A target motion state classification method, apparatus and electronic device

Abstract

Embodiments of the present application disclose a target motion state classification method and device and electronic equipment, and relate to the technical field of radar and the technical field of automatic driving, which comprises: obtaining original target feature data of a current radar frame collected by a radar for a target; determining motion information corresponding to the current radar frame of the target by using a Kalman filtering algorithm; preliminarily determining a motion state of the target based on the motion information as an initial motion state; assigning an instantaneous classification confidence to each preset motion state according to a preset confidence assignment strategy, wherein the instantaneous classification confidence of the initial motion state is greater than the instantaneous classification confidence of each other preset motion state; and calculating the confidence of each preset motion state by using a Bayesian algorithm based on the instantaneous classification confidence of the plurality of preset motion states and the confidence of the plurality of preset motion states calculated for a previous radar frame. The present application improves the accuracy and stability of the motion state classification of the target.

Description

Technical Field

[0001] This invention relates to the fields of radar technology and autonomous driving technology, and in particular to a target motion state classification method, device and electronic equipment. Background Technology

[0002] Vehicle-mounted radar is one of the key sensors in autonomous driving systems. Currently, millimeter-wave radar is widely used in practical applications. Millimeter-wave radar can perform range measurement, Doppler velocity detection, azimuth angle and other applications for targets. In millimeter-wave radar target tracking, it has advantages such as strong anti-interference ability and all-weather operation in judging the motion state of targets.

[0003] Currently, in practical applications based on millimeter-wave radar, the traditional Kalman filter algorithm is usually used to estimate the speed of the tracked target, and the target's motion state is determined by comparing the speed with a threshold.

[0004] However, when using the above-mentioned methods in the existing technology, for targets that brake suddenly, the Kalman filter algorithm has a certain lag in estimating the target's speed, which may result in the target actually stopping, but the estimated result still having a certain speed; and, due to the instability of long-range radar echoes, the detection point may have a certain jitter, which may cause stationary targets to be misidentified as moving targets.

[0005] In summary, existing technologies suffer from inaccurate judgment of the target's motion state. Summary of the Invention

[0006] In view of this, embodiments of the present invention provide a target motion state classification method, apparatus and electronic device to solve the problem of inaccurate judgment of the motion state of a target in the prior art.

[0007] In a first aspect, embodiments of the present invention provide a target motion state classification method, including:

[0008] Acquire the raw target feature data of the current radar frame collected by the radar for the target;

[0009] Based on the original target feature data, the Kalman filter algorithm is used to determine the motion information corresponding to the current radar frame of the target;

[0010] Based on the motion information, the motion state of the target is initially determined as the initial motion state, which is one of a variety of preset motion states;

[0011] According to a preset confidence allocation strategy, an instantaneous classification confidence score is assigned to each preset motion state, wherein the instantaneous classification confidence score of the initial motion state is greater than the instantaneous classification confidence scores of other preset motion states, and the sum of the instantaneous classification confidence scores of the multiple preset motion states is 1.

[0012] Based on the instantaneous classification confidence of the various preset motion states, and the confidence of the various preset motion states calculated for the previous radar frame, a Bayesian algorithm is used to calculate the confidence of each preset motion state, which serves as the basis for determining the motion state classification result of the target for the current radar frame.

[0013] According to a specific implementation of an embodiment of the present invention, the parameter values ​​of the process noise used in the Kalman filter algorithm are adaptively adjusted using the following steps:

[0014] After determining the target's velocity and acceleration each time based on a radar frame, the process noise is adjusted based on the relationship between the velocity and acceleration and threshold values. The process noise is adjusted by using a preset low-noise matrix when the velocity is trending low and a preset high-noise matrix when the acceleration is trending high.

[0015] According to a specific implementation of the present invention, after determining the velocity and acceleration of the target based on each radar frame, the process noise is adjusted based on the magnitude relationship between the velocity and acceleration and threshold values, according to an adjustment strategy that uses a preset low-noise matrix when the velocity is trending low and a preset high-noise matrix when the acceleration is trending high. This adjustment includes:

[0016] When the process noise currently uses a preset high noise matrix, after the speed of the target is determined based on a radar frame each time, if the speed of the target determined based on a preset number of consecutive radar frames is less than a first preset speed threshold, the process noise is switched to a preset low noise matrix.

[0017] When the process noise currently uses the preset low-noise matrix, after determining the velocity of the target based on each radar frame, the Mahalanobis distance between the predicted position and the actual observed position of the target is calculated. The predicted position is determined based on the velocity and acceleration corresponding to the previous radar frame of the target.

[0018] When the Mahalanobis distance is less than a preset distance threshold and the target's speed is not greater than a second preset speed threshold, for the next radar frame, it is determined whether the speed of the target in the next radar frame is not greater than the second preset speed threshold, where the second preset speed threshold is less than the first preset speed threshold.

[0019] If the velocity of the target in the next radar frame is not greater than the second preset velocity threshold, the velocity of the target in the next radar frame remains unchanged.

[0020] If the velocity corresponding to the next radar frame of the target is greater than the second preset velocity threshold, and the difference between the acceleration corresponding to the next radar frame of the target and the acceleration corresponding to the current radar frame is less than the preset acceleration difference threshold, the velocity corresponding to the next radar frame of the target is adjusted to the second preset velocity threshold.

[0021] If the velocity corresponding to the next radar frame of the target is greater than the second preset velocity threshold, and the difference between the acceleration corresponding to the next radar frame of the target and the acceleration corresponding to the current radar frame is not less than the preset acceleration difference threshold, the process noise is switched to the preset high noise matrix.

[0022] According to a specific implementation of an embodiment of the present invention, the motion information corresponding to the current radar frame of the target includes: the target's x-axis velocity, y-axis velocity, x-axis acceleration, and y-axis acceleration;

[0023] The preliminary determination of the target's motion state based on the motion information includes:

[0024] Based on the target's x-axis velocity, y-axis velocity, x-axis acceleration, and y-axis acceleration, determine the target's acceleration / deceleration state, as well as the predicted x-axis velocity and y-axis velocity;

[0025] Based on the target's x-axis velocity, y-axis velocity, predicted x-axis velocity, and predicted y-axis velocity, as well as the target's acceleration / deceleration state and the obtained motion state of the target's associated points, the motion state of the target is preliminarily determined.

[0026] According to a specific implementation of an embodiment of the present invention, the step of assigning instantaneous classification confidence to each preset motion state according to a preset confidence allocation strategy includes:

[0027] Obtain the confidence allocation strategy corresponding to the initial motion state, which includes instantaneous classification confidence levels pre-set for each preset motion state;

[0028] According to the obtained confidence allocation strategy, an instantaneous classification confidence level is assigned to each of the preset motion states.

[0029] According to a specific implementation of an embodiment of the present invention, the plurality of preset motion states include: a plurality of defined motion states and an unknown motion state;

[0030] The instantaneous classification confidence based on the multiple preset motion states, and the confidence of the multiple preset motion states calculated for the previous radar frame, are used to calculate the confidence of each preset motion state using a Bayesian algorithm, including:

[0031] For various preset motion states other than the unknown motion state, the confidence level of each preset motion state is calculated using the following formula:

[0032] ;

[0033] ;

[0034] in, This represents the confidence level of the i-th preset motion state of target j in the radar frame at time k. Let represent the confidence level of the i-th preset motion state of target j in radar frame at time k-1. Let represent the instantaneous classification confidence of the i-th preset motion state of target j in radar frame at time k. This represents the instantaneous classification confidence of the unknown motion state of target j in radar frame at time k. This represents the confidence level of the unknown motion state of target j in the radar frame at time k-1;

[0035] For the unknown motion state, calculate the sum of the confidence levels of 1 minus the confidence levels of other preset motion states, and use the result as the confidence level of the unknown motion state of target j in the radar frame at time k.

[0036] Secondly, embodiments of the present invention also provide a target motion state classification device, comprising:

[0037] The data acquisition module is used to acquire the raw target feature data of the current radar frame collected by the radar for the target;

[0038] The motion information determination module is used to determine the motion information of the target corresponding to the current radar frame based on the original target feature data and using a Kalman filter algorithm.

[0039] The initial state determination module is used to initially determine the motion state of the target based on the motion information, which is an initial motion state, and the initial motion state is one of a variety of preset motion states.

[0040] The confidence allocation module is used to allocate instantaneous classification confidence to each preset motion state according to a preset confidence allocation strategy, wherein the instantaneous classification confidence of the initial motion state is greater than the instantaneous classification confidence of other preset motion states, and the sum of the instantaneous classification confidence of the multiple preset motion states is 1.

[0041] The confidence calculation module is used to calculate the confidence of each preset motion state based on the instantaneous classification confidence of the multiple preset motion states and the confidence of the multiple preset motion states calculated for the previous radar frame, using a Bayesian algorithm, as the basis for determining the motion state classification result of the target for the current radar frame.

[0042] According to a specific implementation of an embodiment of the present invention, the motion information determination module is further configured to adaptively adjust the parameter values ​​of the process noise used in the Kalman filter algorithm using the following steps:

[0043] After determining the target's velocity and acceleration each time based on a radar frame, the process noise is adjusted based on the relationship between the velocity and acceleration and threshold values. The process noise is adjusted by using a preset low-noise matrix when the velocity is trending low and a preset high-noise matrix when the acceleration is trending high.

[0044] According to a specific implementation of an embodiment of the present invention, the motion information determination module is specifically used to switch the process noise to a preset low noise matrix when the process noise is currently using a preset high noise matrix, after determining the speed of the target based on a radar frame each time, and when the speed of the target determined based on a preset number of consecutive radar frames is less than a first preset speed threshold.

[0045] When the process noise currently uses the preset low-noise matrix, after determining the velocity of the target based on each radar frame, the Mahalanobis distance between the predicted position and the actual observed position of the target is calculated. The predicted position is determined based on the velocity and acceleration corresponding to the previous radar frame of the target.

[0046] When the Mahalanobis distance is less than a preset distance threshold and the target's speed is not greater than a second preset speed threshold, for the next radar frame, it is determined whether the speed of the target in the next radar frame is not greater than the second preset speed threshold, where the second preset speed threshold is less than the first preset speed threshold.

[0047] If the velocity of the target in the next radar frame is not greater than the second preset velocity threshold, the velocity of the target in the next radar frame remains unchanged.

[0048] If the velocity corresponding to the next radar frame of the target is greater than the second preset velocity threshold, and the difference between the acceleration corresponding to the next radar frame of the target and the acceleration corresponding to the current radar frame is less than the preset acceleration difference threshold, the velocity corresponding to the next radar frame of the target is adjusted to the second preset velocity threshold.

[0049] If the velocity corresponding to the next radar frame of the target is greater than the second preset velocity threshold, and the difference between the acceleration corresponding to the next radar frame of the target and the acceleration corresponding to the current radar frame is not less than the preset acceleration difference threshold, the process noise is switched to the preset high noise matrix.

[0050] According to a specific implementation of an embodiment of the present invention, the motion information corresponding to the current radar frame of the target includes: the target's x-axis velocity, y-axis velocity, x-axis acceleration, and y-axis acceleration;

[0051] The preliminary state determination module is specifically used to determine the acceleration / deceleration state of the target, as well as the predicted x-axis velocity and predicted y-axis velocity, based on the target's x-axis velocity, y-axis velocity, x-axis acceleration, and y-axis acceleration.

[0052] Based on the target's x-axis velocity, y-axis velocity, predicted x-axis velocity, and predicted y-axis velocity, as well as the target's acceleration / deceleration state and the obtained motion state of the target's associated points, the motion state of the target is preliminarily determined.

[0053] According to a specific implementation of an embodiment of the present invention, the confidence allocation module is specifically used to obtain the confidence allocation strategy corresponding to the initial motion state, wherein the confidence allocation strategy includes instantaneous classification confidence levels preset for each preset motion state.

[0054] According to the obtained confidence allocation strategy, an instantaneous classification confidence level is assigned to each of the preset motion states.

[0055] According to a specific implementation of an embodiment of the present invention, the plurality of preset motion states include: a plurality of defined motion states and an unknown motion state;

[0056] The confidence calculation module is specifically used to calculate the confidence level of each preset motion state, excluding the unknown motion state, using the following formula:

[0057] ;

[0058] ;

[0059] in, This represents the confidence level of the i-th preset motion state of target j in the radar frame at time k. Let represent the confidence level of the i-th preset motion state of target j in radar frame at time k-1. Let represent the instantaneous classification confidence of the i-th preset motion state of target j in radar frame at time k. This represents the instantaneous classification confidence of the unknown motion state of target j in radar frame at time k. This represents the confidence level of the unknown motion state of target j in the radar frame at time k-1;

[0060] For the unknown motion state, calculate the sum of the confidence levels of 1 minus the confidence levels of other preset motion states, and use the result as the confidence level of the unknown motion state of target j in the radar frame at time k.

[0061] Thirdly, embodiments of the present invention also provide an electronic device, the electronic device comprising: a housing, a processor, a memory, a circuit board, and a power supply circuit, wherein the circuit board is disposed within the space enclosed by the housing, and the processor and the memory are disposed on the circuit board; the power supply circuit is used to supply power to various circuits or devices of the above-mentioned electronic device; the memory is used to store executable program code; the processor runs a program corresponding to the executable program code by reading the executable program code stored in the memory, for executing the method described in any of the preceding claims.

[0062] Fourthly, embodiments of the present invention also provide a computer-readable storage medium storing one or more programs that can be executed by one or more processors to implement the method described in any of the foregoing implementations.

[0063] Fifthly, embodiments of the present invention also provide an application program that is executed to implement the method described in any embodiment of the present invention.

[0064] This invention provides a target motion state classification method, apparatus, and electronic device. For a tracked target, the method first acquires the original target feature data of the current radar frame collected by the radar. A Kalman filter algorithm is used to determine the motion information corresponding to the current radar frame. Then, based on the motion information, the initial motion state of the target is preliminarily determined. According to a pre-set confidence allocation strategy, an instantaneous classification confidence level is assigned to each preset motion state. This instantaneous classification confidence level is also calculated based on multiple preset motion states, as well as the confidence level of these multiple preset motion states calculated for the previous radar frame. A Bayesian algorithm is used to calculate the confidence level of each preset motion state, which serves as the basis for determining the target's motion state classification result for the current radar frame. Compared to existing technologies that determine the target's motion state based solely on the original target feature data of a single radar frame, this method, in addition to considering the target's motion features corresponding to the current radar frame, also incorporates the target's historical motion states. A Bayesian algorithm is used recursively to update the confidence levels of each preset motion state of the target in the current radar frame, thereby improving the accuracy and stability of the target's motion state classification. Attached Figure Description

[0065] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0066] Figure 1 A flowchart of a target motion state classification method provided in an embodiment of the present invention;

[0067] Figure 2 This is a flowchart of the process noise adjustment algorithm in the adaptive Kalman filter embodiment of the present invention;

[0068] Figure 3 This is a flowchart illustrating the preliminary determination of the target's motion state in an embodiment of the present invention;

[0069] Figure 4 This is a schematic diagram of the target motion state classification device provided in an embodiment of the present invention;

[0070] Figure 5 This is a schematic diagram of the structure of an electronic device provided in an embodiment of the present invention. Detailed Implementation

[0071] The embodiments of the present invention will now be described in detail with reference to the accompanying drawings. It should be understood that the described embodiments are merely some, not all, of the embodiments of the present invention. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0072] This invention provides a method for classifying the motion state of a target, such as... Figure 1 As shown, it includes:

[0073] Step 11: Obtain the raw target feature data of the current radar frame collected by the radar for the target;

[0074] Step 12: Based on the original target feature data, use the Kalman filter algorithm to determine the motion information corresponding to the current radar frame of the target;

[0075] Step 13: Based on the motion information, preliminarily determine the target's motion state as the initial motion state, which is one of several preset motion states;

[0076] Step 14: According to the preset confidence allocation strategy, assign instantaneous classification confidence to each preset motion state. The instantaneous classification confidence of the initial motion state is greater than the instantaneous classification confidence of other preset motion states, and the sum of the instantaneous classification confidence of multiple preset motion states is 1.

[0077] Step 15: Based on the instantaneous classification confidence of multiple preset motion states and the confidence of multiple preset motion states calculated for the previous radar frame, the Bayesian algorithm is used to calculate the confidence of each preset motion state, which serves as the basis for determining the motion state classification result of the target for the current radar frame.

[0078] The target motion state classification method provided in this embodiment of the invention, compared with the prior art which determines the motion state of a target based only on the original target feature data of a single radar frame, not only uses the motion features of the target corresponding to the current radar frame, but also combines the historical motion states of the target. It uses a Bayesian algorithm to recursively update the confidence of each preset motion state of the target in the current radar frame, thereby improving the accuracy and stability of the target motion state classification.

[0079] The method provided in this embodiment of the invention can be applied to scenarios where millimeter-wave radar is used as a sensor in an autonomous driving system, and further, it can also be applied to scenarios where 4D millimeter-wave radar is used as a sensor.

[0080] In practical applications, this method can correspond to a radar frame, calculate the confidence level of a target motion state classification based on a single collection of target feature data, and determine the target motion state classification result.

[0081] After acquiring the original target feature data of the radar frame at each moment through steps 11 and 12, the Kalman filter algorithm can be used to determine the motion information of the target corresponding to the current radar frame. However, the traditional Kalman filter algorithm has a certain lag in estimating the target's velocity, which may lead to inaccurate estimation results. To solve this problem, this embodiment of the invention proposes an adaptive Kalman filter algorithm to provide a more reliable estimate of the target's motion information.

[0082] The adaptive Kalman filter algorithm proposed in this embodiment of the invention can adaptively adjust the process noise parameter used in the Kalman filter algorithm. It pre-sets high noise parameter values ​​and low noise parameter values, and makes adaptive adjustments based on the motion characteristics of the target during the continuous estimation of the target's motion information based on continuous radar frames.

[0083] One embodiment of the present invention proposes that, for process noise in the Kalman filter algorithm, after determining the target's velocity and acceleration based on each radar frame, an adjustment strategy is proposed based on the magnitude relationship between the velocity and acceleration and the threshold, using a preset low-noise matrix for process noise when the velocity is trending low and a preset high-noise matrix for process noise when the acceleration is trending high, to adjust the process noise.

[0084] Furthermore, such as Figure 2 As shown, it may include the following steps:

[0085] Step 21: After determining that the Kalman filter algorithm is used to estimate the motion information of a target, set the initial value of the process noise Q of the Kalman filter algorithm to a preset high-noise matrix. .

[0086] Step 22: When the process noise Q is currently using a preset high-noise matrix At that time, after determining the target's speed (one of the target's motion information) based on each radar frame, it is determined whether the target's speed determined based on a preset number of radar frames is less than a first preset speed threshold.

[0087] If all values ​​are less than the first preset speed threshold, proceed to step 23;

[0088] If the speed is not less than the first preset speed threshold, the process noise Q is kept at the preset high noise matrix. If unchanged, return to step 22 and proceed to the judgment for the next radar frame.

[0089] Step 23: Switch the process noise Q to a preset low-noise matrix. .

[0090] When the velocities corresponding to consecutive radar frames are all less than a first preset velocity threshold, it indicates that the target's velocity is trending towards low speed. At this time, the adjustment process noise Q is a preset low-noise matrix. This allows subsequent Kalman filtering algorithms to estimate the target's motion information more accurately.

[0091] The adjusted process noise Q will be used to estimate the target's motion information based on subsequent radar frames.

[0092] Step 24: When the process noise Q is currently using a preset low-noise matrix... At that time, after determining the target's velocity based on each radar frame, the predicted position of the target is calculated. Compared with the actual observation location Mahalanobis distance between Predicted location It is determined based on the velocity and acceleration corresponding to the previous radar frame of the target, where k represents time k, and one time corresponds to one radar frame.

[0093] In this step, the Mahalanobis distance can be calculated using the following formula. :

[0094] ;

[0095] in, The Mahalanobis distance is calculated for the radar frame at time k. The predicted position of the target is calculated based on the radar frame at time k. The actual observation position is determined based on the radar frame at time k. This is the covariance matrix, which can be flexibly set according to the needs of actual applications.

[0096] Step 25: Determine whether the calculated Mahalanobis distance is less than the preset distance threshold;

[0097] If the distance is less than the preset distance threshold, proceed to step 26;

[0098] If the distance is not less than the preset distance threshold, the process noise Q is kept at the preset low noise matrix. If unchanged, return to step 24 and proceed to the judgment for the next radar frame.

[0099] Step 26: Determine whether the target's speed is not greater than the second preset speed threshold;

[0100] If the speed is not greater than the second preset speed threshold, proceed to step 27;

[0101] If the speed exceeds the second preset speed threshold, the process noise Q is kept at a preset low noise matrix. If unchanged, return to step 24 and proceed to the judgment for the next radar frame.

[0102] In this step, the second preset speed threshold is less than the first preset speed threshold. The second preset speed threshold can be set to a value close to 0, such as 0.5 m / s or 0.3 m / s.

[0103] Step 27: For the next radar frame, determine whether the velocity of the target in the next radar frame is not greater than the second preset velocity threshold.

[0104] If the velocity is not greater than the second preset velocity threshold, the velocity of the target in the next radar frame remains unchanged, and the process noise Q is kept at the preset low noise matrix. If unchanged, return to step 24 and proceed to the judgment for the next radar frame;

[0105] If the speed exceeds the second preset speed threshold, proceed to step 28.

[0106] In this embodiment of the invention, when the Mahalanobis distance is less than a preset distance threshold and the target speed is not greater than a second preset speed threshold, it indicates that the target is stationary. In this case, in order to avoid errors in speed estimation, the update of speed estimation can be restricted. That is, the update of speed estimation is restricted through steps 27, 28 and 29, thereby avoiding misjudging a stationary target as a moving target.

[0107] Step 28: Determine whether the difference between the acceleration corresponding to the next radar frame of the target and the acceleration corresponding to the current radar frame is less than a preset acceleration difference threshold.

[0108] If the difference is less than the preset acceleration difference threshold, proceed to step 29.

[0109] If the difference is not less than the preset acceleration difference threshold, proceed to step 210.

[0110] Step 29: Adjust the velocity of the target in the next radar frame to the second preset velocity threshold.

[0111] Step 210: Switch the process noise Q to a preset high-noise matrix. Then return to step 22 to proceed with the judgment for the next radar frame.

[0112] When the velocity of the target in the next radar frame is greater than the second preset velocity threshold, and the difference between the acceleration of the target in the next radar frame and the acceleration of the target in the current radar frame is not less than the preset acceleration difference threshold, it indicates that the target's velocity is increasing and the acceleration is trending towards higher acceleration, i.e., the acceleration has suddenly increased. At this time, the adjustment process noise Q is a preset high noise matrix. This allows subsequent Kalman filtering algorithms to estimate the target's motion information more accurately.

[0113] In this embodiment of the invention, a high-noise matrix is ​​preset. Preset low-noise matrix The preset number of times threshold, the first preset speed threshold, the second preset speed threshold, the preset distance threshold, and the preset acceleration difference threshold can be flexibly set through testing based on the actual application scenario and requirements. No further examples or detailed descriptions will be provided here.

[0114] Using the above-described embodiments of the present invention Figure 2 The process shown adaptively adjusts the noise of the Kalman filter algorithm, which can improve the accuracy of estimating the target's motion information, including the accuracy of estimating the target's velocity and acceleration.

[0115] In this embodiment of the invention, based on the original target feature data of the radar frame, the Kalman filter algorithm is used to determine the target's motion information, which may include the target's x-axis velocity, y-axis velocity, x-axis acceleration, and y-axis acceleration. Correspondingly, for step 13 above, the target's motion state is initially determined based on the motion information corresponding to the current radar frame, such as... Figure 3 As shown, it may include the following steps:

[0116] Step 31: Based on the target's x-axis velocity y-axis velocity x-axis acceleration and y-axis acceleration Determine the target's acceleration / deceleration state and predict the velocity along the x-axis. and y-axis prediction speed .

[0117] Step 32: Obtain the motion state of the associated points of the target.

[0118] In this step, the motion state of the target's associated points can be determined based on the feature data of the pixels or pixel blocks contained in the target in the original target feature data of the current radar frame. The original target feature data includes information such as the position, velocity, height, and noise of the pixels or pixel blocks contained in the target point cloud acquired by the radar.

[0119] There is no strict order between steps 31 and 32 above.

[0120] Through steps 31 and 32 above, the observation data corresponding to the target's current radar frame (radar frame at time k) is obtained. .

[0121] Step 33: Based on the target's x-axis velocity, y-axis velocity, x-axis predicted velocity, y-axis predicted velocity, target's acceleration / deceleration state, and the obtained motion state of the target's associated points, preliminarily determine the target's motion state.

[0122] In this step, the initially determined motion state of the target is the initial motion state, which is one of a variety of preset motion states. In one embodiment, it may include: a stationary state, a moving state, a stopped state, and an unknown motion state, of which the first three are definite motion states.

[0123] Stationary state: The target remains in the same position (velocity close to 0 and no acceleration);

[0124] Motion state: The target has a significant velocity (exceeding the system's detection velocity threshold);

[0125] Stopped state: The target changes from a moving state to a stationary state.

[0126] After determining the initial motion state of the target for the current radar frame, step 14 is executed, that is, according to the preset confidence allocation strategy, an instantaneous classification confidence is assigned to each preset motion state. The instantaneous classification confidence of the initial motion state must be greater than the instantaneous classification confidence of other preset motion states, and the sum of the instantaneous classification confidence of multiple preset motion states must be 1.

[0127] In one embodiment of the present invention, the following steps may be adopted:

[0128] First, obtain the confidence allocation strategy corresponding to the initial motion state. This confidence allocation strategy includes the instantaneous classification confidence level pre-set for each preset motion state. That is, each preset motion state corresponds to a specific confidence allocation strategy. In practical applications, the corresponding instantaneous classification confidence level can be pre-set for each preset motion state based on actual needs.

[0129] Then, according to the obtained confidence allocation strategy, an instantaneous classification confidence level is assigned to each preset motion state.

[0130] In this embodiment of the invention, the instantaneous classification confidence level preset for each preset motion state can be flexibly set based on the test results and actual needs in actual road tests, and will not be described in detail here.

[0131] In this embodiment of the invention, there are multiple preset motion states, including multiple defined motion states and unknown motion states. Defined motion states can be understood as motion states that are definite and have been clearly defined. For example, as mentioned above, they can include static states, motion states, and stopped states.

[0132] After assigning instantaneous classification confidence to each preset motion state for the current radar frame according to the preset confidence allocation strategy, step 15 above can be executed. That is, based on the instantaneous classification confidence of the multiple preset motion states and the confidence of the multiple preset motion states calculated for the previous radar frame, the Bayesian algorithm is used to calculate the confidence of each preset motion state.

[0133] In one embodiment of the present invention, for various preset motion states other than unknown motion states, the confidence level of each preset motion state can be calculated using the following formula:

[0134] ;

[0135] ;

[0136] in, This represents the confidence level of the i-th preset motion state of target j in the radar frame at time k. Let represent the confidence level of the i-th preset motion state of target j in radar frame at time k-1. Let represent the instantaneous classification confidence of the i-th preset motion state of target j in radar frame at time k. This represents the instantaneous classification confidence of the unknown motion state of target j in radar frame at time k. This represents the confidence level of the unknown motion state of target j in the radar frame at time k-1;

[0137] For an unknown motion state, calculate the sum of the confidence levels of 1 minus the confidence levels of other preset motion states, and use the result as the confidence level of the unknown motion state of target j in the radar frame at time k.

[0138] The confidence level of an unknown motion state is calculated using the following formula:

[0139] ;

[0140] in, Let n be the confidence level of the unknown motion state of target j in the radar frame at time k, and n be the number of other definite motion states besides the positional motion state.

[0141] The confidence levels of the various preset motion states calculated using the method provided in the embodiments of the present invention can be used as the basis for determining the motion state classification result of the target for the current radar frame.

[0142] For example, the preset motion state corresponding to the highest confidence level can be selected as the motion state of the target determined for the current radar frame.

[0143] Based on the same inventive concept, and according to the target pitch classification confidence calculation method provided in the above embodiments of the present invention, another embodiment of the present invention also provides a target pitch classification confidence calculation method, the structural schematic diagram of which is shown below. Figure 4 As shown, it specifically includes:

[0144] The data acquisition module 41 is used to acquire the original target feature data of the current radar frame collected by the radar for the target;

[0145] The motion information determination module 42 is used to determine the motion information of the target corresponding to the current radar frame based on the original target feature data and using a Kalman filter algorithm.

[0146] The preliminary state determination module 43 is used to preliminarily determine the motion state of the target based on the motion information, as an initial motion state, wherein the initial motion state is one of a variety of preset motion states;

[0147] The confidence allocation module 44 is used to allocate instantaneous classification confidence to each preset motion state according to a preset confidence allocation strategy, wherein the instantaneous classification confidence of the initial motion state is greater than the instantaneous classification confidence of other preset motion states, and the sum of the instantaneous classification confidence of the multiple preset motion states is 1.

[0148] The confidence calculation module 45 is used to calculate the confidence of each preset motion state based on the instantaneous classification confidence of the multiple preset motion states and the confidence of the multiple preset motion states calculated for the previous radar frame, using a Bayesian algorithm, as the basis for determining the motion state classification result of the target for the current radar frame.

[0149] According to a specific implementation of an embodiment of the present invention, the motion information determination module 42 is further configured to adaptively adjust the parameter values ​​of the process noise used in the Kalman filter algorithm using the following steps:

[0150] After determining the target's velocity and acceleration each time based on a radar frame, the process noise is adjusted based on the relationship between the velocity and acceleration and threshold values. The process noise is adjusted by using a preset low-noise matrix when the velocity is trending low and a preset high-noise matrix when the acceleration is trending high.

[0151] According to a specific implementation of an embodiment of the present invention, the motion information determination module 42 is specifically used to switch the process noise to a preset low noise matrix when the process noise is currently using a preset high noise matrix, after determining the speed of the target based on a radar frame each time, and when the speed of the target determined based on a preset number of consecutive radar frames is less than a first preset speed threshold.

[0152] When the process noise currently uses the preset low-noise matrix, after determining the velocity of the target based on each radar frame, the Mahalanobis distance between the predicted position and the actual observed position of the target is calculated. The predicted position is determined based on the velocity and acceleration corresponding to the previous radar frame of the target.

[0153] When the Mahalanobis distance is less than a preset distance threshold and the target's speed is not greater than a second preset speed threshold, for the next radar frame, it is determined whether the speed of the target in the next radar frame is not greater than the second preset speed threshold, where the second preset speed threshold is less than the first preset speed threshold.

[0154] If the velocity of the target in the next radar frame is not greater than the second preset velocity threshold, the velocity of the target in the next radar frame remains unchanged.

[0155] If the velocity corresponding to the next radar frame of the target is greater than the second preset velocity threshold, and the difference between the acceleration corresponding to the next radar frame of the target and the acceleration corresponding to the current radar frame is less than the preset acceleration difference threshold, the velocity corresponding to the next radar frame of the target is adjusted to the second preset velocity threshold.

[0156] If the velocity corresponding to the next radar frame of the target is greater than the second preset velocity threshold, and the difference between the acceleration corresponding to the next radar frame of the target and the acceleration corresponding to the current radar frame is not less than the preset acceleration difference threshold, the process noise is switched to the preset high noise matrix.

[0157] According to a specific implementation of an embodiment of the present invention, the motion information corresponding to the current radar frame of the target includes: the target's x-axis velocity, y-axis velocity, x-axis acceleration, and y-axis acceleration;

[0158] The preliminary state determination module 43 is specifically used to determine the acceleration / deceleration state of the target, as well as the predicted x-axis velocity and predicted y-axis velocity, based on the target's x-axis velocity, y-axis velocity, x-axis acceleration, and y-axis acceleration.

[0159] Based on the target's x-axis velocity, y-axis velocity, predicted x-axis velocity, and predicted y-axis velocity, as well as the target's acceleration / deceleration state and the obtained motion state of the target's associated points, the motion state of the target is preliminarily determined.

[0160] According to a specific implementation of an embodiment of the present invention, the confidence allocation module 44 is specifically used to obtain the confidence allocation strategy corresponding to the initial motion state, wherein the confidence allocation strategy includes instantaneous classification confidence levels preset for each preset motion state.

[0161] According to the obtained confidence allocation strategy, an instantaneous classification confidence level is assigned to each of the preset motion states.

[0162] According to a specific implementation of an embodiment of the present invention, the plurality of preset motion states include: a plurality of defined motion states and an unknown motion state;

[0163] The confidence calculation module 45 is specifically used to calculate the confidence level of each preset motion state, excluding the unknown motion state, using the following formula:

[0164] ;

[0165] ;

[0166] in, This represents the confidence level of the i-th preset motion state of target j in the radar frame at time k. Let represent the confidence level of the i-th preset motion state of target j in radar frame at time k-1. Let represent the instantaneous classification confidence of the i-th preset motion state of target j in radar frame at time k. This represents the instantaneous classification confidence of the unknown motion state of target j in radar frame at time k. This represents the confidence level of the unknown motion state of target j in the radar frame at time k-1;

[0167] For the unknown motion state, calculate the sum of the confidence levels of 1 minus the confidence levels of other preset motion states, and use the result as the confidence level of the unknown motion state of target j in the radar frame at time k.

[0168] The target motion state classification device described in this embodiment of the invention can be used to perform... Figures 1 to 3 The technical solutions of the method embodiments shown are similar in principle and in effect, and will not be described again here.

[0169] This invention also provides an electronic device, which includes the apparatus described in any of the foregoing embodiments.

[0170] Figure 5 This is a schematic diagram of the structure of an embodiment of the electronic device of the present invention, which can realize the present invention. Figure 1-3 The process of the illustrated embodiment is as follows: Figure 5 As shown, the aforementioned electronic device may include: a housing 51, a processor 52, a memory 53, a circuit board 54, and a power supply circuit 55. The circuit board 54 is disposed inside the space enclosed by the housing 51, and the processor 52 and the memory 53 are disposed on the circuit board 54. The power supply circuit 55 is used to supply power to the various circuits or devices of the aforementioned electronic device. The memory 53 is used to store executable program code. The processor 52 runs the program corresponding to the executable program code by reading the executable program code stored in the memory 53, and is used to execute the target motion state classification method described in any of the foregoing embodiments.

[0171] For details on the specific execution process of the above steps by the processor 52, and the steps further executed by the processor 52 through running executable program code, please refer to the present invention. Figure 1-3 The description of the illustrated embodiments will not be repeated here.

[0172] This electronic device exists in various forms, including but not limited to: (1) Mobile communication devices: These devices are characterized by their mobile communication capabilities and are primarily designed to provide voice and data communication. These terminals include smartphones (such as iPhones), multimedia phones, feature phones, and low-end phones.

[0173] (2) Ultra-mobile personal computer devices: These devices fall under the category of personal computers, possessing computing and processing capabilities, and generally also have mobile internet access features. These terminals include PDAs, MIDs, and UMPCs, such as the iPad.

[0174] (3) Portable entertainment devices: These devices can display and play multimedia content. This category includes: audio and video players (such as iPods), handheld game consoles, e-books, as well as smart toys and portable car navigation devices.

[0175] (4) Server: A device that provides computing services. The components of a server include a processor, hard disk, memory, system bus, etc. Servers are similar to general computer architectures, but because they need to provide highly reliable services, they have higher requirements in terms of processing power, stability, reliability, security, scalability, and manageability.

[0176] (5) Other electronic devices with data interaction functions.

[0177] Embodiments of the present invention also provide a computer-readable storage medium storing one or more programs that can be executed by one or more processors to implement the method described in any of the foregoing implementations.

[0178] Embodiments of the present invention also provide an application program that is executed to implement the methods provided in any embodiment of the present invention.

[0179] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, 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. Without further limitations, 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 said element.

[0180] The various embodiments in this specification are described in a related manner. The same or similar parts between the various embodiments can be referred to each other. Each embodiment focuses on describing the differences from other embodiments.

[0181] In particular, the device embodiment is basically similar to the method embodiment, so the description is relatively simple. For relevant details, please refer to the description of the method embodiment.

[0182] For ease of description, the above apparatus is described by dividing it into various functional units / modules. Of course, in implementing this invention, the functions of each unit / module can be implemented in one or more software and / or hardware.

[0183] Those skilled in the art will understand that all or part of the processes in the above embodiments can be implemented by a computer program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, it can include the processes of the embodiments of the above methods. The storage medium can be a magnetic disk, optical disk, read-only memory (ROM), or random access memory (RAM), etc.

[0184] The above description is merely a specific embodiment of the present invention, but the scope of protection of the present invention is not limited thereto. Any variations or substitutions that can be easily conceived by those skilled in the art within the technical scope disclosed in the present invention should be included within the scope of protection of the present invention. Therefore, the scope of protection of the present invention should be determined by the scope of the claims.

Claims

1. A method for classifying the motion state of a target, characterized in that, include: Acquire the raw target feature data of the current radar frame collected by the radar for the target; Based on the original target feature data, the Kalman filter algorithm is used to determine the motion information corresponding to the current radar frame of the target; Based on the motion information, the motion state of the target is initially determined as the initial motion state, which is one of a variety of preset motion states; According to a preset confidence allocation strategy, an instantaneous classification confidence score is assigned to each preset motion state, wherein the instantaneous classification confidence score of the initial motion state is greater than the instantaneous classification confidence scores of other preset motion states, and the sum of the instantaneous classification confidence scores of the multiple preset motion states is 1. Based on the instantaneous classification confidence of the various preset motion states and the confidence of the various preset motion states calculated for the previous radar frame, a Bayesian algorithm is used to calculate the confidence of each preset motion state, which serves as the basis for determining the motion state classification result of the target for the current radar frame. The following steps are used to adaptively adjust the parameter values ​​of the process noise used in the Kalman filter algorithm: When the process noise currently uses a preset high noise matrix, after the speed of the target is determined based on a radar frame each time, if the speed of the target determined based on a preset number of consecutive radar frames is less than a first preset speed threshold, the process noise is switched to a preset low noise matrix. When the process noise currently uses the preset low-noise matrix, after determining the velocity of the target based on each radar frame, the Mahalanobis distance between the predicted position and the actual observed position of the target is calculated. The predicted position is determined based on the velocity and acceleration corresponding to the previous radar frame of the target. When the Mahalanobis distance is less than a preset distance threshold and the target's speed is not greater than a second preset speed threshold, for the next radar frame, it is determined whether the speed of the target in the next radar frame is not greater than the second preset speed threshold, where the second preset speed threshold is less than the first preset speed threshold. If the velocity of the target in the next radar frame is not greater than the second preset velocity threshold, the velocity of the target in the next radar frame remains unchanged. If the velocity corresponding to the next radar frame of the target is greater than the second preset velocity threshold, and the difference between the acceleration corresponding to the next radar frame of the target and the acceleration corresponding to the current radar frame is less than the preset acceleration difference threshold, the velocity corresponding to the next radar frame of the target is adjusted to the second preset velocity threshold. If the velocity corresponding to the next radar frame of the target is greater than the second preset velocity threshold, and the difference between the acceleration corresponding to the next radar frame of the target and the acceleration corresponding to the current radar frame is not less than the preset acceleration difference threshold, the process noise is switched to the preset high noise matrix.

2. The method according to claim 1, characterized in that, The motion information corresponding to the current radar frame of the target includes: the target's x-axis velocity, y-axis velocity, x-axis acceleration, and y-axis acceleration; The preliminary determination of the target's motion state based on the motion information includes: Based on the target's x-axis velocity, y-axis velocity, x-axis acceleration, and y-axis acceleration, determine the target's acceleration / deceleration state, as well as the predicted x-axis velocity and y-axis velocity; Based on the target's x-axis velocity, y-axis velocity, predicted x-axis velocity, and predicted y-axis velocity, as well as the target's acceleration / deceleration state and the obtained motion state of the target's associated points, the motion state of the target is preliminarily determined.

3. The method according to claim 1, characterized in that, The step of assigning instantaneous classification confidence to each preset motion state according to a preset confidence allocation strategy includes: Obtain the confidence allocation strategy corresponding to the initial motion state, which includes instantaneous classification confidence levels pre-set for each preset motion state; According to the obtained confidence allocation strategy, an instantaneous classification confidence level is assigned to each of the preset motion states.

4. The method according to claim 1, characterized in that, The various preset motion states include: various defined motion states and unknown motion states; The instantaneous classification confidence based on the multiple preset motion states, and the confidence of the multiple preset motion states calculated for the previous radar frame, are used to calculate the confidence of each preset motion state using a Bayesian algorithm, including: For various preset motion states other than the unknown motion state, the confidence level of each preset motion state is calculated using the following formula: ； ； in, This represents the confidence level of the i-th preset motion state of target j in the radar frame at time k. Let represent the confidence level of the i-th preset motion state of target j in radar frame at time k-1. Let represent the instantaneous classification confidence of the i-th preset motion state of target j in radar frame at time k. This represents the instantaneous classification confidence of the unknown motion state of target j in radar frame at time k. This represents the confidence level of the unknown motion state of target j in the radar frame at time k-1; For the unknown motion state, calculate the sum of the confidence levels of 1 minus the confidence levels of other preset motion states, and use the result as the confidence level of the unknown motion state of target j in the radar frame at time k.

5. A target motion state classification device, characterized in that, include: The data acquisition module is used to acquire the raw target feature data of the current radar frame collected by the radar for the target; The motion information determination module is used to determine the motion information of the target corresponding to the current radar frame based on the original target feature data and using a Kalman filter algorithm. The initial state determination module is used to initially determine the motion state of the target based on the motion information, which is an initial motion state, and the initial motion state is one of a variety of preset motion states. The confidence allocation module is used to allocate instantaneous classification confidence to each preset motion state according to a preset confidence allocation strategy, wherein the instantaneous classification confidence of the initial motion state is greater than the instantaneous classification confidence of other preset motion states, and the sum of the instantaneous classification confidence of the multiple preset motion states is 1. The confidence calculation module is used to calculate the confidence of each preset motion state based on the instantaneous classification confidence of the multiple preset motion states and the confidence of the multiple preset motion states calculated for the previous radar frame. The module uses a Bayesian algorithm to calculate the confidence of each preset motion state as the basis for determining the motion state classification result of the target for the current radar frame. The motion information determination module is further configured to adaptively adjust the parameter values ​​of the process noise used in the Kalman filter algorithm using the following steps: When the process noise currently uses a preset high noise matrix, after the speed of the target is determined based on a radar frame each time, if the speed of the target determined based on a preset number of consecutive radar frames is less than a first preset speed threshold, the process noise is switched to a preset low noise matrix. When the process noise currently uses the preset low-noise matrix, after determining the velocity of the target based on each radar frame, the Mahalanobis distance between the predicted position and the actual observed position of the target is calculated. The predicted position is determined based on the velocity and acceleration corresponding to the previous radar frame of the target. When the Mahalanobis distance is less than a preset distance threshold and the target's speed is not greater than a second preset speed threshold, for the next radar frame, it is determined whether the speed of the target in the next radar frame is not greater than the second preset speed threshold, where the second preset speed threshold is less than the first preset speed threshold. If the velocity of the target in the next radar frame is not greater than the second preset velocity threshold, the velocity of the target in the next radar frame remains unchanged. If the velocity corresponding to the next radar frame of the target is greater than the second preset velocity threshold, and the difference between the acceleration corresponding to the next radar frame of the target and the acceleration corresponding to the current radar frame is less than the preset acceleration difference threshold, the velocity corresponding to the next radar frame of the target is adjusted to the second preset velocity threshold. If the velocity corresponding to the next radar frame of the target is greater than the second preset velocity threshold, and the difference between the acceleration corresponding to the next radar frame of the target and the acceleration corresponding to the current radar frame is not less than the preset acceleration difference threshold, the process noise is switched to the preset high noise matrix.

6. An electronic device, characterized in that, The electronic device includes: a housing, a processor, a memory, a circuit board, and a power supply circuit, wherein the circuit board is disposed inside the space enclosed by the housing, and the processor and the memory are disposed on the circuit board; the power supply circuit is used to supply power to various circuits or devices of the electronic device; the memory is used to store executable program code; the processor runs a program corresponding to the executable program code by reading the executable program code stored in the memory, for executing the method described in any one of claims 1-4.

7. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores one or more programs, which can be executed by one or more processors to implement the method described in any one of claims 1-4.

Images