Target tracking method and device and domain controller

Abstract

The invention discloses a target tracking method and device and a domain controller, and belongs to the technical field of computer vision. According to the target tracking scheme provided by the invention, the threshold parameter used for target tracking can be dynamically adjusted according to the target state and the external environment, even if the environmental complexity is improved or the target state is changed, dynamic adaptation can be realized, and the problem that the tracking precision is reduced does not occur. In other words, according to the scheme, the accuracy and robustness of target tracking are improved, and the probability of target loss and error tracking is greatly reduced. Therefore, powerful data support is provided for subsequent decision making of the system, the decision making accuracy of the system is improved, and then the safety and reliability of automatic driving are ensured.

Description

Technical Field

[0001] The present application relates to the field of computer vision technology, and in particular to a target tracking method, device, and domain controller. Background Art

[0002] Object tracking is a core task in computer vision and is currently being applied in numerous fields, such as autonomous driving and robotic vision. In autonomous driving, object tracking uses multi-sensor fusion, including cameras, lidar, and millimeter-wave radar, to capture the position, velocity, and acceleration of targets (such as vehicles, pedestrians, or non-motorized vehicles) in the road environment in real time, predict their future behavior, and provide data support for path planning and obstacle avoidance decisions.

[0003] Based on the above description, it can be seen that the quality of tracking directly affects the system's subsequent decision-making. Therefore, in order to improve the safety and reliability of autonomous driving, how to improve the accuracy and robustness of target tracking and reduce the probability of target loss or mistracking has become a focus of technical personnel in this field. Summary of the Invention

[0004] The present invention provides a target tracking method, device, and domain controller. The technical solution is as follows:

[0005] In one aspect, a target tracking method is provided, the method comprising:

[0006] Acquiring sensor data, performing target detection based on the sensor data, and obtaining a target detection result;

[0007] In the initial tracking phase, the state of the potential target is initialized according to the target detection result, the threshold parameters for target tracking are initialized, and the potential target is tracked and verified based on the initialized threshold parameters; wherein the potential target is the target identified during the target detection process;

[0008] During the stable tracking phase, the threshold parameter is adjusted based on at least one of the current motion parameters of the target being tracked or the external environment; wherein the target being tracked is a target among the potential targets that has passed tracking verification; for any target, the target passing tracking verification means that the target matches the established tracking track and matches the tracking track for multiple consecutive frames;

[0009] According to the adjusted threshold parameters, the disappearance determination and reappearance detection of the tracked target are performed, and when the tracked target does not disappear or reappear, the state of the tracked target is updated.

[0010] In some embodiments, the threshold parameters include a distance threshold, an appearance threshold, and a disappearance threshold; wherein the distance threshold is used to determine whether the target being tracked is pointing to the same object; the appearance threshold is used to determine whether the detected new target is in a valid tracking state; and the disappearance threshold is used to determine whether the target being tracked is lost.

[0011] The adjusting the threshold parameter according to at least one of the motion state data of the target being tracked or the external environment includes:

[0012] Adjusting an initially set distance threshold according to a current speed of the target being tracked and a maximum speed limit in an environment in which the target being tracked is currently located;

[0013] Adjust the initially set appearance threshold and disappearance threshold according to the complexity coefficient of the current environment and the set maximum environment complexity coefficient.

[0014] In some other embodiments, the external environment includes road conditions and weather conditions; and the method further includes:

[0015] According to preset weights, complexity linear weighting is performed on the indicators included in the road conditions and the indicators included in the weather conditions to obtain a complexity coefficient of the current environment of the target being tracked.

[0016] In some other embodiments, updating the status of the tracked target includes:

[0017] According to a preset tracking strategy, the status of the target being tracked is updated; wherein the preset tracking strategy includes a status update frequency.

[0018] In some other embodiments, the method further comprises:

[0019] Determining a risk level corresponding to the tracked target based on a positional relationship between the tracked target and the lane in which the vehicle is located, a distance between the tracked target and the vehicle, and a speed relative to the vehicle;

[0020] A tracking strategy corresponding to the risk level is determined as the preset tracking strategy.

[0021] In some other embodiments, if there are multiple targets being tracked, updating the status of the targets being tracked includes:

[0022] Determine the tracking priority of each target according to the risk level of each target being tracked;

[0023] Update the status of each tracked target in sequence according to the determined priority.

[0024] In some other embodiments, during the initial tracking phase, the method further includes:

[0025] configuring a motion prediction model for the potential target;

[0026] If the potential target passes the tracking verification, a tracking identifier is assigned to the potential target, a timestamp of the first detection of the potential target is recorded, and the potential target is added to the tracking object pool as a target that needs to be tracked.

[0027] In some other embodiments, updating the status of the tracked target includes:

[0028] Predicting the state of the target being tracked at a current moment using a motion prediction model according to the state of the target being tracked at a previous moment to obtain a predicted state;

[0029] The predicted state is corrected according to real-time sensor data related to the tracked target to obtain the final state of the tracked target at the current moment.

[0030] In another aspect, a target tracking device is provided, comprising:

[0031] a target detection module configured to acquire sensor data, perform target detection based on the sensor data, and obtain a target detection result;

[0032] a first tracking module configured to, in an initial tracking phase, initialize a state of a potential target according to the target detection result, initialize a threshold parameter for target tracking, and track and verify the potential target based on the initialized threshold parameter; wherein the potential target is a target identified during the target detection process;

[0033] a threshold adjustment module configured to adjust the threshold parameter during a stable tracking phase based on at least one of a current motion parameter of a target being tracked or an external environment; wherein the target being tracked is a target among the potential targets that has passed tracking verification; for any target, passing tracking verification means that the target matches a created tracking track and matches the tracking track for multiple consecutive frames;

[0034] The second tracking module is configured to perform disappearance determination and reappearance detection on the tracked target according to the adjusted threshold parameter, and update the status of the tracked target if the tracked target does not disappear or reappear.

[0035] In some embodiments, the threshold parameters include a distance threshold, an appearance threshold, and a disappearance threshold; wherein the distance threshold is used to determine whether the target being tracked is pointing to the same object; the appearance threshold is used to determine whether the detected new target is in a valid tracking state; and the disappearance threshold is used to determine whether the target being tracked is lost.

[0036] The threshold adjustment module is configured to:

[0037] Adjusting an initially set distance threshold according to a current speed of the target being tracked and a maximum speed limit in an environment in which the target being tracked is currently located;

[0038] Adjust the initially set appearance threshold and disappearance threshold according to the complexity coefficient of the current environment and the set maximum environment complexity coefficient.

[0039] In other embodiments, the external environment includes road conditions and weather conditions;

[0040] The threshold adjustment module is configured to perform complexity linear weighting on the indicators included in the road conditions and the indicators included in the weather conditions according to preset weights to obtain a complexity coefficient of the current environment of the target being tracked.

[0041] In some other embodiments, the second tracking module is configured to:

[0042] According to a preset tracking strategy, the status of the target being tracked is updated; wherein the preset tracking strategy includes a status update frequency.

[0043] In some other embodiments, the second tracking module is further configured to:

[0044] Determining a risk level corresponding to the tracked target based on a positional relationship between the tracked target and the lane in which the vehicle is located, a distance between the tracked target and the vehicle, and a speed relative to the vehicle;

[0045] A tracking strategy corresponding to the risk level is determined as the preset tracking strategy.

[0046] In some other embodiments, the second tracking module is configured to:

[0047] If there are multiple targets being tracked, determining the tracking priority of each target being tracked according to the risk level corresponding to each target being tracked;

[0048] Update the status of each tracked target in sequence according to the determined priority.

[0049] In some other embodiments, the first tracking module is further configured to:

[0050] In the initial tracking stage, a motion prediction model is configured for the potential target; if the potential target passes the tracking verification, a tracking identifier is assigned to the potential target, the timestamp of the first detection of the potential target is recorded, and the potential target is added to the tracking object pool as a target that needs to be tracked.

[0051] In some other embodiments, the second tracking module is configured to:

[0052] Predicting the state of the target being tracked at a current moment using a motion prediction model according to the state of the target being tracked at a previous moment to obtain a predicted state;

[0053] The predicted state is corrected according to real-time sensor data related to the tracked target to obtain the final state of the tracked target at the current moment.

[0054] On the other hand, a domain controller is provided. The device includes a processor and a memory. The memory stores at least one program code. The at least one program code is loaded and executed by the processor to implement the above-mentioned target tracking method.

[0055] On the other hand, a computer-readable storage medium is provided, wherein at least one program code is stored in the storage medium, and the at least one program code is loaded and executed by a processor to implement the above-mentioned target tracking method.

[0056] On the other hand, a computer program product or computer program is provided, which includes computer program code, which is stored in a computer-readable storage medium. A processor of a domain controller reads the computer program code from the computer-readable storage medium, and the processor executes the computer program code, so that the domain controller performs the above-mentioned target tracking method.

[0057] The target tracking solution provided by the embodiment of the present application can dynamically adjust the threshold parameters used for target tracking according to the target state and the external environment. Even if the complexity of the environment increases or the target state changes, it can adapt dynamically and there will be no problem of decreased tracking accuracy. In other words, the solution improves the accuracy and robustness of target tracking and greatly reduces the probability of target loss and mistracking. This provides strong data support for subsequent decision-making of the system, improves the accuracy of system decision-making, and thus ensures the safety and reliability of autonomous driving. Among them, the solution can maintain efficient target tracking in various environments by adaptively adjusting the threshold, whether it is a fast-moving target on a highway or a slow-moving target on a complex urban road, accurate tracking can be achieved. In addition, the solution adaptively adjusts the threshold parameters in combination with the target state and the external environment, which can provide more stable tracking results after the target is temporarily lost, reducing tracking failures caused by occlusion, harsh environment, etc. BRIEF DESCRIPTION OF THE DRAWINGS

[0058] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0059] Figure 1 This is a schematic diagram of an implementation environment of a target tracking method provided in an embodiment of the present application;

[0060] Figure 2 This is a flow chart of a target tracking method provided by an embodiment of the present application;

[0061] Figure 3 This is a schematic structural diagram of a target tracking device provided in an embodiment of the present application;

[0062] Figure 4 This is a structural diagram of a domain controller provided in an embodiment of the present application. DETAILED DESCRIPTION

[0063] In order to make the objectives, technical solutions and advantages of this application clearer, the implementation methods of this application will be further described in detail below with reference to the accompanying drawings.

[0064] In this application, the terms "first," "second," and the like are used to distinguish identical or similar items having substantially the same role and function. It should be understood that "first," "second," and "nth" do not have a logical or temporal dependency, nor do they limit the quantity or execution order. It should also be understood that although the following description uses the terms "first," "second," and the like to describe various elements, these elements should not be limited by these terms.

[0065] These terms are simply used to distinguish one element from another. For example, a first element can be referred to as a second element, and similarly, a second element can be referred to as a first element, without departing from the scope of various examples. Both the first element and the second element can be elements, and in some cases, can be separate and different elements.

[0066] Here, at least one refers to one or more than one. For example, at least one element can be one element, two elements, three elements, or any other integer greater than or equal to one. And multiple refers to two or more than two. For example, multiple elements can be two elements, three elements, or any other integer greater than or equal to two.

[0067] The term "and / or" used in this document indicates that three possible relationships exist. For example, "A and / or B" can mean: A exists alone, A and B exist simultaneously, or B exists alone. The character " / " generally indicates that the related objects are in an "or" relationship.

[0068] It should be noted that the information (including but not limited to user device information, user personal information, etc.), data (including but not limited to data used for analysis, stored data, displayed data, etc.) and signals involved in this application are all authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant laws, regulations and standards in the relevant regions.

[0069] The embodiment of the present application provides a target tracking solution based on adaptive threshold adjustment, which is applicable to fields such as autonomous driving or robot vision. The solution automatically adjusts the threshold parameters (including distance threshold, appearance threshold, and disappearance threshold) used in the target tracking process by dynamically analyzing the target's motion state and sensing changes in the external environment, thereby improving the robustness and accuracy of target tracking. Among them, the external environment includes road conditions (such as highways, urban roads or rural roads, etc.) and weather conditions (such as night, rainy days or foggy days, etc.), which are not limited by this application.

[0070] Specifically, the target tracking solution provided by the embodiment of the present application mainly includes the following aspects:

[0071] 1. Dynamic threshold adjustment mechanism: This solution can dynamically adjust the distance threshold, appearance threshold, and disappearance threshold used for target tracking based on changes in the target state (such as position, velocity, or acceleration) and the external environment, thereby adapting to different working environments and avoiding the use of fixed threshold parameters that limit the system's adaptability and lead to target loss or mistracking.

[0072] In other words, in different working environments, fixed threshold parameters cannot cope with the dynamic changes of target status and environment, and it is easy to cause target loss or mistracking.

[0073] 2. Environmental adaptability: The solution has good environmental adaptability and can perceive the external environment in real time. It can then dynamically adjust the threshold parameters according to the perceived external environment, thereby improving the tracking effect.

[0074] 3. Intelligent Target Classification and Tracking Strategy Optimization: This solution classifies targets based on sensor fusion perception results. It also assesses target risk and assigns priority to each target in multi-target tracking scenarios. Furthermore, it optimizes target tracking strategies. See below for details.

[0075] In summary, this solution dynamically adjusts target tracking threshold parameters based on the target state and external environment. Even if environmental complexity increases or the target state changes, it can adapt dynamically without causing a drop in tracking accuracy. In other words, this solution improves the accuracy and robustness of target tracking, significantly reducing the probability of target loss and mistracking. This provides strong data support for subsequent system decisions, improving their accuracy and ultimately ensuring the safety and reliability of autonomous driving.

[0076] This solution adaptively adjusts thresholds to maintain efficient target tracking in a variety of environments, enabling accurate tracking of both fast-moving targets on highways and slow-moving targets on complex urban roads. Furthermore, by adaptively adjusting threshold parameters based on target status and the external environment, the solution provides more stable tracking results after a brief target loss, reducing tracking failures due to occlusion, harsh environments, and other factors. Furthermore, the solution can assess the target's risk level and dynamically adjust the tracking strategy accordingly, providing safer and more reliable functional support for autonomous driving scenarios.

[0077] For example, Figure 1 This is a schematic diagram of an implementation environment of a target tracking method provided by an embodiment of the present application. Figure 1 The implementation environment includes: the vehicle 1, pedestrians 2, other vehicles 3 and non-motor vehicles 4.

[0078] The execution subject of this method is the perception module or domain controller on the vehicle. Taking the autonomous driving domain controller to perform target tracking as an example, the autonomous driving domain controller can use the target tracking algorithm to capture targets in the road environment (such as Figure 1The position, speed or acceleration of pedestrians 2, other vehicles 3 and non-motor vehicles 4) in the vehicle are obtained, and their future behavior is predicted, thereby providing data support for the path planning and obstacle avoidance decision-making of the vehicle 1.

[0079] It should be noted that Figure 1 The number of pedestrians, other vehicles or non-motor vehicles can be greater. Figure 1 This is just an illustrative example and does not constitute any limitation on the number of pedestrians, other vehicles or non-motor vehicles. Figure 2 The target tracking solution provided in the embodiments of the present application is introduced in detail.

[0080] Figure 2 This is a flow chart of a target detection method provided by an embodiment of the present application. The execution subject of this method is the perception module or domain controller on the vehicle. Taking the autonomous driving domain controller to perform target detection as an example, see Figure 2 , the method flow includes the following steps.

[0081] 201. Acquire sensor data, and perform target detection based on the acquired sensor data to obtain a target detection result.

[0082] In the target tracking task of autonomous driving, target detection results are the basic input for tracking, and their core output provides key information for subsequent trajectory association and motion prediction. In the embodiment of the present application, the autonomous driving domain controller obtains data collected by sensors (such as lidar, radar, or cameras) in real time and performs target detection on the acquired sensor data. It uses the target detection algorithm to identify potential targets from the data source and obtain the initial information of the target, namely the target detection result.

[0083] As an example, the target detection results include but are not limited to: the position and size of the target, the category of the target, the confidence level, the preliminary speed and acceleration, etc., which are not limited in this application.

[0084] As another example, target classification is based on sensor fusion perception results, and the classification types mainly include vehicles, non-motor vehicles, and pedestrians. Furthermore, target classification methods can adopt either deep learning-based classification methods or methods based on regular shape, size, and speed, which are not limited in this application.

[0085] It should be noted that in this article, the targets identified during the target detection process are referred to as potential targets.

[0086] 202. In the initial tracking phase, the state of the potential target is initialized according to the target detection result, a motion prediction model is configured for the potential target, the threshold parameters for target tracking are initialized, and the potential target is tracked and verified based on the initialized threshold parameters; if the potential target passes the tracking verification, a tracking identifier is assigned to the potential target, the timestamp of the first detection of the potential target is recorded, and the potential target is added to the tracking object pool as a target that needs to be tracked.

[0087] In an embodiment of the present application, the threshold parameters used in the target tracking process include a distance threshold, an appearance threshold, and a disappearance threshold. Among them, the distance threshold is used to determine whether the target being tracked points to the same object. In other words, the distance threshold is used to limit the maximum matching distance between the target being tracked and the existing trajectory. The appearance threshold is used to determine whether the detected new target is in a valid tracking state. That is, the appearance threshold is used to limit the number of consecutive frames after which the detected new target will be confirmed. The disappearance threshold is used to determine whether the target being tracked is lost. That is, the disappearance threshold is used to limit how many consecutive frames after which the target being tracked is marked as lost. For example, the value of the distance threshold is 1 meter, the value of the appearance threshold is 3 frames, and the value of the disappearance threshold is 5 frames, which are not limited in this application.

[0088] It should be noted that fixed threshold parameters are used in the initial tracking stage, that is, the distance threshold, appearance threshold, and disappearance threshold are fixed in the initial tracking stage rather than dynamically changing.

[0089] Among them, initial tracking is a key starting stage in the target tracking process. Taking the initial tracking process based on Kalman filtering as an example, the initial tracking stage usually involves the following aspects:

[0090] 1. State initialization

[0091] The Kalman filter is a commonly used state estimation tool for target tracking. It combines motion models to predict the target state and uses observations to update the state, making tracking more stable. During initialization, the filter is given an initial state, which is defined as the filter's state vector (including the target's initial position, velocity, or acceleration, depending on the motion model). Furthermore, the covariance matrix is ​​typically initialized to reflect the uncertainty of the state estimate.

[0092] 2. Configure the motion prediction model

[0093] This step is used to select a suitable motion model for the target, such as a constant speed model or an acceleration model, etc., which is not limited in this application.

[0094] 3. Threshold parameter initialization

[0095] Among them, the initially set distance threshold, appearance threshold and disappearance threshold are generally empirical values.

[0096] 4. Data association and trajectory confirmation

[0097] This step involves tracking and verifying potential targets based on initialized threshold parameters. Data association is the process of matching the target observed by the sensor in real time with existing trajectories. Trajectory verification, on the other hand, verifies the authenticity of the target's trajectory using multiple frames of observation data.

[0098] It should be noted that for any target, passing tracking verification means that the target matches the established tracking track and that multiple consecutive frames have matched to this tracking track. For example, in the initial frame, an attempt is made to match the newly detected target with the existing track (for example, based on a distance threshold or IOU calculation). The newly detected target enters the active tracking state only after multiple consecutive frames of matching are successful; otherwise, it is ignored.

[0099] 5. Trajectory management

[0100] If a potential target passes the tracking verification, a tracking identifier (ID) will be assigned to this target to facilitate the distinction between different targets. The target's movement path will be recorded throughout the entire tracking process. In addition, the timestamp when the target is first detected will be recorded to mark the time point when the target enters the tracking process. This is used for timing management (such as determining the duration of the target's appearance). It can also be combined with subsequent timestamps to analyze the temporal characteristics of the target's movement (such as speed, duration of stay, etc.). In addition, this target will be added to the tracking object pool as a target that needs to be tracked for subsequent management and updates. The tracking object pool is a data container. When each frame is processed subsequently, the target is taken out of the tracking object pool, the target state is predicted using the Kalman filter, and then matched with the new target detection result, the state is updated, and the continuity of tracking is maintained. At the same time, invalid tracking trajectories (such as targets that have not been matched for a long time) can be cleaned up.

[0101] 203. In the stable tracking phase, adjust the threshold parameter according to the current motion parameter of the target being tracked or at least one item in the external environment; wherein the target being tracked is a target among potential targets that has passed tracking verification.

[0102] This step is used to adaptively adjust the threshold parameters used in the target tracking process. Adaptation refers to environmental adaptation and target state adaptation. For example, the target's motion trend is estimated based on its speed and acceleration. If the target moves faster, the distance threshold will be reduced; otherwise, the distance threshold will be increased. For another example, environmental sensors or map data are used to identify the current road type (such as highways, urban roads, or rural roads). In high-speed scenarios, the appearance threshold and disappearance threshold are reduced to quickly identify and track targets. In complex urban environments, the appearance threshold and disappearance threshold are increased to improve tracking stability. For another example, in complex weather conditions (such as night, rainy days, foggy days, etc.), the threshold is adjusted to adapt to the situation of reduced perception ability, thereby ensuring the robustness of tracking.

[0103] In some embodiments, the threshold parameter is adjusted based on at least one of the motion state data of the target being tracked or the external environment, including but not limited to the following methods:

[0104] Method 1: Adjust the initially set distance threshold based on the current speed of the target being tracked and the maximum speed limit in the target's current environment.

[0105] The adjustment formula of the distance threshold is shown in the following formula (1):

[0106]

[0107] Among them, d thr is the distance threshold after adaptive adjustment, d0 is the distance threshold set initially, and v target is the current speed of the target, v max The maximum speed in the current environment (such as the speed limit on a highway).

[0108] Method 2: According to the complexity coefficient of the current environment and the set maximum environment complexity coefficient, the initially set appearance threshold and disappearance threshold are adjusted.

[0109] The adjustment formulas for the appearance threshold and disappearance threshold are shown in the following formula (2):

[0110]

[0111] Among them, a thr is the appearance or disappearance threshold after adaptive adjustment, a0 is the initial appearance threshold or disappearance threshold, c complexity is the complexity coefficient of the target’s current environment, c max is the maximum complexity coefficient of the environment (such as the complexity coefficient corresponding to the urban environment). max The value of is 1.

[0112] In other embodiments, the present application determines the complexity coefficient of the target's current environment in the following manner:

[0113] According to the preset weights, the indicators included in the road conditions and the indicators included in the weather conditions are linearly weighted in complexity to obtain the complexity coefficient of the target's current environment.

[0114] For example, the complexity corresponding to various indicators included in road conditions and various indicators included in weather conditions is shown in Table 1 below.

[0115] Table 1

[0116] factor index Value Examples Lighting conditions Day, night, backlight, tunnel Good (0.1), Poor (0.9) visibility Rain, fog, snow, strong reflections Clear (0.1), Blurry (0.8) Scene structure complexity Urban roads, rural roads, highways High (1.0), Low (0.3) Number of dynamic jamming targets The number of moving targets at the same time Less (<5, 0.2), More (>20, 0.9) Road geometry Multiple forks, obstructions, and bends Simple (0.2), Complex (0.8)

[0117] Taking the complexity corresponding to the lighting conditions of the target's current environment as 0.7 and the weight as 0.2, the complexity corresponding to the visibility as 0.8 and the weight as 0.3, the complexity corresponding to the complexity of the light scene structure as 0.9 and the weight as 0.3, and the complexity corresponding to the number of dynamic interference targets as 0.6 and the weight as 0.2 as an example, the complexity coefficient of the target's current environment = 0.7*0.2+0.8*0.3+0.9*0.3+0.6*0.2=0.76.

[0118] In the stable tracking phase, the disappearance determination and reappearance detection of the tracked target are performed according to the adjusted threshold parameters, and the state of the tracked target is updated if the tracked target does not disappear or reappear.

[0119] After adjusting the threshold parameters, the solution tracks the target in real time based on the adaptively adjusted threshold. If the target is not detected within a certain period of time, the system enters the disappearance determination process; if the target reappears, the system enters the redetection process. Furthermore, if the target is not detected for multiple consecutive times and exceeds the disappearance threshold, it is marked as lost. If the target reappears and meets the adjusted appearance threshold, it is re-marked as a tracked target.

[0120] In some embodiments, the present application also assesses the target's risk level, priority, and tracking strategy. The risk level assessment of a target includes, but is not limited to, the following methods:

[0121] The target's corresponding risk level is determined based on the positional relationship between the target and the vehicle's lane (e.g., 1 if the target is in the lane and 0 if it is not), the distance between the target and the vehicle, and the speed relative to the vehicle. For example, the risk level of the target is determined by linearly weighting these three factors according to preset weights. For example, the corresponding relationship between risk level, priority, and tracking strategy is shown in Table 2 below.

[0122] Table 2

[0123] Risk Level Priority Tracking Strategy Greater than 0.7 Priority 1 High-frequency updates + continuous tracking 0.4 to 0.7 Priority 2 Normal frequency + moderate updates Less than 0.4 Priority 3 Delayed updates or on-demand processing

[0124] After determining the risk level corresponding to the target, the priority and tracking strategy corresponding to the target can be automatically determined based on the above Table 2. Accordingly, the status of the tracked target is updated, including: updating the status of the tracked target according to the preset tracking strategy.

[0125] The preset tracking strategy refers to a tracking strategy that matches the risk level of the target. The preset tracking strategy includes at least the status update frequency. Furthermore, the preset tracking strategy may also include whether to retain the track, which is not limited in this application. Furthermore, the status update frequency may be every frame, every two frames, or every three frames, etc., which is also not limited in this application.

[0126] In other embodiments, if there are multiple targets being tracked, status updates for the targets may include: determining a tracking priority for each target based on its corresponding risk level; and sequentially updating the status of each target according to the determined priority. Specifically, when tracking multiple targets, status updates are performed in order of priority, ensuring that high-priority targets are updated first, while low-priority targets are updated later or processed on demand.

[0127] In other embodiments, updating the state of the tracked target means: predicting the state of the tracked target at the current moment (such as the current frame) through a motion prediction model based on the state of the tracked target at the previous moment (such as the previous frame) to obtain a predicted state; then, correcting the predicted state based on real-time sensor data related to the tracked target to obtain the final state of the tracked target at the current moment (such as including position, velocity or acceleration, etc.).

[0128] In summary, the target tracking solution provided by the embodiments of the present application can dynamically adjust the threshold parameters used for target tracking based on the target state and external environment. Even if the environmental complexity increases or the target state changes, it can adapt dynamically without the problem of reduced tracking accuracy. In other words, this solution improves the accuracy and robustness of target tracking, significantly reducing the probability of target loss and mistracking. This provides strong data support for subsequent system decisions, improves the accuracy of system decisions, and thus ensures the safety and reliability of autonomous driving. Specifically, by adaptively adjusting the threshold, this solution can maintain efficient target tracking in various environments, whether it is a fast-moving target on a highway or a slow-moving target on a complex urban road. In addition, this solution adaptively adjusts the threshold parameters based on the target state and external environment, providing more stable tracking results after a temporary target loss, reducing tracking failures caused by occlusion, harsh environment, etc. In addition, this solution can also assess the risk level of the target and dynamically adjust the tracking strategy accordingly, providing safer and more reliable functional support for autonomous driving scenarios.

[0129] The following example illustrates the application of the adaptive threshold adjustment mechanism in a practical scenario.

[0130] Example 1: Application of the adaptive threshold adjustment mechanism in highway scenarios

[0131] In highway scenarios, targets (such as vehicles) move at high speeds. The sensor detects a target moving at 120 km / h. The system dynamically adjusts the distance threshold based on this speed, keeping it relatively small. This ensures faster vehicle identification and tracking at high speeds.

[0132] Example 2: Application of adaptive threshold adjustment mechanism in complex urban environments

[0133] In urban environments, targets (such as pedestrians and cyclists) move slowly, but the environment is complex and often obstructed. In such scenarios, the system analyzes the complexity of the environment and sets the appearance and disappearance thresholds to larger values ​​to ensure that the target is not lost due to brief occlusions, thereby achieving stable tracking.

[0134] Figure 3 This is a schematic diagram of the structure of a target tracking device provided by an embodiment of the present application. Figure 3 , the device comprises:

[0135] The target detection module 301 is configured to obtain sensor data, perform target detection based on the sensor data, and obtain a target detection result;

[0136] The first tracking module 302 is configured to initialize the state of a potential target according to the target detection result, initialize the threshold parameters for target tracking, and track and verify the potential target based on the initialized threshold parameters during the initial tracking phase; wherein the potential target is a target identified during the target detection process;

[0137] The threshold adjustment module 303 is configured to adjust the threshold parameter according to the current motion parameter of the target being tracked or at least one of the external environment during the stable tracking phase; wherein the target being tracked is a target among the potential targets that has passed tracking verification; for any target, the target passing tracking verification means that the target matches the established tracking track and matches the tracking track for multiple consecutive frames;

[0138] The second tracking module 304 is configured to perform disappearance determination and reappearance detection on the tracked target according to the adjusted threshold parameter, and update the status of the tracked target if the tracked target does not disappear or reappear.

[0139] The target tracking solution provided by the embodiment of the present application can dynamically adjust the threshold parameters used for target tracking according to the target state and the external environment. Even if the complexity of the environment increases or the target state changes, it can adapt dynamically and there will be no problem of decreased tracking accuracy. In other words, the solution improves the accuracy and robustness of target tracking and greatly reduces the probability of target loss and mistracking. This provides strong data support for subsequent decision-making of the system, improves the accuracy of system decision-making, and thus ensures the safety and reliability of autonomous driving. Among them, the solution can maintain efficient target tracking in various environments by adaptively adjusting the threshold, whether it is a fast-moving target on a highway or a slow-moving target on a complex urban road, accurate tracking can be achieved. In addition, the solution adaptively adjusts the threshold parameters in combination with the target state and the external environment, which can provide more stable tracking results after the target is temporarily lost, reducing tracking failures caused by occlusion, harsh environment, etc.

[0140] In some embodiments, the threshold parameters include a distance threshold, an appearance threshold, and a disappearance threshold; wherein the distance threshold is used to determine whether the target being tracked is pointing to the same object; the appearance threshold is used to determine whether the detected new target is in a valid tracking state; and the disappearance threshold is used to determine whether the target being tracked is lost.

[0141] The threshold adjustment module is configured to:

[0142] Adjusting an initially set distance threshold according to a current speed of the target being tracked and a maximum speed limit in an environment in which the target being tracked is currently located;

[0143] Adjust the initially set appearance threshold and disappearance threshold according to the complexity coefficient of the current environment and the set maximum environment complexity coefficient.

[0144] In other embodiments, the external environment includes road conditions and weather conditions;

[0145] The threshold adjustment module is configured to perform complexity linear weighting on the indicators included in the road conditions and the indicators included in the weather conditions according to preset weights to obtain a complexity coefficient of the current environment of the target being tracked.

[0146] In some other embodiments, the second tracking module is configured to:

[0147] According to a preset tracking strategy, the status of the target being tracked is updated; wherein the preset tracking strategy includes a status update frequency.

[0148] In some other embodiments, the second tracking module is further configured to:

[0149] Determining a risk level corresponding to the tracked target based on a positional relationship between the tracked target and the lane in which the vehicle is located, a distance between the tracked target and the vehicle, and a speed relative to the vehicle;

[0150] A tracking strategy corresponding to the risk level is determined as the preset tracking strategy.

[0151] In some other embodiments, the second tracking module is configured to:

[0152] If there are multiple targets being tracked, determining the tracking priority of each target being tracked according to the risk level corresponding to each target being tracked;

[0153] Update the status of each tracked target in sequence according to the determined priority.

[0154] In some other embodiments, the first tracking module is further configured to:

[0155] In the initial tracking stage, a motion prediction model is configured for the potential target; if the potential target passes the tracking verification, a tracking identifier is assigned to the potential target, the timestamp of the first detection of the potential target is recorded, and the potential target is added to the tracking object pool as a target that needs to be tracked.

[0156] In some other embodiments, the second tracking module is configured to:

[0157] Predicting the state of the target being tracked at a current moment using a motion prediction model according to the state of the target being tracked at a previous moment to obtain a predicted state;

[0158] The predicted state is corrected according to real-time sensor data related to the tracked target to obtain the final state of the tracked target at the current moment.

[0159] All of the above optional technical solutions can be combined in any way to form optional embodiments of the present application, and will not be described in detail here.

[0160] It should be noted that the target tracking device provided in the above embodiment only uses the division of the above functional modules as an example to illustrate target tracking. In actual applications, the above functions can be assigned to different functional modules as needed, that is, the internal structure of the device can be divided into different functional modules to complete all or part of the functions described above. In addition, the target tracking device provided in the above embodiment and the target tracking method embodiment are of the same concept. The specific implementation process is detailed in the method embodiment and will not be repeated here.

[0161] Figure 4 This is a schematic diagram of the structure of a domain controller provided according to an embodiment of the present application.

[0162] Typically, the domain controller 400 includes a main control module 401, a CAN interface 402, a hard-line input interface 403, and a hard-line output interface 404. The main control module 401 is connected to the CAN interface 402, the hard-line input interface 403, and the hard-line output interface 404, respectively.

[0163] The main control module 401 typically includes a processor and memory. The processor may include one or more processing cores, such as a quad-core processor or an octa-core processor. The processor may be implemented in hardware using at least one of the following: a DSP (Digital Signal Processing), an FPGA (Field-Programmable Gate Array), or a PLA (Programmable Logic Array). The processor may also include a main processor and a coprocessor. The main processor is used to process data in the awake state, also known as a CPU (Central Processing Unit); the coprocessor is a low-power processor used to process data in the standby state. In some embodiments, the processor may integrate a GPU (Graphics Processing Unit), which is responsible for rendering and drawing content required to be displayed on the vehicle display screen. In some embodiments, the processor may also include an AI (Artificial Intelligence) processor, which is used to handle computational operations related to machine learning. The memory may include one or more computer-readable storage media, which may be non-transitory. The memory may also include high-speed random access memory and non-volatile memory, such as one or more disk storage devices or flash memory storage devices. In some embodiments, the non-transitory computer-readable storage medium in the memory is used to store at least one computer program, which is used to be executed by the processor to implement the target detection method provided by the method embodiment of the present application.

[0164] In some embodiments, the main control module 401 can communicate with the vehicle's power system module, motor controller and diagnostic equipment through the CAN interface 402, and generate control instructions based on the hard-wired control signal received by the hard-wired input interface 403 to send control instructions to the vehicle's electronic control components through the hard-wired output interface 404.

[0165] Those skilled in the art will understand that Figure 4 The structure shown in the figure does not constitute a limitation to the domain controller 400, and the domain controller 400 may include more or fewer components than shown in the figure, or combine certain components, or adopt a different component arrangement.

[0166] In an exemplary embodiment, a computer-readable storage medium is also provided, such as a memory including program code, which can be executed by a processor in a domain controller to perform the target tracking method. For example, the computer-readable storage medium can be a read-only memory (ROM), a random access memory (RAM), a compact disc read-only memory (CD-ROM), a magnetic tape, a floppy disk, an optical data storage device, or the like.

[0167] In an exemplary embodiment, a computer program product or computer program is also provided. The computer program product or computer program includes computer program code, which is stored in a computer-readable storage medium. A processor of a domain controller reads the computer program code from the computer-readable storage medium, and the processor executes the computer program code, so that the domain controller performs the above-mentioned target tracking method.

[0168] Those skilled in the art will understand that all or part of the steps to implement the above embodiments may be accomplished by hardware, or by a program to instruct the relevant hardware, and the program may be stored in a computer-readable storage medium, which may be a read-only memory, a disk, or an optical disk, etc.

[0169] The above description is merely an optional embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present application shall be included in the scope of protection of the present application.

Claims

1. A target tracking method, characterized in that: The method comprises: Acquiring sensor data, performing target detection based on the sensor data, and obtaining a target detection result; In the initial tracking phase, the state of the potential target is initialized according to the target detection result, the threshold parameters for target tracking are initialized, and the potential target is tracked and verified based on the initialized threshold parameters; wherein the potential target is the target identified during the target detection process; During the stable tracking phase, the threshold parameter is adjusted based on at least one of the current motion parameters of the target being tracked or the external environment; wherein the target being tracked is a target among the potential targets that has passed tracking verification; for any target, the target passing tracking verification means that the target matches the established tracking track and matches the tracking track for multiple consecutive frames; According to the adjusted threshold parameters, the disappearance determination and reappearance detection of the tracked target are performed, and when the tracked target does not disappear or reappear, the state of the tracked target is updated.

2. The method according to claim 1, characterized in that The threshold parameters include a distance threshold, an appearance threshold, and a disappearance threshold; wherein the distance threshold is used to determine whether the target being tracked is pointing to the same object; the appearance threshold is used to determine whether the detected new target is in a valid tracking state; and the disappearance threshold is used to determine whether the target being tracked is lost. The adjusting the threshold parameter according to at least one of the motion state data of the target being tracked or the external environment includes: Adjusting an initially set distance threshold according to a current speed of the target being tracked and a maximum speed limit in an environment in which the target being tracked is currently located; Adjust the initially set appearance threshold and disappearance threshold according to the complexity coefficient of the current environment and the set maximum environment complexity coefficient.

3. The method according to claim 2, characterized in that The external environment includes road conditions and weather conditions; the method further includes: According to preset weights, complexity linear weighting is performed on the indicators included in the road conditions and the indicators included in the weather conditions to obtain a complexity coefficient of the current environment of the target being tracked.

4. The method according to claim 1, wherein The updating of the status of the tracked target includes: According to a preset tracking strategy, the status of the target being tracked is updated; wherein the preset tracking strategy includes a status update frequency.

5. The method according to claim 4, characterized in that The method further comprises: Determining a risk level corresponding to the tracked target based on a positional relationship between the tracked target and the lane in which the vehicle is located, a distance between the tracked target and the vehicle, and a speed relative to the vehicle; The tracking strategy corresponding to the risk level is determined as the preset tracking strategy.

6. The method according to claim 1, wherein If there are multiple targets being tracked, updating the status of the targets being tracked includes: Determine the tracking priority of each target according to the risk level of each target; Update the status of each tracked target in sequence according to the determined priority.

7. The method according to any one of claims 1 to 6, characterized in that During the initial tracking phase, the method further includes: configuring a motion prediction model for the potential target; If the potential target passes the tracking verification, a tracking identifier is assigned to the potential target, a timestamp of the first detection of the potential target is recorded, and the potential target is added to the tracking object pool as a target that needs to be tracked.

8. The method according to claim 7, characterized in that The updating of the status of the tracked target includes: According to the state of the target being tracked at the previous moment, predicting the state of the target being tracked at the current moment by using a motion prediction model to obtain a predicted state; The predicted state is corrected according to real-time sensor data related to the tracked target to obtain the final state of the tracked target at the current moment.

9. A target tracking device, characterized in that: The device comprises: a target detection module configured to acquire sensor data, perform target detection based on the sensor data, and obtain a target detection result; a first tracking module configured to, in an initial tracking phase, initialize a state of a potential target according to the target detection result, initialize a threshold parameter for target tracking, and track and verify the potential target based on the initialized threshold parameter; wherein the potential target is a target identified during the target detection process; a threshold adjustment module configured to adjust the threshold parameter during a stable tracking phase based on at least one of a current motion parameter of a target being tracked or an external environment; wherein the target being tracked is a target among the potential targets that has passed tracking verification; for any target, passing tracking verification means that the target matches a created tracking track and matches the tracking track for multiple consecutive frames; The second tracking module is configured to perform disappearance determination and reappearance detection on the tracked target according to the adjusted threshold parameter, and update the status of the tracked target if the tracked target does not disappear or reappear.

10. A domain controller, characterized in that: The device includes a processor and a memory, wherein at least one program code is stored in the memory, and the at least one program code is loaded and executed by the processor to implement the target tracking method according to any one of claims 1 to 8.

Images