Target detection method for vehicle-mounted millimeter wave radar and driving assistance system

Abstract

The embodiment of the invention provides a target detection method for a vehicle-mounted millimeter wave radar and a driving assistance system. The target detection method for the vehicle-mounted millimeter-wave radar comprises the following steps: processing radar measurement data of a plurality of reflection points detected by the vehicle-mounted millimeter-wave radar to obtain a main reflection point and at least one associated reflection point; the confidence degree of the at least one associated reflection point is determined based on a preset confidence degree evaluation logic, and the confidence degree of the at least one associated reflection point represents the matching confidence degree between the position of the reflection point and the real target position; determining the contribution weight of the radar measurement data of the main reflection point in the determined target state based on the confidence of the at least one associated reflection point; and determining a target state based on the determined contribution weight.

Description

Technical Field

[0001] The present invention generally relates to the technical field of vehicle-mounted millimeter-wave radar sensors. Specifically, the present invention relates to a target detection method for a vehicle-mounted millimeter-wave radar and a driving assistance system. Background Art

[0002] Millimeter-wave radar is widely used for target detection in vehicle ACC systems due to its all-weather capability and ranging accuracy. However, in complex scenarios (such as curves, near guardrails, or in tunnels), radar signals may generate false detection points due to multipath reflection. For example, when the ego vehicle (Ego Vehicle) travels along a guardrail, the radar wave may reflect from the guardrail to the ACC target (e.g., the target vehicle) and then return to the radar sensor, forming a mirror reflection point that deviates from the actual ACC target position. This phenomenon significantly reduces the ACC system's estimation accuracy of the ACC target position (especially the lateral position of the ACC target), leading to target loss (TOL) or mistracking issues. This not only affects driving comfort, but also may cause safety hazards due to the failure to timely correct the actual ACC target position. Summary of the Invention

[0003] In view of the above-mentioned problems in the prior art, according to an embodiment of the first aspect of the present invention, a target detection method for a vehicle-mounted millimeter-wave radar is provided, comprising: processing radar measurement data of multiple reflection points detected by the vehicle-mounted millimeter-wave radar to obtain a main reflection point and at least one associated reflection point; determining the confidence of the at least one associated reflection point based on a preset confidence assessment logic, wherein the confidence of the at least one associated reflection point represents the reliability of the match between the position of the reflection point and the true target position; determining the contribution weight of the radar measurement data of the main reflection point in determining the target state based on the confidence of the at least one associated reflection point; and determining the target state based on the determined contribution weight; wherein the preset confidence assessment logic includes one or more of the following:

[0004] - a first evaluation logic based on the scattering cross section of the at least one associated reflection point;

[0005] - a second evaluation logic based on whether there are static reflection points within a predetermined area around the at least one associated reflection point; and

[0006] - A third evaluation logic based on whether the at least one associated reflection point is shared by two adjacent moving objects.

[0007] According to an embodiment of the second aspect of the present invention, a driving assistance method is provided, which includes: a detection process, in which the target state is determined using the method described above, which includes the position and speed of the target; a determination process, in which the relative motion relationship between the vehicle and the target is determined based on the target state; and a decision process, in which a vehicle control instruction is generated based on the relative motion relationship to adjust the vehicle's driving parameters.

[0008] According to an embodiment of a third aspect of the present invention, a driving assistance system for a vehicle is provided, comprising: a millimeter-wave radar for detecting an environment surrounding the vehicle and outputting a reflection point dataset, the dataset comprising radar measurement data of a plurality of reflection points; and a processing unit configured to: process the radar measurement data to obtain a main reflection point and at least one associated reflection point; determine a confidence level of the at least one associated reflection point based on a preset confidence evaluation logic, wherein the confidence level of the at least one associated reflection point represents a reliability level of a match between a position of the reflection point and a true target position; determine a contribution weight of the radar measurement data of the main reflection point in determining a target state based on the confidence level of the at least one associated reflection point; and determine the target state based on the determined contribution weight.

[0009] The preset confidence assessment logic includes one or more of the following:

[0010] - a first evaluation logic based on a scattering cross section (RCS) of the at least one associated reflection point;

[0011] - a second evaluation logic based on whether there are static reflection points within a predetermined area around the at least one associated reflection point;

[0012] - A third evaluation logic based on whether the at least one associated reflection point is shared by two adjacent moving objects.

[0013] According to an embodiment of a fourth aspect of the present invention, there is provided an electronic device comprising one or more processors for executing the method of the first or second aspect as described above.

[0014] According to an embodiment of the fifth aspect of the present invention, a computer program product is provided, which includes instructions, which, when executed by one or more processors, cause the one or more processors to perform the method described in the first or second aspect above.

[0015] According to an embodiment of the sixth aspect of the present invention, a machine-readable storage medium is provided, which stores executable instructions, which, when executed, cause one or more processors to perform the method of the first or second aspect described above.

[0016] According to various embodiments of the present invention, by intelligently identifying multipath reflection interference points, the position estimation accuracy (particularly lateral position accuracy) of vehicle-mounted millimeter-wave radar in complex scenarios is improved, thereby optimizing the target tracking performance of the ACC system. This enables the vehicle to more stably track the ACC target ahead, effectively reducing sudden acceleration or deceleration caused by misjudgment. This not only improves driving safety but also enhances driving comfort, providing users with a smoother and more reliable driver assistance experience. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 is a schematic block diagram of a vehicle driving assistance system according to an embodiment of the present invention.

[0018] Figure 2 4 is a flow chart of a target detection method for a vehicle-mounted millimeter-wave radar according to one embodiment of the present invention.

[0019] Figures 3A-3C Is used to illustrate Figure 2 Schematic diagram of the target detection method in . DETAILED DESCRIPTION

[0020] In a vehicle's Advanced Driver Assistance System (ADAS), millimeter-wave radar serves as a sensor for environmental perception, and its detection accuracy directly impacts the performance of features like Adaptive Cruise Control (ACC). However, in scenarios like highways, urban elevated roads, and tunnels, multipath reflections of radar signals can cause fluctuations in target detection results. This technical pain point not only causes ACC to experience target jumps and misjudgments during vehicle following, but also impacts the driving experience and system reliability, becoming a factor hindering the performance improvement of ADAS products.

[0021] To address this issue, this paper innovatively proposes an intelligent identification method for radar reflection interference points based on multi-dimensional feature analysis. By analyzing the physical characteristics of multipath reflection and its impact on radar detection, an evaluation logic for interference reflection points is proposed. Furthermore, by adjusting the weight of interference reflection points in target status updates, the negative impact of interference reflection points on ACC target tracking is reduced. This significantly improves the detection stability of millimeter-wave radar in complex environments, providing more reliable environmental perception capabilities for ADAS systems.

[0022] The specific embodiments of the present invention are described below with reference to the accompanying drawings.

[0023] One aspect of the present invention relates to a vehicle driving assistance system. Figure 1A vehicle driving assistance system 100 according to an embodiment of the present invention is shown, and is sometimes referred to as system 100. System 100 is provided on a vehicle and is an onboard system. System 100 includes an onboard millimeter-wave radar sensor 10 and a processing unit 20.

[0024] The onboard millimeter-wave radar sensor 10 (hereinafter referred to as the onboard radar or radar) senses the vehicle's surroundings and outputs a reflection point dataset. Specifically, during each scanning cycle, the radar 10 processes the echo signal to obtain multiple radar reflection points (hereinafter referred to as reflection points). Each reflection point corresponds to a target scatterer (which may be a real ACC target or a false ACC target), and characterizes the spatial position and dynamic properties of the target scatterer.

[0025] The processing unit 20 is in communication with the radar 10 to receive and process the reflection point dataset outputted by the radar 10. The processing unit 20 is configured to execute the ACC target detection method according to an embodiment of the present invention. Specifically, the processing unit 20 employs a preset confidence assessment logic to determine the confidence level of each reflection point and reduces the contribution of reflection points with confidence levels below a corresponding confidence threshold to the ACC target state calculation. The implementation of the ACC target detection method is described in detail in the methods section below.

[0026] According to an embodiment of the present invention, multiple millimeter-wave radars 10 may be deployed on a vehicle. For example, millimeter-wave radars 10 may be deployed at the left and right front corners of the vehicle. When multiple millimeter-wave radars 10 are deployed on a vehicle, the processing unit 20 fuses the reflection point datasets from the multiple millimeter-wave radar sensors and executes the ACC target detection method according to an embodiment of the present invention based on the fused reflection point dataset.

[0027] The processing unit 20 may be implemented using software, hardware (such as a customized integrated circuit), or a combination of software and hardware.

[0028] In one embodiment, the processing unit 20 can be deployed in one of the following ways: 1) set in the processing chip of the millimeter wave radar sensor 10; 2) integrated in the vehicle's driving assistance system controller (ADAS ECU); 3) set in an on-board chip (such as an autonomous driving dedicated SoC); 4) set in a vehicle controller (VCU); 5) set in a domain controller (such as an autonomous driving domain controller).

[0029] In one embodiment, the processing unit 20 can be divided into multiple functional modules based on their functions, and these functional modules can be deployed in multiple ECUs of the vehicle using a distributed architecture design. This distributed deployment approach can improve functional scalability and fully utilize the electronic and electrical architecture resources of the vehicle.

[0030] Another aspect of the present invention relates to a target detection method for a vehicle-mounted millimeter-wave radar. Figure 2 FIG2 is a flow chart of a target detection method 200 for a vehicle-mounted millimeter-wave radar according to an embodiment of the present invention. The method 200 may be executed by the processing unit 20 described above.

[0031] See also Figure 2 In block 210, the processing unit 20 processes the radar measurement data of the multiple emission points detected by the radar 10 to obtain a primary reflection point and one or more associated reflection points (i.e., associated reflection points). These reflection points (including the primary reflection point and associated reflection points) can be acquired by the radar 10 within a scanning cycle based on the received echo signals. Various methods can be used to process the radar measurement data to determine the primary reflection point from the multiple reflection points. For example, the signal strength of the primary reflection point can be significantly higher than the signal strengths of the other reflection points. It is worth noting that the present invention is not limited to the specific implementation method for determining the primary reflection point from the multiple reflection points.

[0032] At block 220, the processing unit 20 determines the confidence level of each associated reflection point based on a preset confidence evaluation logic (hereinafter, the confidence evaluation logic is referred to as the evaluation logic). The confidence level of each associated reflection point represents the confidence level of the match between the position of the reflection point and the true ACC target position. In other words, the confidence level of each associated reflection point quantitatively represents the confidence level of the match (consistency) between the position of the reflection point and the true ACC target position. For example, the higher the confidence level of a reflection point, the higher the match between the position of the reflection point and the true ACC target position; while reflection points with low confidence levels may be due to interference from virtual targets caused by multipath effects.

[0033] The preset confidence evaluation logic includes one or more of the following 1) to 3).

[0034] 1) The first evaluation logic related to the radar cross-section (RCS) of the reflection point. The radar cross-section (RCS) is a physical quantity that measures the target's ability to scatter radar waves. It represents the effective reflection area of ​​the target under radar detection, and its unit is square meters (m2). 2 ).

[0035] 2) Second evaluation logic related to whether there are static reflection points within a predetermined area around the reflection point. A static reflection point is a reflection point corresponding to a radar-detected reflection signal from a stationary target in the vehicle's surroundings. The target scatterer corresponding to a static reflection point has a velocity of zero or nearly zero. These target scatterers are typically fixed objects in the vehicle's surroundings, such as lane edge guardrails, traffic signs, or buildings.

[0036] 3) A third evaluation logic related to whether the reflection point is a shared reflection point (Shared Reflection Point, SRP) shared by two adjacent moving objects. In an embodiment of the present invention, a "shared reflection point" refers to: a position point detected by the radar 10, whose echo signal is simultaneously associated with two adjacent moving objects. Such a shared reflection point may be caused by the superposition effect of multi-target reflection signals. For example, 1) Ghost Point: a non-real target reflection point formed by signal phase interference after the radar beam is reflected by multiple adjacent targets (such as densely packed vehicles or pedestrians); 2) Mirror reflection superposition: when a smooth surface (such as the exterior wall of a building) exists, the superposition interference of the target reflection signal and the mirror reflection signal; 3) Multipath propagation effect: in a complex urban environment, the radar wave is mistakenly judged as a single target reflection after multiple reflections.

[0037] For clarity, Figures 3A-3C The principle of the above multiple evaluation logics is shown in Figures 3A-3C In the figure, the vehicle (marked as EGO) is equipped with the above-mentioned driving assistance system 100. The millimeter wave radar 10 is installed in the front of the vehicle, and the processing unit 20 is not shown in these figures. The vehicle is currently traveling on a road consisting of two parallel lanes (marked as R1 and R2) and is located in lane R1. The boundary lines on both sides of the lane R1 are marked as L1 and L2 respectively; the boundary lines on both sides of the adjacent lane R2 are marked as L2 (shared with lane R1) and L3 respectively. Figures 3A-3C In FIG. 1 , the arrowed ray emitted from the radar 10 represents a detection electromagnetic wave beam (detection beam) radiated outward by the radar 10 .

[0038] Next, combine Figures 3A to 3C , introduces the specific implementation of each of the above preset evaluation logics (boxes 221 to 223). It should be noted that, Figures 3A to 3C The schematic diagrams of the associated reflection points corresponding to each evaluation logic are shown separately, which is only for the sake of clarity. In actual driving scenarios, there may be Figures 3A to 3C There are two or three cases of associated reflection points.

[0039] At block 221, see Figure 3A, P0 is the main reflection point, P1 is the associated reflection point, and the processing unit 20 uses a first evaluation logic to determine the confidence level of the associated reflection point P1. In one embodiment, the processing unit 20 compares the RCS value of the associated reflection point P1 with a first RCS threshold and a second RCS threshold, respectively, where the first RCS threshold is greater than the second RCS threshold. It should be understood that the RCS value of the associated reflection point P1 is included in the radar measurement data of the reflection point, that is, the radar 10 directly outputs the RCS value of the reflection point when detecting the reflection point P1. If the comparison result shows that the RCS value of the associated reflection point P1 is greater than or equal to the first RCS threshold, the processing unit 20 determines the confidence level of the reflection point P1 to be the highest level. If the comparison result shows that the RCS value of the reflection point P1 is greater than the second RCS threshold and less than the first RCS threshold, the processing unit 20 determines the confidence level of the reflection point P1 to be a first degree lower than the highest level. When the comparison result shows that the RCS value of the reflection point P1 is less than or equal to the second RCS threshold, the processing unit 20 determines the confidence level of the associated reflection point P1 to be lowered by a second degree relative to the highest level, wherein the second degree is greater than the first degree.

[0040] In an embodiment of the present invention, the confidence level of an associated reflection point represents the confidence level of the match between the position of the reflection point and the true ACC target position. A higher confidence level indicates a greater degree of confidence. If the confidence level of a reflection point is at the highest level, the position of the reflection point matches the position of the true ACC target. As the confidence level of a reflection point decreases from the highest level, the confidence level of the match between the position of the reflection point and the true ACC target decreases. The greater the degree of decrease, the greater the deviation between the position of the reflection point and the true ACC target.

[0041] In the above description, the RCS detection thresholds (i.e., the first and second RCS thresholds) and the predetermined adjustment range of the confidence level (i.e., the first and second degrees) are involved. They are all pre-calibrated based on actual vehicle test data and / or theoretical model calculations and are scenario-adaptive.

[0042] In one embodiment, different first and second RCS thresholds are pre-calibrated for different target types. Specifically, the first and second RCS thresholds are calibrated based on the physical reflective properties of the target type. For example, a vehicle, as a metal target, has an RCS value higher than that of a guardrail. Based on this, different first and second RCS thresholds are calibrated for vehicle targets and guardrail targets, respectively. When the vehicle is currently traveling along a guardrail, the processing unit 20 uses the first and second RCS thresholds calibrated for the guardrail. When there is a queue of vehicles around the vehicle, the processing unit 20 uses the first and second RCS thresholds calibrated for the vehicle.

[0043] In one embodiment, different confidence adjustment ranges are pre-calibrated for different driving scenarios. Specifically, differentiated confidence adjustment ranges are pre-calibrated based on the dynamic characteristics of the driving scenario. An implementation of this embodiment is described below.

[0044] In one embodiment, driving scenes are divided into two categories, namely, dynamic scenes and static scenes. If the number of moving targets in the driving scene is greater than a predetermined number (the predetermined number is, for example, 1, 2, 3, or 5), the driving scene is classified as a dynamic scene. For dynamic scenes, it is necessary to prioritize the detection stability of high RCS targets to avoid unnecessary deceleration of the vehicle due to false filtering. An example of a dynamic scene is a highway scene. If the number of moving targets in the driving scene is less than or equal to a predetermined number or the density of static targets in the driving scene is greater than a dense threshold, the driving scene is classified as a static scene. For static scenes, it is necessary to prioritize suppressing low RCS interference to prevent false triggering of brakes. An example of a static scene is a parking lot scene.

[0045] In this embodiment, different first and second degrees are pre-calibrated for dynamic and static scenes. For example, the first and second degrees for dynamic scenes are denoted as D11 and D12, respectively, and the first and second degrees for static scenes are denoted as D21 and D22, respectively. The pre-calibrated confidence adjustment amplitudes satisfy the following relationship: D11 < D21 < D12 < D22. The advantages of this calibration are: in dynamic scenes, the confidence reduction (D11) for medium RCS targets is small, but the reduction (D12) for low RCS targets is large, thereby strictly filtering low RCS targets; in static scenes, the confidence reduction (D21 and D22) for all suspicious reflection points is large, especially the confidence reduction (D22) for low RCS targets.

[0046] According to this embodiment, the processing unit 20 first determines whether the current driving scene of the vehicle is a dynamic scene or a static scene, and then determines the first and second degrees of calibration for the dynamic scene or the static scene based on the determination result.

[0047] At block 222, see Figure 3B, P0 is the primary reflection point, P2 and P2' are associated reflection points, and processing unit 20 uses a second evaluation logic to determine the confidence level of associated reflection point P2. In one embodiment, processing unit 20 determines a detection area A centered on associated reflection point P2 and determines whether there are any static reflection points within detection area A. For the definition and examples of static reflection points, please refer to the above description. If it is determined that there are no static reflection points within detection area A, the confidence level of associated reflection point P2 is determined to be the highest level. If it is determined that one or more static reflection points exist within detection area A, the confidence level of associated reflection point P2 is determined to be a third level lower than the highest level.

[0048] According to an embodiment of the present invention, the third degree is pre-calibrated based on actual vehicle test results in different driving scenarios (e.g., highways, urban roads, tunnels, etc.) and / or model calculations. The processing unit 20 may also adjust the third degree based on the density or number of static reflection points in detection area A. For example, the third degree may be increased as the density or number of static reflection points in detection area A increases. That is, the degree of reduction in confidence increases as the density or number of static reflection points in detection area A increases.

[0049] Next, a method for determining the detection area A is described.

[0050] The radar 10 measures the position of the associated reflection point P2 as P2=(x d ,y d ). Then, with P2 as the center, an elliptical area around it is determined as the above-mentioned detection area A. The elliptical area A (hereinafter referred to as area A) can be expressed by the following formula (1):

[0051]

[0052] Among them, (x, y) represents the coordinates of any point in area A, (x d ,y d ) is the coordinate of point P2 (i.e., the center point of area A), r x is the lateral radius of area A, r y is the longitudinal radius of area A.

[0053] According to an embodiment of the present invention, by defining the lateral radius and longitudinal radius of area A as functions of the relative distance d between the associated reflection point P2 and the radar 10 (the distance d is, for example, the straight-line distance between the associated reflection point P2 and the radar 10), the range and size of area A can be dynamically adjusted according to the relative distance d. For example, the lateral radius r of area A is x The following formula (2) can be used to express the longitudinal radius r of area A: yIt can be expressed by the following formula (3):

[0054] r x (d) = α x +β x ·d (2)

[0055] r y (d) = α y +β y ·d (3)

[0056] Among them, r x (d) represents the lateral radius of area A; r y (d) represents the longitudinal radius of area A; αx represents the lateral static offset control amount; αy represents the longitudinal state offset control amount; βx represents the lateral offset control coefficient; βy represents the longitudinal offset control coefficient.

[0057] According to an embodiment of the present invention, the lateral static offset control value αx is predetermined based on the radar detection radial error. For example, the lateral static offset control value αx is determined to be equal to three times the radar detection radial error. The longitudinal state offset control value αy is predetermined based on the radar detection azimuth error. For example, the longitudinal state offset control value αy is determined to be equal to the product of three times the radar detection azimuth error and the distance d.

[0058] According to an embodiment of the present invention, the values ​​of the lateral offset control coefficient βx and the longitudinal offset control coefficient βy are both positively correlated with the relative distance d. For example, the values ​​of the lateral offset control coefficient βx and the longitudinal offset control coefficient βy both increase linearly with increasing relative distance d. Furthermore, the lateral offset control coefficient βx is greater than the longitudinal offset control coefficient βy. For example, the lateral offset control coefficient βx is 2-4 times the longitudinal offset control coefficient βy. This configuration of lateral and longitudinal offset control coefficients is advantageous, for example, in order to demonstrate that the lateral offset is greater than the longitudinal offset in the multipath reflection offset phenomenon.

[0059] Thus, the detection area A can be obtained by the above formulas (1)-(3). Figure 3B , assuming that there is a static reflection point P2 ′ in the detection area A, the processing unit 20 determines the confidence level associated with the reflection point P2 as being lowered by a third degree relative to the highest level.

[0060] At block 223, see Figure 3C, P0 is the main reflection point, P3 is the associated reflection point, and the processing unit 20 uses the third evaluation logic to determine the confidence of the associated reflection point P3. P3 is a shared reflection point of two adjacent moving targets "TAR" and "TAR'". In one embodiment, the processing unit 20 determines whether the two adjacent moving targets have similar longitudinal motion attributes based on the longitudinal distance difference, longitudinal velocity difference, longitudinal acceleration difference and lateral distance difference between the two adjacent moving targets. The longitudinal distance difference, longitudinal velocity difference, longitudinal acceleration difference and lateral distance difference between the two adjacent moving targets can be obtained in a Cartesian coordinate system or in other coordinate systems, and the present invention is not limited to this. If the judgment result is positive, that is, the two adjacent moving targets have similar longitudinal motion attributes, the processing unit 20 calculates the similarity between the two adjacent moving targets. If the calculated similarity is less than the similarity threshold, it means that although the two adjacent moving targets have similar longitudinal motion attributes, their similarity is low, and thus the probability that the associated reflection point P3 is an interference reflection point is low. In this case, the processing unit 20 determines the confidence level of the associated reflection point P3 to be the highest level. If the calculated similarity is greater than or equal to the similarity threshold, it means that the two adjacent moving targets have similar longitudinal motion attributes and the similarity is high, and thus the probability that the associated reflection point is an interference reflection point is high. In this case, the processing unit 20 determines the confidence level of the associated reflection point P3 to be a fourth level lower than the highest level.

[0061] According to an embodiment of the present invention, the fourth degree is pre-calibrated based on test results of actual vehicles in different driving scenarios (e.g., highways, urban roads, tunnels, etc.) and / or model calculations. Processing unit 20 may also adjust the fourth degree based on the calculated similarity. For example, the fourth degree may be increased as the similarity value increases, i.e., the degree of confidence reduction increases as the similarity value increases.

[0062] Next, a method for calculating the above similarity is introduced.

[0063] The processing unit 20 calculates the cosine value between the velocity vectors of the two adjacent moving targets in the Cartesian coordinate system and normalizes the calculated cosine value to obtain a first similarity index. The first similarity index is used to evaluate the similarity of the two adjacent moving targets in the direction of motion. The processing unit 20 calculates the Mahalanobis distance between the two adjacent moving targets and normalizes the calculated Mahalanobis distance to obtain a second similarity index. The second similarity index is used to evaluate the degree of proximity between the two adjacent moving targets. Then, the processing unit 20 performs a weighted average calculation based on the first and second similarity indices and their predetermined weights to obtain the similarity between the two adjacent moving targets.

[0064] According to an embodiment of the present invention, the weight of the first similarity index and the weight of the second similarity index are both predetermined. These two weights are used to quantify the role played by the first similarity index and the second similarity index in calculating the similarity between two adjacent moving objects, that is, to characterize their respective contributions to the similarity calculation. Because the Mahalanobis distance has superior stability and effectiveness in measuring the similarity between two moving objects and can more reliably contribute to the similarity calculation, the weight of the second similarity index is set to be greater than the weight of the first similarity index.

[0065] In one embodiment, the processing unit 20 calculates the similarity between two adjacent moving targets according to the following formula (4):

[0066] S=0.3*F1+0.7*(5-F2) / 5 (4)

[0067] Where S represents the similarity between two adjacent moving targets; F1 and F2 represent the normalized values ​​of the first and second similarity indices, respectively; and 0.3 and 0.7 correspond to the weight values ​​of the first and second similarity indices, respectively.

[0068] It is worth noting that in the above description, for the sake of clarity, Figures 3A-3C Each of the diagrams in Figure 1 shows only one type of interference reflection point. However, in actual driving scenarios, multiple types of interference reflection points may coexist. To address this, the first through third evaluation logics can be executed in parallel or serially for each associated reflection point to ensure effective evaluation and processing of each interference reflection point.

[0069] return Figure 2 At block 230 , processing unit 20 determines a contribution weight of the radar measurement data of the primary reflection point in determining the ACC target state based on the confidence levels of the associated reflection points, and determines the ACC target state based on the determined contribution weight. Here, determining the ACC target state refers to determining the current state of the ACC target. For example, determining the current state of the ACC target includes updating the ACC target state or maintaining the ACC target state unchanged.

[0070] In one embodiment, the processing unit 20 calculates the position variance of the main reflection point based on the confidence level of each associated reflection point, and determines the contribution weight of the radar measurement data of the main reflection point in the ACC target state determination based on the position variance. When the contribution weight is greater than a weight threshold, the ACC target state (e.g., position, velocity, etc.) is generated based on the radar measurement data of the main reflection point. When the contribution weight is less than or equal to the weight threshold, the ACC target state is generated using the state prediction result based on the ACC target state of the previous radar cycle. For example, the processing unit 20 derives the ACC target state based on the ACC target state of the previous radar cycle and using a kinematic model.

[0071] The following describes an example of calculating the position variance of a main reflection point based on the confidence levels of each associated reflection point. Table 1 below lists the measured distances and confidence levels of associated reflection points P1-P3. The measured distance of a reflection point refers to the straight-line distance between the reflection point and radar 10. The confidence level of a reflection point is the confidence level of the reflection point determined using the above method. Table 1 and the calculated values ​​are provided for illustrative purposes only and are not intended to limit the present invention.

[0072] Table 1

[0073] Reflection Points Measuring distance (m) Confidence (0-1) P1 50.2 0.9 P2 50.5 0.7 P3 49.8 0.8

[0074] First, the processing unit 20 calculates the average value of the measured distance of the main reflection point according to the measured distance and confidence level of each associated reflection point. For example, the processing unit calculates the weighted average value of the measured distance of the main reflection point using the following formula (5):

[0075] D_P0=∑(D_Pi×CL_i) / ∑CL_i (5)

[0076] Wherein, D_P0 represents the weighted average of the measured distances of the main reflection point; D_Pi represents the measured distances of each associated reflection point; and CL_i represents the confidence level of each associated reflection point.

[0077] Substitute the values ​​listed in Table 1 into formula (5):

[0078]

[0079] Next, the processing unit 20 calculates the position variance of the main reflection point based on the weighted average of the measured distances of the main reflection point and the measured distances and confidence levels of each associated reflection point. For example, the processing unit calculates the position variance of the main reflection point using the following formula (6).

[0080] σ 2 =∑[CL_i×(D_Pi-D_P0) 2 ] / ∑CL_i (6)

[0081] Among them, σ 2 represents the position variance of the main reflection point; D_P0 represents the weighted average of the measured distances of the main reflection point; D_Pi represents the measured distances of each associated reflection point; CL_i represents the confidence level of each associated reflection point.

[0082] Substitute the weighted average of the measured distances of the main reflection points calculated according to formula (5) and the values ​​listed in Table 1 into formula (6) to calculate the denominator of formula (6) for each associated reflection point:

[0083] Reflection point P1: 0.9×(50.2-50.15) 2 =0.9×0.0025=0.00225

[0084] Reflection point P2: 0.7×(50.5-50.15) 2 =0.7×0.1225≈0.08575

[0085] Reflection point P3: 0.8×(49.8-50.15) 2 =0.8×0.1225≈0.098

[0086] Next, the denominator value calculated for each associated reflection point is substituted into formula (6) to calculate the position variance of the main reflection point:

[0087]

[0088] The position variance of the main reflection point quantifies the degree of dispersion (deviation) between multiple associated reflection points and the main reflection point. Its value can be used to characterize the reliability of radar measurement data. A small variance indicates high consistency in the positions of the reflection points (e.g., a large Kalman filter gain) and high radar measurement reliability. In this case, radar measurement data from the main reflection point should be prioritized for updating the ACC target state. A large variance indicates a dispersed distribution of reflection points (e.g., a small Kalman filter gain), possibly due to multipath effects, leading to low measurement reliability. In this case, the reliability of the radar measurement data should be reduced, and the ACC target state should be determined more reliant on the vehicle's driver assistance system's predictions based on historical data (e.g., the Kalman filter's state prediction value). By dynamically adjusting whether to use radar measurements or model predictions, the ACC system can accurately track the target when the radar measurement data is reliable and maintain stable output when the radar measurement data is disturbed, thereby improving the robustness of the ACC system.

[0089] According to an embodiment of the present invention, a driving assistance method is also provided, which includes: a detection process, in which the method 200 described above is used to determine the target state, the target state including the position and speed of the target; a decision process, in which the relative motion relationship between the vehicle and the target is determined based on the determined target state; and a generation process, in which a vehicle control instruction is generated based on the relative motion relationship to adjust the vehicle's driving parameters.

[0090] In one embodiment, generating the vehicle control instruction includes: when the relative motion relationship satisfies a first condition, generating an acceleration instruction to control the vehicle to accelerate so that the vehicle maintains a preset safety distance from the target (for example, the following distance in the ACC system or a safety distance set based on the following distance); when the relative motion relationship satisfies a second condition, generating a deceleration instruction to control the vehicle to decelerate to avoid collision with the target; when the relative motion relationship satisfies a third condition, generating a steering instruction to control the vehicle to adjust the driving direction to avoid collision with the target in a detour.

[0091] In one embodiment, the target determined in method 200 is the vehicle's forward ACC target. The first condition is that the forward ACC target's speed is greater than the vehicle's current speed and the distance between them is greater than a predetermined safety distance. The second condition is that the forward ACC target's speed is less than the vehicle's current speed or the distance between them is less than a predetermined safety distance. The third condition is that the forward ACC target is within the vehicle's planned travel path and collision avoidance is unavoidable by deceleration.

[0092] According to the driving assistance method of an embodiment of the present invention, by analyzing the relative motion relationship between the vehicle and the target, acceleration, deceleration or steering instructions are intelligently generated to achieve dynamic safe distance maintenance and collision avoidance. An exemplary application scenario is the adaptive cruise control (ACC) and emergency obstacle avoidance function in the intelligent driving system (advanced driver assistance system), which is particularly suitable for complex road conditions such as highway cruising and following vehicles in congestion. Compared with the traditional physical fixed-distance control strategy, it has smoother speed adjustment capabilities and active obstacle avoidance path optimization capabilities, and realizes multi-modal collaborative control, that is, longitudinal speed control and lateral path planning, with higher road adaptability and driving comfort.

[0093] According to an embodiment of the present invention, an electronic device is further provided, which includes one or more processors for executing the target detection method or driving assistance method described above. The electronic device is, for example, a vehicle-mounted millimeter-wave radar sensor.

[0094] According to an embodiment of the present invention, a computer program product is further provided, which includes instructions. When the instructions are executed by one or more processors, the one or more processors execute the target detection method or driving assistance method as described above.

[0095] According to an embodiment of the present invention, a computer-readable storage medium is further provided, which includes instructions that, when executed, cause one or more processors to perform the object detection method or driving assistance method described above.

[0096] Should be understood that processor can use electronic hardware, computer software or its any combination to implement.These processors are embodied as hardware or software and will depend on specific application and the overall design constraint that is imposed on the system.As an example, the processor that provides in the present invention, any part of processor or any combination of processor can be embodied as microprocessor, microcontroller, digital signal processor (DSP), field programmable gate array (FPGA), programmable logic device (PLD), state machine, gate logic, discrete hardware circuit and be configured for carrying out other suitable processing unit of the various functions described in the present invention.The function of the processor that the present invention provides, any part of processor or any combination of processor can be embodied as the software performed by microprocessor, microcontroller, DSP or other suitable platform.

[0097] Software should be broadly considered to mean instructions, instruction sets, codes, code segments, program codes, programs, subroutines, software modules, applications, software applications, software packages, routines, subroutines, objects, running threads, processes, functions, etc. Software can reside in a computer-readable medium. A computer-readable medium can include, for example, a memory, which can be, for example, a magnetic storage device (e.g., a hard disk, a floppy disk, a magnetic stripe), an optical disk, a smart card, a flash memory device, a random access memory (RAM), a read-only memory (ROM), a programmable ROM (PROM), an erasable PROM (EPROM), an electrically erasable PROM (EEPROM), a register, or a removable disk. Although memory is shown as being separated from a processor in the various aspects provided herein, memory can also be located inside the processor (e.g., a cache or register).

[0098] Although some embodiments have been described above, these embodiments are given by way of example only and are not intended to limit the scope of the invention. The appended claims and their equivalents are intended to cover all modifications, substitutions and changes made within the scope and spirit of the invention.

Claims

1. A target detection method for a vehicle-mounted millimeter-wave radar, comprising: processing radar measurement data of a plurality of reflection points detected by the vehicle-mounted millimeter-wave radar to obtain a main reflection point and at least one associated reflection point; Determining the confidence of the at least one associated reflection point based on a preset confidence evaluation logic, wherein the confidence of the at least one associated reflection point represents the reliability of the match between the position of the reflection point and the true target position; determining a contribution weight of the radar measurement data of the main reflection point in determining the target state based on the confidence level of the at least one associated reflection point; and determining a target state based on the determined contribution weights; The preset confidence assessment logic includes one or more of the following: - a first evaluation logic based on the scattering cross section of the at least one associated reflection point; - a second evaluation logic based on whether there are static reflection points within a predetermined area around the at least one associated reflection point; and - A third evaluation logic based on whether the at least one associated reflection point is shared by two adjacent moving objects.

2. The target detection method according to claim 1, wherein: The preset confidence evaluation logic includes the first evaluation logic, and at least one of the second evaluation logic and the third evaluation logic.

3. The target detection method according to claim 1, wherein: The preset confidence evaluation logic includes at least one of the second evaluation logic and the third evaluation logic.

4. The target detection method according to claim 1, wherein: Determining the target state based on the determined contribution weights includes: When the contribution weight is greater than a weight threshold, generating a target state according to the radar measurement data of the main reflection point; and When the contribution weight is less than or equal to the weight threshold, the target state is generated using a state prediction result based on the target state in the previous radar cycle.

5. The target detection method according to claim 1, wherein: The first evaluation logic includes: comparing the RCS of the at least one associated reflection point with a first RCS threshold and a second RCS threshold, respectively, wherein the first RCS threshold is greater than the second RCS threshold; When the comparison result shows that the RCS of the at least one associated reflection point is greater than or equal to the first RCS threshold, determining the confidence level of the at least one associated reflection point as the highest level; When the comparison result shows that the RCS of the at least one associated reflection point is greater than the second RCS threshold and less than the first RCS threshold, determining the confidence level of the at least one associated reflection point to be lowered by a first degree relative to the highest level; and When the comparison result shows that the RCS of the at least one associated reflection point is less than or equal to the second RCS threshold, the confidence level of the at least one associated reflection point is determined to be lowered by a second degree relative to the highest level, wherein the second degree is greater than the first degree.

6. The target detection method according to claim 5, wherein: The values ​​of the first and second RCS thresholds depend on the physical reflection characteristics of different target types.

7. The target detection method according to claim 5, wherein: The values ​​of the first and second degrees depend on whether the current driving scene of the vehicle is a dynamic scene or a static scene. A dynamic scene refers to a scene in which the number of moving targets is greater than a predetermined number, and a static scene refers to a scene in which the number of moving targets is less than or equal to the predetermined number or the density of static targets is greater than a density threshold.

8. The target detection method according to claim 1, wherein: The second evaluation logic includes: Determining a detection area with the at least one associated reflection point as the center; Determine whether there is a static reflection point in the detection area; If it is determined that there is no static reflection point in the detection area, determining the confidence level of at least one associated reflection point to be the highest level; and If it is determined that one or more static reflection points exist within the detection area, the confidence level of at least one associated reflection point is determined to be lowered by a third degree relative to the highest level.

9. The target detection method according to claim 8, wherein: The detection area is elliptical, and determining the detection area includes: Determining a lateral radius of the detection area based on a relative distance between the at least one associated reflection point and the vehicle-mounted radar, a lateral offset control amount associated with a radial detection error of the vehicle-mounted millimeter-wave radar, and a lateral adjustment coefficient; and The longitudinal radius of the detection area is determined based on the relative distance between the at least one associated reflection point and the vehicle-mounted radar, the longitudinal offset control amount associated with the detection azimuth error of the vehicle-mounted radar, and the longitudinal adjustment coefficient.

10. The target detection method according to claim 9, wherein: The lateral adjustment coefficient and the longitudinal adjustment coefficient are both positively correlated with the relative distance, and the lateral adjustment coefficient is 2-4 times the longitudinal adjustment coefficient.

11. The target detection method according to claim 1, wherein: The third evaluation logic includes: Determining whether the two adjacent moving targets have similar longitudinal motion attributes based on a longitudinal distance difference, a longitudinal velocity difference, a longitudinal acceleration difference, and a lateral distance difference between the two adjacent moving targets; If the judgment result is positive, calculating the similarity between the two adjacent moving targets; If the calculated similarity is less than the similarity threshold, determining the confidence level of the at least one associated reflection point as the highest level; and If the calculated similarity is greater than or equal to the similarity threshold, the confidence level of the at least one associated reflection point is determined to be lowered by a fourth degree relative to the highest level.

12. The target detection method according to claim 11, wherein: Calculating the similarity between the two adjacent moving targets includes: Calculating a first similarity index based on the cosine values ​​of the velocity vectors of the two adjacent moving objects in a Cartesian coordinate system; Calculating a second similarity index based on the Mahalanobis distance between the two adjacent moving objects; and A weighted average calculation is performed according to the first and second similarity indicators and their preset weights to obtain the similarity, wherein the preset weight of the first similarity indicator is less than the preset weight of the second similarity indicator.

13. A driving assistance method, comprising: A detection process, in which the target state, including the position and velocity of the target, is determined using the method according to any one of claims 1 to 12; a determination process, in which a relative motion relationship between the vehicle and the target is determined based on the target state; as well as A decision-making process, in which a vehicle control instruction is generated according to the relative motion relationship to adjust the vehicle's driving parameters.

14. The driving assistance method according to claim 13, wherein: Generating the vehicle control instruction includes: When the relative motion relationship satisfies a first condition, generating an acceleration instruction to control the vehicle to accelerate and maintain a preset safety distance between the vehicle and the target; When the relative motion relationship satisfies a second condition, generating a deceleration instruction to control the vehicle to decelerate to avoid collision with the target; When the relative motion relationship satisfies the third condition, a steering instruction is generated to control the vehicle to adjust the driving direction to avoid collision with the target in a detour manner.

15. The driving assistance method according to claim 14, wherein: The target is the vehicle's front ACC target, and The first condition is that the speed of the front ACC target is greater than the current speed of the vehicle and the distance between the two is greater than the preset safety distance; The second condition is that the speed of the front ACC target is less than the current speed of the vehicle or the distance between the two is less than the preset safety distance; The third condition is that the front ACC target is on the planned driving path of the vehicle and a collision cannot be avoided by deceleration.

16. A driving assistance system for a vehicle, comprising: a millimeter-wave radar for detecting the environment around the vehicle and outputting a reflection point dataset comprising radar measurement data from multiple reflection points; as well as The processing unit is configured as follows: processing the radar measurement data to obtain a main reflection point and at least one associated reflection point; Determining the confidence of the at least one associated reflection point based on a preset confidence evaluation logic, wherein the confidence of the at least one associated reflection point represents the reliability of the match between the position of the reflection point and the true target position; determining a contribution weight of the radar measurement data of the main reflection point in determining the target state based on the confidence level of the at least one associated reflection point; and determining a target state based on the determined contribution weights; The preset confidence assessment logic includes one or more of the following: - a first evaluation logic based on a scattering cross section (RCS) of the at least one associated reflection point; - a second evaluation logic based on whether there are static reflection points within a predetermined area around the at least one associated reflection point; - A third evaluation logic based on whether the at least one associated reflection point is shared by two adjacent moving objects.

17. An electronic device comprising one or more processors, configured to execute the method according to any one of claims 1-12 or 13-15.

18. The electronic device according to claim 17, wherein: The electronic device is a millimeter wave radar.

19. A computer program product comprising instructions which, when executed by one or more processors, cause the one or more processors to perform the method of any one of claims 1-12 or 13-15.

20. A machine-readable storage medium storing executable instructions that, when executed, cause one or more processors to perform the method of any one of claims 1-12 or 13-15.

Images