State estimation method, device, storage medium and system

Abstract

The application discloses a state estimation method and device, a storage medium and an electronic device, and relates to the technical field of data processing and automatic driving. The method comprises the following steps: acquiring a state information sequence of a scene object; based on the state information sequence, a position change sequence corresponding to the scene object is calculated, wherein the position change sequence comprises a target change amount at a target moment and a plurality of historical change amounts at a plurality of historical moments before the target moment; statistical distribution characteristics of the plurality of historical change amounts are analyzed to determine an abnormality evaluation condition; the target change amount is evaluated by using the abnormality evaluation condition to obtain an evaluation result; the observation noise corresponding to the target moment is determined according to the evaluation result; and Kalman filtering processing is performed by using the observation noise to obtain a state estimation result of the scene object at the target moment. The application solves the technical problem of low accuracy of state estimation results in the related art.

Description

Technical Field

[0001] This application relates to the fields of data processing technology and autonomous driving technology, and more specifically, to a state estimation method, apparatus, storage medium and system. Background Technology

[0002] In fields such as vehicle-to-everything (V2X) and autonomous driving, Kalman filtering algorithms are commonly used to estimate the state of perceived targets to improve the stability and accuracy of tracking. With the development of intelligent transportation systems, higher demands are placed on the accuracy and reliability of target state estimation, especially in complex scenarios where effective handling of abnormal sensor data is required. However, traditional Kalman filtering methods typically assume that observation noise follows a fixed distribution and cannot adaptively identify and suppress abnormal positional information caused by factors such as occlusion. This leads to the accumulation of state estimation biases, affecting the effectiveness of subsequent tracking and control decisions.

[0003] There is currently no effective solution to the above problems. Summary of the Invention

[0004] This application provides a state estimation method, apparatus, storage medium, and system to at least solve the technical problem of low accuracy of state estimation results in related technologies.

[0005] According to one aspect of the embodiments of this application, a state estimation method is provided, which acquires a state information sequence of a scene object, wherein the state information sequence is used to characterize the position coordinates and heading angle of the scene object at multiple consecutive moments; based on the state information sequence, a position change sequence corresponding to the scene object is calculated, wherein the position change sequence includes: a target change amount at a target moment, and multiple historical change amounts at multiple historical moments prior to the target moment; statistical distribution characteristic analysis is performed on the multiple historical change amounts to determine anomaly evaluation conditions; the target change amount is evaluated using the anomaly evaluation conditions to obtain an evaluation result; the observation noise corresponding to the target moment is determined based on the evaluation result; and Kalman filtering is performed on the observation noise to obtain the state estimation result of the scene object at the target moment.

[0006] Optionally, the position change sequence includes: a longitudinal sequence corresponding to a first direction and a lateral sequence corresponding to a second direction, wherein the first direction is determined by the heading angle and the second direction is perpendicular to the first direction; the target change includes: a longitudinal target change component and a lateral target change component; each of the multiple historical change components includes: a longitudinal historical change component and a lateral historical change component; the anomaly assessment conditions include: a longitudinal anomaly condition corresponding to the first direction and a lateral anomaly condition corresponding to the second direction.

[0007] Optionally, the position change sequence includes: a longitudinal sequence corresponding to the first direction and a lateral sequence corresponding to the second direction; the position change sequence calculated based on the state information sequence includes: for each group of adjacent moments in multiple consecutive moments, extracting the first position coordinates and the first heading angle corresponding to the earlier moment from the state information sequence, and extracting the second position coordinates corresponding to the later moment from the state information sequence; calculating the position distance corresponding to the adjacent moments based on the first position coordinates and the second position coordinates; determining the first direction and the second direction corresponding to the adjacent moments based on the first heading angle; calculating the projection of the position distance onto the first direction to obtain the longitudinal change amount corresponding to the later moment in the longitudinal sequence; and calculating the projection of the position distance onto the second direction to obtain the lateral change amount corresponding to the later moment in the lateral sequence.

[0008] Optionally, statistical distribution characteristic analysis of multiple historical changes is performed to determine anomaly assessment conditions, including: extracting multiple longitudinal historical changes and multiple horizontal historical changes from multiple historical changes; calculating the statistical distribution characteristics of multiple longitudinal historical changes to obtain the longitudinal mean and longitudinal standard deviation; determining the longitudinal anomaly conditions based on the longitudinal mean and longitudinal standard deviation; calculating the statistical distribution characteristics of multiple horizontal historical changes to obtain the horizontal mean and horizontal standard deviation; and determining the horizontal anomaly conditions based on the horizontal mean and horizontal standard deviation.

[0009] Optionally, the target change is evaluated using anomaly assessment conditions, and the assessment results include: when the longitudinal target change component is detected to meet the longitudinal anomaly condition, or when the lateral target change component is detected to meet the lateral anomaly condition, the assessment result is determined to be: the target has an abnormal positional change at any given time.

[0010] Optionally, the vertical target change component satisfying the vertical anomaly condition includes: the vertical target change component exceeding the vertical threshold range, wherein the vertical threshold range is determined by the vertical mean and vertical standard deviation corresponding to the vertical anomaly condition; the horizontal target change component satisfying the horizontal anomaly condition includes: the horizontal target change component exceeding the horizontal threshold range, wherein the horizontal threshold range is determined by the horizontal mean and horizontal standard deviation corresponding to the horizontal anomaly condition.

[0011] Optionally, determining the observation noise corresponding to the target time based on the evaluation results includes: when it is determined from the evaluation results that there is an abnormal position change at the target time, adjusting the initial observation noise according to the noise amplification factor and updating the observation noise, wherein the noise amplification factor is determined based on the target change amount and the abnormal threshold range corresponding to the abnormal evaluation conditions.

[0012] According to one aspect of the embodiments of this application, a state estimation apparatus is provided, comprising: an acquisition module, configured to acquire a state information sequence of a scene object, wherein the state information sequence is used to characterize the position coordinates and heading angle of the scene object at multiple consecutive time points; a calculation module, configured to calculate a position change sequence corresponding to the scene object based on the state information sequence, wherein the position change sequence includes: a target change amount at a target time, and multiple historical change amounts at multiple historical time points prior to the target time; an analysis module, configured to perform statistical distribution characteristic analysis on the multiple historical change amounts to determine anomaly evaluation conditions; an anomaly evaluation module, configured to evaluate the target change amount using the anomaly evaluation conditions to obtain an evaluation result; a determination module, configured to determine the observation noise corresponding to the target time based on the evaluation result; and an estimation module, configured to perform Kalman filtering processing on the observation noise to obtain a state estimation result of the scene object at the target time.

[0013] According to one aspect of the embodiments of this application, a computer-readable storage medium is provided, the computer-readable storage medium including a stored program, wherein, when the program is running, it controls the device where the storage medium is located to perform any of the above-mentioned state estimation methods.

[0014] According to one aspect of the embodiments of this application, an electronic device is provided, including a memory and a processor, wherein the memory stores a computer program and the processor is configured to run the computer program to perform any of the above-described state estimation methods.

[0015] In this embodiment, a state information sequence of a scene object is obtained, wherein the state information sequence is used to characterize the position coordinates and heading angle of the scene object at multiple consecutive moments; based on the state information sequence, a position change sequence corresponding to the scene object is calculated, wherein the position change sequence includes: the target change amount at the target moment, and multiple historical change amounts at multiple historical moments before the target moment; statistical distribution characteristic analysis is performed on the multiple historical change amounts to determine anomaly evaluation conditions; the target change amount is evaluated using the anomaly evaluation conditions to obtain an evaluation result; the observation noise corresponding to the target moment is determined based on the evaluation result; and Kalman filtering is performed using the observation noise to obtain the state estimation result of the scene object at the target moment.

[0016] It is noteworthy that this embodiment of the application deeply analyzes the changes in the heading angle and position coordinates of scene objects at continuous time intervals, and uses statistical distribution feature analysis methods to identify abnormal information in the target changes. By comparing and evaluating the target changes with historical changes, abnormal position changes along the heading angle direction can be accurately identified. Based on the anomaly evaluation results, this embodiment of the application can adaptively adjust the observation noise in the Kalman filter, effectively reducing the impact of abnormal position information on subsequent state estimation, and improving the accuracy and robustness of state estimation. The dynamic and adaptive noise adjustment strategy in this embodiment of the application does not require manual intervention and automatically adapts to the complexity of scene object motion under different scenarios, ensuring estimation accuracy in highly uncertain environments, thereby significantly improving the overall performance and reliability of the vehicle-road cooperative system. In other words, this embodiment of the application achieves the goal of accurate real-time estimation of scene object state information by combining anomaly detection and adaptive noise adjustment, thereby achieving the technical effect of reducing the impact of abnormal positions on target state estimation and improving the accuracy of state estimation, and thus solving the technical problem of low accuracy of state estimation results in related technologies. Attached Figure Description

[0017] The accompanying drawings, which are included to provide a further understanding of this application and form part of this application, illustrate exemplary embodiments and are used to explain this application, but do not constitute an undue limitation of this application. In the drawings:

[0018] Figure 1 This is a hardware structure block diagram of an optional terminal device for implementing a state estimation method according to an embodiment of this application;

[0019] Figure 2 This is a flowchart of a state estimation method according to an embodiment of this application;

[0020] Figure 3 This is a schematic diagram of an optional method for calculating position change according to an embodiment of this application;

[0021] Figure 4 This is a structural block diagram of a state estimation device according to an embodiment of this application. Detailed Implementation

[0022] To enable those skilled in the art to better understand the present application, the technical solutions in the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present application, and not all embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present application.

[0023] It should be noted that the terms "first," "second," etc., in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of this application described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0024] According to an embodiment of this application, an embodiment of a state estimation method is provided. It should be noted that the steps shown in the flowchart in the accompanying drawings can be executed in a computer system such as a set of computer-executable instructions. Furthermore, although a logical order is shown in the flowchart, in some cases, the steps shown or described may be executed in a different order than that shown here.

[0025] Figure 1 This is a hardware structure block diagram of an optional terminal device for implementing a state estimation method according to an embodiment of this application, such as... Figure 1 As shown, the terminal device may include one or more processors 102 (processor 102 may include, but is not limited to, a microprocessor (MCU) or a field-programmable gate array (FPGA), etc.), a memory 104 for storing data, and a transmission device 106 for communication functions. In addition, it may also include: a display device 110, an input / output device 108, a Universal Serial Bus (USB) port (which may be included as one of the ports of a computer bus, not shown in the figure), a network interface (not shown in the figure), a power supply (not shown in the figure), and / or a camera (not shown in the figure). Those skilled in the art will understand that... Figure 1 The structure shown is for illustrative purposes only and does not limit the structure of the terminal device described above. For example, the terminal device may also include components that are more... Figure 1 The more or fewer components shown, or having the same Figure 1 The different configurations shown.

[0026] It should be noted that the aforementioned one or more processors 102 and / or other data processing circuits may be embodied, in whole or in part, as software, hardware, firmware, or any other combination thereof. Furthermore, the data processing circuitry may be a single, independent processing module, or may be integrated, in whole or in part, into any other element within the terminal device (or mobile device).

[0027] The memory 104 can be used to store software programs and modules of application software, such as the program instructions / data storage device corresponding to the state estimation method in this embodiment. The processor 102 executes various functional applications and data processing by running the software programs and modules stored in the memory 104, thereby realizing the aforementioned state estimation method. The memory 104 may include high-speed random access memory, and may also include non-volatile memory, such as one or more magnetic storage devices, flash memory, or other non-volatile solid-state memory. In some instances, the memory 104 may further include memory remotely located relative to the processor 102, and these remote memories can be connected to terminal devices via a network. Examples of such networks include, but are not limited to, the Internet, corporate intranets, local area networks, mobile communication networks, and combinations thereof.

[0028] The transmission device 106 is used to receive or send data via a network. Specific examples of the network described above may include a wireless network provided by the communication provider of the terminal device. In one example, the transmission device 106 includes a Network Interface Controller (NIC), which can connect to other network devices via a base station to communicate with the Internet. In another example, the transmission device 106 may be a Radio Frequency (RF) module, used for wireless communication with the Internet.

[0029] Under the above operating environment, the embodiments of this application provide the following: Figure 2 The state estimation method shown, Figure 2 This is a flowchart of a state estimation method according to an embodiment of this application, such as... Figure 2 As shown, the method includes the following implementation steps S201 to S206.

[0030] Step S201: Obtain the state information sequence of the scene object, wherein the state information sequence is used to characterize the position coordinates and heading angle of the scene object at multiple consecutive moments.

[0031] The aforementioned scene objects can be various targets perceived and tracked in vehicle-to-everything (V2X) systems or autonomous driving environments, such as vehicles, pedestrians, and non-motorized vehicles. The aforementioned state information sequence refers to the set of state data of the scene objects at different times, typically continuously sensed and output by sensors (such as cameras, millimeter-wave radar, and lidar) installed on the roadside or vehicle-mounted. Each state data point in the state information sequence includes at least the scene object's position coordinates in two-dimensional or three-dimensional space and its heading angle. By acquiring state information sequences at consecutive moments, the system can track the motion trajectory and behavioral intentions of scene objects, providing a data foundation for subsequent state estimation and anomaly detection. In specific applications, the aforementioned state information sequence can be preprocessed and time-aligned temporal data, ensuring data consistency and accuracy.

[0032] For example, in one application scenario, the state information sequence corresponding to the above scenario object can be represented as: posHis={pos1,pos2,…,pos k}, where pos k This represents the state parameters of the scene object at time k. Specifically, the state parameters can include multiple parameter items, such as pos. k =[x k ,y k ,θ k ], where (x k ,y k ) represents the position coordinates of the scene object at time k, θ k This represents the heading angle of the scene object at time k.

[0033] Step S202: Based on the state information sequence, calculate the position change sequence corresponding to the scene object. The position change sequence includes: the target change amount at the target time, and multiple historical change amounts at multiple historical times before the target time.

[0034] The aforementioned position change sequence refers to the set of positional movements of scene objects between consecutive moments, used to quantify changes in the motion state of scene objects. When calculating the position change sequence, the system calculates the positional change between adjacent moments based on the position coordinates and heading angles of adjacent moments in the state information sequence. The aforementioned target change refers to the positional change of the current target moment relative to its previous moment, while the historical change refers to the positional changes of multiple historical moments prior to the target moment relative to their respective previous moments. In specific implementations, the calculation of positional changes can be further distinguished into longitudinal changes along the heading angle and lateral changes perpendicular to the heading angle, to more precisely capture abnormal patterns in the motion of scene objects. By constructing the position change sequence, the system can analyze the statistical characteristics of the motion of scene objects, providing a basis for subsequent anomaly detection.

[0035] Step S203: Analyze the statistical distribution characteristics of multiple historical changes to determine the anomaly assessment conditions.

[0036] The aforementioned statistical distribution characteristic analysis refers to the statistical analysis of historical change sequences to characterize their normal fluctuation range. Specifically, the system calculates the statistical characteristics of the historical change sequences, such as the mean (denoted as μ) and standard deviation (denoted as σ). These characteristics reflect the central tendency and dispersion of changes in scene objects under normal motion conditions. Based on these statistical characteristics, the system determines anomaly assessment conditions. For example, the anomaly assessment condition is set to whether the change exceeds the range of (μ ± 3σ) (i.e., the 3σ criterion). The above anomaly assessment conditions are used to distinguish between normal motion changes and abnormal position jumps caused by occlusion, sensor noise, or other interference. By dynamically updating the historical change sequences and corresponding statistical characteristics, the system can adapt to changes in the motion patterns of scene objects, ensuring the accuracy and adaptability of anomaly detection conditions.

[0037] Step S204: Using the anomaly assessment conditions, assess the target change and obtain the assessment results.

[0038] Evaluating the target change using anomaly assessment conditions involves comparing the target change at the current time with the aforementioned anomaly assessment conditions to determine whether the target change is abnormal, i.e., whether there is an anomaly in the target's position at that time. For example, if the target change exceeds a threshold range determined based on historical statistical characteristics, the assessment result indicates an anomaly; otherwise, the assessment result indicates normality. The assessment result can be a binary decision output or a multi-valued decision output, used to indicate whether the position observation value of the scene object at the current time is reliable. Step S204 is a key step in anomaly detection. By comparing real-time data with historical statistical benchmarks, potential anomalies are effectively identified, providing a decision basis for subsequent adaptive filtering.

[0039] Step S205: Based on the evaluation results, determine the observation noise corresponding to the target time.

[0040] In Kalman filtering, observation noise (essentially the covariance matrix) reflects the confidence level in the uncertainty of sensor measurements. Step S205 dynamically adjusts the observation noise based on the evaluation results: if the evaluation results indicate anomalies, the observation noise is increased to reduce the weight of anomalous observations in state updates; if the evaluation results are normal, the observation noise remains at its default value. A specific adjustment strategy could be to calculate a noise amplification factor based on the degree of anomaly (e.g., the magnitude exceeding a threshold), and then multiply this amplification factor by the default observation noise (which could be the initial observation noise) to obtain the adjusted observation noise. Through this adaptive mechanism, the system can suppress the negative impact of unreliable observations on state estimation when anomalies occur, improving the robustness of the filter.

[0041] Step S206: Perform Kalman filtering on the observation noise to obtain the state estimation result of the scene object at the target time.

[0042] In application scenarios, Kalman filtering can be divided into a prediction phase and an update phase. In the prediction phase, the system predicts the current state based on the state estimate from the previous time step and the motion model. In the update phase, the system calculates the Kalman gain using the current observations and adjusted observation noise, and then weights and fuses the predicted and observed states to obtain the optimal state estimate. By using adjusted observation noise, the Kalman filter can adaptively adjust the confidence level of the observations based on anomaly detection results, effectively suppressing the influence of outliers and outputting more accurate and stable state estimates. This result includes state variables such as the position and velocity of scene objects, which can be used for subsequent tracking, prediction, and decision-making tasks.

[0043] Through steps S201 to S206 described above, this embodiment of the application implements a scheme for scene object state estimation based on an adaptive Kalman filter method combined with anomaly detection. First, by acquiring the state information sequence of scene objects and calculating the position change sequence, a data foundation for anomaly detection is provided. Then, based on the statistical distribution characteristics of historical changes, anomaly evaluation conditions are determined, and the current change is evaluated, achieving real-time detection of positional anomalies. The observation noise is dynamically adjusted according to the anomaly evaluation results, enabling the Kalman filter to adaptively respond to abnormal observations. Finally, the adjusted observation noise is used for Kalman filtering to obtain accurate state estimation results. The above method effectively reduces the negative impact of abnormal position information on state estimation, improving the accuracy and robustness of target tracking in vehicle-road cooperative and autonomous driving systems.

[0044] It is noteworthy that this embodiment of the application deeply analyzes the changes in the heading angle and position coordinates of scene objects at continuous time intervals, and uses statistical distribution feature analysis methods to identify abnormal information in the target changes. By comparing and evaluating the target changes with historical changes, abnormal position changes along the heading angle direction can be accurately identified. Based on the anomaly evaluation results, this embodiment of the application can adaptively adjust the observation noise in the Kalman filter, effectively reducing the impact of abnormal position information on subsequent state estimation, and improving the accuracy and robustness of state estimation. The dynamic and adaptive noise adjustment strategy in this embodiment of the application does not require manual intervention and automatically adapts to the complexity of scene object motion under different scenarios, ensuring estimation accuracy in highly uncertain environments, thereby significantly improving the overall performance and reliability of the vehicle-road cooperative system. In other words, this embodiment of the application achieves the goal of accurate real-time estimation of scene object state information by combining anomaly detection and adaptive noise adjustment, thereby achieving the technical effect of reducing the impact of abnormal positions on target state estimation and improving the accuracy of state estimation, and thus solving the technical problem of low accuracy of state estimation results in related technologies.

[0045] Other embodiments of the above-described methods in the present application will be further described below.

[0046] As an optional implementation, in the above state estimation method, the position change sequence includes: a longitudinal sequence corresponding to a first direction and a lateral sequence corresponding to a second direction, wherein the first direction is determined by the heading angle and the second direction is perpendicular to the first direction; the target change includes: a longitudinal target change component and a lateral target change component; each of the multiple historical change components includes: a longitudinal historical change component and a lateral historical change component; the anomaly assessment conditions include: a longitudinal anomaly condition corresponding to the first direction and a lateral anomaly condition corresponding to the second direction.

[0047] The aforementioned position change sequence includes position changes corresponding to multiple consecutive moments. The last moment (which can be understood as the current moment or the latest moment) among these consecutive moments can be used as the target moment, while the other moments, excluding the last moment, can be used as multiple historical moments relative to the target moment. Based on this, the position change sequence can include: multiple historical changes and the target change.

[0048] In some special application scenarios, in the above position change sequence, the position change corresponding to the first moment of multiple consecutive moments is determined independently by the state parameter corresponding to the first moment, or jointly by the first moment and the preset initial parameter.

[0049] When analyzing positional changes, the heading direction (i.e., the first direction) and the vertical heading direction (i.e., the second direction) at each moment are considered. This is because, at any given moment, the direction of motion of a scene object is usually consistent with its real-time heading angle. Furthermore, considering these two directions helps handle some extreme cases. For example, if occlusion causes a small positional change in a scene object, but this change is mainly reflected in the vertical heading direction (i.e., the second direction), this situation has a significant impact on the state estimation of the scene object, and the positional change represented by Euclidean distance cannot be used alone to assess whether anomalies exist.

[0050] In the above implementation, the target change is structured to include two directional components: a longitudinal target change component and a lateral target change component. The longitudinal target change component represents the degree of change of the scene object's current position in the heading angle direction, and the lateral target change component represents the degree of change of the scene object's current position in the vertical heading angle direction. Similarly, each of the multiple historical change quantities also includes two directional components: a longitudinal historical change component and a lateral historical change component. These components together constitute the historical data basis of the longitudinal and lateral sequences.

[0051] Furthermore, the anomaly assessment conditions also employ a directional processing strategy, including longitudinal anomaly conditions corresponding to the first direction and lateral anomaly conditions corresponding to the second direction. The longitudinal anomaly conditions are established based on the statistical characteristics of historical longitudinal variation components and are used to assess whether the current longitudinal target variation component is abnormal; the lateral anomaly conditions are established based on the statistical characteristics of historical lateral variation components and are used to assess whether the current lateral target variation component is abnormal. This directional anomaly assessment mechanism enables the system to perform accurate anomaly detection based on the characteristics of different motion directions.

[0052] Through the optional implementation methods described above, this application embodiment achieves refined modeling and analysis of the motion state of scene objects. By orthogonally decomposing motion changes along the heading angle and its perpendicular direction, and establishing corresponding change sequences and anomaly evaluation conditions, the system can more accurately capture motion characteristics and anomaly patterns in different directions. This processing method significantly improves the detection sensitivity of lateral anomaly changes, effectively solves the problem of missed detection caused by neglecting motion direction characteristics in traditional methods, and provides a more accurate and reliable basis for anomaly judgment for adaptive Kalman filtering, ultimately significantly improving the accuracy and robustness of the state estimation system in complex environments.

[0053] As an optional implementation, the position change sequence includes: a longitudinal sequence corresponding to the first direction and a transverse sequence corresponding to the second direction; in step S202 above, calculating the position change sequence based on the state information sequence may further include the following execution steps:

[0054] Step S221: For each group of adjacent moments in multiple consecutive moments, extract the first position coordinates and the first heading angle corresponding to the earlier moment from the state information sequence, and extract the second position coordinates corresponding to the later moment from the state information sequence.

[0055] Step S222: Based on the first position coordinates and the second position coordinates, calculate the position distance corresponding to adjacent time points;

[0056] Step S223: Based on the first heading angle, determine the first direction and the second direction corresponding to adjacent times;

[0057] Step S224: Calculate the projection of the position distance in the first direction to obtain the longitudinal change in the longitudinal sequence corresponding to the later time.

[0058] Step S225: Calculate the projection of the position distance in the second direction to obtain the lateral change in the lateral sequence corresponding to the later time step.

[0059] The aforementioned adjacent moments refer to two moments that are sequentially adjacent among multiple consecutive moments, including the earlier moment (e.g., moment k-1) and the later moment (e.g., moment k). The state information of the earlier and later moments are used together to calculate the position change corresponding to the adjacent moments.

[0060] In one exemplary application scenario, such as Figure 3 One optional method for calculating position change is shown, such as Figure 3 As shown, assuming a scene object is at point A at a previous time and at point B at a later time, extract the first position coordinate (x, y) corresponding to the earlier time from the state information sequence. A ,y A And the first heading angle θ, and the second position coordinates (x, y) extracted from the state information sequence at later times. B ,y B By extracting these basic state data, a data foundation is provided for subsequent calculations of position changes and directional decomposition.

[0061] The aforementioned positional distance refers to the Euclidean distance that a scene object moves between adjacent moments. This positional distance is obtained by calculating the linear difference between the first and second positional coordinates. Thus, the positional change of the scene object is transformed into a scalar distance, providing a basis for subsequent projection decomposition in a specific direction.

[0062] The first direction mentioned above refers to the principal direction of motion of the scene object at a given moment, i.e., the heading angle direction, representing the direction of the scene object's motion trend. The second direction mentioned above refers to the direction perpendicular to the first direction, usually representing the lateral or sideways motion direction of the scene object. Determining these two orthogonal directions by the heading angle allows the system to decompose the position change into two components: one consistent with the motion trend (longitudinal) and the other perpendicular to it (lateral), i.e., longitudinal change and lateral change. This is crucial for identifying anomalies (such as lateral drift) of scene objects in specific directions.

[0063] Still as Figure 3 As shown, the direction determined by the first heading angle θ is the AD direction, point C is the projection of point B onto the AD direction, and AB represents the position distance. Based on this, the longitudinal change dis is calculated according to the following formula (1). lon And calculate the lateral variation dis according to the following formula (2). lat .

[0064] dis lon =abs((x B -x A )*cosθ+(y B -y A )*sinθ) Formula (1)

[0065] dis lat =abs((x B -x A )*sinθ+(y B -y A Equation (2) is given by (cosθ).

[0066] Longitudinal variation dis lon This reflects the degree of positional change of scene objects along the heading angle. (Longitudinal change dis) lon It is calculated by projecting the position distance vector onto the first direction. The longitudinal change is recorded in the longitudinal sequence for subsequent analysis of the normal or abnormal behavior of scene objects in the direction of motion.

[0067] Horizontal variation dis lat This reflects the degree of positional change of scene objects in the vertical heading angle direction. Lateral change (dis) lat The lateral change is calculated by projecting the position distance vector onto a second direction (a unit vector perpendicular to the first direction). The lateral change is recorded in a lateral sequence. Because lateral movement is generally less associated with the normal motion patterns of an object, the lateral change is particularly sensitive to detecting anomalous positional shifts caused by factors such as sensor false alarms or the re-emergence of occluded targets.

[0068] Through steps S221 to S225 described above, this embodiment of the application achieves a refined decomposition of positional changes. Compared with the traditional method that only uses Euclidean distance, this embodiment of the application, by introducing a heading angle and decomposing the lateral and longitudinal changes, can more accurately characterize the motion patterns of scene objects, especially effectively detecting anomalies in the vertical direction that may have small amplitudes but have significant actual impacts. This provides a more accurate data foundation for subsequent anomaly detection based on statistical features, ultimately improving the robustness and accuracy of the state estimation system in complex environments.

[0069] As an optional implementation, step S203 above, which involves statistically analyzing the distribution characteristics of multiple historical changes to determine anomaly assessment conditions, may further include the following steps:

[0070] Step S231: Extract multiple vertical historical changes and multiple horizontal historical changes from multiple historical changes.

[0071] Step S232: Calculate the statistical distribution characteristics of multiple longitudinal historical changes to obtain the longitudinal mean and longitudinal standard deviation;

[0072] Step S233: Determine longitudinal outlier conditions based on the longitudinal mean and longitudinal standard deviation;

[0073] Step S234: Calculate the statistical distribution characteristics of multiple horizontal historical changes to obtain the horizontal mean and horizontal standard deviation;

[0074] Step S235: Based on the horizontal mean and horizontal standard deviation, determine the horizontal outlier conditions.

[0075] The aforementioned historical change quantities include the positional changes of scene objects at multiple historical moments, and each positional change quantity contains both vertical and horizontal components. This step extracts all vertical components from these historical data to form multiple vertical historical change quantities, and all horizontal components to form multiple horizontal historical change quantities. By separating and extracting motion changes by direction, a data foundation is provided for subsequent independent statistical analysis of the characteristics of different motion directions, enabling the system to establish statistical feature models for vertical and horizontal motion respectively.

[0076] The longitudinal mean is the arithmetic average of multiple historical longitudinal variations, reflecting the average magnitude of positional changes of scene objects in the heading direction. The longitudinal standard deviation measures the dispersion of multiple historical longitudinal variations relative to the longitudinal mean, reflecting the stability of longitudinal motion. By calculating the longitudinal mean and longitudinal standard deviation, the system can quantify the normal fluctuation range of longitudinal motion, establishing a dynamic statistical benchmark for judging longitudinal anomalies.

[0077] Vertical anomaly criteria are quantitative standards used to determine whether the current vertical variation is abnormal. This criterion constructs a dynamic threshold range based on the vertical mean and standard deviation. For example, a vertical anomaly criterion can be defined as whether the current vertical variation exceeds the range of the vertical mean plus or minus three times the vertical standard deviation. This means that, assuming the historical vertical variation follows a normal distribution, approximately 99.7% of normal data will fall within this range, and data points exceeding this range are considered abnormal. This dynamic threshold can adapt to the motion characteristics of objects in different scenarios.

[0078] The lateral mean is the arithmetic average of multiple lateral historical changes, reflecting the average magnitude of positional changes of scene objects in the vertical heading direction. The lateral standard deviation measures the dispersion of multiple lateral historical changes relative to the lateral mean, reflecting the stability of lateral motion. Since lateral motion is usually perpendicular to the main direction of motion of scene objects, its magnitude and range of variation are typically smaller than those of longitudinal motion.

[0079] Horizontal anomaly criteria are quantitative standards used to determine whether a current horizontal change is abnormal. These criteria construct a dynamic threshold range based on the horizontal mean and standard deviation. For example, a horizontal anomaly criterion can be defined as whether the current horizontal change exceeds the range of the horizontal mean plus or minus three times the horizontal standard deviation. Since small horizontal changes can indicate serious anomalies, independent statistical analysis of horizontal movement and the setting of anomaly criteria are crucial for improving the sensitivity of anomaly detection.

[0080] Still in Figure 3 In the example scenario shown, the change in longitudinal distance corresponding to each group of adjacent time points is stored in the variable sequence diffList. lon In this process, the change in lateral distance corresponding to each group of adjacent time points is stored in the variable sequence diffList. lat In China. Based on the variable sequence diffList. lon Calculate the longitudinal mean μ lon and longitudinal standard deviation σ lon Based on the variable sequence diffList lat Calculate the horizontal mean μ lat and lateral standard deviation σ lat Based on this, the longitudinal anomaly condition is determined to be: the longitudinal change dis. lon Satisfy dis lon -μ lon >3×σ lon The horizontal anomaly condition is determined as: the horizontal change dis lat Satisfy dis lat -μ lat >3×σ lat .

[0081] Through steps S231 to S235 described above, this embodiment of the application achieves refined independent statistical analysis of historical motion changes. By calculating the statistical characteristics of the longitudinal and lateral motion components separately and determining the corresponding anomaly conditions, the system can establish a more accurate anomaly assessment benchmark that better conforms to the laws of physical motion. This independent analysis approach not only significantly improves the detection capability of anomalies in the heading angle direction and its perpendicular direction, but also effectively reduces the probability of misjudgment caused by differences in motion direction. This provides a scientific and reliable decision-making basis for subsequent adaptive adjustment of filter parameters, ultimately improving the accuracy and robustness of the state estimation system in complex real-world environments.

[0082] As an optional implementation, step S204 above, which uses anomaly evaluation conditions to evaluate the target change and obtain the evaluation result, may further include the following execution steps:

[0083] Step S241: When the longitudinal target change component is detected to meet the longitudinal anomaly condition, or when the lateral target change component is detected to meet the lateral anomaly condition, the evaluation result is determined to be: the target has an abnormal position change at any given time.

[0084] The aforementioned longitudinal target change component refers to the change in the target's position along the heading angle at any given time, while the lateral target change component refers to the change in the target's position along the direction perpendicular to the heading angle at any given time. Longitudinal anomaly conditions are judgment criteria set based on the statistical characteristics (such as longitudinal mean and longitudinal standard deviation) of the longitudinal historical change component sequence, used to identify longitudinal motion anomalies. Lateral anomaly conditions are judgment criteria set based on the statistical characteristics (such as lateral mean and lateral standard deviation) of the lateral historical change component sequence, used to identify lateral motion anomalies.

[0085] When the system detects that the longitudinal target change component meets the longitudinal anomaly condition, it indicates that the scene object has experienced a positional change beyond the normal statistical range in the heading angle direction. This change may be caused by sensor noise, target occlusion, or other interference factors. When the system detects that the lateral target change component meets the lateral anomaly condition, it indicates that the scene object has experienced a positional change beyond the normal statistical range in the vertical heading angle direction. This change is usually highly correlated with anomalies because normal object motion typically maintains a small amplitude of change in the lateral direction.

[0086] The above evaluation results are determined using "OR" logic. That is, if either the longitudinal or lateral target change component satisfies the corresponding anomaly condition, the system will determine that the target has an abnormal positional change at any given time. This determination strategy ensures high sensitivity to anomalies, effectively capturing both longitudinal and lateral anomalies.

[0087] Through step S241 described above, this embodiment of the application achieves efficient and accurate detection and judgment of positional anomalies. By independently evaluating the longitudinal and lateral target change components and using "OR" logic to comprehensively determine the abnormal state, the system can comprehensively capture abnormal change patterns in different directions. This detection mechanism significantly improves the ability to identify abnormal locations, especially enhancing the detection sensitivity to small but important lateral anomalies. This provides accurate decision input for subsequent adaptive adjustment of observation noise, ultimately ensuring the robustness and reliability of the state estimation system in complex environments.

[0088] As an optional implementation, in the above state estimation method, the longitudinal target change component satisfies the longitudinal anomaly condition as follows: the longitudinal target change component exceeds the longitudinal threshold range, wherein the longitudinal threshold range is determined by the longitudinal mean and longitudinal standard deviation corresponding to the longitudinal anomaly condition; the lateral target change component satisfies the lateral anomaly condition as follows: the lateral target change component exceeds the lateral threshold range, wherein the lateral threshold range is determined by the lateral mean and lateral standard deviation corresponding to the lateral anomaly condition.

[0089] The aforementioned vertical and horizontal threshold ranges are dynamic ranges set based on statistical principles. The vertical threshold range is determined by the vertical mean and vertical standard deviation corresponding to the vertical outlier conditions, and the horizontal threshold range is determined by the horizontal mean and horizontal standard deviation corresponding to the horizontal outlier conditions.

[0090] For example, the longitudinal threshold range can be defined as [μ lon -3×σ lon μ lon +3×σ lon ], where μ lon σ represents the longitudinal mean. lon This represents the longitudinal standard deviation. The lateral threshold range can be defined as [μ]. lat -3×σ lat μ lat +3×σ lat ], where μ lat σ represents the horizontal mean. lat It represents the lateral standard deviation.

[0091] The threshold range setting method described above is based on the 3σ principle in mathematical statistics. Assuming that the data follows a normal distribution, it can ensure that 99.7% of normal data falls within the threshold range. When the vertical or horizontal target change component exceeds its corresponding threshold range, it indicates that the change has deviated from the normal fluctuation range, which is a low-probability event and can therefore be judged as abnormal.

[0092] Through the optional implementation methods described above, this application embodiment achieves the quantification and standardization of anomaly detection. By dynamically calculating the vertical and horizontal threshold ranges based on statistical features, the system can adapt to changes in the motion characteristics of objects in different scenarios, avoiding over-detection or under-detection problems that may occur with fixed thresholds. This anomaly detection mechanism based on statistical principles not only improves the scientificity and accuracy of anomaly detection but also enhances the system's adaptability to different motion modes and environments, providing a reliable technical basis for subsequent adaptive adjustment of observation noise, and ultimately significantly improving the performance and robustness of the state estimation system in complex scenarios.

[0093] As an optional implementation, step S205 above, determining the observation noise corresponding to the target time based on the evaluation results, may further include the following execution steps:

[0094] Step S251: When it is determined from the evaluation results that there is an abnormal position change of the target at any time, the initial observation noise is adjusted according to the noise amplification factor, and the observation noise is updated. The noise amplification factor is determined based on the target change amount and the abnormal threshold range corresponding to the abnormal evaluation conditions.

[0095] The initial observation noise mentioned above refers to the system's preset default observation noise covariance matrix, which reflects the confidence level of sensor measurements under normal conditions. When the evaluation results determine that there is an abnormal position change at the target time, it indicates that the current observation value may be unreliable, and it is necessary to increase the observation noise to reduce the weight of abnormal observation values ​​in the Kalman filter state update.

[0096] The noise amplification factor is a scaling factor greater than 1 used to quantitatively control the adjustment magnitude of observation noise. This noise amplification factor is determined based on the target change and the anomaly threshold range corresponding to the anomaly assessment conditions. Specifically, the noise amplification factor can be calculated based on the degree to which the target change exceeds the anomaly threshold range: the greater the exceedance, the larger the noise amplification factor, indicating higher uncertainty about the current observation and a greater need to reduce its weight.

[0097] Adjusting the initial observation noise according to the noise amplification factor involves multiplying the initial observation noise matrix by the noise amplification factor to obtain an updated observation noise matrix. This adjustment effectively increases the covariance of the observation noise, thereby reducing the confidence in outlier observations during Kalman filtering calculations and suppressing the impact of outliers on the state estimation results.

[0098] For example, in a specific application scenario, based on the Kalman filter principle, the observation noise matrix is ​​used to characterize the confidence level of the sensor observations. When an anomaly in the position of a scene object is detected at the target time, the system increases the observation noise to reduce the confidence in that anomaly observation; if no anomaly is detected at the target time, the observation noise remains at the initial set value. In this way, an adaptive noise adjustment mechanism based on anomaly detection results is achieved, thereby effectively reducing the impact of target anomaly position information on the state estimation results.

[0099] Through step S251 above, this embodiment of the application achieves intelligent adaptive adjustment of observation noise. By dynamically calculating the noise amplification coefficient based on the relative relationship between the target change and the anomaly threshold range, and precisely adjusting the observation noise accordingly, the system can automatically reduce the confidence level of unreliable observations when an anomaly is detected, effectively suppressing the negative impact of outliers on state estimation. This adaptive mechanism significantly improves the robustness of the Kalman filter algorithm under abnormal conditions, ensures the accuracy and stability of the state estimation system in complex environments, and provides more reliable technical support for vehicle-road cooperative and autonomous driving applications.

[0100] In this embodiment, a state estimation device is also provided, which is used to implement the above embodiments and preferred embodiments; details already described will not be repeated. As used below, a "module" is a combination of software and / or hardware that can perform a predetermined function. Although the device described in the following embodiments is preferably implemented in software, hardware implementation, or a combination of software and hardware, is also possible and contemplated.

[0101] Figure 4 This is a structural block diagram of a state estimation device according to an embodiment of this application, such as... Figure 4 As shown, the device includes: an acquisition module 401, used to acquire a state information sequence of scene objects, wherein the state information sequence is used to characterize the position coordinates and heading angle of the scene objects at multiple consecutive moments; a calculation module 402, used to calculate the position change sequence corresponding to the scene objects based on the state information sequence, wherein the position change sequence includes: the target change amount at the target moment, and multiple historical change amounts at multiple historical moments before the target moment; an analysis module 403, used to perform statistical distribution characteristic analysis on multiple historical change amounts to determine anomaly evaluation conditions; an anomaly evaluation module 404, used to evaluate the target change amount using the anomaly evaluation conditions to obtain an evaluation result; a determination module 405, used to determine the observation noise corresponding to the target moment based on the evaluation result; and an estimation module 406, used to perform Kalman filtering processing on the observation noise to obtain the state estimation result of the scene objects at the target moment.

[0102] Optionally, in the aforementioned state estimation device, the position change sequence includes: a longitudinal sequence corresponding to a first direction and a lateral sequence corresponding to a second direction, wherein the first direction is determined by the heading angle and the second direction is perpendicular to the first direction; the target change includes: a longitudinal target change component and a lateral target change component; each of the multiple historical change components includes: a longitudinal historical change component and a lateral historical change component; the anomaly assessment conditions include: a longitudinal anomaly condition corresponding to the first direction and a lateral anomaly condition corresponding to the second direction.

[0103] Optionally, the position change sequence includes: a longitudinal sequence corresponding to the first direction and a lateral sequence corresponding to the second direction; the calculation module 402 is further configured to: for each group of adjacent moments in multiple consecutive moments, extract the first position coordinates and the first heading angle corresponding to the earlier moment from the state information sequence, and extract the second position coordinates corresponding to the later moment from the state information sequence; calculate the position distance corresponding to the adjacent moments based on the first position coordinates and the second position coordinates; determine the first direction and the second direction corresponding to the adjacent moments based on the first heading angle; calculate the projection of the position distance in the first direction to obtain the longitudinal change amount corresponding to the later moment in the longitudinal sequence; calculate the projection of the position distance in the second direction to obtain the lateral change amount corresponding to the later moment in the lateral sequence.

[0104] Optionally, the analysis module 403 is further configured to: extract multiple longitudinal historical changes and multiple horizontal historical changes from multiple historical changes; calculate the statistical distribution characteristics of the multiple longitudinal historical changes to obtain the longitudinal mean and longitudinal standard deviation; determine the longitudinal outlier conditions based on the longitudinal mean and longitudinal standard deviation; calculate the statistical distribution characteristics of the multiple horizontal historical changes to obtain the horizontal mean and horizontal standard deviation; and determine the horizontal outlier conditions based on the horizontal mean and horizontal standard deviation.

[0105] Optionally, the above-mentioned anomaly assessment module 404 is further configured to: when the longitudinal target change component is detected to meet the longitudinal anomaly condition, or when the lateral target change component is detected to meet the lateral anomaly condition, determine the assessment result as: the target has an abnormal position change at any given time.

[0106] Optionally, in the above-mentioned state estimation device, the longitudinal target change component satisfies the longitudinal anomaly condition as follows: the longitudinal target change component exceeds the longitudinal threshold range, wherein the longitudinal threshold range is determined by the longitudinal mean and longitudinal standard deviation corresponding to the longitudinal anomaly condition; the lateral target change component satisfies the lateral anomaly condition as follows: the lateral target change component exceeds the lateral threshold range, wherein the lateral threshold range is determined by the lateral mean and lateral standard deviation corresponding to the lateral anomaly condition.

[0107] Optionally, the determination module 405 is further configured to: when it is determined from the evaluation results that there is an abnormal position change of the target at a given time, adjust the initial observation noise according to the noise amplification factor and update the observation noise, wherein the noise amplification factor is determined based on the target change amount and the abnormal threshold range corresponding to the abnormal evaluation conditions.

[0108] It should be noted that the above-mentioned acquisition module 401, calculation module 402, analysis module 403, anomaly assessment module 404, determination module 405 and estimation module 406 correspond to steps S201 to S206 in the method embodiment. The six modules are the same as the instances and application scenarios implemented by the corresponding steps, but are not limited to the content disclosed in the above method embodiment.

[0109] It should be noted that the modules mentioned in the above device embodiments can be implemented by software, hardware, or a combination of both. For example, when the modules are implemented by hardware, they can be placed in the same processor, or they can be placed in different processors in any combination. As another example, the modules can be hardware or software components stored in memory and processed by one or more processors; they can also run as part of a computing terminal.

[0110] According to another aspect of the embodiments of this application, a computer-readable storage medium is also provided, the storage medium including a stored program, wherein, when the program is running, it controls the device where the storage medium is located to execute any of the aforementioned state estimation methods.

[0111] Optionally, in this embodiment, the storage medium may be configured to store a computer program for performing the following steps: acquiring a state information sequence of a scene object, wherein the state information sequence is used to characterize the position coordinates and heading angle of the scene object at multiple consecutive moments; calculating a position change sequence corresponding to the scene object based on the state information sequence, wherein the position change sequence includes: the target change amount at the target moment, and multiple historical change amounts at multiple historical moments prior to the target moment; performing statistical distribution characteristic analysis on the multiple historical change amounts to determine anomaly evaluation conditions; evaluating the target change amount using the anomaly evaluation conditions to obtain an evaluation result; determining the observation noise corresponding to the target moment based on the evaluation result; and performing Kalman filtering on the observation noise to obtain the state estimation result of the scene object at the target moment.

[0112] Optionally, in this embodiment, the storage medium may include, but is not limited to, various media capable of storing computer programs, such as USB flash drives, read-only memory (ROM), random access memory (RAM), portable hard drives, magnetic disks, or optical disks.

[0113] According to another aspect of the embodiments of this application, an electronic device is also provided, including a memory and a processor, wherein the memory stores a computer program and the processor is configured to run the computer program to perform any of the aforementioned state estimation methods.

[0114] Optionally, in this embodiment, the processor can be configured to perform the following steps via a computer program: acquiring a state information sequence of a scene object, wherein the state information sequence is used to characterize the position coordinates and heading angle of the scene object at multiple consecutive moments; calculating a position change sequence corresponding to the scene object based on the state information sequence, wherein the position change sequence includes: the target change amount at the target moment, and multiple historical change amounts at multiple historical moments prior to the target moment; performing statistical distribution characteristic analysis on the multiple historical change amounts to determine anomaly evaluation conditions; evaluating the target change amount using the anomaly evaluation conditions to obtain an evaluation result; determining the observation noise corresponding to the target moment based on the evaluation result; and performing Kalman filtering on the observation noise to obtain the state estimation result of the scene object at the target moment.

[0115] Optionally, specific examples in this embodiment can refer to the examples described in the above embodiments and their optional implementations, and will not be repeated here.

[0116] In this application, the descriptions of the various embodiments have different focuses. For parts not described in detail in a certain embodiment, please refer to the relevant descriptions of other embodiments.

[0117] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of units can be a logical functional division, and in actual implementation, there may be other division methods. For instance, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual coupling, direct coupling, or communication connection can be through some interfaces; the indirect coupling or communication connection between units or modules can be electrical or other forms.

[0118] The units described as separate components may or may not be physically separate. The components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple units. Some or all of the units can be selected to achieve the purpose of this embodiment according to actual needs.

[0119] Furthermore, the functional units in the various embodiments of this application can be integrated into one processing unit, or each unit can exist physically separately, or two or more units can be integrated into one unit. The integrated unit can be implemented in hardware or as a software functional unit.

[0120] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, ROM, RAM, portable hard drives, magnetic disks, or optical disks.

[0121] The above description is only a preferred embodiment of this application. It should be noted that for those skilled in the art, several improvements and modifications can be made without departing from the principle of this application, and these improvements and modifications should also be considered within the scope of protection of this application.

Claims

1. A state estimation method, characterized in that, include: Obtain a sequence of state information of a scene object, wherein the sequence of state information is used to characterize the position coordinates and heading angle of the scene object at multiple consecutive moments; Based on the state information sequence, the position change sequence corresponding to the scene object is calculated, wherein the position change sequence includes: the target change amount at the target time, and multiple historical change amounts at multiple historical times before the target time; Statistical distribution characteristic analysis was performed on the multiple historical changes to determine the anomaly assessment conditions; The target change is evaluated using the aforementioned anomaly evaluation conditions to obtain the evaluation result; Based on the evaluation results, the observation noise corresponding to the target time is determined; The observation noise is processed by Kalman filtering to obtain the state estimation result of the scene object at the target time. The position change sequence includes: a longitudinal sequence corresponding to a first direction and a lateral sequence corresponding to a second direction, wherein the first direction is determined by the heading angle and the second direction is perpendicular to the first direction; The target change includes: the longitudinal target change component and the lateral target change component; Each of the multiple historical change quantities includes: a longitudinal historical change component and a horizontal historical change component. The anomaly assessment conditions include: longitudinal anomaly conditions corresponding to the first direction, and lateral anomaly conditions corresponding to the second direction; The statistical distribution characteristic analysis of the multiple historical changes to determine the anomaly assessment conditions includes: extracting multiple longitudinal historical changes and multiple horizontal historical changes from the multiple historical changes; calculating the statistical distribution characteristics of the multiple longitudinal historical changes to obtain the longitudinal mean and longitudinal standard deviation; determining the longitudinal anomaly conditions based on the longitudinal mean and the longitudinal standard deviation; calculating the statistical distribution characteristics of the multiple horizontal historical changes to obtain the horizontal mean and the horizontal standard deviation; and determining the horizontal anomaly conditions based on the horizontal mean and the horizontal standard deviation.

2. The state estimation method according to claim 1, characterized in that, The position change sequence includes: a longitudinal sequence corresponding to a first direction, and a transverse sequence corresponding to a second direction; based on the state information sequence, the position change sequence is calculated as follows: For each group of adjacent moments in a series of consecutive moments, extract the first position coordinates and the first heading angle corresponding to the earlier moment from the state information sequence, and extract the second position coordinates corresponding to the later moment from the state information sequence; Based on the first position coordinates and the second position coordinates, the position distance corresponding to the adjacent time moments is calculated; Based on the first heading angle, determine the first direction and the second direction corresponding to the adjacent time points; Calculate the projection of the position distance onto the first direction to obtain the longitudinal change in the longitudinal sequence corresponding to the later time step; Calculate the projection of the position distance onto the second direction to obtain the lateral change in the lateral sequence corresponding to the later time step.

3. The state estimation method according to claim 1, characterized in that, Using the aforementioned anomaly assessment conditions, the target change is evaluated, and the assessment results include: When the longitudinal target change component is detected to meet the longitudinal anomaly condition, or when the lateral target change component is detected to meet the lateral anomaly condition, the evaluation result is determined to be: the target has an abnormal positional change at any given time.

4. The state estimation method according to claim 3, characterized in that, The longitudinal target change component satisfies the longitudinal anomaly condition as follows: the longitudinal target change component exceeds the longitudinal threshold range, wherein the longitudinal threshold range is determined by the longitudinal mean and the longitudinal standard deviation corresponding to the longitudinal anomaly condition; The lateral target variation component satisfies the lateral anomaly condition as follows: the lateral target variation component exceeds the lateral threshold range, wherein the lateral threshold range is determined by the lateral mean and the lateral standard deviation corresponding to the lateral anomaly condition.

5. The state estimation method according to claim 1, characterized in that, Based on the evaluation results, the observation noise corresponding to the target time is determined to include: When it is determined from the evaluation results that there is an abnormal position change at the target time, the initial observation noise is adjusted according to the noise amplification factor to update the observation noise, wherein the noise amplification factor is determined based on the target change amount and the abnormal threshold range corresponding to the abnormal evaluation conditions.

6. A state estimation device, characterized in that, include: The acquisition module is used to acquire a sequence of state information of a scene object, wherein the sequence of state information is used to characterize the position coordinates and heading angle of the scene object at multiple consecutive moments; The calculation module is used to calculate the position change sequence corresponding to the scene object based on the state information sequence, wherein the position change sequence includes: the target change amount at the target time, and multiple historical change amounts at multiple historical times before the target time; The analysis module is used to perform statistical distribution characteristic analysis on the multiple historical changes and determine the anomaly assessment conditions; An anomaly assessment module is used to assess the target change using the anomaly assessment conditions and obtain an assessment result; The determination module is used to determine the observation noise corresponding to the target time based on the evaluation results; The estimation module is used to perform Kalman filtering on the observation noise to obtain the state estimation result of the scene object at the target time; The position change sequence includes: a longitudinal sequence corresponding to a first direction and a lateral sequence corresponding to a second direction, wherein the first direction is determined by the heading angle and the second direction is perpendicular to the first direction; The target change includes: the longitudinal target change component and the lateral target change component; Each of the multiple historical change quantities includes: a longitudinal historical change component and a horizontal historical change component. The anomaly assessment conditions include: longitudinal anomaly conditions corresponding to the first direction, and lateral anomaly conditions corresponding to the second direction; The analysis module is further configured to: extract multiple longitudinal historical changes and multiple horizontal historical changes from the multiple historical changes; calculate the statistical distribution characteristics of the multiple longitudinal historical changes to obtain the longitudinal mean and the longitudinal standard deviation; determine the longitudinal anomaly conditions based on the longitudinal mean and the longitudinal standard deviation; calculate the statistical distribution characteristics of the multiple horizontal historical changes to obtain the horizontal mean and the horizontal standard deviation; and determine the horizontal anomaly conditions based on the horizontal mean and the horizontal standard deviation.

7. A computer-readable storage medium, characterized in that, The computer-readable storage medium includes a stored program, wherein, when the program is executed, it controls the device on which the storage medium is located to perform the state estimation method of any one of claims 1 to 5.

8. An electronic device, characterized in that, It includes a memory and a processor, wherein the memory stores a computer program and the processor is configured to run the computer program to perform the state estimation method of any one of claims 1 to 5.

Images