Calibration method and device of vehicle collision detection dynamic threshold line and storage medium
By constructing a dynamic threshold line for vehicle collision detection and utilizing the two-dimensional distribution of acceleration and velocity changes and anchor point fitting, the problem of misjudgment under complex working conditions with fixed threshold parameters is solved, achieving higher detection accuracy and adaptability.
Patent Information
- Application Number
- CN202611067215.4
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2026-07-17
- Publication Date
- 2026-11-06
AI Technical Summary
In existing vehicle collision detection methods, fixed threshold parameters are difficult to adapt to the complex working conditions of different vehicle models, sensor installation locations, and different collision types, resulting in inaccurate collision judgment boundaries, especially in scenarios such as minor collisions, bumpy roads, and sudden braking, which are prone to misjudgment.
By acquiring collision sample data and non-collision sample data, a two-dimensional distribution of acceleration and velocity changes is constructed. Anchor points are selected and fitted to establish a dynamic threshold line. Combined with safety margin offset, a functional relationship adapted to different collision types is constructed.
It improves the accuracy and reliability of vehicle collision detection, reduces the false positive rate under complex conditions such as minor collisions, sudden braking, and bumpy roads, and enhances the adaptability of the dynamic threshold line to different vehicle conditions.
Smart Images
Figure CN123045657A_ABST
Abstract
Description
Technical Field
[0001] This application relates to the field of vehicle safety technology, and in particular to a method, apparatus and storage medium for calibrating a dynamic threshold line for vehicle collision detection. Background Technology
[0002] With the development of active safety technologies for vehicles, vehicle collision detection technology has been widely used in airbag deployment, accident alarms, and vehicle-to-everything (V2X) security systems. Existing collision detection methods typically rely on acceleration thresholds, velocity change thresholds, or a combination of both to determine collisions.
[0003] However, vehicle dynamic responses vary across different vehicle models, sensor installation locations, and collision types, making it difficult to adapt fixed threshold parameters to the demands of collision detection under complex conditions. Especially in scenarios such as minor collisions, bumpy roads, speed bumps, and sudden braking, vehicle operating data can easily overlap, making it difficult to accurately determine collision detection boundaries and thus affecting the reliability of collision detection results.
[0004] Furthermore, most threshold parameters in existing technologies rely on empirical settings or simple statistical methods for determination, lacking a dynamic threshold line construction method based on systematic calibration of collision and non-collision sample data. This results in weak threshold adaptability under different collision types and poor parameter repeatability and verifiability.
[0005] Therefore, existing technologies lack a method for calibrating the dynamic threshold line for vehicle collision detection to construct dynamic threshold line parameters for different collision types. Summary of the Invention
[0006] In view of this, embodiments of this application provide a method, apparatus and storage medium for calibrating a dynamic threshold line for vehicle collision detection, in order to address the problem of the lack of a method for calibrating dynamic threshold lines for different collision types.
[0007] In a first aspect, embodiments of this application provide a method for calibrating a dynamic threshold line for vehicle collision detection, the method comprising: Acquire collision sample data and non-collision sample data; Based on the distribution boundary between the collision sample data and the non-collision sample data, at least two anchor points corresponding to the collision type are selected, and the anchor points include acceleration coordinate values and velocity change coordinate values. Based on the at least two anchor points, determine the functional relationship between the velocity change threshold and acceleration, and construct a dynamic threshold line corresponding to the collision type.
[0008] Preferably, the step of selecting at least two anchor points corresponding to the collision type based on the distribution boundary between the collision sample data and the non-collision sample data includes: Construct a two-dimensional sample distribution of acceleration and velocity change based on the collision sample data and the non-collision sample data; By fitting the boundary region between collision sample data and non-collision sample data, the boundary line corresponding to the collision type is obtained; Select at least two anchor points according to the boundary line corresponding to the collision type.
[0009] Preferably, constructing a two-dimensional sample distribution of acceleration and velocity changes based on the collision sample data and the non-collision sample data includes: Determine the acceleration data and velocity change data in the corresponding direction based on the corresponding collision type; A two-dimensional sample distribution for the corresponding collision type is constructed based on the acceleration data and velocity change data in the corresponding direction.
[0010] Preferably, after selecting at least two anchor points of the corresponding collision type according to the boundary line of the corresponding collision type, the process includes: A safety margin offset is applied to the at least two anchor points to obtain at least two target anchor points for constructing the dynamic threshold line.
[0011] Preferably, the offset corresponding to the safety margin offset is a preset ratio value, which is determined according to the sensitivity requirements of collision detection and the false alarm control requirements.
[0012] Preferably, the preset ratio value ranges from 5% to 30%.
[0013] Preferably, the safety margin offset of the at least two anchor points includes: Offset the at least two anchor points in a direction away from the collision sample data region or the non-collision sample data region.
[0014] Preferably, the at least two anchor points include a first anchor point located in a first acceleration region and a second anchor point located in a second acceleration region; the acceleration in the first acceleration region is less than the acceleration in the second acceleration region. The threshold for the first velocity change corresponding to the first anchor point is greater than the threshold for the second velocity change corresponding to the second anchor point.
[0015] Preferably, the functional relationship is a linear function, and the linear function satisfies: Δv_threshold = k×acc + b Wherein, Δv_threshold represents the threshold of velocity change, acc represents acceleration, k represents slope, and b represents intercept. The slope k and the intercept b are determined by at least two anchor points of the corresponding collision type based on a two-point linear equation.
[0016] Preferably, the functional relationship is a quadratic function, and the quadratic function satisfies: Δv_threshold = a×acc 2 + b×acc + c Where Δv_threshold represents the threshold of velocity change, acc represents acceleration, and a, b, and c are coefficients determined by curve fitting based on at least three anchor points of the corresponding collision type; The at least three anchor points include anchor points distributed across different acceleration ranges.
[0017] Preferably, the functional relationship is a piecewise linear function, and the piecewise linear function satisfies: When acc < acc1, Δv_threshold = k1×acc + b1; When acc1 ≤ acc < acc2, Δv_threshold = k2×acc + b2; When acc ≥ acc2, Δv_threshold = k3×acc + b3; Wherein, Δv_threshold represents the threshold of velocity change, acc represents acceleration, acc1 and acc2 are segmentation points, which are determined based on the distribution of anchor points within the corresponding acceleration interval, and the slope k and intercept b of each segment are determined based on the anchor points within the corresponding acceleration interval.
[0018] Preferably, the functional relationship is an exponential function, and the exponential function satisfies: Δv_threshold = A×exp(B×acc) + C Where Δv_threshold represents the threshold of velocity change, acc represents acceleration, and A, B, and C are coefficients determined by curve fitting based on at least three anchor points of the corresponding collision type.
[0019] Secondly, embodiments of this application provide a calibration device for a dynamic threshold line of vehicle collision detection, the device comprising: The acquisition module is used to acquire collision sample data and non-collision sample data; The selection module is used to select at least two anchor points corresponding to the collision type based on the distribution boundary between the collision sample data and the non-collision sample data. The anchor points include acceleration coordinate values and velocity change coordinate values. The function determination module is used to determine the functional relationship between the velocity change threshold and the acceleration based on the at least two anchor points, and to construct a dynamic threshold line corresponding to the collision type.
[0020] Thirdly, embodiments of this application provide a storage medium storing computer program instructions, which, when executed by a processor, implement the method of the first aspect described above.
[0021] This application provides a method, apparatus, and storage medium for calibrating a dynamic threshold line for vehicle collision detection, comprising: acquiring collision sample data and non-collision sample data; selecting at least two anchor points corresponding to the collision type based on the distribution boundary between the collision sample data and the non-collision sample data, the anchor points including acceleration coordinate values and velocity change coordinate values; determining the functional relationship between the velocity change threshold and acceleration based on the at least two anchor points, and constructing a dynamic threshold line corresponding to the collision type. This application determines the dynamic threshold line parameters by using collision sample data and non-collision sample data, enabling the dynamic threshold lines corresponding to different collision types to be constructed based on the relationship between acceleration and velocity change, thereby improving the adaptability of the dynamic threshold line parameters to different vehicle operating conditions. Attached Figure Description
[0022] To more clearly illustrate the technical solutions of the embodiments of this application, the drawings used in the embodiments of this application will be briefly introduced below. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort, and these are all within the protection scope of this application.
[0023] Figure 1 This is a schematic diagram of the overall process of the vehicle collision detection dynamic threshold line calibration method according to the embodiments of this application; Figure 2 This is a detailed flowchart of step S2 of the vehicle collision detection dynamic threshold line calibration method in the embodiments of this application; Figure 3 This is a schematic diagram of the specific process of step S21 of the vehicle collision detection dynamic threshold line calibration method in the embodiments of this application; Figure 4 This is a schematic diagram illustrating the impact of safety margin offset on the false alarm rate and false negative rate of collision detection in the embodiments of this application; Figure 5 This is a schematic diagram of the structure of the calibration device for the dynamic threshold line of vehicle collision detection according to an embodiment of this application. Detailed Implementation
[0024] The features and exemplary embodiments of various aspects of this application will now be described in detail. To make the objectives, technical solutions, and advantages of this application clearer, the application will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are only configured to explain this application and are not configured to limit this application. For those skilled in the art, this application can be implemented without some of these specific details. The following description of the embodiments is merely to provide a better understanding of this application by illustrating examples of this application.
[0025] It should be noted that, in this document, relational terms such as "first" and "second" are used merely to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus. Without further limitations, an element defined by the phrase "comprising..." does not exclude the presence of additional identical elements in the process, method, article, or apparatus that includes said element.
[0026] It should be noted that all actions involving the acquisition of signals, information, or data in this application are carried out in compliance with the relevant data protection laws and regulations of the locality and with authorization from the owner of the relevant device.
[0027] Firstly, please see Figure 1 This application provides a method for calibrating a dynamic threshold line for vehicle collision detection, the method comprising: S1. Obtain collision sample data and non-collision sample data; Collision sample data refers to data collected during a vehicle collision. The collision sample data includes at least one of the following: acceleration data, velocity change data, direction data, and duration data during the collision.
[0028] Non-collision sample data refers to data collected by a vehicle under non-collision conditions. The non-collision sample data includes at least one of the following: emergency braking condition data, speed bump driving condition data, bumpy road condition data, and normal driving condition data.
[0029] In one specific implementation, corresponding collision sample data can be obtained by conducting multiple sets of standard collision tests, including at least one of frontal collision test, rear collision test, side collision test, and offset collision test.
[0030] S2. Based on the distribution boundary between the collision sample data and the non-collision sample data, select at least two anchor points corresponding to the collision type. The at least two anchor points include acceleration coordinate values and velocity change coordinate values. Specifically, at least two anchor points can be selected based on the distribution boundaries of collision sample data and non-collision sample data in the two-dimensional sample distribution of acceleration and velocity change.
[0031] In one implementation, such as Figure 2 As shown, S2 includes: S21. Construct a two-dimensional distribution of acceleration and velocity change based on the collision sample data and the non-collision sample data; The two-dimensional distribution of acceleration and velocity change refers to a data distribution map constructed with acceleration as one of the coordinate axes and velocity change as the other coordinate axis. It is used to represent the distribution of collision sample data and non-collision sample data in the dimensions of acceleration and velocity change.
[0032] In some implementations, all experimental data are plotted in an acceleration-Δv two-dimensional coordinate system, where the horizontal axis represents acceleration acc and the vertical axis represents the velocity change Δv. Collision sample data and non-collision sample data form different clustering regions in the two-dimensional coordinate system, and corresponding boundary trends emerge between the two types of data.
[0033] In some implementations, different collision types correspond to different combinations of data in different directions. For example, frontal collisions and rear collisions correspond to vehicle longitudinal direction data, side collisions correspond to vehicle lateral direction data, and offset collisions correspond to a combination of vehicle longitudinal and lateral direction data.
[0034] In one embodiment, such as Figure 3 As shown, constructing a two-dimensional distribution of acceleration and velocity changes based on the collision sample data and the non-collision sample data includes: S211. Determine the acceleration data and velocity change data in the corresponding direction according to the corresponding collision type; In some implementations, acceleration data and velocity change data in the corresponding direction are extracted from vehicle operation data according to the collision type. Specifically, the longitudinal direction data reflects the vehicle's motion changes in the forward and backward direction, while the lateral direction data reflects the vehicle's motion changes in the left and right direction.
[0035] For example, when the collision type is a frontal or rear-end collision, the acceleration data and velocity change data in the longitudinal direction of the vehicle are extracted; when the collision type is a side collision, the acceleration data and velocity change data in the lateral direction of the vehicle are extracted; when the collision type is an offset collision, the acceleration data and velocity change data in both the longitudinal and lateral directions of the vehicle are extracted simultaneously.
[0036] S212. Construct a two-dimensional sample distribution for the corresponding collision type based on the acceleration data and velocity change data in the corresponding direction.
[0037] In some implementations, the acceleration data in the corresponding direction is used as the horizontal axis, and the velocity change data in the corresponding direction is used as the vertical axis. The collision sample data and the non-collision sample data are mapped to obtain a two-dimensional sample distribution for the corresponding collision type.
[0038] In some implementations, different collision types correspond to independent two-dimensional sample distributions, and different two-dimensional sample distributions are used to reflect the distribution boundary between collision sample data and non-collision sample data under the corresponding collision type.
[0039] This embodiment extracts data from corresponding directions for different collision types and constructs a two-dimensional sample distribution, enabling motion change features under different collision types to participate in dynamic threshold line calibration. This improves the adaptability of the dynamic threshold line to different collision conditions, reduces interference from non-collision conditions on the collision determination boundary, and enhances the accuracy and reliability of collision recognition in subsequent vehicle collision detection processes.
[0040] S22. Fit the boundary region between collision sample data and non-collision sample data to obtain the boundary line of the corresponding collision type. For different collision types, the boundary region between collision sample data and non-collision sample data is fitted to obtain the boundary line of the corresponding collision type. In some implementations, linear fitting, piecewise linear fitting, curve fitting, or other boundary fitting methods can be used to fit the boundary region between collision sample data and non-collision sample data for different collision types, thereby obtaining the boundary line of the corresponding collision type and determining the boundary change trend between collision sample data and non-collision sample data.
[0041] In one specific implementation, such as Figure 3 As shown, fitting the boundary region between collision sample data and non-collision sample data to obtain the boundary line corresponding to the collision type includes the following sub-steps: S221. Based on the range of values of the acceleration coordinate axis in the two-dimensional sample distribution of acceleration and velocity change, the two-dimensional sample distribution is divided into multiple acceleration sub-intervals. An acceleration sub-interval refers to a series of consecutive value intervals divided according to the acceleration coordinate axis. Each acceleration sub-interval corresponds to a portion of sample data in a two-dimensional sample distribution.
[0042] In one feasible implementation, the two-dimensional sample distribution can be divided into multiple acceleration sub-intervals according to the range of acceleration coordinate axis values, using a fixed interval length, a fixed number of samples, or an adaptive interval width.
[0043] In one feasible implementation, the acceleration sub-intervals are arranged sequentially along the acceleration coordinate axis, and adjacent acceleration sub-intervals can be connected end to end or partially overlap to ensure the continuity of local boundary regions.
[0044] S222. Within each acceleration sub-interval, determine the first edge sample set in the collision sample data that is closer to the non-collision sample data, and the second edge sample set in the non-collision sample data that is closer to the collision sample data. The edge sample set refers to the sample set located near the boundary of the corresponding sample distribution, used to represent the boundary between collision sample data and non-collision sample data.
[0045] In one feasible implementation, the set of samples with the smallest velocity change in the collision sample data can be selected as the first edge sample set according to the direction of velocity change, and the set of samples with the largest velocity change in the non-collision sample data can be selected as the second edge sample set.
[0046] In other feasible implementations, the first edge sample set and the second edge sample set can also be determined based on the local density of the samples, the nearest neighbor distance, or the convex hull boundary.
[0047] S223. Determine the local boundary zone of the corresponding acceleration sub-interval based on the interval region between the first edge sample set and the second edge sample set within the same acceleration sub-interval; The local boundary zone refers to the region located between the first edge sample set and the second edge sample set, used to represent the local boundary between collision sample data and non-collision sample data.
[0048] In one feasible implementation, the corresponding local boundary zone can be determined based on the minimum distance, average distance, or preset ratio range between the first edge sample set and the second edge sample set.
[0049] S224. Determine candidate boundary points within each of the aforementioned local boundary zones; Boundary candidate points refer to candidate boundary locations located within the local boundary zone, which are used to participate in the subsequent overall boundary line fitting.
[0050] In one feasible implementation, the center position, median position, or position equidistant from the sample sets on both sides of the local boundary zone can be determined as a candidate boundary point.
[0051] S225. Based on the coordinate change amplitude between candidate boundary points in adjacent acceleration sub-intervals, outlier points are removed from the candidate boundary points to obtain multiple target boundary points; In one feasible implementation, the coordinate change between adjacent boundary candidate points can be calculated. When the coordinate change exceeds a preset change threshold, the corresponding boundary candidate point is identified as an anomaly and removed.
[0052] In one feasible implementation, outlier removal can also be accomplished using a sliding window, median filtering, or local continuity constraints.
[0053] S226. Fit multiple target boundary points to obtain the boundary line corresponding to the collision type.
[0054] In one feasible implementation, linear fitting, piecewise linear fitting, polynomial fitting, or spline curve fitting can be used to fit multiple target boundary points to obtain the boundary line corresponding to the collision type.
[0055] This implementation divides the two-dimensional sample distribution into multiple acceleration sub-intervals and extracts the local boundary positions between collision sample data and non-collision sample data. Then, by combining outlier removal and overall fitting, the boundary lines for the corresponding collision types are determined, enabling the boundary lines to more accurately reflect the true distribution boundaries between the two types of samples. Compared to directly fitting all samples, this reduces the impact of sample overlap and local outliers on the boundary line position under conditions such as minor collisions, sudden braking, speed bumps, and bumpy roads. It improves the stability and accuracy of dynamic threshold line calibration, providing a more reliable data foundation for subsequent anchor point selection and collision determination.
[0056] S23. Select at least two anchor points of the corresponding collision type according to the boundary line of the corresponding collision type.
[0057] In some implementations, at least two control points can be selected as anchor points for the corresponding collision type on the boundary line obtained from the fitting. For example, two anchor points can be selected: a first anchor point and a second anchor point. The first anchor point is used to control the collision recognition range under low acceleration conditions, and the second anchor point is used to control the collision recognition range under high acceleration conditions. The at least two anchor points are used together to determine the changing trend and position parameters of the dynamic threshold line.
[0058] This implementation constructs a two-dimensional distribution based on collision sample data and non-collision sample data, and selects at least two anchor points based on the boundary line. It can establish a corresponding dynamic threshold line by combining the actual data distribution under different collision conditions, thereby reducing the possibility of misjudgment caused by the fixed threshold method under conditions such as sudden braking, road bumps and speed bumps, and improving the detection capability in minor collision and compound collision scenarios.
[0059] In other embodiments, to improve the calibration accuracy of the dynamic threshold line in complex boundary regions between collision sample data and non-collision sample data, at least two anchor points corresponding to the collision type are selected based on the boundary line of the corresponding collision type, including: S231. Based on the range of values of the acceleration coordinate axis in the two-dimensional sample distribution of acceleration and velocity change, the two-dimensional sample distribution is divided into multiple candidate acceleration intervals.
[0060] In one feasible implementation, the acceleration coordinate axis is divided into intervals according to a preset acceleration step size or a preset acceleration range, so that collision sample data and non-collision sample data in different acceleration ranges participate in subsequent local analysis respectively.
[0061] S232. Determine the local interval width between collision sample data and non-collision sample data in each of the candidate acceleration intervals.
[0062] Local interval width refers to the distance along the direction of velocity change between the edge position of the collision sample data (closer to the non-collision sample data) and the edge position of the non-collision sample data (closer to the collision sample data) within the same candidate acceleration interval.
[0063] In one feasible implementation, the smaller the local interval width, the closer the local boundary between the collision sample data and the non-collision sample data is, and the higher the uncertainty of the corresponding collision determination boundary.
[0064] S233. Determine at least one boundary sensitive interval based on the local interval width corresponding to each candidate acceleration interval; The boundary-sensitive interval refers to the acceleration interval in which the local interval width between collision sample data and non-collision sample data is less than a preset interval threshold, making it easy for the distribution boundaries of the two types of samples to intersect or be confused, and having a significant impact on the position of the dividing line.
[0065] When the local interval width is less than a preset interval threshold, the corresponding candidate acceleration interval can be determined as a boundary sensitive interval. Multiple adjacent boundary sensitive intervals can form a continuous boundary sensitive region.
[0066] S234. Determine at least one intermediate anchor point based on the location of the boundary line corresponding to the boundary sensitive interval; An intermediate anchor point refers to an anchor point located between the first and second anchor points, used to reflect the trend of local boundary changes.
[0067] In one feasible implementation, the center position on the boundary line corresponding to the boundary-sensitive interval, the position with large local curvature, or the representative position of the local boundary can be selected as the intermediate anchor point.
[0068] S235. Based on the first anchor point, the second anchor point, and the at least one intermediate anchor point, determine a set of anchor points for constructing the dynamic threshold line.
[0069] In one feasible implementation, the set of anchor points includes a first anchor point corresponding to a low acceleration region, a second anchor point corresponding to a high acceleration region, and at least one intermediate anchor point located within the boundary sensitive region.
[0070] This implementation identifies boundary-sensitive regions where the boundaries between collision sample data and non-collision sample data are relatively close, and adds intermediate anchor points on the corresponding boundary lines. This expands the construction of the dynamic threshold line from being jointly determined by the first and second anchor points to being jointly determined by the set of anchor points. This improves the ability of the dynamic threshold line to represent the trend of local boundary changes, reduces the impact of local sample distribution on the construction of the dynamic threshold line under complex conditions such as minor collisions, sudden braking, speed bumps, and bumpy roads, improves the adaptability of the dynamic threshold line to different collision intensities and different vehicle conditions, and improves the accuracy and stability of the collision determination boundary in the subsequent vehicle collision detection process.
[0071] In a further embodiment, at least two anchor points corresponding to the collision type are selected based on the boundary line of the corresponding collision type, including: S24. Apply a safety margin offset to the at least two anchor points to obtain at least two target anchor points for constructing the dynamic threshold line.
[0072] Safety margin offset refers to the offset adjustment towards the non-collision area based on the fitted boundary line or corresponding anchor point, in order to establish a safety buffer between the collision sample area and the non-collision sample area and reduce the risk of false judgment in collision detection.
[0073] In some implementations, the purpose of safety margin offset is to dynamically adjust the collision detection decision boundary by adjusting the relative position of the anchor point in the boundary area, thereby achieving a balance between false alarm risk and false alarm risk.
[0074] Specifically, when shifting away from the collision sample data area, the collision determination threshold is increased to reduce the false positive rate, which is suitable for scenarios sensitive to false alarms, such as avoiding false airbag triggering; when shifting away from the non-collision sample data area, the collision determination threshold is decreased to reduce the false negative rate, which is suitable for scenarios sensitive to false negatives, such as automatic accident alarm systems.
[0075] False Positive indicates a situation where a non-collision condition is mistakenly identified as a collision, while False Negative indicates a situation where a true collision condition is not identified as a collision.
[0076] In some implementations, the offset corresponding to the safety margin offset is a preset ratio value, which is determined based on the sensitivity requirements of collision detection and the false alarm control requirements.
[0077] Sensitivity requirements refer to the response requirements of a collision detection system to real collision events. They are used to determine the ability of a dynamic threshold line to trigger minor, moderate, and severe collision scenarios, in order to avoid missed detections due to real collision events not meeting the collision judgment conditions.
[0078] False alarm control requirements refer to the control requirements of the collision detection system for false triggering under non-collision conditions. They are used to limit the probability of false collision detection when the vehicle brakes suddenly, encounters speed bumps, road bumps, and other non-collision conditions. The false alarm control requirements can be set according to the system's allowable false alarm tolerance.
[0079] False alarm tolerance refers to the degree to which a collision detection system is allowed to trigger false alarms, and it can be set according to the vehicle application scenario, safety level requirements, and collision detection strategy.
[0080] In some implementations, the preset ratio value ranges from 5% to 30%.
[0081] This implementation sets the safety margin offset to 5% to 30%, allowing the safety margin to be configured according to different vehicle platforms and collision detection requirements. This ensures collision detection sensitivity while also taking into account false alarm control, and improves the adaptability of the dynamic threshold line to different vehicle operating conditions.
[0082] In some implementations, the value of the safety margin offset can be set according to the false alarm tolerance and false negative tolerance in different application scenarios.
[0083] Specifically, when the system is sensitive to false alarms and has a low tolerance for false alarms, the safety margin offset can be set to 15% to 30% to raise the collision determination threshold and thus reduce the probability of false alarms; when the system is sensitive to missed alarms and has a low tolerance for missed alarms, the safety margin offset can be set to 5% to 15% to lower the collision determination threshold and thus reduce the probability of missed alarms; in general vehicle application scenarios, the safety margin offset can be set to 10% to 20% to achieve a balance between the risk of false alarms and the risk of missed alarms; in automotive-grade safety control systems, the safety margin offset can be set to 15% to 25% to appropriately raise the collision detection threshold while meeting safety level requirements; in automatic accident alarm application scenarios, the safety margin offset can be set to 5% to 10% to improve detection sensitivity and ensure that no accident events are missed.
[0084] In the above embodiments, a value range of 10% to 20% is considered a preferred implementation method under the condition of balancing false alarm risk and false alarm risk. By setting the preset ratio value to 10% to 20%, the dynamic threshold line has an appropriate safety margin on the basis of the boundary line, which can reduce false triggering caused by non-collision conditions, while avoiding false alarms in minor collision scenarios due to excessive safety margin, thereby improving the applicability of the dynamic threshold line under different collision conditions.
[0085] In some implementations, the safety margin offset of the at least two anchor points includes: S241. Offset the at least two anchor points in a direction away from the collision sample data region or the non-collision sample data region.
[0086] Specifically, for anchor point (acc_i, Δv_i), the offset target anchor point (acc_i', Δv_i') satisfies the following relationship: acc_i' = acc_i +δ×acc_i Δv_i' = Δv_i + δ×Δv_i Where δ is the safety margin offset ratio, used to characterize the offset intensity; for example, when δ = 0.1, it means 10% offset.
[0087] In some implementations, when δ is positive, the anchor point is shifted away from the collision sample data area to increase the collision detection threshold. When δ takes a negative value, the anchor point shifts away from the non-collision sample data area to lower the collision detection threshold.
[0088] like Figure 4As shown, the horizontal axis represents the risk characteristic value, and the vertical axis represents the sample density. By adjusting the threshold offset δ, a dynamic balance can be achieved between reducing the false alarm rate (right shift) and reducing the false negative rate (left shift).
[0089] By using the aforementioned safety margin offset method, a buffer zone is formed between collision sample data and non-collision sample data, thereby reducing the probability of misjudgment of boundary area samples in the collision determination process and improving the adaptability of the dynamic threshold line to complex vehicle conditions.
[0090] In some implementations, the selected target anchor point satisfies the following relationship: target anchor point = boundary fitting point + safety margin offset.
[0091] By offsetting at least two anchor points with a safety margin, false triggering under conditions such as sudden braking, speed bumps, and road bumps can be reduced, and the ability to distinguish between real collision scenarios and non-collision scenarios can be improved.
[0092] Among them, the method of shifting towards the non-collision sample data area is suitable for improving the detection sensitivity in the case of slight collision; the method of shifting towards the collision sample data area is suitable for reducing false triggering in the case of non-collision.
[0093] In some implementations, the at least two anchor points include a first anchor point located in a first acceleration region and a second anchor point located in a second acceleration region; the acceleration in the first acceleration region is less than the acceleration in the second acceleration region; the first velocity change threshold corresponding to the first anchor point is greater than the second velocity change threshold corresponding to the second anchor point; the first anchor point and the second anchor point are used to determine the slope and position of the dynamic threshold line.
[0094] In one specific implementation, the first anchor point can be set in a region with lower acceleration and a higher velocity change threshold to increase the collision determination threshold under low acceleration conditions; the second anchor point can be set in a region with higher acceleration and a lower velocity change threshold as a large collision identification point to adapt to collision identification under high acceleration conditions; the first anchor point and the second anchor point together determine the slope and position of the dynamic threshold line.
[0095] This implementation improves the rationality and adaptability of the dynamic threshold line calibration by setting a first anchor point with a higher velocity change threshold in the low acceleration region and a second anchor point with a lower velocity change threshold in the high acceleration region. This allows the dynamic threshold line to reflect the relationship between acceleration and velocity change under different collision intensities. Furthermore, by assigning a higher velocity change threshold to low acceleration conditions and a lower velocity change threshold to high acceleration conditions, the collision determination threshold can be dynamically adjusted according to the vehicle's operating state. This reduces the risk of misjudgment under low-intensity interference conditions and improves detection sensitivity under high-intensity collision scenarios.
[0096] S3. Determine the functional relationship between the velocity change threshold and the acceleration based on the at least two anchor points, and construct the corresponding dynamic threshold line.
[0097] The functional relationship is used to represent the relationship between the velocity change threshold under different acceleration conditions, so as to dynamically adjust the collision determination boundary according to the acceleration change.
[0098] In some implementations, the functional relationship can be at least one of a linear functional relationship, a quadratic functional relationship, a piecewise linear functional relationship, or an exponential functional relationship.
[0099] In some implementations, the functional relationship is a linear functional relationship, which satisfies: Δv_threshold = k×acc + b Wherein, Δv_threshold represents the threshold of velocity change, acc represents acceleration, k represents slope, and b represents intercept; the slope k and the intercept b are determined by at least two anchor points of the corresponding collision type based on a two-point linear equation.
[0100] Specifically, the functional relationship can be determined by the first anchor point (x1, y1) and the second anchor point (x2, y2), where x represents acceleration and y represents the change in velocity. The slope k and the intercept can be determined based on the first anchor point and the second anchor point, and the calculation formula is as follows: k = (y2 - y1) / (x2 - x1) b = y1 - k × x1 Where x1 and x2 represent the acceleration coordinates of the first anchor point and the second anchor point, respectively, and y1 and y2 represent the velocity change coordinates of the first anchor point and the second anchor point, respectively.
[0101] In some implementations, the functional relationship is a quadratic function, which satisfies: Δv_threshold = a×acc 2+ b×acc + c Wherein, Δv_threshold represents the velocity change threshold, acc represents acceleration, and a, b, and c are coefficients determined by curve fitting based on at least three anchor points of the corresponding collision type; the at least three anchor points include anchor points distributed in different acceleration ranges.
[0102] In some implementations, the functional relationship is a piecewise linear function, which satisfies: When acc < acc1, Δv_threshold = k1×acc + b1; When acc1 ≤ acc < acc2, Δv_threshold = k2×acc + b2; When acc ≥ acc2, Δv_threshold = k3×acc + b3; Wherein, Δv_threshold represents the threshold of velocity change, acc represents acceleration, acc1 and acc2 are segmentation points, which are determined based on the distribution of anchor points within the corresponding acceleration interval, and the slope k and intercept b of each segment are determined based on the anchor points within the corresponding acceleration interval.
[0103] In some implementations, the functional relationship is an exponential function, which satisfies: Δv_threshold = A×exp(B×acc) + C Where Δv_threshold represents the threshold of velocity change, acc represents acceleration, and A, B, and C are coefficients determined by curve fitting based on at least three anchor points of the corresponding collision type.
[0104] This application does not limit the specific functional form of the dynamic threshold line. In addition to the above functional relationship, the construction of the dynamic threshold line can also be achieved by using polynomial functions, spline functions, lookup table mapping relationships, interpolation functions, empirical fitting functions, mapping relationships generated by machine learning models, or other methods that can represent the correspondence between acceleration and velocity change threshold. As long as the corresponding velocity change threshold can be determined based on acceleration, it can be applied to the vehicle collision detection method based on the dynamic threshold line in this application, and is not limited here.
[0105] The vehicle collision detection dynamic threshold line calibration method in this application adopts an anchor point-based functional relationship determination method. By modeling the distribution relationship between collision sample data and non-collision sample data, a corresponding dynamic threshold line is constructed, thereby realizing an adaptive mapping relationship between acceleration and velocity change under different collision intensities. This enables the collision judgment process to combine motion changes under different working conditions for judgment, and reduces the possibility of misjudgment or missed judgment in complex working conditions caused by the fixed threshold method.
[0106] When using a linear function relationship, the slope parameter and intercept parameter can be directly determined based on the first anchor point and the second anchor point, thereby reducing the complexity of solving the function parameters and improving the adjustment efficiency of the dynamic threshold line under different collision types.
[0107] When using a nonlinear functional relationship, the nonlinear mapping relationship between the velocity change threshold and acceleration can be fitted based on at least three anchor points, thereby enhancing the ability of the dynamic threshold line to express the data distribution characteristics under complex collision conditions and improving the adaptability of the collision determination boundary.
[0108] In some implementations, different function forms can be used to adapt to different acceleration ranges or different collision intensity scenarios, thereby achieving a better judgment balance between low acceleration interference conditions and high acceleration collision conditions and improving the stability of collision detection results.
[0109] In some implementations, different collision types correspond to at least two different anchor points. These collision types include, but are not limited to, frontal collisions, rear-end collisions, side collisions, and offset collisions. Different collision types can have corresponding dynamic threshold lines established based on acceleration data and velocity change data in the corresponding directions.
[0110] In some implementations, different collision types correspond to acceleration data and velocity change data in different directions.
[0111] For example, frontal and rear collisions are based on dynamic threshold lines constructed from vehicle longitudinal data, side collisions are based on vehicle lateral data, and offset collisions are based on dynamic threshold lines constructed from both vehicle longitudinal and lateral data.
[0112] After determining the corresponding collision type, the acceleration data and velocity change data in the corresponding direction are extracted, and at least two anchor points are determined based on the data distribution in the corresponding direction.
[0113] Subsequently, the functional relationship between the velocity change threshold and acceleration is determined based on the at least two anchor points, and a dynamic threshold line corresponding to the collision type is constructed.
[0114] The vehicle collision detection dynamic threshold line calibration method of this application includes: acquiring collision sample data and non-collision sample data; selecting at least two anchor points corresponding to the collision type based on the distribution boundary between the collision sample data and the non-collision sample data, wherein the at least two anchor points include acceleration coordinate values and velocity change coordinate values; determining the functional relationship between the velocity change threshold and acceleration based on the at least two anchor points, and constructing a dynamic threshold line corresponding to the collision type. This application determines the dynamic threshold line parameters by using collision sample data and non-collision sample data, enabling the dynamic threshold lines corresponding to different collision types to be constructed based on the relationship between acceleration and velocity change, thereby improving the adaptability of the dynamic threshold line parameters to different vehicle operating conditions and providing corresponding collision judgment boundaries for subsequent vehicle collision detection.
[0115] In some implementations, the dynamic threshold line for the corresponding collision type constructed by the above-described vehicle collision detection dynamic threshold line calibration method can be used for collision determination during the vehicle collision detection process. In this case, the vehicle collision detection method includes the following steps: S4. Acquire vehicle operation data, wherein the vehicle operation data includes acceleration data in at least one direction; Vehicle operating data can be acquired through inertial sensors on the vehicle, including accelerometers or inertial measurement units, which are used to collect acceleration data of the vehicle in at least one direction.
[0116] To construct dynamic threshold lines for corresponding collision types, multiple sets of standard collision tests can be conducted in advance to obtain acceleration and velocity change data for different collision types. Collision types include frontal collisions, rear-end collisions, side collisions, and offset collisions. Simultaneously, vehicle operation data in non-collision scenarios, such as emergency braking, crossing speed bumps, and bumpy road surfaces, are also acquired. Based on the distribution differences between collision scenario data and non-collision scenario data, collision determination boundaries for corresponding collision types are determined, which are used for subsequent construction of dynamic threshold lines and collision judgment.
[0117] The acceleration data includes triaxial acceleration data, namely longitudinal acceleration, lateral acceleration, and vertical acceleration.
[0118] S5. Determine the corresponding velocity change based on the acceleration data; During a collision, sudden deceleration, or road impact, the acceleration data of a vehicle changes over time. By processing the acceleration data over a continuous time period, the corresponding change in vehicle speed can be obtained, which indicates the change in the vehicle's motion state during the current time period.
[0119] S51. Perform calculations on the acceleration data according to a preset time window to determine the corresponding velocity change.
[0120] In one embodiment, an integral operation can be used to determine the corresponding velocity change based on the acceleration data.
[0121] The preset event window can process the acceleration data using a sliding window method to obtain the vehicle's velocity change over a continuous time period. The velocity change is calculated using a sliding window integration method, and the calculation formula is as follows: Δv(t) =∫acc(τ)dτ,τ∈ [tT, t] Where Δv(t) represents the change in velocity at time t, acc(τ) represents the acceleration data at time τ, T represents the preset time window length, and τ∈[tT,t].
[0122] Under discrete sampling conditions, the velocity change can be calculated as follows: Δv(t)≈Σacc(i)×Δt Where acc(i) represents the acceleration data corresponding to the i-th sampling time, and Δt represents the sampling time interval.
[0123] In some implementations, the preset time window length is 30ms to 200ms, corresponding to 2 to 10 sampling periods.
[0124] In some implementations, different collision types correspond to different preset time window lengths. For example, the preset time window length for a frontal collision is 50ms-100ms, the preset time window length for a side collision is 30ms-80ms, and the preset time window length for a rear collision is 80ms-150ms.
[0125] In some implementations, the duration of the sampling period is 0.5 ms to 5 ms, preferably 1 ms. During continuous sampling, the preset time window slides continuously along the time axis, and after each sampling period, the corresponding velocity change is calculated based on the acceleration data within the updated time window.
[0126] In the discrete implementation process, the acceleration data can be processed using numerical integration methods, including but not limited to trapezoidal integration, rectangular integration, or other discrete integration methods.
[0127] In some implementations, the acceleration data is sampled at a frequency of 100 Hz to 10 kHz.
[0128] This embodiment employs a time-window-based method for calculating speed changes, which continuously characterizes the vehicle's motion state by combining the vehicle's acceleration changes over a short period, thereby reducing the impact of instantaneous acceleration fluctuations on collision assessment. Furthermore, by continuously updating the speed change using a sliding window, the collision assessment process corresponding to the dynamic threshold line can incorporate real-time motion changes, facilitating the differentiation of motion characteristics under different conditions such as emergency braking, road bumps, and actual collisions.
[0129] S6. Based on the acceleration data, the velocity change, and the dynamic threshold line, a collision judgment is made to determine whether a collision of the corresponding collision type has occurred.
[0130] The dynamic threshold line is used to represent the boundary relationship for determining the velocity change with acceleration. The acceleration data includes acceleration data in a first direction and acceleration data in a second direction, and the velocity change includes the velocity change in the first direction and the velocity change in the second direction. In this embodiment, the first direction is the longitudinal direction of the vehicle, used to indicate the vehicle's forward and backward movement; the second direction is the lateral direction of the vehicle, used to indicate the vehicle's left and right movement.
[0131] In some implementations, different collision types correspond to different detection directions, basic threshold conditions, and dynamic threshold line relationships. Each collision type corresponds to an independent dynamic threshold line, used to adjust the collision determination boundary under different collision scenarios, including the following steps: S61. Based on the preset collision type, determine the corresponding direction data processing method, where different collision types correspond to different directions of acceleration data and velocity change. In some implementations, different collision types correspond to different combinations of motion directions, wherein: frontal collisions and rear collisions are judged only based on the first direction of the vehicle; side collisions are judged based on the second direction of the vehicle; and offset collisions are judged based on both the first and second directions.
[0132] S62. Based on the collision type, select the acceleration data, velocity change, and corresponding dynamic threshold line in the corresponding direction to perform collision judgment.
[0133] In some implementations, a corresponding dynamic threshold line function is selected for different collision types, where each collision type corresponds to an independent curve relating acceleration and velocity change thresholds, which is used to adjust the judgment boundary under different collision intensities.
[0134] This implementation maps different collision types to data processing paths in different directions and combines them with corresponding dynamic threshold lines for judgment, enabling the collision detection process to adapt to acceleration change characteristics under different motion conditions, thereby reducing the risk of misjudgment in complex road conditions by using fixed threshold judgment methods.
[0135] In some implementations, the collision determination is based on a comprehensive judgment of a basic threshold condition and a dynamic threshold line condition. The basic threshold condition is used to filter low-intensity interference signals, and the dynamic threshold line condition is used to identify collisions based on the relationship between acceleration and velocity changes. This includes the following steps: S621. Input the acceleration data in the corresponding direction and the velocity change in the corresponding direction to the dynamic threshold line of the corresponding collision type for matching and judgment; S622. When the acceleration data and the velocity change satisfy the corresponding basic threshold conditions, and the correspondence between the acceleration data in the corresponding direction and the velocity change satisfies the dynamic threshold line conditions, it is determined that a collision of the corresponding collision type has occurred.
[0136] In some implementations, collision determination combines basic threshold conditions and dynamic threshold line conditions for comprehensive judgment. The basic threshold is used to eliminate weak interference conditions, while the dynamic threshold line is used to characterize the relationship between acceleration and velocity changes, thereby achieving precise identification of collision events.
[0137] In some implementations, the collision determination satisfies the following conditions: the absolute value of the acceleration data is greater than or equal to the basic acceleration threshold of the corresponding collision type; the absolute value of the velocity change is greater than or equal to the basic velocity change threshold of the corresponding collision type; the functional relationship between the velocity change and acceleration satisfies the threshold condition corresponding to the dynamic threshold line, that is, the velocity change is greater than the threshold output value of the corresponding acceleration under the dynamic threshold line function; when the above three conditions are satisfied at the same time, a collision of the corresponding type is determined to have occurred.
[0138] In some implementations, the collision type includes a frontal collision, the detection of which is based on a first direction of the vehicle; The first direction is the longitudinal direction of the vehicle, used to indicate the deceleration state of the vehicle during forward travel.
[0139] When the acceleration data in the first direction is negative and meets the first basic acceleration threshold, and the velocity change in the first direction is negative and meets the first basic velocity change threshold, and the correspondence between the acceleration data and the velocity change in the first direction meets the dynamic threshold line condition for a frontal collision, it is determined to be a frontal collision.
[0140] The negative direction indicates the vehicle's deceleration trend in the longitudinal direction. The positive and negative directions of the acceleration data characterize the change in the vehicle's direction of motion in the preset coordinate system; the positive and negative directions of the velocity change characterize the trend of the vehicle's speed increasing or decreasing.
[0141] In some implementations, the collision type includes a rear-end collision, the detection of which is based on a first direction of the vehicle; When the acceleration data in the first direction is positive and meets the second basic acceleration threshold, and the velocity change in the first direction is positive and meets the second basic velocity change threshold, and the correspondence between the acceleration data and the velocity change in the first direction meets the dynamic threshold line condition for a subsequent collision, it is determined to be a subsequent collision.
[0142] Positive is used to indicate the acceleration trend of a vehicle in the longitudinal direction, such as forward acceleration caused by a rear-end collision.
[0143] In some implementations, the collision type includes a side collision, the detection of which is based on a second direction of the vehicle; The second direction is the lateral direction of the vehicle, used to indicate the vehicle's left and right lateral movement.
[0144] When the absolute value of the acceleration data in the second direction meets the third basic acceleration threshold, and the absolute value of the velocity change in the second direction meets the third basic velocity change threshold, and the correspondence between the acceleration data in the second direction and the velocity change meets the dynamic threshold line condition for a side collision, it is determined to be a side collision.
[0145] The absolute value is used to characterize the bidirectional impact characteristics that may exist during a lateral collision.
[0146] In some implementations, the collision type includes an offset collision, the detection of which is based on a first direction and a second direction of the vehicle; The first direction is used to represent the longitudinal motion state, and the second direction is used to represent the lateral motion state.
[0147] When the acceleration data and velocity change in the first direction and the acceleration data and velocity change in the second direction respectively satisfy the dynamic threshold line relationship corresponding to the offset collision, an offset collision is determined to have occurred.
[0148] In some implementations, offset collision is used to represent the combined collision behavior of a vehicle under non-orthogonal force conditions, which involves both longitudinal and lateral motion characteristics.
[0149] In some implementations, after a collision is determined to have occurred, the vehicle's status is monitored over a preset time period, including but not limited to: whether the vehicle's travel distance is less than a preset distance threshold (e.g., 30m); whether the vehicle's hazard lights are on; whether the vehicle doors are open; and whether the seat belts are unfastened.
[0150] This implementation method detects vehicle behavior after a collision, which can be used to assist in confirming collision events and subsequent safety measures.
[0151] In the vehicle collision detection method based on dynamic threshold lines according to the embodiments of this application, by introducing dynamic threshold lines, the collision judgment conditions are adjusted with changes in acceleration, maintaining stable judgment capability under different collision intensities and reducing the false alarms and false negatives caused by fixed threshold methods. By constructing independent judgment paths corresponding to different collision types, differentiated identification of frontal collisions, rear collisions, side collisions, and offset collisions is achieved, improving the collision type differentiation capability. By introducing a combined judgment mechanism of basic thresholds and dynamic threshold lines, the system can filter weak interference signals while retaining sensitivity to real collisions. Through multi-dimensional data processing based on the first and second directions, reliable identification can still be achieved under combined longitudinal and lateral forces on the vehicle. By constructing dynamic threshold lines based on real vehicle operation data, the parameters can be adapted to different vehicle models, different sensor installation methods, and different road conditions.
[0152] Secondly, please refer to Figure 5 This application provides a calibration device for a dynamic threshold line of vehicle collision detection, the device 200 comprising: The acquisition module 201 is used to acquire collision sample data and non-collision sample data; The selection module 202 is used to select at least two anchor points corresponding to the collision type based on the distribution boundary between the collision sample data and the non-collision sample data. The anchor points include acceleration coordinate values and velocity change coordinate values. The function determination module 203 is used to determine the functional relationship between the velocity change threshold and the acceleration based on the at least two anchor points, and to construct the corresponding dynamic threshold line.
[0153] It should be noted that each module in the vehicle collision detection dynamic threshold line calibration device in this embodiment corresponds one-to-one with each step in the vehicle collision detection dynamic threshold line calibration method in the first aspect embodiment above. Therefore, the specific implementation of this embodiment can refer to the implementation of the aforementioned vehicle collision detection dynamic threshold line calibration method, and will not be repeated here.
[0154] Thirdly, embodiments of this application provide an electronic device, including: at least one processor, at least one memory, and computer program instructions stored in the memory, which, when executed by the processor, implement the method described in the first aspect above.
[0155] Specifically, the processor may include a central processing unit (CPU), an application-specific integrated circuit (ASIC), or one or more integrated circuits that can be configured to implement the embodiments of this application.
[0156] Memory may include non-persistent memory in computer-readable media, such as random access memory (RAM) and / or non-volatile memory, such as read-only memory (ROM) or flash RAM. Memory is an example of computer-readable media.
[0157] The processor implements the method described in the first aspect embodiment by reading and executing computer program instructions stored in memory.
[0158] In one example, the electronic device may also include a communication interface and a bus. The processor, memory, and communication interface are connected via the bus to communicate with each other. The communication interface is primarily used to enable communication between modules, devices, units, and / or equipment in the embodiments of this application. The bus includes hardware, software, or both, coupling components of the electronic device together. For example, and not limitingly, the bus may include an Accelerated Graphics Port (AGP) or other graphics bus, an Enhanced Industry Standard Architecture (EISA) bus, a Front Side Bus (FSB), HyperTransport (HT) interconnect, an Industry Standard Architecture (ISA) bus, an Infinite Bandwidth Interconnect, a Low Pin Count (LPC) bus, a memory bus, a Microchannel Architecture (MCA) bus, a Peripheral Component Interconnect (PCI) bus, a PCI-Express (PCI-X) bus, a Serial Advanced Technology Attachment (SATA) bus, a Video Electronics Standards Association Local (VLB) bus, or other suitable buses, or combinations of two or more of these. Where appropriate, the bus may include one or more buses. Although specific buses are described and illustrated in the embodiments of this application, this application contemplates any suitable bus or interconnect.
[0159] Fourthly, embodiments of this application provide a computer-readable storage medium having computer program instructions stored thereon, which, when executed by a processor, implement the method described in the first aspect above.
[0160] It should be clarified that this application is not limited to the specific configurations and processes described above and shown in the figures. For the sake of brevity, detailed descriptions of known methods are omitted here. In the above embodiments, several specific steps are described and shown as examples. However, the method process of this application is not limited to the specific steps described and shown. Those skilled in the art can make various changes, modifications, and additions, or change the order of steps, after understanding the spirit of this application.
[0161] Those skilled in the art will understand that embodiments of this application can be provided as methods, systems, or computer program products. Therefore, this application can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, this application can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.
[0162] The above description is merely a specific implementation of this application. Those skilled in the art will clearly understand that, for the sake of convenience and brevity, the specific working processes of the systems, modules, and units described above can be referred to the corresponding processes in the foregoing method embodiments, and will not be repeated here. It should be understood that the protection scope of this application is not limited thereto. Any person skilled in the art can easily conceive of various equivalent modifications or substitutions within the technical scope disclosed in this application, and these modifications or substitutions should all be covered within the protection scope of this application.
Claims
1. A method for calibrating a dynamic threshold line for vehicle collision detection, characterized in that, The method includes: Acquire collision sample data and non-collision sample data; Based on the distribution boundary between the collision sample data and the non-collision sample data, at least two anchor points corresponding to the collision type are selected, and the anchor points include acceleration coordinate values and velocity change coordinate values. Based on the at least two anchor points, determine the functional relationship between the velocity change threshold and acceleration, and construct a dynamic threshold line corresponding to the collision type.
2. The method according to claim 1, characterized in that, The step of selecting at least two anchor points corresponding to the collision type based on the distribution boundary between the collision sample data and the non-collision sample data includes: Construct a two-dimensional sample distribution of acceleration and velocity change based on the collision sample data and the non-collision sample data; By fitting the boundary region between collision sample data and non-collision sample data, the boundary line corresponding to the collision type is obtained; Select at least two anchor points according to the boundary line corresponding to the collision type.
3. The method according to claim 2, characterized in that, The step of constructing a two-dimensional sample distribution of acceleration and velocity change based on the collision sample data and the non-collision sample data includes: Determine the acceleration data and velocity change data in the corresponding direction based on the corresponding collision type; A two-dimensional sample distribution for the corresponding collision type is constructed based on the acceleration data and velocity change data in the corresponding direction.
4. The method according to claim 3, characterized in that, After selecting at least two anchor points of the corresponding collision type according to the boundary line of the corresponding collision type, the process includes: A safety margin offset is applied to the at least two anchor points to obtain at least two target anchor points for constructing the dynamic threshold line.
5. The method according to claim 4, characterized in that, The offset corresponding to the safety margin offset is a preset ratio value, which is determined according to the sensitivity requirements of collision detection and the false alarm control requirements.
6. The method according to claim 5, characterized in that, The preset ratio value ranges from 5% to 30%.
7. The method according to claim 4, characterized in that, The safety margin offset of the at least two anchor points includes: Offset the at least two anchor points in a direction away from the collision sample data region or the non-collision sample data region.
8. The method according to any one of claims 1-7, characterized in that, The at least two anchor points include a first anchor point located in a first acceleration region and a second anchor point located in a second acceleration region; the acceleration in the first acceleration region is less than the acceleration in the second acceleration region; The threshold for the first velocity change corresponding to the first anchor point is greater than the threshold for the second velocity change corresponding to the second anchor point.
9. The method according to any one of claims 1-7, characterized in that, The functional relationship is a linear function, and the linear function satisfies: Δv_threshold = k×acc + b Wherein, Δv_threshold represents the threshold of velocity change, acc represents acceleration, k represents slope, and b represents intercept. The slope k and the intercept b are determined by at least two anchor points of the corresponding collision type based on a two-point linear equation.
10. The method according to any one of claims 1-7, characterized in that, The functional relationship is a quadratic function, and the quadratic function satisfies: Δv_threshold = a×acc 2 + b×acc + c Where Δv_threshold represents the threshold of velocity change, acc represents acceleration, and a, b, and c are coefficients determined by curve fitting based on at least three anchor points of the corresponding collision type; The at least three anchor points include anchor points distributed across different acceleration ranges.
11. The method according to any one of claims 1-7, characterized in that, The functional relationship is a piecewise linear function, and the piecewise linear function satisfies: When acc < acc1, Δv_threshold = k1×acc + b1; When acc1 ≤ acc < acc2, Δv_threshold = k2×acc + b2; When acc ≥ acc2, Δv_threshold = k3×acc + b3; Wherein, Δv_threshold represents the threshold of velocity change, acc represents acceleration, acc1 and acc2 are segmentation points, which are determined based on the distribution of anchor points within the corresponding acceleration interval, and the slope k and intercept b of each segment are determined based on the anchor points within the corresponding acceleration interval.
12. The method according to any one of claims 1-7, characterized in that, The functional relationship is an exponential function, and the exponential function satisfies: Δv_threshold = A×exp(B×acc) + C Where Δv_threshold represents the threshold of velocity change, acc represents acceleration, and A, B, and C are coefficients determined by curve fitting based on at least three anchor points of the corresponding collision type.
13. A calibration device for a dynamic threshold line for vehicle collision detection, characterized in that, The device includes: The acquisition module is used to acquire collision sample data and non-collision sample data; The selection module is used to select at least two anchor points corresponding to the collision type based on the distribution boundary between the collision sample data and the non-collision sample data. The anchor points include acceleration coordinate values and velocity change coordinate values. The function determination module is used to determine the functional relationship between the velocity change threshold and the acceleration based on the at least two anchor points, and to construct a dynamic threshold line corresponding to the collision type.
14. A computer-readable storage medium having computer program instructions stored thereon, characterized in that, When the computer program instructions are executed by a processor, the method as described in any one of claims 1-12 is implemented.