures the drone reconnaissance target entity; define the grid cell c where there is measurement but no drone reconnaissance target entity exists, that is, the probability of sensor false alarm is The definition is based on the state x in grid cell c k+1 The likelihood function of the UAV reconnaissance target entity generating measurement data is The association probability between the measurement and the UAV reconnaissance target entity is defined as It can indicate whether the measurement value near the grid cell c is related to the UAV reconnaissance target entity in the grid cell. The probability value decreases as the distance between the measurement and the grid cell increases. The density function of the clutter is defined as p cl (z), for different situations, the likelihood function has different forms, as follows:
[0026] when When there is no entity in the grid cell and no measurement occurs, the likelihood function is Among them, X k+1 Indicates the state of the drone reconnaissance target entity at the k+1th step, Z k+1 Represents the sensor measurement data at step k+1;
[0027] when X k+1 ={x k+1}, there is an entity in the grid cell, but no measurement occurs, indicating that a missed detection has occurred. At this time, the likelihood function is Among them, x k+1 Represents the entity state model of the drone reconnaissance target;
[0028] When Z k+1 ={z k+1}, When , there is no entity in the grid cell, but a measurement is generated, indicating that a false alarm has occurred and the sensor measures clutter. At this time, the likelihood function is
[0029]
[0030] When Z k+1 ={z k+1},X k+1 ={x k+1}, there is an entity in the grid cell and a measurement is generated. At this time, the sensor measures the entity or the clutter, and the likelihood function is
[0031]
[0032] In one embodiment, the predicted state of the drone reconnaissance target entity of each grid unit is corrected and updated using a pre-derived likelihood function to obtain the updated drone reconnaissance target entity state of each grid unit, including:
[0033] The pre-derived likelihood function is used to correct and update the prediction results of the UAV reconnaissance target entity state of each grid unit, and the updated UAV reconnaissance target entity state of each grid unit is obtained as
[0034]
[0035] in, represents the pre-derived likelihood function, represents the new UAV reconnaissance target entity state prediction result of each grid cell, X k+1It represents the state of the drone reconnaissance target entity at the k+1th step, and c represents the sequence number of the grid unit.
[0036] In one embodiment, the UAV reconnaissance target entity states of all grid cells are summed to obtain a UAV reconnaissance target state generation model, including:
[0037] The final UAV reconnaissance target state is obtained by summing up the UAV reconnaissance target entity states of all grid cells.
[0038]
[0039] Among them, c represents the sequence number of the grid unit, C represents the number of grid units, and x k+1 Represents the entity state model of the drone reconnaissance target, X k+1 Indicates the state of the drone reconnaissance target entity at the k+1th step.
[0040] The above-mentioned real-time generation method of UAV reconnaissance target state based on PHD filter, this application describes the state of UAV reconnaissance target entity in UAV reconnaissance environment by establishing the association relationship between grid map model and UAV reconnaissance target, and jointly describes the state of UAV reconnaissance target entity in grid map model with the state set of multiple UAV reconnaissance target entity state points, and uses random finite sets to represent the state of UAV reconnaissance target, which can realize accurate modeling of target entities of different shapes and dynamically changing entity measurements, which is more in line with the dynamic and random characteristics of UAV reconnaissance simulation, can improve the authenticity of entity state representation, adapt to the dynamic changes in the number of entities in real scenes, clutter interference, etc., which is conducive to improving the authenticity, real-time and accuracy of subsequent UAV reconnaissance target entity state generation, and then according to the PHD filter in the grid map model The state of the entity is estimated in the process. The entity state prediction equation and update equation can be derived according to the UAV reconnaissance target entity state model and the sensor measurement model. The UAV reconnaissance target entity state prediction result of each grid unit is corrected and updated using the pre-derived likelihood function, which can obtain more reliable results and thus improve the accuracy of the UAV reconnaissance entity state generation. Finally, the updated UAV reconnaissance target entity states of all grid units are summed to obtain the UAV reconnaissance target state generation model representing the state of the UAV reconnaissance target entity in the grid map model; the UAV reconnaissance target state generation model is solved according to the Gaussian sum approximation algorithm to obtain the real-time UAV reconnaissance target state. The real-time generated current UAV reconnaissance target state can provide strong data support for the simulation deduction of subsequent decision-making behaviors such as path planning. BRIEF DESCRIPTION OF THE DRAWINGS
[0041] Figure 11 is a flow chart of a method for real-time generation of UAV reconnaissance target status based on a PHD filter in one embodiment;
[0042] Figure 2 FIG. 1 is a diagram showing the internal structure of a computer device in one embodiment. DETAILED DESCRIPTION
[0043] In order to make the purpose, technical solutions and advantages of this application more clear, the following further describes this application in detail with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only used to explain this application and are not intended to limit this application.
[0044] In one embodiment, Figure 1 As shown, a method for real-time generation of UAV reconnaissance target status based on PHD filter is provided, which includes the following steps:
[0045] Step 102: Obtain a drone reconnaissance area; model the drone reconnaissance area to obtain a grid map model; establish an association relationship between a drone reconnaissance target entity and one or more grids in the grid map model, where each grid corresponds to a drone reconnaissance target entity state point, and the state set of multiple drone reconnaissance target entity state points jointly describes the state of the drone reconnaissance target entity in the grid map model.
[0046] The present application is used to generate the real-time status of drone reconnaissance target entities in a drone reconnaissance environment. That is, real-time measurement data of entities in a real reconnaissance environment is obtained by sensors to update the result status of the drone reconnaissance target simulation model. The real-time generated current entity status information can provide data support for the simulation deduction of subsequent decision-making behaviors such as path planning. First, it is necessary to model the situation of the drone reconnaissance area. The modeling process is existing technology and will not be described in detail in this application. The drone reconnaissance entity is mapped to a dynamic network map, and the state of the drone reconnaissance target is represented using a random finite set. This can achieve accurate modeling of target entities of different shapes and dynamically changing entity numbers, which is more in line with the dynamic and random characteristics of drone reconnaissance simulation. The state of the drone reconnaissance target entity in the drone reconnaissance environment is described by establishing an association relationship between a grid map model and the entity target, including establishing an association relationship between the drone reconnaissance target entity and one or more grids, where each grid corresponds to a drone reconnaissance target entity state point, and the state set of multiple drone reconnaissance target entity state points jointly describes the state of the drone reconnaissance target entity in the grid map model.
[0047] Step 104 : Modeling the UAV reconnaissance target entity state and sensor measurement data using random finite sets to obtain a UAV reconnaissance target entity state model and a sensor measurement model.
[0048] Traditional entity states are mostly described using vectors. In order to improve the authenticity of entity state representation and adapt to the dynamic changes in the number of entities and clutter interference in real scenes, this application proposes to use random finite sets to describe the target entity state. Each grid corresponds to a random finite set of state points. The entities and target entities in the following description both represent drone reconnaissance target entities. In the grid map, the target entity state at step k+1 can be represented by the following finite set:
[0049] X k+1 =X p,k+1 ∪X b,k+1 (1)
[0050] Among them, X k+1 is the set of predicted states, X p,k+1 is the target set of entities that persist compared to the previous moment, X b,k+1 It is a new set of battlefield target entity states compared to the previous moment.
[0051] This application focuses on the position and velocity information of entities in the drone reconnaissance environment, so a four-dimensional vector is used to represent the state corresponding to each entity.
[0052] x=[p x ,p y ,v x ,v y ] T (2)
[0053] Thus, the finite set of predicted states can be expressed as
[0054] X k+1 ={x1,x2...,x n} (3)
[0055] Where n is the maximum possible number of grids corresponding to the target entity.
[0056] The simulation model in which the state of the target entity increases with time steps is defined as
[0057] x k+1 =f k+1|k (x k+1 |x k )x k +ξ k (4)
[0058] Among them, f k+1| k(x k+1 |x k ) is the state transfer function, which is defined in this application as a uniform linear motion model, ξ kis the zero-mean process noise of the Gaussian distribution.
[0059] In the process of measuring the UAV reconnaissance environment, the sensor may have false alarms, missed detections, clutter, etc., that is, the measurement process has uncertainty, which is similar to the state model of the entity. Therefore, this application uses random finite sets to model the measurement data model. First, the k+1-th step measurement in each grid is independently modeled as a Bernoulli random finite set. In each grid cell, there is either a measurement z k+1 , or no measurement, the probability density function (PDF) in any grid cell c can be expressed as
[0060]
[0061] Among them, ε represents the probability of measurement existence, p(z k+1 ) represents z k+1 spatial distribution.
[0062] Step 106, estimate the state of the entity in the grid map model according to the PHD filter, use the UAV reconnaissance target entity state model and the sensor measurement model to predict the persistent UAV reconnaissance target entity and the newly generated UAV reconnaissance target entity, and obtain the persistent UAV reconnaissance target entity state prediction result and the newly generated UAV reconnaissance target entity state prediction result for each grid unit.
[0063] This application uses a PHD / MIB filter to estimate the state of an entity in a grid map, including an estimation of the occupancy state of a grid cell. The proposed filter represents the state of the posterior entity through its PHD, and the prediction step simply applies the standard PHD filter prediction. In order to update the predicted random finite set state with a measurement grid, the PHD / MIB filter approximates the entity state in each grid cell as a random finite set of Bernoulli distribution, and performs the update step independently for each grid cell in combination with the measurement of each grid. Finally, the PHD / MIB filter transfers all instances of the Bernoulli set to a joint PHD to represent the posterior state. In step 104, the drone reconnaissance target entity state model and sensor measurement model of the entity state in the drone reconnaissance environment have been established, and the entity state prediction equation and update equation can be derived based on the drone reconnaissance target entity state model and sensor measurement model.
[0064] First, the situation of entities that persist in the UAV reconnaissance environment is predicted. The PHD / MIB filter is used to predict the k (x k ) represents the posterior state of the entity at step k. This application treats new objects in a Bernoulli form, so only the PHD filter is used to predict persistent entities, i.e.
[0065] D p,k+1|k (x k+1 )=p S ∫f k+1|k (x k+1 |x k )D k (x k )dx k (6)
[0066] Among them, p S is the probability that the entity continues to exist compared to the previous moment, f k+1|k (x k+1 |x k ) is the state transfer function.
[0067] Since each grid needs to be updated using the MIB filter, the predicted persistent entity state PHD needs to be converted into a random finite set (RFS) of Bernoulli distribution in each grid, which can be understood as each grid is either occupied by an entity or idle.
[0068] The prediction of the state of a persistent entity in a grid cell c is represented by a Bernoulli RFS, whose PDF is
[0069]
[0070] in, To predict the probability of an entity persisting, The PDF of the state of the entity that is predicted to persist.
[0071] According to the results predicted by the PHD filter, we can get
[0072]
[0073] Among them, integrating the PHD of the entity state in the grid represents the number of entities in the grid. Since it is approximately a Bernoulli distribution, each grid cell cannot be occupied by more than one entity, so the maximum value is limited.
[0074] Then, the situation of the newly emerged entities (i.e., the newly emerged UAV reconnaissance targets) in the UAV reconnaissance environment is predicted, and the following situations are defined: if there is an entity at step k, then the probability of a new entity being generated at step k+1 is zero; if there is no entity in grid cell c at step k, then the new entity will be generated at step p at step k+1. B The probability of generation, the spatial distribution PDF of the new entity is p b (x k+1 ). Therefore, the predicted probability of entity creation in step k+1 is
[0075]
[0076] Considering both the case of persistent entity and newly created entity, the PDF of the predicted Bernoulli RFS is
[0077]
[0078] The predicted probability of grid cell c being occupied is
[0079]
[0080] Step 108 , using the pre-derived likelihood function, correct and update the prediction result of the drone reconnaissance target entity state of each grid unit to obtain the updated drone reconnaissance target entity state of each grid unit.
[0081] This application uses It represents the likelihood function, which mainly describes the likelihood of the measurement data corresponding to the entity and can generate the corresponding measurement data of the entity according to different situations such as false alarm and missed detection.
[0082] Defined in a grid cell c, there is a measurement, and the probability of generating measurement data when there is an entity is That is, the probability that the sensor measures an entity; define the probability that there is a measurement in grid cell c, but no entity exists, that is, the probability of a sensor false alarm is The definition is based on the state x in grid cell c k+1 The likelihood function of the entity generating the measurement data is Define the probability of association between a measurement and an entity as It can indicate whether the measurement value near the grid cell c is related to the entity in the grid cell, and the probability value decreases as the distance between the measurement and the grid cell increases; the density function of the clutter is defined as p cl (z). The following is a categorized discussion of different situations.
[0083] Scenario 1:
[0084] In this case, there is no entity in the grid cell and no measurement occurs, so the likelihood function is
[0085]
[0086] Scenario 2: X k+1 ={x k+1}
[0087] In this case, there is an entity in the grid cell, but no measurement occurs, indicating that a missed detection has occurred. The likelihood function is
[0088]
[0089] Case 3: Z k+1 ={z k+1},
[0090] In this case, there is no entity in the grid cell, but a measurement is generated, indicating that a false alarm has occurred and the sensor measures clutter. The likelihood function is
[0091]
[0092] Case 4: Z k+1 ={z k+1},X k+1 ={x k+1}
[0093] In this case, there are entities in the grid cells and measurements are generated. At this time, the sensor may have measured the entity or the clutter. The likelihood function is
[0094]
[0095] The pre-derived likelihood function is used to correct and update the prediction results of the UAV reconnaissance target entity state of each grid unit, and the updated UAV reconnaissance target entity state of each grid unit is obtained as
[0096]
[0097] Then we can get the posterior occupancy probability of the grid cell as
[0098]
[0099] The purpose of the update is to use the k+1 step measurement to correct the k+1 step prediction to obtain more reliable results and improve the accuracy of the UAV reconnaissance entity state generation.
[0100] Step 110 , summing the updated UAV reconnaissance target entity states of all grid cells to obtain a UAV reconnaissance target state generation model; solving the UAV reconnaissance target state generation model according to a Gaussian sum approximation algorithm to obtain the UAV reconnaissance target state.
[0101] After the update step, the PHD / MIB filter converts all Bernoulli RFS instances into joint PHD again, that is, summing the Bernoulli RFS of all grids
[0102]
[0103] Where c is the sequence number of the grid unit, and C is the total number of grid units.
[0104] The process of solving the UAV reconnaissance target state generation model based on the Gaussian sum approximation algorithm is an existing technology and will not be described in detail in this application.
[0105] In the above-mentioned real-time generation method of UAV reconnaissance target state based on PHD filter, the present application describes the state of UAV reconnaissance target entity in UAV reconnaissance environment by establishing the association relationship between grid map model and UAV reconnaissance target, and jointly describes the state of UAV reconnaissance target entity in grid map model with the state set of multiple UAV reconnaissance target entity state points, and uses random finite sets to represent the state of UAV reconnaissance target, which can realize accurate modeling of target entities of different shapes and dynamically changing entity measurements, which is more in line with the dynamic and random characteristics of UAV reconnaissance simulation, can improve the authenticity of entity state representation, adapt to the dynamic changes in the number of entities in real scenes, clutter interference, etc., which is conducive to improving the authenticity, real-time and accuracy of subsequent UAV reconnaissance target entity state generation, and then according to the PHD filter in the grid map model, the real-time and accuracy of the subsequent UAV reconnaissance target entity state generation are improved. The state of the entity is estimated in the model. The entity state prediction equation and update equation can be derived according to the UAV reconnaissance target entity state model and the sensor measurement model. The pre-derived likelihood function is used to correct and update the UAV reconnaissance target entity state prediction result of each grid unit, which can obtain more reliable results and thus improve the accuracy of UAV reconnaissance entity state generation. Finally, the updated UAV reconnaissance target entity states of all grid units are summed to obtain a UAV reconnaissance target state generation model representing the state of the UAV reconnaissance target entity in the grid map model; the UAV reconnaissance target state generation model is solved according to the Gaussian sum approximation algorithm to obtain the real-time UAV reconnaissance target state. The real-time generated current UAV reconnaissance target state can provide strong data support for the simulation deduction of subsequent decision-making behaviors such as path planning.
[0106] In one embodiment, the state and measurement data of the drone reconnaissance target entity are modeled according to a random finite set to obtain a drone reconnaissance target entity state model and a measurement model, including:
[0107] The UAV reconnaissance target entity state is modeled according to the random finite set, and the UAV reconnaissance target entity state model is obtained as follows:
[0108] x k+1 =f k+1|k (x k+1 |x k )x k +ξ k
[0109] Among them, f k+1|k (x k+1 |x k ) is the state transfer function, x k represents the target entity state at step k, ξ k is the zero-mean process noise of the Gaussian distribution.
[0110] In one embodiment, the sensor measurement data is modeled according to a random finite set, and the sensor measurement model is obtained as follows:
[0111]
[0112] Among them, ε represents the probability of unexpected conditions in the measurement, which include false alarms, missed detections, and clutter, and p(z k+1 ) represents the sensor measurement data z k+1 spatial distribution.
[0113] In one embodiment, the state of an entity is estimated in a grid map model based on a PHD filter, and a persistent UAV reconnaissance target entity and a newly emerged UAV reconnaissance target entity are predicted using a UAV reconnaissance target entity state model and a sensor measurement model, thereby obtaining a persistent UAV reconnaissance target entity state prediction result and a newly emerged UAV reconnaissance target entity state prediction result for each grid cell, including:
[0114] The state of the entity is estimated in the grid map model according to the PHD filter. The persistent UAV reconnaissance target entity state model and the sensor measurement model are used to predict the persistent UAV reconnaissance target entity. The prediction result of the persistent UAV reconnaissance target entity state of each grid unit is obtained as follows:
[0115]
[0116] Among them, c represents the grid unit number, k represents the time step number, X k+1 Indicates the state of the drone reconnaissance target entity at the k+1th step, x k+1 Represents the entity state model of the drone reconnaissance target, To predict the probability of the continued existence of the target entity in the drone reconnaissance, The probability density function for predicting the state of the persistent drone reconnaissance target entity.
[0117] In one embodiment, the state of the entity is estimated in the grid map model according to the PHD filter, and the newly generated drone reconnaissance target entity is predicted using the drone reconnaissance target entity state model and the sensor measurement model, and the newly generated drone reconnaissance target entity state prediction result for each grid unit is obtained as follows:
[0118]
[0119] in, It represents the probability of predicting the new generation of the target entity during the k+1-step drone reconnaissance.
[0120] In one embodiment, a likelihood function is used to describe the likelihood of measurement data corresponding to a target entity detected by a UAV. The derivation process of the likelihood function includes:
[0121] Define that there is a measurement in a certain grid cell c, and the probability of generating measurement data when there is a UAV reconnaissance target entity is: That is, the probability that the sensor measures the drone reconnaissance target entity; define the grid cell c where there is measurement but no drone reconnaissance target entity exists, that is, the probability of sensor false alarm is The definition is based on the state x in grid cell c k+1 The likelihood function of the UAV reconnaissance target entity generating measurement data is The association probability between the measurement and the UAV reconnaissance target entity is defined as It can indicate whether the measurement value near the grid cell c is related to the UAV reconnaissance target entity in the grid cell. The probability value decreases as the distance between the measurement and the grid cell increases. The density function of the clutter is defined as p cl (z), for different situations, the likelihood function has different forms, as follows:
[0122] when When there is no entity in the grid cell and no measurement occurs, the likelihood function is Among them, X k+1 Indicates the state of the drone reconnaissance target entity at the k+1th step, Z k+1 Represents the sensor measurement data at step k+1;
[0123] when X k+1 ={x k+1}, there is an entity in the grid cell, but no measurement occurs, indicating that a missed detection has occurred. At this time, the likelihood function is Among them, x k+1 Represents the entity state model of the drone reconnaissance target;
[0124] When Z k+1 ={z k+1}, When , there is no entity in the grid cell, but a measurement is generated, indicating that a false alarm has occurred and the sensor measures clutter. At this time, the likelihood function is
[0125]
[0126] When Z k+1 ={z k+1},X k+1 ={x k+1}, there is an entity in the grid cell and a measurement is generated. At this time, the sensor measures the entity or the clutter, and the likelihood function is
[0127]
[0128] In one embodiment, the predicted state of the drone reconnaissance target entity of each grid unit is corrected and updated using a pre-derived likelihood function to obtain the updated drone reconnaissance target entity state of each grid unit, including:
[0129] The pre-derived likelihood function is used to correct and update the prediction results of the UAV reconnaissance target entity state of each grid unit, and the updated UAV reconnaissance target entity state of each grid unit is obtained as
[0130]
[0131] in, represents the pre-derived likelihood function, represents the new UAV reconnaissance target entity state prediction result of each grid cell, X k+1 It represents the state of the drone reconnaissance target entity at the k+1th step, and c represents the sequence number of the grid unit.
[0132] In one embodiment, the UAV reconnaissance target entity states of all grid cells are summed to obtain a UAV reconnaissance target state generation model, including:
[0133] The final UAV reconnaissance target state is obtained by summing up the UAV reconnaissance target entity states of all grid cells.
[0134]
[0135] Among them, c represents the sequence number of the grid unit, C represents the number of grid units, and x k+1 Represents the entity state model of the drone reconnaissance target, X k+1 Indicates the state of the drone reconnaissance target entity at the k+1th step.
[0136] It should be understood that although Figure 1 The steps in the flowchart are shown in sequence as indicated by the arrows, but these steps are not necessarily executed in the order indicated by the arrows. Unless otherwise specified in this application, there is no strict order restriction for the execution of these steps, and these steps can be executed in other orders. Figure 1At least part of the steps may include multiple sub-steps or multiple stages. These sub-steps or stages are not necessarily executed at the same time, but can be executed at different times. The execution order of these sub-steps or stages is not necessarily sequential, but can be executed in turn or alternately with other steps or at least part of the sub-steps or stages of other steps.
[0137] In one embodiment, a computer device is provided. The computer device may be a terminal, and its internal structure diagram may be as follows: Figure 2 As shown. The computer device includes a processor, a memory, a network interface, a display screen and an input device connected via a system bus. The processor of the computer device is used to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and the computer program in the non-volatile storage medium. The network interface of the computer device is used to communicate with an external terminal via a network connection. When the computer program is executed by the processor, a method for real-time generation of the target state of a drone reconnaissance based on a PHD filter is implemented. The display screen of the computer device can be a liquid crystal display or an electronic ink display screen, and the input device of the computer device can be a touch layer covering the display screen, or a key, trackball or touchpad provided on the computer device housing, or an external keyboard, touchpad or mouse.
[0138] Those skilled in the art will understand that Figure 2 The structure shown in the figure is only a block diagram of a part of the structure related to the solution of the present application, and does not constitute a limitation on the computer device to which the solution of the present application is applied. The specific computer device may include more or fewer components than shown in the figure, or combine certain components, or have a different component arrangement.
[0139] Those skilled in the art will appreciate that all or part of the processes in the above-mentioned embodiments can be implemented by instructing the relevant hardware through a computer program. The computer program can be stored in a non-volatile computer-readable storage medium. When the computer program is executed, it can include the processes of the embodiments of the above-mentioned methods. Among them, any reference to memory, storage, database or other media used in the embodiments provided in this application can include non-volatile and / or volatile memory. Non-volatile memory can include read-only memory (ROM), programmable ROM (PROM), electrically programmable ROM (EPROM), electrically erasable programmable ROM (EEPROM) or flash memory. Volatile memory can include random access memory (RAM) or external cache memory. By way of illustration and not limitation, RAM is available in various forms, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDRSDRAM), enhanced SDRAM (ESDRAM), synchronous link (Synchlink) DRAM (SLDRAM), memory bus (Rambus) direct RAM (RDRAM), direct memory bus dynamic RAM (DRDRAM), and memory bus dynamic RAM (RDRAM).
[0140] The technical features of the above embodiments can be combined arbitrarily. To make the description concise, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, they should be considered to be within the scope of this specification.
[0141] The above-described embodiments merely represent several implementation methods of the present application. While the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the present invention. It should be noted that a person skilled in the art could make various modifications and improvements without departing from the spirit of the present application, all of which fall within the scope of protection of the present application. Therefore, the scope of protection of the present patent application shall be determined by the appended claims.
Claims
1. A real-time generation method for UAV reconnaissance target status based on PHD filter, characterized in that: The method comprises: Acquire a drone reconnaissance area; model the drone reconnaissance area to obtain a grid map model; establish an association relationship between a drone reconnaissance target entity and one or more grids in the grid map model, wherein each grid corresponds to a drone reconnaissance target entity state point, and a state set of multiple drone reconnaissance target entity state points collectively describes the state of the drone reconnaissance target entity in the grid map model; The UAV reconnaissance target entity state and sensor measurement data are modeled using random finite sets to obtain the UAV reconnaissance target entity state model and sensor measurement model; The state of the entity is estimated in the grid map model according to the PHD filter, and the persistent UAV reconnaissance target entity and the newly-generated UAV reconnaissance target entity are predicted using the UAV reconnaissance target entity state model and the sensor measurement model to obtain the persistent UAV reconnaissance target entity state prediction result and the newly-generated UAV reconnaissance target entity state prediction result for each grid cell; Correcting and updating the prediction result of the UAV reconnaissance target entity state of each grid unit using the pre-derived likelihood function to obtain the updated UAV reconnaissance target entity state of each grid unit; The updated UAV reconnaissance target entity states of all grid cells are summed to obtain a UAV reconnaissance target state generation model; the UAV reconnaissance target state generation model is solved according to a Gaussian sum approximation algorithm to obtain the UAV reconnaissance target state.
2. The method according to claim 1, characterized in that The entity state and measurement data of the UAV reconnaissance target are modeled according to random finite sets, and the entity state model and measurement model of the UAV reconnaissance target are obtained, including: The UAV reconnaissance target entity state is modeled according to the random finite set, and the UAV reconnaissance target entity state model is obtained as follows: x k+1 =f k+1|k (x k+1 |x k )x k +ξ k Among them, f k+1|k (x k+1 |x k ) is the state transfer function, x k represents the target entity state at step k, ξ k is the zero-mean process noise of the Gaussian distribution.
3. The method according to claim 1, characterized in that The method further comprises: The sensor measurement data is modeled according to random finite sets, and the sensor measurement model is obtained as follows: Among them, ε represents the probability of unexpected conditions in the measurement, which include false alarms, missed detections, and clutter, and p(z k+1 ) represents the sensor measurement data z k+1 spatial distribution.
4. The method according to any one of claims 1 to 3, characterized in that The state of the entity is estimated in the grid map model according to the PHD filter, and the persistent UAV reconnaissance target entity and the newly emerged UAV reconnaissance target entity are predicted using the UAV reconnaissance target entity state model and the sensor measurement model, thereby obtaining the persistent UAV reconnaissance target entity state prediction result and the newly emerged UAV reconnaissance target entity state prediction result for each grid cell, including: The state of the entity is estimated in the grid map model according to the PHD filter, and the persistent UAV reconnaissance target entity is predicted using the UAV reconnaissance target entity state model and the sensor measurement model. The prediction result of the persistent UAV reconnaissance target entity state of each grid unit is obtained as follows: Among them, c represents the grid unit number, k represents the time step number, X k+1 Indicates the state of the drone reconnaissance target entity at the k+1th step, x k+1 Represents the entity state model of the drone reconnaissance target, To predict the probability of the continued existence of the target entity in the drone reconnaissance, The probability density function for predicting the state of the persistent drone reconnaissance target entity.
5. The method according to claim 4, characterized in that The method further comprises: The state of the entity is estimated in the grid map model according to the PHD filter, and the newly generated UAV reconnaissance target entity is predicted using the UAV reconnaissance target entity state model and the sensor measurement model. The newly generated UAV reconnaissance target entity state prediction result of each grid unit is obtained as follows: in, It represents the probability of predicting the new generation of the target entity during the k+1-step drone reconnaissance.
6. The method according to claim 1, characterized in that The likelihood function is used to describe the likelihood of the measurement data corresponding to the target entity reconnaissance by the UAV; the derivation process of the likelihood function includes: Define that there is a measurement in a certain grid cell c, and the probability of generating measurement data when there is a UAV reconnaissance target entity is: That is, the probability that the sensor measures the drone reconnaissance target entity; define the grid cell c where there is measurement but no drone reconnaissance target entity exists, that is, the probability of sensor false alarm is The definition is based on the state x in grid cell c k+1 The likelihood function of the UAV reconnaissance target entity generating measurement data is The association probability between the measurement and the UAV reconnaissance target entity is defined as It can indicate whether the measurement value near the grid cell c is related to the UAV reconnaissance target entity in the grid cell. The probability value decreases as the distance between the measurement and the grid cell increases. The density function of the clutter is defined as p cl (z), for different situations, the likelihood function has different forms, as follows: when When there is no entity in the grid cell and no measurement occurs, the likelihood function is Among them, X k+1 Indicates the state of the drone reconnaissance target entity at the k+1th step, Z k+1 represents the sensor measurement data set at the k+1th step; when When , there is an entity in the grid cell, but no measurement occurs, indicating that a missed detection has occurred. At this time, the likelihood function is Among them, x k+1 Represents the entity state model of the drone reconnaissance target; when When , there is no entity in the grid cell, but a measurement is generated, indicating that a false alarm has occurred and the sensor measures clutter. At this time, the likelihood function is When Z k+1 ={z k+1 },X k+1 ={x k+1 }, there is an entity in the grid cell and a measurement is generated. At this time, the sensor measures the entity or the clutter, and the likelihood function is 7. The method according to claim 6, characterized in that The predicted state of the UAV reconnaissance target entity of each grid unit is corrected and updated using the pre-derived likelihood function to obtain the updated UAV reconnaissance target entity state of each grid unit, including: The predicted state of the UAV reconnaissance target entity of each grid unit is corrected and updated using the pre-derived likelihood function, and the updated state of the UAV reconnaissance target entity of each grid unit is obtained as in, represents the pre-derived likelihood function, represents the new UAV reconnaissance target entity state prediction result of each grid cell, X k+1 It represents the state of the drone reconnaissance target entity at the k+1th step, and c represents the sequence number of the grid unit.
8. The method according to claim 7, characterized in that The UAV reconnaissance target entity states of all grid cells are summed to obtain the UAV reconnaissance target state generation model, including: The final UAV reconnaissance target state is obtained by summing up the UAV reconnaissance target entity states of all grid cells. Among them, c represents the sequence number of the grid unit, C represents the number of grid units, and x k+1 Represents the entity state model of the drone reconnaissance target, X k+1 Indicates the state of the drone reconnaissance target entity at the k+1th step.