Target state determination method and device

Abstract

The invention discloses a target state determination method and device, and relates to the technical field of automatic driving. A specific embodiment of the method comprises the steps of obtaining a detection result of a detector on a first type of motion features of a target at the current moment, wherein the detection result comprises a detection state value at the current moment and a detection score representing detection reliability; adjusting an initial value of the measurement noise data according to the detection score to obtain a current value of the measurement noise data; obtaining a state optimal estimation value and a state covariance optimal estimation value of the second type of motion features of the target at the previous moment, and determining a Kalman coefficient at the current moment according to the state covariance optimal estimation value and the current value of the measurement noise data; and determining the optimal state estimation value of the second type of motion features of the target at the current moment by using the optimal state estimation value, the Kalman coefficient at the current moment and the detection state value at the current moment. According to the embodiment, the calculation accuracy of the Kalman filtering algorithm can be improved.

Description

Technical Field

[0001] This invention relates to the field of autonomous driving technology, and in particular to a method and apparatus for determining a target state. Background Technology

[0002] In multi-object tracking scenarios for autonomous driving, the Kalman filter algorithm is frequently used to estimate the state of moving objects. Traditional Kalman filters typically treat the detector's measurement noise as a constant for state estimation at each time step. However, in practical applications, detectors are susceptible to external interference. If the measurement noise is still treated as a constant in the calculation, it affects the accuracy of the Kalman filter algorithm, and consequently, the judgment results of the autonomous driving perception system. Summary of the Invention

[0003] In view of this, embodiments of the present invention provide a target state determination method and apparatus that can adjust the measurement noise data of the detector according to the detection score at the current moment, thereby improving the calculation accuracy of the Kalman filter algorithm.

[0004] To achieve the above objectives, according to one aspect of the present invention, a method for determining a target state is provided.

[0005] The target state determination method of this invention includes: acquiring the detection result of the detector on the first type of motion feature of the target at the current moment, the detection result including: the detection state value at the current moment and the detection score characterizing the detection reliability; adjusting the preset initial value of the measurement noise data according to the detection score to obtain the current value of the measurement noise data; acquiring the optimal state estimate and the optimal state covariance estimate of the second type of motion feature of the target at the previous moment; determining the Kalman coefficient at the current moment according to the optimal state covariance estimate and the current value of the measurement noise data; and determining the optimal state estimate of the second type of motion feature of the target at the current moment using the optimal state estimate, the Kalman coefficient at the current moment, and the detection state value at the current moment.

[0006] Optionally, for the same detector, the current value of the measurement noise data is negatively correlated with the detection reliability characterized by the detection score.

[0007] Optionally, adjusting the preset initial value of the measurement noise data according to the detection score to obtain the current value of the measurement noise data includes: obtaining an opposition number based on the detection score at the current moment, determining an adjustment coefficient based on the opposition number, and multiplying the initial value of the measurement noise data by the adjustment coefficient to obtain the current value of the measurement noise data.

[0008] Optionally, obtaining the counter-number based on the detection score at the current moment and determining the adjustment coefficient based on the counter-number includes: subtracting the detection score at the current moment from a preset detection score threshold to obtain the counter-number; determining the product of the sign function value of the counter-number and the counter-number as a first value; determining the sum of the first value and 1 as a second value; and determining the product of the second value and a preset proportional coefficient as the adjustment coefficient.

[0009] Optionally, determining the Kalman coefficients at the current moment based on the optimal estimate of the state covariance and the current value of the measurement noise data includes: determining the prior estimate of the state covariance at the current moment based on the optimal estimate of the state covariance at the previous moment and preset process noise data; and determining the Kalman coefficients at the current moment using the prior estimate of the state covariance at the current moment, the current value of the measurement noise data, and the transformation matrix from the second type of motion feature to the first type of motion feature.

[0010] Optionally, determining the optimal state estimate of the second type of motion feature of the target at the current moment using the optimal state estimate, the Kalman coefficient at the current moment, and the detection state value at the current moment includes: determining the prior state estimate at the current moment based on the optimal state estimate at the previous moment; and determining the optimal state estimate of the second type of motion feature of the target at the current moment based on the prior state estimate, the Kalman coefficient at the current moment, the detection state value at the current moment, and the transformation matrix.

[0011] Optionally, the method further includes: determining the optimal estimate of the state covariance of the second type of motion feature of the target at the current time based on the Kalman coefficients at the current time, the prior estimate of the state covariance at the current time, and the transformation matrix.

[0012] To achieve the above objectives, according to another aspect of the present invention, a target state determination apparatus is provided.

[0013] The target state determination device according to the present invention includes: the target state determination device provided in the present invention may include: a detection result acquisition unit, a measurement noise adjustment unit, and a calculation unit.

[0014] The detection result acquisition unit is used to acquire the detection result of the detector on the first type of motion feature of the target at the current moment. The detection result includes: the detection state value at the current moment and the detection score characterizing the detection reliability. The measurement noise adjustment unit is used to adjust the preset initial value of the measurement noise data according to the detection score to obtain the current value of the measurement noise data. The calculation unit is used to acquire the optimal state estimate and the optimal state covariance estimate of the second type of motion feature of the target at the previous moment, determine the Kalman coefficient at the current moment according to the optimal state covariance estimate and the current value of the measurement noise data, and determine the optimal state estimate of the second type of motion feature of the target at the current moment using the optimal state estimate, the Kalman coefficient at the current moment, and the detection state value at the current moment.

[0015] Optionally, for the same detector, the current value of the measurement noise data is negatively correlated with the detection reliability characterized by the detection score.

[0016] Optionally, the measurement noise adjustment unit is further configured to: obtain an opposition number based on the detection score at the current moment, determine an adjustment coefficient based on the opposition number, and multiply the initial value of the measurement noise data by the adjustment coefficient to obtain the current value of the measurement noise data.

[0017] Optionally, the measurement noise adjustment unit is further configured to: subtract the detection score at the current moment from a preset detection score threshold to obtain an anti-value; determine the product of the sign function value of the anti-value and the anti-value as a first value; determine the sum of the first value and 1 as a second value; and determine the product of the second value and a preset proportional coefficient as an adjustment coefficient.

[0018] Optionally, the calculation unit is further configured to: determine the prior estimate of the state covariance at the current moment based on the optimal estimate of the state covariance at the previous moment and the preset process noise data; and determine the Kalman coefficients at the current moment using the prior estimate of the state covariance at the current moment, the current value of the measurement noise data, and the transformation matrix from the second type of motion feature to the first type of motion feature.

[0019] Optionally, the computing unit is further configured to: determine the state prior estimate at the current moment based on the state optimal estimate at the previous moment; and determine the state optimal estimate of the second type of motion feature of the target at the current moment based on the state prior estimate at the current moment, the Kalman coefficient at the current moment, the detection state value at the current moment, and the transformation matrix.

[0020] Optionally, the calculation unit is also used to: determine the optimal estimate of the state covariance of the second type of motion characteristics of the target at the current time based on the Kalman coefficients at the current time, the prior estimate of the state covariance at the current time, and the transformation matrix.

[0021] To achieve the above objectives, according to another aspect of the present invention, an electronic device is provided.

[0022] An electronic device according to the present invention includes: one or more processors; and a storage device for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the target state determination method provided by the present invention.

[0023] To achieve the above objectives, according to another aspect of the present invention, a computer-readable storage medium is provided.

[0024] The present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the target state determination method provided by the present invention.

[0025] According to the technical solution of the present invention, the embodiments described above have the following advantages or beneficial effects: After obtaining the detector's current state value and the detection score representing detection reliability, the initial value of the measurement noise data is adjusted based on the detection score to obtain the current value of the measurement noise data. For the same detector, this adjustment makes the current value of the measurement noise data negatively correlated with the detection reliability represented by the detection score. Then, a Kalman filter algorithm is executed based on the current detection state value and the current value of the measurement noise data to determine the optimal state estimate for the current moment. This allows for adaptive adjustment of the measurement noise data based on the detection score at different times. When the detection reliability represented by the detection score is high, the measurement noise data is reduced to increase the Kalman coefficient value, thereby increasing the weight of the detection component in the final calculation result. Conversely, when the detection reliability represented by the detection score is low, the measurement noise data is increased to decrease the Kalman coefficient value, thereby reducing the weight of the detection component in the final calculation result. Ultimately, this avoids the influence of external interference on the state estimation of the detector, improving the accuracy and stability of the Kalman filter algorithm.

[0026] The further effects of the aforementioned unconventional alternative methods will be explained below in conjunction with specific implementation methods. Attached Figure Description

[0027] The accompanying drawings are provided to better understand the invention and are not intended to unduly limit the scope of the invention. Wherein: Figure 1 This is a schematic diagram of the main steps of the target state determination method in an embodiment of the present invention; Figure 2 This is a schematic diagram illustrating the execution of the Kalman filtering algorithm according to an embodiment of the present invention; Figure 3 This is a schematic diagram of the components of the target state determination device in an embodiment of the present invention; Figure 4 This is an exemplary system architecture diagram that can be applied thereto according to embodiments of the present invention; Figure 5 This is a schematic diagram of the electronic device structure used to implement the target state determination method in the embodiments of the present invention. Detailed Implementation

[0028] The following description, in conjunction with the accompanying drawings, illustrates exemplary embodiments of the present invention, including various details to aid understanding. These details should be considered merely exemplary. Therefore, those skilled in the art will recognize that various changes and modifications can be made to the embodiments described herein without departing from the scope and spirit of the invention. Similarly, for clarity and brevity, descriptions of well-known functions and structures are omitted in the following description.

[0029] It should be noted that, unless otherwise specified, the embodiments of the present invention and the technical features thereof can be combined with each other.

[0030] Figure 1 This is a schematic diagram of the main steps of the target state determination method in an embodiment of the present invention.

[0031] like Figure 1 As shown, the target state determination method of this invention can be executed by a computing device, and the specific execution steps are as follows: Step S101: Obtain the detection results of the detector for the first type of motion features of the target at the current moment. The detection results include: the detection state value at the current moment and the detection score characterizing the detection reliability.

[0032] In this step, the detector can be various sensors such as LiDAR, image acquisition devices, etc. The detection state value can be detection state data of any dimension, such as position, velocity, direction, size, etc. The detection score is used to represent the reliability and confidence level of this detection. The detection score can be positively or negatively correlated with the detection reliability. The following explanation uses the example of the detection score being positively correlated with the detection reliability. For example, the detection score can take a value between zero and 1. The target can be any object to be detected. The first type of motion feature includes at least one motion feature that the detector is targeting. Motion features refer to physical quantities related to motion, such as position, time, velocity, acceleration, etc.

[0033] Step S102: Adjust the preset initial value of the measured noise data according to the detection score to obtain the current value of the measured noise data.

[0034] In practical applications, the measurement noise data can be the variance or covariance data of the measurement noise. The initial value of the measurement noise data can be a fixed value or a value generated according to preset rules. In this step, the computing device can adjust the initial value of the measurement noise data according to the detection score to obtain the current value of the measurement noise data. Preferably, for the same detector, the computing device can make the current value of the measurement noise data negatively correlated with the detection reliability represented by the detection score through the above adjustments. That is, in the same detector, the higher the detection reliability represented by the detection score, the smaller the corresponding current value of the measurement noise data. This makes higher detection reliability correspond to a smaller current value of the measurement noise data, i.e., a larger Kalman coefficient, which means that the detection part has a higher weight in the final Kalman filter calculation result. Similarly, lower detection reliability corresponds to a larger current value of the measurement noise data, i.e., a smaller Kalman coefficient, which means that the detection part has a lower weight in the final Kalman filter calculation result. This improves the accuracy of the Kalman filter algorithm and the stability of the autonomous driving perception system.

[0035] In practical applications, computing devices can be adjusted in various ways to make the current value of the measured noise data negatively correlated with the detection reliability represented by the detection score. These methods can be achieved through various applicable types of functions, such as the following functions:

[0036]

[0037] in, To measure the current value of the noise data, To provide the initial values ​​for measuring noise data, To test the score, The preset detection score threshold, This is a preset scaling factor.

[0038] The following provides a superior method for adjusting measurement noise data, which is not intended to limit the possible adjustment methods. Specifically, the computing device obtains an offset number based on the detection score at the current moment, determines an adjustment coefficient based on the offset number, and then multiplies the initial value of the measurement noise data by the adjustment coefficient to obtain the current value of the measurement noise data. Thus, the initial value of the measurement noise data is adjusted using the offset number of the detection score and the adjustment coefficient in the above negative correlation form. In an optional implementation, the computing device first subtracts the detection score at the current moment from a preset detection score threshold to obtain an offset number; then, the product of the sign function value of the offset number and the offset number is determined as a first value; the sum of the first value and 1 is determined as a second value; the product of the second value and a preset proportionality coefficient is determined as the adjustment coefficient; finally, the initial value of the measurement noise data is multiplied by the adjustment coefficient to obtain the current value of the measurement noise data. This can be expressed as a formula as follows:

[0039]

[0040]

[0041]

[0042] in, For the function used to calculate the complement, For symbolic functions, The first value, This is the second value.

[0043] In the above adjustment methods, the detection score threshold is... The configuration can be adjusted based on the detector's performance. If the detection performance is good, the setting can be increased; if the performance is poor, the setting can be decreased. This adjustment method ensures that the current value of the measured noise data is negatively correlated with the detection reliability represented by the detection score, while allowing the current value of the measured noise data to fluctuate around its initial value, facilitating practical Kalman filter calculations. It is understood that the above-mentioned opposite number and adjustment coefficient can be calculated using other methods that achieve the same effect. For example, the ratio of a fixed value to the detection score can be used as the opposite number, or the cube of the opposite number can be used as the adjustment coefficient.

[0044] Step S103: Obtain the optimal state estimate and the optimal state covariance estimate of the second type of motion feature of the target at the previous time. Determine the Kalman coefficient at the current time based on the optimal state covariance estimate and the current value of the measurement noise data. Use the optimal state estimate, the Kalman coefficient at the current time, and the detection state value at the current time to determine the optimal state estimate of the second type of motion feature of the target at the current time.

[0045] The second type of motion features includes at least one motion feature used to describe the target state. These second-type motion features are related to the first type of motion features. In practical applications, the second-type motion features can be completely identical or completely different from the first-type motion features, and they may also overlap. For example, if the second-type motion features are selected as position and velocity, then the first-type motion features can be only position, only velocity, only acceleration, or a combination of position and velocity, position and acceleration, or a combination of all three. In specific applications, a transformation matrix can be used to transform the data of the second-type motion features into the data of the first-type motion features.

[0046] After obtaining the current detection state value and the current measurement noise data value, the computing device can execute the Kalman filter algorithm to calculate the optimal state estimate and the optimal state covariance estimate at the current moment. Specifically, the computing device can determine the prior state estimate at the current moment based on the optimal state estimate from the previous moment, i.e.:

[0047]

[0048] in, This represents the prior estimate of the state at the current moment. Let F represent the optimal state estimate at the previous time step, and B represent the state transition matrix and the control matrix. This indicates the control quantity.

[0049] The computing device determines the prior estimate of the state covariance at the current moment based on the optimal estimate of the state covariance at the previous moment and the preset process noise data, that is:

[0050]

[0051] in, This represents the prior estimate of the state covariance at the current moment. This represents the optimal estimate of the state covariance at the previous time step. Denotes the transpose of F. This represents the process noise data (i.e., the initial value of the process noise data).

[0052] The computing device uses the prior estimate of the state covariance at the current moment, the current value of the measurement noise data, and the transformation matrix from the second type of motion feature to the first type of motion feature to determine the Kalman coefficients at the current moment, that is:

[0053]

[0054] in, Let H be the Kalman coefficients at the current time step, and H be the transformation matrix.

[0055] The computing device determines the optimal state estimate of the target's second type of motion feature at the current moment based on the prior state estimate, the Kalman coefficient, the detection state value, and the transformation matrix.

[0056]

[0057] in, This is the optimal estimate of the state at the current moment. This represents the current detection status value.

[0058] The computing device determines the optimal estimate of the state covariance of the target's second type of motion feature at the current moment based on the Kalman coefficients at the current moment, the prior estimate of the state covariance at the current moment, and the transformation matrix, that is:

[0059]

[0060] in, Let I be the optimal estimate of the state covariance at the current moment, and let I be the identity matrix.

[0061] In this way, the optimal state estimate and the optimal state covariance estimate at the current moment can be calculated. These two data are passed to the next moment to perform the Kalman filter calculation for the next moment. The above process is repeated to achieve Kalman filter calculation and state estimation for the entire time period.

[0062] In one alternative technical solution, if the actual motion model is close to the ideal model, process noise can be disregarded. The computing device first determines the prior state estimate for the current moment based on the optimal state estimate from the previous moment, and then determines the prior state covariance estimate for the current moment based on the optimal state covariance estimate from the previous moment. Subsequently, it uses the prior state covariance estimate, the current value of the measured noise data, and the transformation matrix from the second type of motion feature to the first type of motion feature to determine the Kalman coefficients for the current moment. After obtaining the Kalman coefficients, the computing device determines the optimal state estimate of the second type of motion feature of the target for the current moment based on the prior state estimate, the Kalman coefficients, the detected state value, and the transformation matrix, and then determines the optimal state covariance estimate of the second type of motion feature of the target for the current moment based on the Kalman coefficients, the prior state covariance estimate, and the transformation matrix.

[0063] Figure 2 The following are the specific execution steps of the Kalman filtering algorithm in this embodiment of the invention. Specifically, after the computing device obtains the detection score in the detector detection result, it can adjust the measurement noise data according to the detection score, that is, form the current value of the measurement noise data at the current moment. Then, the Kalman filtering calculation at the current moment is completed based on the current value of the measurement noise data. The specific calculation process is the same as described above and will not be repeated here.

[0064] The following describes a specific embodiment of the present invention.

[0065] Multi-object tracking is a crucial module in autonomous driving applications. Its role is to estimate the position, velocity, orientation, and size of all objects in the environment over time, providing a reliable dynamic world representation for the planning module. In multi-object tracking, the industry-standard solution is to use the Kalman filter algorithm to estimate the state of moving objects and identify the trajectories of different categories of moving objects, such as pedestrians, bicycles, and cars. The obtained trajectories can be used to infer the motion patterns and driving behaviors of autonomous driving systems, improving predictions and, in turn, aiding in planning for autonomous driving. The main principle of the Kalman filter algorithm is to probabilistically fuse the predicted and measured values ​​of the object's motion state, with the fused state serving as the object's final motion state. Therefore, the performance of the Kalman filter algorithm directly determines the accuracy of multi-object tracking, thus affecting the final result of the autonomous driving perception system and consequently impacting the stability and safety of vehicle operation.

[0066] The Kalman filter algorithm mainly consists of two steps: prediction and update. In the prediction step, the predicted state estimate is typically derived from a kinematic model, with the state value linearly recursively over time, while also incorporating process noise whose variance follows a Gaussian distribution. The measured state estimate is obtained from sensor readings, also incorporating measurement noise whose variance follows a Gaussian distribution. In the update step, the predicted and measured state estimates are fused to obtain the final motion state. Therefore, process noise and measurement noise are the main factors affecting the performance of the filtering algorithm. Lower process noise indicates a more reliable prediction result, and the Kalman filter result will be more confident in the predicted state. Similarly, lower measurement noise indicates a more reliable measurement result, and the Kalman filter result will be closer to the measured state.

[0067] Traditional Kalman filtering algorithms typically set process noise and measurement noise as constants. Process noise is the error of the filter's motion model relative to the actual motion in the physical world. In the field of autonomous driving, the filter's motion model is generally selected as uniform or uniformly accelerated motion, so process noise can be considered to be unchanging over time. Measurement noise is the error between sensor readings or upstream detector results and the actual motion state. Generally, the results of sensors or detectors are considered stable, so the optimal measurement noise can be determined through repeated experiments.

[0068] The drawback of traditional Kalman filtering algorithms is that they treat measurement noise as a constant. When the results from sensors or upstream detectors are affected by external interference, making the results unreliable, they cannot dynamically adjust the measurement noise, thus affecting the accuracy and stability of the Kalman filtering algorithm. This problem is particularly evident in the field of autonomous driving. When the detection results from upstream detectors are inaccurate, the target tracking algorithm uses this unreliable result for state updates, leading to inaccurate trajectory tracking and ultimately affecting the final result of the autonomous driving perception system.

[0069] In autonomous driving solutions, LiDAR and cameras are commonly used data acquisition devices. Point cloud data and image data acquired by LiDAR and cameras are input into a detector to obtain object detection results, including their position information, velocity information, and target detection score. Because different objects differ in shape, visual scale, and size in point cloud data or image data, the detector's target detection score for different objects will also vary. This embodiment proposes a method that adaptively updates the Kalman filter measurement noise using the target detection score.

[0070] To consider the performance of different detection results in the Kalman filter algorithm, this embodiment differentiates the measurement noise based on the object's target detection score. For objects with higher target detection scores, the measurement noise is appropriately reduced, indicating a more reliable result and allowing them to have a greater weight in the Kalman filter algorithm's update step. For objects with lower target detection scores, the measurement noise is appropriately increased, allowing them to have a smaller weight in the Kalman filter algorithm's update step. The measurement noise is dynamically updated based on the upstream detector's detection results, ensuring that reliable results dominate the Kalman filter algorithm, thereby improving the accuracy of the Kalman filter algorithm and the stability of the autonomous driving perception system. The specific execution steps of this embodiment are as follows.

[0071] (1) The Kalman filter algorithm is an optimal estimation algorithm, which consists of two stages: prediction and update. In the prediction stage, the filter uses the estimation result of the previous state to make an estimate of the current state. In the update stage, the filter uses the observation of the current state to optimize the estimate obtained in the prediction stage to obtain a new estimate that is more accurate for the current stage.

[0072] (2) In the prediction phase of the Kalman filter algorithm, the predicted state value is derived through a kinematic model, which is generally a uniform or uniformly accelerated motion. Based on the state value at the previous moment, combined with the kinematic model and the running time, the state value at the current moment (i.e., the prior estimate of the state at the current moment) is obtained. Specifically, if it is the first time to predict the state value, the state value is the initial value set by the system; if it is not the first prediction, the state value is the updated value of the state value obtained at the previous moment. While calculating the state value, the prediction phase also calculates the state estimation covariance (i.e., the prior estimate of the state covariance at the current moment). The calculation of the state estimation covariance mainly depends on the state estimation covariance at the previous moment and the process noise. Similarly, if it is the first time to calculate the state estimation covariance, the state estimation covariance is the initial value set by the system; if it is not the first calculation, the covariance is the updated value of the state estimation covariance obtained at the previous moment.

[0073] (3) In the update phase of the Kalman filter algorithm, in addition to relying on the state value and state estimation covariance obtained in (2), it also relies on the detection results output by the upstream detector, such as the detected state value and the corresponding detection score. Both serve as data support for the update phase of the Kalman filter algorithm. Generally, the detected state value and detection score are directly input into the Kalman filter algorithm as data sources for state updates. The method proposed in this embodiment for adaptively updating the Kalman filter measurement noise based on the target detection score dynamically updates the Kalman filter measurement noise by combining the information from the upstream detector. It integrates the target detection score and the initial value of the measurement noise data as the current value of the measurement noise data, making the results more accurate. The formula is shown below:

[0074]

[0075]

[0076]

[0077] The parameters in the formula have already been explained above and will not be repeated here. This formula dynamically adjusts the measurement noise data based on the detection score, while ensuring the improved measurement noise data is accurate. Measuring noise constant The magnitude of the fluctuation depends on the detection score. A higher detection score means a lower improvement in the measurement noise compared to the measurement noise constant, indicating a more reliable detection result. In the Kalman filter update phase, the detection result carries more weight. Conversely, a lower detection score means an increase in the improvement in the measurement noise compared to the measurement noise constant, indicating a less reliable detection result. In the Kalman filter update phase, the detection result carries less weight.

[0078] (4) In the update phase of the Kalman filter algorithm, the state value and the state estimation covariance are updated. The update process mainly depends on the Kalman coefficients. The Kalman gain is calculated using the state estimation covariance and the dynamically updated measurement noise data in (3).

[0079] (5) The Kalman filter algorithm is an iterative algorithm. The updated state value and state estimate covariance obtained in (4) are used as the initial values ​​for the prediction stage, and the state value and covariance at the next moment are calculated iteratively.

[0080] This embodiment improves the accuracy of the Kalman filter algorithm and the stability of the autonomous driving perception system by dynamically updating the measurement noise data of the Kalman filter by combining information from the upstream detector.

[0081] It should be noted that the technical solutions of this invention, including the collection, updating, analysis, processing, use, transmission, and storage of user personal information, all comply with relevant laws and regulations, are used for legitimate purposes, and do not violate public order and good morals. Necessary measures are taken to prevent unauthorized access to user personal information data and to safeguard user personal information security, network security, and national security.

[0082] For the foregoing method embodiments, they are described as a series of actions for ease of description. However, those skilled in the art should understand that the present invention is not limited to the described order of actions, and some steps may actually be performed in other orders or simultaneously. Furthermore, those skilled in the art should also understand that the embodiments described in the specification are preferred embodiments, and the actions and modules involved are not necessarily essential for implementing the present invention.

[0083] To facilitate better implementation of the above-described solutions of the embodiments of the present invention, related apparatus for implementing the above-described solutions is also provided below.

[0084] Please see Figure 3 As shown, the target state determination device provided in this embodiment of the invention may include: a detection result acquisition unit 301, a measurement noise adjustment unit 302, and a calculation unit 303.

[0085] The detection result acquisition unit 301 is used to acquire the detection result of the detector on the first type of motion feature of the target at the current moment. The detection result includes: the detection state value at the current moment and the detection score characterizing the detection reliability. The measurement noise adjustment unit 302 is used to adjust the preset initial value of the measurement noise data according to the detection score to obtain the current value of the measurement noise data. The calculation unit 303 is used to acquire the optimal state estimate and the optimal state covariance estimate of the second type of motion feature of the target at the previous moment, determine the Kalman coefficient at the current moment according to the optimal state covariance estimate and the current value of the measurement noise data, and determine the optimal state estimate of the second type of motion feature of the target at the current moment using the optimal state estimate, the Kalman coefficient at the current moment, and the detection state value at the current moment.

[0086] In this embodiment of the invention, for the same detector, the current value of the measurement noise data is negatively correlated with the detection reliability represented by the detection score.

[0087] As a preferred embodiment, the measurement noise adjustment unit 302 is further configured to: obtain an opposition number based on the detection score at the current moment, determine an adjustment coefficient based on the opposition number, and multiply the initial value of the measurement noise data by the adjustment coefficient to obtain the current value of the measurement noise data.

[0088] Preferably, the measurement noise adjustment unit 302 is further configured to: subtract the detection score at the current moment from a preset detection score threshold to obtain an offset number; determine the product of the sign function value of the offset number and the offset number as a first value; determine the sum of the first value and 1 as a second value; and determine the product of the second value and a preset proportional coefficient as an adjustment coefficient.

[0089] In one embodiment, the calculation unit 303 is further configured to: determine the prior estimate of the state covariance at the current moment based on the optimal estimate of the state covariance at the previous moment and the preset process noise data; and determine the Kalman coefficients at the current moment using the prior estimate of the state covariance at the current moment, the current value of the measurement noise data, and the transformation matrix from the second type of motion feature to the first type of motion feature.

[0090] In an optional implementation, the computing unit 303 is further configured to: determine the state prior estimate at the current moment based on the state optimal estimate at the previous moment; and determine the state optimal estimate of the second type of motion feature of the target at the current moment based on the state prior estimate at the current moment, the Kalman coefficient at the current moment, the detection state value at the current moment, and the transformation matrix.

[0091] Furthermore, in this embodiment of the invention, the calculation unit 303 is also used to: determine the optimal estimate of the state covariance of the second type of motion feature of the target at the current time based on the Kalman coefficients at the current time, the prior estimate of the state covariance at the current time, and the transformation matrix.

[0092] According to the technical solution of this invention, after obtaining the detection state value of the detector at the current moment and the detection score characterizing the detection reliability, the initial value of the measurement noise data is adjusted according to the detection score to obtain the current value of the measurement noise data. For the same detector, the above adjustment can make the current value of the measurement noise data negatively correlated with the detection reliability characterized by the detection score. Subsequently, a Kalman filtering algorithm is executed based on the detection state value and the current value of the measurement noise data at the current moment to determine the optimal state estimate at the current moment. In this way, the measurement noise data can be adaptively adjusted based on the detection score at different moments. When the detection reliability characterized by the detection score is high, the measurement noise data is reduced to increase the Kalman coefficient value, thereby increasing the proportion of the detection part in the final calculation result. When the detection reliability characterized by the detection score is low, the measurement noise data is increased to reduce the Kalman coefficient value, thereby reducing the proportion of the detection part in the final calculation result. Ultimately, the influence of external interference on the state estimation of the detector is avoided, and the accuracy and stability of the Kalman filtering algorithm are improved.

[0093] Figure 4 An exemplary system architecture 400 is shown that can be applied to the target state determination method or target state determination apparatus of the present invention.

[0094] like Figure 4 As shown, system architecture 400 may include terminal devices 401, 402, and 403, network 404, and server 405 (this architecture is merely an example; the components included in a specific architecture may be adjusted according to the specific application). Network 404 serves as the medium for providing a communication link between terminal devices 401, 402, and 403 and server 405. Network 404 may include various connection types, such as wired or wireless communication links or fiber optic cables.

[0095] Users can use terminal devices 401, 402, and 403 to interact with server 405 via network 404 to receive or send messages, etc. Various communication client applications can be installed on terminal devices 401, 402, and 403, such as multi-target tracking applications (for example only).

[0096] Terminal devices 401, 402, and 403 can be various electronic devices with displays and web browsing capabilities, including but not limited to smartphones, tablets, laptops, desktop computers, and autonomous vehicles.

[0097] Server 405 can be a server that provides various services, such as a backend server that supports multi-target tracking applications operated by users using terminal devices 401, 402, and 403 (for example only). The backend server can process received state estimation requests, etc., and feed back the processing results (such as optimal state estimates - for example only) to terminal devices 401, 402, and 403.

[0098] It should be noted that the target state determination method provided in the embodiments of the present invention is generally executed by server 405, and correspondingly, the target state determination device is generally set in server 405.

[0099] It should be understood that Figure 4 The number of terminal devices, networks, and servers shown is merely illustrative. Depending on implementation needs, any number of terminal devices, networks, and servers can be included.

[0100] The present invention also provides an electronic device. The electronic device according to an embodiment of the present invention includes: one or more processors; and a storage device for storing one or more programs, wherein when the one or more programs are executed by the one or more processors, the one or more processors implement the target state determination method provided by the present invention.

[0101] The following is for reference. Figure 5 It shows a schematic diagram of the structure of a computer system 500 suitable for implementing an electronic device according to embodiments of the present invention. Figure 5The electronic device shown is merely an example and should not be construed as limiting the functionality and scope of use of the embodiments of the present invention.

[0102] like Figure 5 As shown, the computer system 500 includes a central processing unit (CPU) 501, which can perform various appropriate actions and processes based on programs stored in read-only memory (ROM) 502 or programs loaded from storage section 508 into random access memory (RAM) 503. The RAM 503 also stores various programs and data required for the operation of the computer system 500. The CPU 501, ROM 502, and RAM 503 are interconnected via a bus 504. An input / output (I / O) interface 505 is also connected to the bus 504.

[0103] The following components are connected to I / O interface 505: an input section 506 including a keyboard, mouse, etc.; an output section 507 including a cathode ray tube (CRT), liquid crystal display (LCD), etc., and speakers, etc.; a storage section 508 including a hard disk, etc.; and a communication section 509 including a network interface card such as a LAN card, modem, etc. The communication section 509 performs communication processing via a network such as the Internet. A drive 510 is also connected to I / O interface 505 as needed. A removable medium 511, such as a disk, optical disk, magneto-optical disk, semiconductor memory, etc., is installed on drive 510 as needed so that computer programs read from it can be installed into storage section 508 as needed.

[0104] In particular, according to the embodiments disclosed in this invention, the processes described in the above main step diagrams can be implemented as computer software programs. For example, embodiments of this invention include a computer program product comprising a computer program carried on a computer-readable medium, the computer program containing program code for performing the methods shown in the main step diagrams. In the above embodiments, the computer program can be downloaded and installed from a network via communication section 509, and / or installed from removable medium 511. When the computer program is executed by central processing unit 501, it performs the functions defined in the system of this invention.

[0105] It should be noted that the computer-readable medium shown in this invention can be a computer-readable signal medium or a computer-readable storage medium, or any combination thereof. A computer-readable storage medium can be, for example,—but not limited to—an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, apparatus, or device, or any combination thereof. More specific examples of a computer-readable storage medium may include, but are not limited to: an electrical connection having one or more wires, a portable computer disk, a hard disk, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fiber, portable compact disk read-only memory (CD-ROM), optical storage device, magnetic storage device, or any suitable combination thereof. In this invention, a computer-readable storage medium can be any tangible medium containing or storing a program that can be used by or in conjunction with an instruction execution system, apparatus, or device. In this invention, a computer-readable signal medium can include a data signal propagated in baseband or as part of a carrier wave, carrying computer-readable program code. Such propagated data signals can take various forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination thereof. A computer-readable signal medium may also be any computer-readable medium other than a computer-readable storage medium, which can send, propagate, or transmit a program for use by or in connection with an instruction execution system, apparatus, or device. The program code contained on the computer-readable medium may be transmitted using any suitable medium, including but not limited to: wireless, wire, optical fiber, RF, etc., or any suitable combination thereof.

[0106] The flowcharts and block diagrams in the accompanying drawings illustrate the architecture, functionality, and operation of possible implementations of systems, methods, and computer program products according to various embodiments of the present invention. In this regard, each block in a flowchart or block diagram may represent a module, segment, or portion of code containing one or more executable instructions for implementing a specified logical function. It should also be noted that in some alternative implementations, the functions indicated in the blocks may occur in a different order than those indicated in the drawings. For example, two consecutively indicated blocks may actually be executed substantially in parallel, and they may sometimes be executed in reverse order, depending on the functions involved. It should also be noted that each block in a block diagram or flowchart, and combinations of blocks in a block diagram or flowchart, may be implemented using a dedicated hardware-based system that performs the specified function or operation, or using a combination of dedicated hardware and computer instructions.

[0107] The units described in the embodiments of the present invention can be implemented in software or hardware. The described units can also be housed in a processor; for example, a processor can be described as including: a detection result acquisition unit, a measurement noise adjustment unit, and a calculation unit. The names of these units do not necessarily limit the specific unit; for example, the detection result acquisition unit can also be described as "a unit that provides detection scores to the measurement noise adjustment unit."

[0108] In another aspect, the present invention also provides a computer-readable medium, which may be included in the device described in the above embodiments; or it may exist independently and not assembled into the device. The computer-readable medium carries one or more programs, which, when executed by the device, cause the device to perform the following steps: acquiring the detection result of a detector on a first type of motion feature of a target at the current moment, the detection result including: a detection state value at the current moment and a detection score characterizing detection reliability; adjusting a preset initial value of measurement noise data according to the detection score to obtain the current value of the measurement noise data; acquiring the optimal state estimate and the optimal state covariance estimate of the second type of motion feature of the target at the previous moment; determining the Kalman coefficient at the current moment according to the optimal state covariance estimate and the current value of the measurement noise data; and determining the optimal state estimate of the second type of motion feature of the target at the current moment using the optimal state estimate, the Kalman coefficient at the current moment, and the detection state value at the current moment.

[0109] In the technical solution of this invention embodiment, after obtaining the detection state value of the detector at the current moment and the detection score characterizing the detection reliability, the initial value of the measurement noise data is adjusted according to the detection score to obtain the current value of the measurement noise data. For the same detector, the above adjustment can make the current value of the measurement noise data negatively correlated with the detection reliability characterized by the detection score. Subsequently, a Kalman filtering algorithm is executed based on the detection state value and the current value of the measurement noise data at the current moment to determine the optimal state estimate at the current moment. In this way, the measurement noise data can be adaptively adjusted based on the detection score at different moments. When the detection reliability characterized by the detection score is high, the measurement noise data is reduced to increase the Kalman coefficient value, thereby increasing the proportion of the detection part in the final calculation result. When the detection reliability characterized by the detection score is low, the measurement noise data is increased to reduce the Kalman coefficient value, thereby reducing the proportion of the detection part in the final calculation result. Ultimately, this avoids the influence of external interference on the state estimation of the detector and improves the accuracy and stability of the Kalman filtering algorithm.

[0110] The specific embodiments described above do not constitute a limitation on the scope of protection of this invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can occur depending on design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this invention should be included within the scope of protection of this invention.

Claims

1. A method for determining a target state, characterized in that, include: Obtain the detection results of the detector for the first type of motion features of the target at the current moment, the detection results including: the detection state value at the current moment and the detection score characterizing the detection reliability; The initial value of the measured noise data is adjusted based on the detection score to obtain the current value of the measured noise data; Obtain the optimal state estimate and optimal state covariance estimate of the second type of motion feature of the target at the previous time step. Determine the Kalman coefficient at the current time step based on the optimal state covariance estimate and the current value of the measurement noise data. Use the optimal state estimate, the Kalman coefficient at the current time step, and the detection state value at the current time step to determine the optimal state estimate of the second type of motion feature of the target at the current time step.

2. The method according to claim 1, characterized in that, For the same detector, the current value of the measured noise data is negatively correlated with the detection reliability characterized by the detection score.

3. The method according to claim 2, characterized in that, The step of adjusting the preset initial value of the measured noise data according to the detection score to obtain the current value of the measured noise data includes: The opposition number is obtained based on the detection score at the current moment, and the adjustment coefficient is determined based on the opposition number; The initial value of the measured noise data is multiplied by the adjustment coefficient to obtain the current value of the measured noise data.

4. The method according to claim 3, characterized in that, The process of obtaining the counter-number based on the detection score at the current moment and determining the adjustment coefficient based on the counter-number includes: Subtract the current detection score from the preset detection score threshold to obtain the opposite number; The product of the sign function value of the opposite number and the opposite number is determined as the first value, the sum of the first value and 1 is determined as the second value, and the product of the second value and a preset proportional coefficient is determined as the adjustment coefficient.

5. The method according to claim 1, characterized in that, The step of determining the Kalman coefficients at the current moment based on the optimal estimate of the state covariance and the current value of the measurement noise data includes: Determine the prior estimate of the state covariance at the current moment based on the optimal estimate of the state covariance at the previous moment and the preset process noise data; The Kalman coefficients at the current moment are determined using the prior estimate of the state covariance at the current moment, the current value of the measured noise data, and the transformation matrix from the second type of motion feature to the first type of motion feature.

6. The method according to claim 5, characterized in that, The process of determining the optimal state estimate of the second type of motion feature of the target at the current moment using the optimal state estimate, the Kalman coefficient at the current moment, and the detection state value at the current moment includes: Determine the prior state estimate for the current time step based on the optimal state estimate from the previous time step; Based on the current state prior estimate, the current Kalman coefficient, the current detection state value, and the transformation matrix, determine the optimal state estimate of the second type of motion features of the target at the current moment.

7. The method according to claim 5, characterized in that, The method further includes: Based on the Kalman coefficients at the current time, the prior estimate of the state covariance at the current time, and the transformation matrix, determine the optimal estimate of the state covariance of the second type of motion characteristics of the target at the current time.

8. A target state determination device, characterized in that, include: The detection result acquisition unit is used to acquire the detection result of the detector on the first type of motion feature of the target at the current moment. The detection result includes: the detection state value at the current moment and the detection score characterizing the detection reliability. A noise adjustment unit is used to adjust the preset initial value of the noise measurement data according to the detection score to obtain the current value of the noise measurement data; The calculation unit is used to obtain the optimal state estimate and the optimal state covariance estimate of the second type of motion feature of the target at the previous time, determine the Kalman coefficient at the current time based on the optimal state covariance estimate and the current value of the measurement noise data, and determine the optimal state estimate of the second type of motion feature of the target at the current time using the optimal state estimate, the Kalman coefficient at the current time, and the detection state value at the current time.

9. An electronic device, characterized in that, include: One or more processors; Storage device for storing one or more programs. When the one or more programs are executed by the one or more processors, the one or more processors implement the method as described in any one of claims 1-7.

10. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the method as described in any one of claims 1-7.

Images