A method for evaluating radar operators' target detection ability
Through reverse sequence filtering tracks and interactive multi-model filtering technology, combined with chi-square inspection and maximum likelihood function method, the accuracy of radar operators in the evaluation of target discovery capabilities in complex radar display images is solved, and the target discovery capabilities are achieved accurate evaluation and accurate judgment of location time.
Patent Information
- Application Number
- CN202411230224.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Patents(China)
- Current Assignee / Owner
- Filing Date
- 2024-09-04
- Publication Date
- 2024-12-06
- Estimated Expiration
- 2044-09-04
AI Technical Summary
The prior art is difficult to accurately evaluate the target discovery capabilities of radar operators in complex radar display images, especially distinguishing civilian targets and military targets, and it is impossible to effectively judge whether the detection points before target approval belong to the same target, resulting in the evaluation of target discovery capabilities. Inaccurate.
The inverse filtering track and interactive multi-model filtering technology are used to determine the initialization state of the target through the inverse filtering track. The uniform speed, uniform acceleration and uniform rate turning models are used for reverse prediction, and the points are combined with the chi-square test and the maximum likelihood function method. track correlation, assess the radar operator's target discovery capabilities.
Accurate assessment of the target discovery capability of the radar operator is achieved, the accuracy of the target position and discovery time is improved, and the accuracy of the target discovery capability assessment of the radar operator is enhanced.
Smart Images

Figure CN118735364B_ABST
Abstract
Description
Technical Field
[0001] The invention relates to a method for evaluating target detection capability of a radar operator, and belongs to the technical field of radar target detection. Background Art
[0002] Target detection capability, especially the ability to accurately distinguish civilian targets and false targets from complex radar display images, and to promptly detect sensitive targets, or even enemy military targets, is a key capability for radar operators.
[0003] Using the historical detection data recorded by the radar, including the original point data and track data, to statistically analyze and evaluate the target detection ability of the radar operator is a good method to evaluate the job ability of the radar operator. Figure 3 As shown in the figure, for the radar automatic start mode, the target track automatically starts to be edited, and the target detection time and judgment time cannot be accurately distinguished. It is only used as an overall indicator to calibrate the operator's ability. Figure 4 As shown in the figure, in the manual start working mode, the target discovery time can be calculated by the batching time and the time when the target is first detected. However, how to judge whether the detection traces recorded by the radar before the target batching belong to the target is an undisclosed technical achievement, which has become the reason why current radar operators have serious doubts about the accuracy of target discovery capability assessment. Summary of the invention
[0004] The purpose of the present invention is to solve the deficiencies in the prior art and to provide a method for evaluating the target detection capability of a radar operator, which can accurately evaluate the target detection capability of the radar operator and detect the target in a timely and accurate manner.
[0005] In order to solve the above problems, the present application is implemented through the following technical solutions:
[0006] A method for evaluating the target detection capability of a radar operator is characterized in that it comprises the following steps:
[0007] Step 1. Select the reverse filter track: For the manually started radar track, according to the experience value, set the number of target track points for stable tracking to , if the target track Number of , then select the entire track Do reverse filtering, otherwise you can choose Perform reverse filtering and select the manual start mark in the radar track data as the manual start point;
[0008] Step 2: Determine the start of the track based on the reverse filtering track in step 1: Select the track in step 1 The first two points in the track are used to start the track and determine the initialization state of the target;
[0009] Step 3: Perform reverse filtering on the reverse filtered track of step 1 based on the interactive multi-model: select motion models such as the uniform velocity model CV, the uniform acceleration model CA, and the uniform rate turning model CT to form a model set. The number of models is , using the interactive multi-model maneuvering target tracking method to reversely filter from the first two points described in step 2 to the manual starting point selected in step 1;
[0010] Step 4, reverse prediction based on interactive multi-model: After filtering to the manual starting point of the track, reverse prediction is performed based on the filter output state vector, covariance matrix and model set of the interactive multi-model tracking filter obtained in step 3 to obtain the predicted value, i.e., the target motion state vector and its corresponding covariance matrix;
[0011] Step 5: Chi-square test judgment: Search using the chi-square test judgment method The point trace that is determined as the candidate reverse vector measurement point trace at the moment, and the first detected point trace is determined;
[0012] Step 6: Correlation of nearest point tracks: Calculate the likelihood function of each point track and use the maximum likelihood function method to correlate point tracks;
[0013] Step 7: Evaluate the target detection capability of the radar operator according to the time interval obtained in step 5: the larger the time interval, the worse the target detection capability of the radar operator; the smaller the time interval, the stronger the target detection capability of the radar operator.
[0014] Further, the step 2 is as follows: For example, take The first two dots and Start the track and determine the initialization state of the target:
[0015] For sea targets , , the target initialization state is: ,
[0016] Assumptions The covariance in the rectangular coordinate system corresponding to the moment measurement is: R ( k ) = [ r 11 r 12 r 21 r 22 ] = A [ s r 2 0 0 s i 2 ] A T ,
[0017] Where: and are the variances of the target radial distance and azimuth measurement errors, respectively;
[0018] A = [ cos i − r sin i sin i r cos i ] ,
[0019] Where: Matrix Represents the motion characteristics of the point trace, and They are respectively the target radial distance and azimuth angle measurement data of the radar in the polar coordinate system;
[0020] The elements of the measurement noise covariance are The initial covariance matrix in the case of a four-dimensional state vector can be obtained: P ( 1 | 1 ) = [ r 11 ( 1 ) r 11 ( 1 ) / T r 12 ( 1 ) r 12 ( 1 ) / T r 11 ( 1 ) / T 2 r 11 ( 1 ) / T 2 r 12 ( 1 ) / T 2 r 12 ( 1 ) / T 2 r 12 ( 1 ) r 12 ( 1 ) / T r 22 ( 1 ) r 22 ( 1 ) / T r 12 ( 1 ) / T 2 r 12 ( 1 ) / T 2 r 22 ( 1 ) / T 2 r 22 ( 1 ) / T 2 ] ,
[0021] In the prediction stage of inverse filtering, the initial covariance matrix is used to predict the uncertainty of the next state estimate under the target initialization state. Each target state corresponds to a covariance matrix to predict the uncertainty of the next target state estimate.
[0022] Furthermore, the step 3 specifically includes the following steps:
[0023] Step 3.1: Determine the state transfer matrix ;
[0024] Step 3.2: After determining the interaction calculation The filter Input of time;
[0025] Step 3.3: Use the extended Kalman filter to calculate the output state vector of each model , the state covariance matrix , ;
[0026] Furthermore, the step 3.1 is specifically as follows: P ˜ = [ P 11 P 12 ⋯ P 1 S N P 21 P 22 ⋯ P 2 S N ⋮ ⋮ ⋱ ⋮ P S N 1 P S N 2 ⋯ P S N S N ] ,
[0027] Where: For Model Transfer to Model The transition probability of
[0028] Furthermore, the step 3.2 is specifically as follows: determining the input state vector and its corresponding state covariance matrix ,
[0029] in, ,
[0030] Where: express Moment The target estimated state predicted by the filter;
[0031] ,
[0032] Where: for Moment Model The probability of is the corresponding state transfer matrix Middle Line The value of the column;
[0033] P oh ( k − 1 | k − 1 ) = ∑ i = 1 S N { P i ( k − 1 | k − 1 ) + [ X ^ i ( k − 1 | k − 1 ) − X ^ oh ( k − 1 | k − 1 )] ⋅ [ X ^ i ( k − 1 | k − 1 ) − X ^ oh ( k − 1 | k − 1 )]'} u k − 1 | k − 1 ( i | j )
[0034] Where: for Moment Model The probability of ;
[0035] The state vector and its state covariance matrix With radar observations Together as Moment The input value of each model;
[0036] Furthermore, the step 3.3 specifically includes the following steps:
[0037] Step 3.3.1. Calculate the nonlinear state transfer function The Jacobian matrix : ,
[0038] Step 3.3.2: According to the Jacobian matrix of the nonlinear state transfer function Update the covariance matrix: ,
[0039] Where: is the predicted covariance matrix, yes The optimal covariance matrix at time , is the noise covariance matrix;
[0040] Step 3.3.3. Determine the Kalman gain : ,
[0041] Where: is a nonlinear measurement function exist The Jacobian matrix of x is: ;
[0042] Step 3.3.4: Determine the output state vector and the state covariance matrix ,in, ,
[0043] ,
[0044] Where: is based on The state prediction at the moment The state vector estimate at time t, is a nonlinear state transfer function based on interactive multiple models, yes The optimal state estimate at time yes Control input at all times;
[0045] .
[0046] Furthermore, the step 4 is specifically as follows: the output state vector of each model determined in step 3.3 , the state covariance matrix Computational Filter Interactive output at all times: ,
[0047] Where: for Time filter The state vector of for Moment Model probability;
[0048] P ^ ( k | k ) = ∑ i = 1 S N u k ( i ){ P i ( k | k ) + [ X ^ i ( k | k ) − X ^ ( k | k )][ X ^ i ( k | k ) − X ^ ( k | k )] ' } .
[0049] Furthermore, the step 5 is specifically as follows:
[0050] Calculate the difference between the radar observation and the filter prediction, i.e., the filter residual and the innovation covariance matrix , , the specific formula is as follows: ,
[0051] Where: for Radar observation value at the time, is the measurement matrix, for The observation error covariance matrix at time t;
[0052] When there is a target motion model Make the measurement points satisfy , determine that the point trace is a reverse vector measurement point trace;
[0053] If only one point track meets the requirements, the point track is judged as the track reverse vector measurement point track, and reverse filtering prediction is performed, that is, step 4 is executed;
[0054] If no point track meets the requirements, the reverse filtering of the track is terminated. The point corresponding to the time is detected for the first time, confirming that the point is the first detected point, and the time k is the time interval between the time when the radar first detects the target and the manual batch start time, and execute step 7;
[0055] If there are multiple points that meet the requirements, all points are included in the point set. , use the nearest neighbor method to determine which point is the reverse filter measurement point, and execute step 6.
[0056] Furthermore, the step 6 is specifically as follows:
[0057] For any point , calculate the corresponding likelihood function under different assumptions of maneuvering target motion model , , L k + 1 i , j = 1 | 2 π S k + 1 j | exp [ − 1 2 ( v k + 1 j )'( S k + 1 j ) − 1 v k + 1 j ] ,
[0058] Where: For Model The filter residual is is the corresponding innovation covariance matrix, which obeys Gaussian distribution; , Calculated by step 5;
[0059] According to the weight corresponding to each maneuvering target motion model in the interactive multi-model algorithm, that is, the transfer probability of each maneuvering target motion model at the current moment is multiplied and summed to calculate the likelihood function corresponding to the point trace ;
[0060] L k + 1 i = ∑ j = 1 S N L k + 1 i , j u k + 1 ( j ) = ∑ j = 1 S N 1 | 2 π S k + 1 j | exp [ − 1 2 ( v k + 1 j )'( S k + 1 j ) − 1 v k + 1 j ] u k + 1 ( j ) ,
[0061] Where: for Moment Model The transition probability, ,
[0062] Where: , ,
[0063] Select the point trace corresponding to the maximum likelihood function as the measurement point trace , proceed to step 4;
[0064] in, .
[0065] This application applies the reverse tracking filtering technology and the idea of point track association to the detection point track judgment before the target is launched; the nearest neighbor method and chi-square test technology are applied to the reverse point track association judgment, so as to obtain a more accurate target position and target discovery time. BRIEF DESCRIPTION OF THE DRAWINGS
[0066] Figure 1 is a flow chart of the present invention;
[0067] Figure 2 This is a schematic diagram of track association judgment of reverse filtering points;
[0068] Figure 3 This is a schematic diagram of target detection time and judgment time in automatic start mode;
[0069] Figure 4 Schematic diagram of target detection time and judgment time in manual start mode. DETAILED DESCRIPTION
[0070] The following is a specific implementation of the present invention with reference to the accompanying drawings to further illustrate the composition of the present invention.
[0071] Related explanation: Radar will automatically initiate batching of aerial targets, so this application mainly studies the method of associating point tracks with targets and performs reverse filtering association on sea surface targets.
[0072] Embodiment 1. A method for evaluating the target detection capability of a radar operator, the process of which is as follows: Figure 1 As shown, the following steps are included:
[0073] Step 1. Select the reverse filter track: For the manually started radar track, according to the experience value, set the number of target track points for stable tracking to , that is, after After point tracking is filtered, the track tracking reaches a certain accuracy and can be used as track prediction. Number of , then select the entire track Do reverse filtering, otherwise you can choose Perform reverse filtering;
[0074] Step 2: Determine the start of the track based on the reverse filtering track in step 1: Select the track in step 1 The first two points in the track are used to start the track and determine the initialization state of the target;
[0075] Step 3: Perform reverse filtering on the reverse filtered track of step 1 based on the interactive multi-model: select the uniform velocity model CV, uniform acceleration model CA, and uniform rate turning model CT (the turning rate uses the typical value coverage , , ) and other motion models form a model set, and the number of models is , using the interactive multi-model maneuvering target tracking method to reversely filter from the first two points described in step 2 to the manual starting point selected in step 1;
[0076] Step 4, reverse prediction based on interactive multi-model: After filtering to the manual starting point of the track, reverse prediction is performed based on the filter output state vector, covariance matrix and model set of the interactive multi-model tracking filter obtained in step 3 to obtain the predicted value, i.e., the target motion state vector and its corresponding covariance matrix;
[0077] Step 5: Chi-square test judgment: Search using the chi-square test judgment method The point trace that is determined as the candidate reverse vector measurement point trace at the moment, and the first detected point trace is determined;
[0078] Step 6: Correlation of nearest point tracks: Calculate the likelihood function of each point track and use the maximum likelihood function method to correlate point tracks;
[0079] The schematic diagram of steps 1 to 6 is as follows Figure 2 As shown;
[0080] Step 7: Evaluate the target detection capability of the radar operator according to the time interval obtained in step 5: the larger the time interval, the worse the target detection capability of the radar operator; the smaller the time interval, the stronger the target detection capability of the radar operator.
[0081] Further, the step 2 is as follows: For example, take The first two dots and Start the track and determine the initialization state of the target:
[0082] For sea targets , , the target initialization state is: ,
[0083] Assumptions The covariance in the rectangular coordinate system corresponding to the moment measurement is: R ( k ) = [ r 11 r 12 r 21 r 22 ] = A [ s r 2 0 0 s i 2 ] A T ,
[0084] Where: and are the variances of the target radial distance and azimuth measurement errors, respectively;
[0085] A = [ cos i − r sin i sin i r cos i ] ,
[0086] Where: Matrix Represents the motion characteristics of the point trace, and They are respectively the target radial distance and azimuth angle measurement data of the radar in the polar coordinate system;
[0087] The elements of the measurement noise covariance are The initial covariance matrix in the case of a four-dimensional state vector can be obtained: P ( 1 | 1 ) = [ r 11 ( 1 ) r 11 ( 1 ) / T r 12 ( 1 ) r 12 ( 1 ) / T r 11 ( 1 ) / T 2 r 11 ( 1 ) / T 2 r 12 ( 1 ) / T 2 r 12 ( 1 ) / T 2 r 12 ( 1 ) r 12 ( 1 ) / T r 22 ( 1 ) r 22 ( 1 ) / T r 12 ( 1 ) / T 2 r 12 ( 1 ) / T 2 r 22 ( 1 ) / T 2 r 22 ( 1 ) / T 2 ] ,
[0088] In the prediction stage of inverse filtering, the initial covariance matrix is used to predict the uncertainty of the next state estimate under the target initialization state. Each target state corresponds to a covariance matrix to predict the uncertainty of the next target state estimate.
[0089] Furthermore, the step 3 specifically includes the following steps:
[0090] Step 3.1: Determine the state transfer matrix , specifically: P ˜ = [ P 11 P 12 ⋯ P 1 S N P 21 P 22 ⋯ P 2 S N ⋮ ⋮ ⋱ ⋮ P S N 1 P S N 2 ⋯ P S N S N ] ,
[0091] Where: For Model Transfer to Model The transition probability is usually set based on experience. It is generally assumed that the elements on the diagonal is a larger value, which can be 0.8, and the other elements are ;
[0092] Step 3.2: After determining the interaction calculation The filter Input at the moment: Determine the input state vector and its corresponding state covariance matrix ,
[0093] in, ,
[0094] Where: express Moment The target estimated state predicted by the filter;
[0095] ,
[0096] Where: for Moment Model The probability of is the state transfer matrix Middle Line The value of the column; P oh ( k − 1 | k − 1 ) = ∑ i = 1 S N { P i ( k − 1 | k − 1 ) + [ X ^ i ( k − 1 | k − 1 ) − X ^ oh ( k − 1 | k − 1 )] ⋅ [ X ^ i ( k − 1 | k − 1 ) − X ^ oh ( k − 1 | k − 1 )]'} u k − 1 | k − 1 ( i | j ) ,
[0097] Where: for Moment Model The probability of ;
[0098] The state vector and its state covariance matrix With radar observations Together as Moment The input value of the model;
[0099] Step 3.3: Use the extended Kalman filter to calculate the output state vector of each model , the state covariance matrix , ,
[0100] The specific steps include:
[0101] Step 3.3.1. Calculate the nonlinear state transfer function The Jacobian matrix : ,
[0102] Step 3.3.2: According to the Jacobian matrix of the nonlinear state transfer function Update the covariance matrix: ,
[0103] Where: is the predicted covariance matrix, yes The optimal covariance matrix at time , is the noise covariance matrix;
[0104] Step 3.3.3. Determine the Kalman gain :
[0105] Where: is a nonlinear measurement function exist The Jacobian matrix of x is: ;
[0106] Step 3.3.4: Determine the output state vector and the state covariance matrix ,in, ,
[0107] ,
[0108] Where: is based on The state prediction at the moment The state vector estimate at time t, is a nonlinear state transfer function based on interactive multiple models, yes The optimal state estimate at time yes Control input at all times;
[0109] .
[0110] Furthermore, the step 4 is specifically as follows: the output state vector of each model determined in step 3.3 , the state covariance matrix Computational Filter Interactive output at all times: ,
[0111] Where: for Time filter The state vector of for Moment Model probability;
[0112] P ^ ( k | k ) = ∑ i = 1 S N u k ( i ){ P i ( k | k ) + [ X ^ i ( k | k ) − X ^ ( k | k )][ X ^ i ( k | k ) − X ^ ( k | k )] ' } .
[0113] Furthermore, the step 5 is specifically as follows:
[0114] Calculate the difference between the radar observation and the filter prediction, i.e., the filter residual and the innovation covariance matrix , , the specific formula is as follows: ,
[0115] Where: for Radar observation value at the time, is the measurement matrix, for The observation error covariance matrix at time t;
[0116] When there is a target motion model Make the measurement points satisfy , is the significance level set in advance by the chi-square test, which is a specific value of 0.05, and it is judged that the point trace is a reverse vector measurement point trace;
[0117] If only one point track meets the requirements, the point track is judged as the track reverse vector measurement point track, and reverse filtering prediction is performed, that is, step 4 is executed;
[0118] If no point track meets the requirements, the reverse filtering of the track is terminated. The point corresponding to the time is detected for the first time, confirming that the point is the first detected point, and the time k is the time interval between the time when the radar first detects the target and the manual batch start time, and execute step 7;
[0119] If there are multiple points that meet the requirements, all points are included in the point set. , use the nearest neighbor method to determine which point is the reverse filter measurement point, and execute step 6.
[0120] Furthermore, the step 6 is specifically as follows: for any point trace , calculate the corresponding likelihood function under different assumptions of maneuvering target motion model , , L k + 1 i , j = 1 | 2 π S k + 1 j | exp [ − 1 2 ( v k + 1 j )'( S k + 1 j ) − 1 v k + 1 j ] ,
[0121] Where: For Model The filter residual is is the corresponding innovation covariance matrix, which obeys Gaussian distribution; , Calculated by step 5;
[0122] According to the weight corresponding to each maneuvering target motion model in the interactive multi-model algorithm, that is, the transfer probability of each maneuvering target motion model at the current moment is multiplied and summed to calculate the likelihood function corresponding to the point trace ;
[0123] L k + 1 i = ∑ j = 1 S N L k + 1 i , j u k + 1 ( j ) = ∑ j = 1 S N 1 | 2 π S k + 1 j | exp [ − 1 2 ( v k + 1 j )'( S k + 1 j ) − 1 v k + 1 j ] u k + 1 ( j ) ,
[0124] Where: for Moment Model The transition probability, ,
[0125] Where: , ,
[0126] Select the point trace corresponding to the maximum likelihood function as the measurement point trace , proceed to step 4;
[0127] in, .
[0128] A method for evaluating the target detection capability of a radar operator is implemented by a point track association judgment method based on reverse filtering. The point track association judgment method based on reverse filtering determines the time point when the target is first detected, and the target detection time is obtained by calculating the time interval between the manual start time of the radar operator and the time when the target is first detected. Finally, the target detection capability of the radar operator is evaluated according to the target detection time of the radar operator.
[0129] The manual start time is the time to start processing target information, which is set manually by the operator during the radar target batching process and is obtained from the recorded data of the radar system.
[0130] The target discovery time is equal to the time interval: the time when the target is first detected is reversely predicted by step 4, and then judged by step 5 to obtain the most likely time when the target track point is first detected.
Claims
1. A method for evaluating a radar operator's target detection capability, characterized in that: The following steps are involved: Step 1. Select the reverse filter track: For the manually started radar track, according to the experience value, set the number of target track points for stable tracking to , if the target track Number of , then select the entire track Do reverse filtering, otherwise select Perform reverse filtering and select the manual start mark in the radar track data as the manual start point; Step 2: Determine the start of the track based on the reverse filtering track in step 1: Select the track in step 1 The first two points in the track are used to start the track and determine the initialization state of the target; Step 3: Perform reverse filtering on the reverse filtered track of step 1 based on the interactive multi-model: select the uniform velocity model CV, uniform acceleration model CA, and uniform rate turning model CT to form a model set. The number of models is , using the interactive multi-model maneuvering target tracking method to reversely filter from the first two points described in step 2 to the manual starting point selected in step 1, including the following steps: Step 3.1, determine the state transfer matrix , , Where: For Model Transfer to Model The transition probability of Step 3.2: After confirming the interactive calculation The filter Input at the moment: Determine the input state vector and its corresponding state covariance matrix , in, , Where: express Moment The target estimated state predicted by the filter; , Where: for Moment Model The probability of is the state transfer matrix Middle Line The value of the column; , Where: for Moment Model The probability of ; The state vector and its state covariance matrix With radar observations Together as Moment The input value of the model; Step 3.3: Use the extended Kalman filter to calculate the output state vector of each model , the state covariance matrix , , specifically including the following steps: Step 3.3.
1. Calculate the nonlinear state transfer function The Jacobian matrix : , Step 3.3.2: According to the Jacobian matrix of the nonlinear state transfer function Update the covariance matrix: , Where: is the predicted covariance matrix, yes The optimal covariance matrix at time , is the noise covariance matrix; Step 3.3.
3. Determine the Kalman gain : , Where: is a nonlinear measurement function exist The Jacobian matrix of x is: ; Step 3.3.4: Determine the output state vector and the state covariance matrix ,in, , , Where: is based on The state prediction at the moment The state vector estimate at time t, is a nonlinear state transfer function based on interactive multiple models, yes The optimal state estimate at time, yes Control input at all times; ; Step 4, reverse prediction based on interactive multi-model: After filtering to the manual starting point of the track, reverse prediction is performed based on the filter output state vector, covariance matrix and model set of the interactive multi-model tracking filter obtained in step 3 to obtain the predicted value, i.e., the target motion state vector and its corresponding covariance matrix; Step 5: Chi-square test judgment: Search using the chi-square test judgment method The point trace that is determined as the candidate reverse vector measurement point trace at the moment, and the first detected point trace is determined, as follows: Calculate the difference between the radar observation and the filter prediction, i.e., the filter residual and the innovation covariance matrix , , the specific formula is as follows: , Where: for Radar observation value at the time, is the measurement matrix, for The observation error covariance matrix at time t; When there is a target motion model Make the measurement points satisfy , determine that the point trace is a reverse vector measurement point trace; If only one point track meets the requirements, the point track is judged as the track reverse vector measurement point track, and reverse filtering prediction is performed, that is, step 4 is executed; If no point track meets the requirements, the reverse filtering of the track is terminated. The point corresponding to the time is detected for the first time, confirming that the point is the first detected point, and the time k is the time interval between the time when the radar first detects the target and the manual batch start time, and execute step 7; If there are multiple points that meet the requirements, all points are included in the point set. , use the nearest neighbor method to determine which point is the reverse filter measurement point, and execute step 6; Step 6: Correlation of nearest point tracks: Calculate the likelihood function of each point track and use the maximum likelihood function method to correlate point tracks; Step 7: Evaluate the target detection capability of the radar operator according to the time interval obtained in step 5: the larger the time interval, the worse the target detection capability of the radar operator; the smaller the time interval, the stronger the target detection capability of the radar operator.
2. The method for evaluating the target detection capability of a radar operator according to claim 1, characterized in that: The step 2 is specifically as follows: by For example, take The first two dots and Start the track and determine the initialization state of the target: For sea targets , , the target initialization state is: , Assumptions The covariance in the rectangular coordinate system corresponding to the moment measurement is: , Where: and are the variances of the target radial distance and azimuth measurement errors, respectively; , Where: Matrix Represents the motion characteristics of the point trace, and They are respectively the target radial distance and azimuth angle measurement data of the radar in the polar coordinate system; The elements of the measurement noise covariance are The initial covariance matrix in the case of a four-dimensional state vector can be obtained: , In the prediction stage of inverse filtering, the initial covariance matrix is used to predict the uncertainty of the next state estimate under the target initialization state. Each target state corresponds to a covariance matrix to predict the uncertainty of the next target state estimate.
3. A method for evaluating the target detection capability of a radar operator according to claim 1 or 2, characterized in that: The step 4 is specifically as follows: the output state vector of each model determined in step 3.3 , the state covariance matrix Computational Filter Interactive output at all times: , Where: for Time filter The state vector of for Moment Model probability; 。 4. The method for evaluating the target detection capability of a radar operator according to claim 3, characterized in that: The step 6 is specifically as follows: for any point trace , calculate the corresponding likelihood function under different assumptions of maneuvering target motion model , , , Where: For Model The filter residual is is the corresponding innovation covariance matrix, which obeys Gaussian distribution; , Calculated by step 5; According to the weight corresponding to each maneuvering target motion model in the interactive multi-model algorithm, that is, the transfer probability of each maneuvering target motion model at the current moment is multiplied and summed to calculate the likelihood function corresponding to the point trace ; , Where: for Moment Model The transition probability, , Where: , , Select the point trace corresponding to the maximum likelihood function as the measurement point trace , proceed to step 4; in, .
