Dynamic target movement tracking method and system for intelligent driving

Abstract

The invention discloses a dynamic target movement tracking method and system for intelligent driving, and relates to the technical field of target tracking, the dynamic change of a moving target is decomposed into a plurality of stages, the time length sequence of each stage is calculated, and the target state of each stage is dynamically adjusted in combination with the predicted movement track of the moving target. Collecting motion data of a moving target in real time, performing target identification and state matching, if the target data in a certain stage is not matched with expected data, adjusting a target tracking strategy according to an error, updating a motion track of the target, and if a deviation exists between a motion behavior of the moving target and the expected track in a target tracking process, adjusting the target tracking strategy according to the error. And if not, continuing to track the moving target after increasing the tracking step length. The tracking method can dynamically decompose the motion change of the moving target and incorporate the motion change into a multi-stage tracking framework, so that the system can adjust the target state at different stages according to real-time feedback, and high-precision target tracking is kept in a complex traffic environment.

Description

Technical Field

[0001] This invention relates to the field of target tracking technology, and specifically to a dynamic target movement tracking method and system for intelligent driving. Background Technology

[0002] Target motion tracking is mainly used to identify, predict, and track the dynamic behavior of targets such as vehicles or pedestrians in complex traffic environments. This technology is a key component of autonomous driving systems, providing vehicles with real-time target perception information and helping to achieve high-level functions such as environmental modeling, path planning, and decision control. With the development of autonomous driving technology, intelligent vehicles are increasingly relying on sensors (such as radar, lidar, and cameras) to obtain information about their surrounding environment. The problem of tracking dynamic targets, especially in complex urban environments, involves multiple challenges.

[0003] The existing technology has the following drawbacks: In the target tracking process, existing technologies fail to fully consider the multi-stage dynamic changes of moving targets, thus failing to accurately predict the target's trajectory and resulting in insufficient prediction accuracy of the target's state. Furthermore, in the process of identifying and matching dynamic targets, they rely on preset target templates or fixed models and fail to adjust the prediction model in real time to cope with complex changes in the traffic environment, such as interference from other targets and changes in road conditions, which increases tracking errors and collision risks. Summary of the Invention

[0004] The purpose of this invention is to provide a dynamic target movement tracking method and system for intelligent driving, which can dynamically decompose the motion changes of a moving target and incorporate them into a multi-stage tracking framework, enabling the system to adjust the target state based on real-time feedback at different stages, thereby maintaining high-precision target tracking in complex traffic environments.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a dynamic target movement tracking method for intelligent driving, the tracking method comprising the following steps: When the tracking system detects a moving target, it combines the target's motion state with a large database to generate a tracking time range. Based on the motion state and tracking time range of the moving target, the trajectory of the moving target is predicted, and the dynamic changes of the moving target are decomposed into multiple stages; Based on the dynamic changes of the moving target and the tracking time range, the time length sequence of each stage is calculated, and the target state of each stage is dynamically adjusted in combination with the predicted moving target trajectory. In the actual target tracking process, the motion data of the moving target is collected in real time, and the target is identified and matched. If the target data does not match the expected data at a certain stage, the target tracking strategy is adjusted according to the error and the target's motion trajectory is updated. If the movement of the moving target deviates from the expected trajectory during target tracking, the tracking step size is increased and the tracking of the moving target continues.

[0006] In a preferred embodiment, based on the dynamic changes of the moving target and the tracking time range, the time length sequence of each stage is calculated, and combined with the predicted trajectory of the moving target, the target state of each stage is dynamically adjusted, including the following steps: Based on the target's acceleration change magnitude, velocity fluctuation characteristics, and traffic conditions ahead in each stage, the time frame for each stage is refined. In the dynamic adjustment mechanism driven by acceleration changes, a stage change rate factor is defined. Used to measure the intensity of movement within a phase; Construct a time adjustment factor model to update the time length of each stage: ,in, This refers to the adjusted stage duration. To adjust the sensitivity coefficient, The values ​​are typically taken as [0.1, 5]. When λ≈0.1~0.5, the system responds mildly to changes in acceleration and is suitable for scenarios where the target motion changes slowly and the path is highly continuous. When λ≈1~3, it is a commonly used range of values, and the system has moderate sensitivity, suitable for general urban traffic or conventional scenarios. When λ>3 (such as taking 4 or 5), the system is extremely sensitive to small fluctuations and is suitable for high-speed scenarios or situations where the target behavior is violent and unpredictable, but it will also bring higher computation frequency and resource consumption. Based on the updated time series, the state vector of the moving target at the end of each stage is recalculated, and the micro-trajectory curve within the stage is constructed.

[0007] In a preferred embodiment, if the target data at a certain stage does not match the expected data, the target tracking strategy is adjusted according to the error, and the target's trajectory is updated, including the following steps: Real-time acquisition of the state vector of the moving target at each moment ,in , , These are the observed target position, velocity, and acceleration, respectively. Compare the observation vector with the predicted state vector To make a comparison, define the stage error index function and obtain the time. State deviation measurement ; When deviation occurs in multiple consecutive frames Exceeding the error threshold In cases where the current target state and the predicted trajectory are mismatched, the recorded error time window data is used as input for a new round of local prediction modeling. A fast recursive filtering algorithm is then used to update the target's motion estimation trajectory. The error threshold is a key reference standard for determining whether the observed state of the target deviates from the predicted state. By statistically analyzing a large amount of historical target tracking data, the average deviation and standard deviation of the observed and predicted states under different operating conditions are calculated to determine the fitness threshold. For example: ,in, The mean of the error. The standard deviation of the error. This represents the confidence factor (e.g., 2 indicates the 95% interval). This method allows the threshold to have a certain tolerance range, avoiding oversensitivity to small perturbations.

[0008] After completing the status update, adjust the phase parameters synchronously. With the predicted state sequence And rebuild the trajectory buffer.

[0009] In a preferred embodiment, the target's motion estimation trajectory is updated using a fast recursive filtering algorithm, as follows: ,in, For the updated state prediction, Estimate the current state. For observation status, For the observation model matrix, The Kalman gain matrix; The observation state is determined by the sensor system (such as camera, millimeter-wave radar, lidar, inertial navigation equipment, etc.) at a given time. The target state vector acquired in real time typically includes: position vector Velocity vector acceleration vector These data are processed by sensor fusion algorithms (such as EKF / UKF front-end prediction) and then combined to form a state vector. , which enters the filter as the observation input; The observation model matrix is ​​used to map state-space variables to the observable quantity space. It depends on: the observable structure defined by the system (i.e., which dimensions the sensors can observe), and the organization of the state vectors. Typically, if the state vectors... If the sensor only observes position and velocity, then the observation matrix will take the following form: That is, only the mapping of position and velocity is retained. If all state variables are observable, It can be an identity matrix or a constant structure, determined during the system modeling phase.

[0010] Kalman gain is the core of weight updates in filtering, reflecting the confidence distribution between predictions and observations. It is jointly determined by the current prediction error covariance and the observation noise covariance. ,in: For the prediction error covariance matrix, To observe the noise covariance matrix, it is typically set by the sensor measurement error model. The observation model matrix; The acquisition logic is as follows: dynamic calculation based on historical observation residuals and system modeling error statistics, or parameter adjustment during the design phase.

[0011] In a preferred embodiment, if the movement behavior of the moving target deviates from the expected trajectory during target tracking, the tracking step size is increased and the tracking of the moving target continues, including the following steps: If continuous time window Internal deviation value Exceeding the dynamic threshold If the behavior of the moving target changes abruptly, the response frequency adjustment mechanism is triggered. Tracking step length The update rules are as follows: ,in, For the current tracking step size, For the updated step size, The minimum sampling interval allowed by the system. To adjust the sensitivity coefficient, From a time period to The cumulative deviation score, after adjusting the tracking step size, synchronously replans the time allocation of each stage, and the dynamic threshold. The acquisition logic is similar to that of the error threshold, and will not be elaborated further in this application.

[0012] In a preferred embodiment, the trajectory of the moving target is predicted based on the moving target's motion state and tracking time range, and the dynamic changes of the moving target are decomposed into multiple stages, including the following steps: Based on the target's initial motion state, within the initial tracking time range An internal trajectory prediction model is built, and a piecewise polynomial trajectory prediction method is used to model each segment of the motion trajectory. After the motion trajectory is established, the entire prediction time interval will be... Divided into There are 1 dynamic phase, and the duration of each phase is 1 / 3. And satisfy: ,in, For the first Duration of the phase This is the initial tracking time range; Calculate the rate of change of velocity function With the rate of change of acceleration function The dividing point is determined based on the changing trends of the rate of change of velocity and the rate of change of acceleration, thus defining the location of the change in motion state; After the phase division is completed, each phase will correspond to a set of predicted state vectors. This represents the starting position, velocity, acceleration, and heading angle of the phase, and constrains the continuity of the state in each phase.

[0013] In a preferred embodiment, the trajectory prediction model expression is: ,in, For a moment The predicted location, The initial position, The initial velocity, For the initial acceleration, It is the variable acceleration of motion.

[0014] In a preferred embodiment, when the tracking system detects a moving target, it generates a tracking time range based on a large database, taking into account the target's motion state, including the following steps: Collect the target's initial state information, including its current location. ,speed acceleration Direction of movement Motion parameters, combined with historical trajectory samples from a large database, are used to select a set of historical trajectory samples corresponding to the current motion state using a similarity matching algorithm. Each trajectory Each sample set includes start and end times, path features, and traffic scene classification information. Based on the sample set, an initial tracking time range for the moving target is constructed. .

[0015] In a preferred embodiment, an initial tracking time range for the moving target is constructed based on the sample set. The expression is: ,in, The average tracking time for the sample trajectory. Standard deviation is the confidence coefficient.

[0016] This application also provides a dynamic target movement tracking system for intelligent driving, including a phase division module, a dynamic adjustment module, and a tracking optimization module; Stage division module: When a moving target is detected, the module combines the target's motion state with a large database to generate a tracking time range, predicts the target's trajectory based on the target's motion state and tracking time range, and decomposes the target's dynamic changes into multiple stages. Dynamic adjustment module: Based on the dynamic changes of the moving target and the tracking time range, calculate the time length sequence of each stage, and combine it with the predicted moving target trajectory to dynamically adjust the target state at each stage; Tracking optimization module: During actual target tracking, the module collects motion data of the moving target in real time, performs target identification and state matching. If the target data does not match the expected data at a certain stage, the module adjusts the target tracking strategy according to the error and updates the target's trajectory. If the moving target's motion behavior deviates from the expected trajectory during target tracking, the module increases the tracking step size and continues tracking the moving target.

[0017] The technical effects and advantages provided by the present invention in the above technical solution are as follows: This invention predicts the trajectory of a moving target based on its motion state and tracking time range. It decomposes the dynamic changes of the moving target into multiple stages, calculates the time length sequence of each stage, and dynamically adjusts the target state at each stage in conjunction with the predicted trajectory. The invention also collects the target's motion data in real time for target identification and state matching. If the target data at a certain stage does not match the expected data, the target tracking strategy is adjusted based on the error, and the target's trajectory is updated. If the moving target's behavior deviates from the expected trajectory during tracking, the tracking step size is increased before continuing to track the target. This tracking method dynamically decomposes the motion changes of the moving target and incorporates them into a multi-stage tracking framework, enabling the system to adjust the target state based on real-time feedback at different stages, thereby maintaining high-precision target tracking in complex traffic environments. Attached Figure Description

[0018] To more clearly illustrate the technical solutions in the embodiments of this application or the prior art, the drawings used in the embodiments will be briefly introduced below. Obviously, the drawings described below are only some embodiments recorded in this invention. For those skilled in the art, other drawings can be obtained based on these drawings.

[0019] Figure 1 This is a flowchart of the present invention.

[0020] Figure 2 This is a diagram of the architecture of the present invention. Detailed Implementation

[0021] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0022] Example: Please refer to Figure 1 As shown in the figure, the dynamic target movement tracking method for intelligent driving described in this embodiment includes the following steps: When the tracking system detects a moving target, it combines the target's motion state with a large database to generate a tracking time range.

[0023] Based on the motion state of the moving target and the tracking time range, the trajectory of the moving target is predicted, and the dynamic changes of the moving target such as speed and position are decomposed into multiple stages.

[0024] Based on the dynamic changes of the moving target and the tracking time range, the time length sequence of each stage is calculated, and combined with the predicted moving target trajectory, the target state of each stage is dynamically adjusted to ensure accurate tracking of the target in the intelligent driving environment.

[0025] In actual target tracking, motion data of the moving target is collected in real time for target identification and state matching. If the target data does not match the expected data at a certain stage, the target tracking strategy is adjusted according to the error, and the target's trajectory is updated.

[0026] If, during target tracking, a significant deviation is found between the target's movement behavior and the expected trajectory (such as a sudden change in path or a drastic change in vehicle speed), the tracking step size is increased, which means increasing the response frequency to the target's dynamic changes to ensure timely response to changes in the target's behavior.

[0027] This application predicts the trajectory of a moving target based on its motion state and tracking time range. It decomposes the dynamic changes of the moving target into multiple stages, calculates the time length sequence of each stage, and dynamically adjusts the target state at each stage in conjunction with the predicted trajectory. The system collects motion data of the moving target in real time for target identification and state matching. If the target data at a certain stage does not match the expected data, the target tracking strategy is adjusted according to the error, and the target's trajectory is updated. If the moving target's behavior deviates from the expected trajectory during tracking, the tracking step size is increased before continuing to track the moving target. This tracking method can dynamically decompose the motion changes of the moving target and incorporate them into a multi-stage tracking framework, enabling the system to adjust the target state based on real-time feedback at different stages, thereby maintaining high-precision target tracking in complex traffic environments.

[0028] When the tracking system detects a moving target, it combines the target's motion state with a large database to generate a tracking time range.

[0029] When the tracking system detects a moving target, in order to achieve efficient and accurate target tracking in an intelligent driving environment, it first needs to generate an initial tracking time range based on the moving target's current motion state and historical motion pattern data in a large database. This process does not involve phase division, but rather predicts and calculates the entire possible tracking path, providing a foundation for subsequent phased modeling and dynamic adjustments.

[0030] In this step, the system first collects the target's initial state information, including its current position. ,speed acceleration Direction of movement Key motion parameters are collected. Combining historical trajectory samples from a large database, the system uses a similarity matching algorithm (e.g., time series comparison methods based on KNN or cosine similarity) to filter out a set of historical trajectory samples corresponding to the current motion state. Each trajectory Each sample set includes complete start and end times, path characteristics, and traffic scenario classifications (such as highway conditions and urban congestion). Based on this sample set, the system constructs the initial tracking time range for the target. This time range can be estimated using the following formula: ,in, To predict the total driving distance by combining historical trajectory data with current traffic conditions, The initial velocity, For acceleration, This is used to estimate the average response interval (e.g., 0.5 seconds) during the process. The function of this formula is to determine the longest possible time period during which the target can be continuously tracked under conditions of no severe external interference, based on the current target's dynamic characteristics and the predicted path. Furthermore, to improve the robustness of the prediction, the system calculates upper and lower time limits for multiple historical trajectory samples and uses a statistical model (e.g., a 95% confidence interval) to generate a reasonable time interval. Finally determined: ,in, The average tracking time for the sample trajectory. Standard deviation This represents the confidence coefficient (e.g., 1.96 corresponds to a 95% confidence level). This statistical modeling process helps avoid distortion in tracking time predictions due to individual outliers.

[0031] In summary, this step focuses on trajectory matching and time modeling driven by a large database. By introducing dynamic modeling and statistical learning methods, it provides a scientific and reasonable initial time frame for the continuous tracking of the target, and serves as the input basis for subsequent staged modeling and dynamic adjustment mechanisms.

[0032] Based on the motion state of the moving target and the tracking time range, the trajectory of the moving target is predicted, and the dynamic changes of the moving target such as speed and position are decomposed into multiple stages.

[0033] After determining the initial tracking timeframe for the moving target, the system needs to further predict its future trajectory based on the target's current motion state, dividing the entire prediction process into several physically consistent dynamic change stages. The core of this process lies in transforming complex, continuous motion behavior into a calculable and controllable staged description through trajectory modeling and dynamic evolution, thereby improving the decision-making accuracy and response efficiency of the intelligent driving system.

[0034] First, the system is based on the target's initial motion state. Within the initial tracking time range An internal trajectory prediction model is built. Considering the target's possible non-uniform velocity and non-uniform acceleration behavior, a piecewise polynomial trajectory prediction method is used to model each segment of the trajectory, whose position change can be expressed as a third-order polynomial in time: ,in, For a moment The predicted location, The initial position, The initial velocity, For the initial acceleration, This refers to the variable acceleration of motion, or jerk. This formula can describe more complex trajectory curves and is used to simulate sudden acceleration, deceleration, or turning behavior, which helps improve the accuracy and robustness of the predicted trajectory. After the motion trajectory is established, the system will predict the entire time interval. Divided into There are 1 dynamic phase, and the duration of each phase is 1 / 3. And satisfy: ,in, For the first The duration of the phase This is the initial tracking time range. The principles for dividing the stages include key indicators such as the rate of change of velocity, the change of acceleration curvature, and predicted events (such as turning and lane changing). A dynamic window adjustment mechanism is adopted to ensure that the motion parameters are relatively stable within each stage, which is conducive to subsequent state matching and anomaly detection.

[0035] To rationally divide the stages, the system calculates the rate of change function of velocity. With the rate of change of acceleration function As shown below: ,in, Let be the absolute rate of change of velocity with time. For a moment velocity function; Let be the absolute rate of change of acceleration with time. The acceleration function is used. The system determines the cut-off points based on the changing trends of these indicators, delineating locations where the motion state changes significantly to form stable phases. After phase division, each phase corresponds to a set of predicted state vectors. This represents the initial position, velocity, acceleration, and heading angle for that phase. The system constrains the continuity of the state in each phase to ensure that: This means that the initial state of each stage is consistent with the final state of the previous stage, thus avoiding inconsistencies in the predicted trajectory.

[0036] In this way, the system accurately divides the dynamic evolution of the target into multiple continuous stages within the overall prediction time range. Each stage has stable physical properties and a controllable evolution trend, providing a solid data and modeling foundation for subsequent dynamic adjustment, state matching and anomaly handling.

[0037] Based on the dynamic changes of the moving target and the tracking time range, the time length sequence of each stage is calculated, and combined with the predicted moving target trajectory, the target state of each stage is dynamically adjusted to ensure accurate tracking of the target in the intelligent driving environment.

[0038] After completing the multi-stage trajectory prediction of the moving target, the system needs to further calculate the time length sequence of each stage based on the dynamic change characteristics of the target at different times and its environment. During the target's operation, the system should dynamically adjust the state of each stage in combination with the trajectory evolution trend to improve the system's response efficiency and trajectory fit in the intelligent driving environment.

[0039] First, the system refines each stage in terms of time based on the target's acceleration change magnitude, velocity fluctuation characteristics, and traffic conditions ahead. In the dynamic adjustment mechanism driven by acceleration changes, a stage change rate factor is defined. Used to measure the intensity of movement within a phase: ,in, For the stage The rate of change of acceleration, For the goal at all times acceleration, The duration of the phase. It reflects the intensity of motion state fluctuations within this stage and is used to determine whether the duration of this stage needs to be shortened to enhance the state response.

[0040] Based on this metric, the system constructs a time adjustment factor model to update the time length of each stage: ,in, This refers to the adjusted stage duration. To adjust the sensitivity coefficient, a trade-off is made between response speed and computational cost. The exponential scaling function is used to adjust the sensitivity coefficient when the target's motion state changes rapidly (i.e., when the target's motion state changes rapidly...). When the time is relatively large, the stage duration is compressed to improve the system's real-time response capability. Next, to achieve continuity in trajectory tracking, the system recalculates the target's state vector at the end of each stage based on the updated time series. Furthermore, micro-trajectory curves within each stage are constructed, and cubic spline interpolation or Bézier curve methods are used to achieve high-precision trajectory fitting and smooth state transition.

[0041] Furthermore, to achieve dynamically adjusted closed-loop control, an online error evaluation mechanism is introduced at the end of each stage, defining a prediction deviation error term: ,in, For the stage The state error, , Predicting position and velocity respectively. For the observed true state, This is the speed error weighting factor. If If the threshold is exceeded, the reconstruction and feedback adjustment of the state in the next stage will be triggered to ensure the dynamic consistency between the state sequence and the actual running trajectory.

[0042] By establishing a time-adaptive adjustment model with the rate of change of acceleration as the core, and combining it with a real-time error closed-loop feedback mechanism, the system can dynamically optimize its time span and state output for each stage, achieving highly robust tracking control of the trajectory of moving targets in complex driving environments.

[0043] In actual target tracking, motion data of the moving target is collected in real time for target identification and state matching. If the target data does not match the expected data at a certain stage, the target tracking strategy is adjusted according to the error, and the target's trajectory is updated.

[0044] In actual operation, the system needs to collect multi-dimensional motion data of moving targets in real time and perform target recognition and state matching operations based on the predicted trajectory model. This process not only requires high-frequency data sampling and high-precision target recognition algorithms, but also relies on a robust state comparison mechanism to dynamically analyze and correct the deviation between the current observed state and the predicted state in order to maintain tracking accuracy and system stability.

[0045] First, the system fuses multiple sensors, such as LiDAR, millimeter-wave radar, visual sensors, or V2X, to acquire the target's state vector at each moment in real time. ,in , , These are the observed target position, velocity, and acceleration, respectively. The system compares this observation vector with the predicted state vector. To make a comparison, define the stage error index function: ,in, For a moment State deviation measure This is a velocity deviation weighting factor used to balance the importance of position and velocity errors; it quantifies the difference between system perception and prediction during tracking and serves as a trigger for error adjustment. When deviations occur over multiple consecutive frames (e.g., a threshold of 3 frames),... Exceeding the error threshold In such cases, the system determines that there is a mismatch between the current target state and the expected predicted trajectory, and enters the error adjustment and trajectory update module. The system uses the data within the recorded error time window as input for a new round of local prediction modeling, and updates the target's estimated motion trajectory through a fast recursive filtering algorithm, such as Extended Kalman Filter (EKF) or Unscented Kalman Filter (UKF). ,in, For the updated state prediction, Estimate the current state. For observation status, For the observation model matrix, This is the Kalman gain matrix. The function of this formula is to use observation errors to correct state predictions in reverse, thereby improving the adaptability and accuracy of subsequent prediction models.

[0046] After completing the status update, the system also needs to adjust the phase parameters synchronously. With the predicted state sequence The system reconstructs the trajectory buffer to ensure that subsequent stages can execute state recognition and control strategies based on more accurate prior states. Furthermore, to improve the sensitivity and stability of anomaly detection, a cumulative deviation scoring mechanism is introduced. ,in, From a time period to The cumulative deviation score is used to capture the trajectory deviation trend. If If the cumulative deviation threshold is exceeded, the target response frequency will be automatically increased and trajectory reconstruction will be initiated, thereby improving the response capability to sudden behavior. Through the two-way fusion of real-time observation and prediction status, closed-loop control is achieved at three levels: deviation detection, error modeling and prediction feedback, thereby ensuring stable and high-precision tracking of the target trajectory in complex and dynamic traffic scenarios.

[0047] If, during target tracking, a significant deviation is found between the target's movement behavior and the expected trajectory (such as a sudden change in path or a drastic change in vehicle speed), the tracking step size is increased, which means increasing the response frequency to the target's dynamic changes to ensure timely response to changes in the target's behavior.

[0048] In the target tracking process of intelligent driving systems, to ensure timely response to sudden changes in target behavior, the system needs to automatically increase the tracking step size when it detects a significant increase in motion deviation. This involves shortening the data sampling period and increasing the frequency of state assessment, thereby enhancing the real-time perception and adaptation capabilities to dynamic trajectory changes. This mechanism is particularly suitable for rapidly capturing nonlinear motion behaviors such as sudden path changes, drastic speed fluctuations, and abnormal acceleration / deceleration.

[0049] If continuous time window Internal deviation value Exceeding the dynamic threshold If the system determines that the target behavior has a sudden change, it triggers the response frequency adjustment mechanism. (Tracking step size) The update rules are as follows: ,in, For the current tracking step size, For the updated step size, The minimum sampling interval allowed by the system. To adjust the sensitivity coefficient, From a time period to The cumulative deviation score is used to adaptively compress the system response frequency: the larger the deviation, the higher the response frequency, thereby reducing the sampling period and improving the real-time performance of prediction and control.

[0050] After adjusting the tracking step size, the system also needs to synchronously replan the time allocation of each stage, so that the duration of each tracking stage is shortened to adapt to the rapidly fluctuating target motion state, and to ensure that state matching and trajectory updates are within the latest perception cycle.

[0051] Meanwhile, to avoid computational resource overload caused by persistently high frequencies, the system introduces a frequency decay control function, which gradually restores the step size to the default level when the target state returns to stability. ,in, The step size recovery rate coefficient is... This sets the initial sampling period. The function of this formula is to gradually reduce the response frequency and return to normal tracking load after the system returns to a stable state, ensuring system efficiency and stability. Finally, to improve the model's response performance to rapid state changes, key covariance parameters in the filtering module (such as extended Kalman filter or particle filter) also need to be adaptively adjusted, for example, dynamically reducing the process noise covariance matrix. The positional term weights are adjusted to enhance sensitivity to abrupt changes in position.

[0052] The tracking step length enhancement mechanism establishes a high-frequency closed-loop tracking system for nonlinear dynamic behavior by introducing multiple methods such as deviation triggering, response frequency reconstruction, stage duration reduction, frequency self-recovery, and dynamic parameter adjustment of the filter. This effectively improves the timeliness of response and tracking stability to uncertain target behavior in complex driving environments.

[0053] Please see Figure 2 As shown in the figure, the dynamic target movement tracking system for intelligent driving described in this embodiment includes a stage division module, a dynamic adjustment module, and a tracking optimization module; Phase division module: When a moving target is detected, the module combines the target's motion state with a large database to generate a tracking time range. Based on the target's motion state and tracking time range, the module predicts the target's trajectory and decomposes the target's dynamic changes into multiple phases. The phase division results are then sent to the dynamic adjustment module. Dynamic adjustment module: Based on the dynamic changes of the moving target and the tracking time range, calculate the time length sequence of each stage, and combine it with the predicted moving target trajectory to dynamically adjust the target state at each stage. The dynamic adjustment results are sent to the tracking optimization module. Tracking optimization module: In the actual target tracking process, the motion data of the moving target is collected in real time, and the target is identified and matched. If the target data does not match the expected data at a certain stage, the target tracking strategy is adjusted according to the error and the target's motion trajectory is updated. If the moving target's motion behavior deviates from the expected trajectory during the target tracking process, the tracking step size is increased and the tracking optimization module continues to track the moving target.

[0054] In the description of this specification, references to terms such as "an embodiment," "example," "specific example," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0055] The preferred embodiments of the present invention disclosed above are merely illustrative of the invention. These preferred embodiments do not exhaustively describe all details, nor do they limit the invention to any specific implementation. Clearly, many modifications and variations can be made based on the content of this specification. This specification selects and specifically describes these embodiments to better explain the principles and practical applications of the invention, thereby enabling those skilled in the art to better understand and utilize the invention. The invention is limited only by the claims and their full scope and equivalents.

Claims

1. A dynamic target movement tracking method for intelligent driving, characterized in that: The tracking method includes the following steps: When the tracking system detects a moving target, it combines the target's motion state with a large database to generate a tracking time range. Based on the motion state and tracking time range of the moving target, the trajectory of the moving target is predicted, and the dynamic changes of the moving target are decomposed into multiple stages; Based on the dynamic changes of the moving target and the tracking time range, the time length sequence of each stage is calculated, and the target state of each stage is dynamically adjusted in combination with the predicted moving target trajectory. In the actual target tracking process, the motion data of the moving target is collected in real time, and the target is identified and matched. If the target data does not match the expected data at a certain stage, the target tracking strategy is adjusted according to the error and the target's motion trajectory is updated. If the movement of the moving target deviates from the expected trajectory during target tracking, the tracking step size is increased and the tracking of the moving target continues.

2. The dynamic target movement tracking method for intelligent driving according to claim 1, characterized in that: Based on the dynamic changes of the moving target and the tracking time range, the time length sequence of each stage is calculated, and combined with the predicted moving target trajectory, the target state of each stage is dynamically adjusted, including the following steps: Based on the target's acceleration change magnitude, velocity fluctuation characteristics, and traffic conditions ahead in each stage, the time frame for each stage is refined. In the dynamic adjustment mechanism driven by acceleration changes, a stage change rate factor is defined. Used to measure the intensity of movement within a phase; Construct a time adjustment factor model to update the time length of each stage: ,in, This refers to the adjusted stage duration. To adjust the sensitivity coefficient; Based on the updated time series, the state vector of the moving target at the end of each stage is recalculated, and the micro-trajectory curve within the stage is constructed.

3. The dynamic target movement tracking method for intelligent driving according to claim 2, characterized in that: If the target data does not match the expected data at a certain stage, the target tracking strategy is adjusted according to the error, and the target's trajectory is updated, including the following steps: Real-time acquisition of the state vector of the moving target at each moment ,in , , These are the observed target position, velocity, and acceleration, respectively. Compare the observation vector with the predicted state vector To make a comparison, define the stage error index function and obtain the time. State deviation measurement ; When deviation occurs in multiple consecutive frames Exceeding the error threshold In such cases, it is determined that there is a mismatch between the current target state and the expected predicted trajectory. The data within the recorded error time window is used as the input for a new round of local prediction modeling, and the target's motion estimation trajectory is updated through a fast recursive filtering algorithm. After completing the status update, adjust the phase parameters synchronously. With the predicted state sequence And rebuild the trajectory buffer.

4. The dynamic target movement tracking method for intelligent driving according to claim 3, characterized in that: The motion estimate trajectory of the target is updated using a fast recursive filtering algorithm, as follows: ,in, For the updated state prediction, Estimate the current state. For observation status, For the observation model matrix, This is the Kalman gain matrix.

5. The dynamic target movement tracking method for intelligent driving according to claim 4, characterized in that: If the movement of the moving target deviates from the expected trajectory during target tracking, the tracking step size is increased and the tracking of the moving target continues, including the following steps: If continuous time window Internal deviation value Exceeding the dynamic threshold If the behavior of the moving target changes abruptly, the response frequency adjustment mechanism is triggered. Tracking step length The update rules are as follows: ,in, For the current tracking step size, For the updated step size, The minimum sampling interval allowed by the system. To adjust the sensitivity coefficient, From a time period to The cumulative deviation score is used to synchronously replan the time allocation of each stage after the tracking step size is adjusted.

6. The dynamic target movement tracking method for intelligent driving according to claim 5, characterized in that: Based on the moving target's motion state and tracking time range, the trajectory of the moving target is predicted, and the dynamic changes of the moving target are decomposed into multiple stages, including the following steps: Based on the target's initial motion state, within the initial tracking time range An internal trajectory prediction model is built, and a piecewise polynomial trajectory prediction method is used to model each segment of the motion trajectory. After the motion trajectory is established, the entire prediction time interval will be... Divided into There are 1 dynamic phase, and the duration of each phase is 1 / 3. And satisfy: ,in, For the first Duration of the phase This is the initial tracking time range; Calculate the rate of change of velocity function With the rate of change of acceleration function The dividing point is determined based on the changing trends of the rate of change of velocity and the rate of change of acceleration, thus defining the location of the change in motion state; After the phase division is completed, each phase will correspond to a set of predicted state vectors. This represents the starting position, velocity, acceleration, and heading angle of the phase, and constrains the continuity of the state in each phase.

7. The dynamic target movement tracking method for intelligent driving according to claim 6, characterized in that: The trajectory prediction model expression is: ,in, For a moment The predicted location, The initial position, The initial velocity, For the initial acceleration, It is the variable acceleration of motion.

8. The dynamic target movement tracking method for intelligent driving according to claim 7, characterized in that: When the tracking system detects a moving target, it combines the target's motion state with a large database to generate a tracking time range, including the following steps: Collect the target's initial state information, including its current location. ,speed acceleration Direction of movement Motion parameters, combined with historical trajectory samples from a large database, are used to select a set of historical trajectory samples corresponding to the current motion state using a similarity matching algorithm. Each trajectory Each sample set includes start and end times, path features, and traffic scene classification information. Based on the sample set, an initial tracking time range for the moving target is constructed. .

9. The dynamic target movement tracking method for intelligent driving according to claim 8, characterized in that: Constructing the initial tracking time range for moving targets based on the sample set. The expression is: ,in, The average tracking time for the sample trajectory. Standard deviation is the confidence coefficient.

10. A dynamic target movement tracking system for intelligent driving, used to implement the tracking method according to any one of claims 1-9, characterized in that: It includes a phase division module, a dynamic adjustment module, and a tracking and optimization module; Stage division module: When a moving target is detected, the module combines the target's motion state with a large database to generate a tracking time range, predicts the target's trajectory based on the target's motion state and tracking time range, and decomposes the target's dynamic changes into multiple stages. Dynamic adjustment module: Based on the dynamic changes of the moving target and the tracking time range, calculate the time length sequence of each stage, and combine it with the predicted moving target trajectory to dynamically adjust the target state at each stage; Tracking optimization module: During actual target tracking, the module collects motion data of the moving target in real time, performs target identification and state matching. If the target data does not match the expected data at a certain stage, the module adjusts the target tracking strategy according to the error and updates the target's trajectory. If the moving target's motion behavior deviates from the expected trajectory during target tracking, the module increases the tracking step size and continues tracking the moving target.

Images