A safety monitoring-based intelligent automobile sensor data correction method and system

Abstract

The application discloses a kind of intelligent car sensor data correction method and system based on safety monitoring, it is related to intelligent driving environment perception technical field, including, based on safety rule base, grade division is carried out to safety feature coding, mode matching is carried out in combination with historical fault case library, and the abnormal diagnosis report of original monitoring data is generated;The physical topology network of intelligent car sensor is constructed, and the node state of physical topology network is dynamically adjusted according to abnormal diagnosis report, and the physical topology network reconstruction instruction is acquired and executed;Based on the node state of reconstructed physical topology network, original monitoring data is corrected twice, and vehicle safety monitoring data is generated.The application is corrected by weighting residual error, combined with iterative reweighted least squares method, based on the node weight distribution table after reconstruction, two-stage correction is carried out to monitoring data, realizes the gradual error compensation under the condition of sensor network dynamic performance degradation, improves the reliability and environmental adaptability of output data.

Description

Technical Field

[0001] This invention relates to the field of intelligent driving environment perception technology, and in particular to a method and system for correcting sensor data of intelligent vehicles based on safety monitoring. Background Technology

[0002] In the field of intelligent vehicle environmental perception, multi-sensor data fusion technology is a core element in ensuring driving safety. Current mainstream sensor data monitoring methods typically employ a combination of multi-source information acquisition, feature extraction, and state assessment. A typical approach uses sensors such as millimeter-wave radar and lidar to acquire raw data on the target's spatial location, motion state, and reflection characteristics. After preprocessing such as denoising, synchronization, and normalization, feature parameters such as point cloud density change rate and signal-to-noise ratio fluctuation are extracted. Existing technologies generally employ a hierarchical analysis framework based on rule bases, classifying feature parameters using predefined safety thresholds and performing pattern matching based on historical fault cases to ultimately output anomaly diagnostic results.

[0003] Conventional methods still have two limitations in engineering practice: First, the dynamic adjustment mechanism of the physical topology network lacks a quantitative evaluation dimension, and node state switching relies on fixed threshold judgments, making it difficult to adapt to the gradual degradation of sensor performance under complex working conditions. Especially in environmental interference scenarios such as rain and fog, static thresholds can easily lead to false isolation or reduced weighting delays. Second, there is a coupling defect between confidence assessment and spatiotemporal verification in the data correction stage. When multiple nodes experience instantaneous anomalies simultaneously, the traditional sliding window statistical method cannot distinguish between systematic biases and random noise, resulting in residual errors remaining in the data after secondary correction. Summary of the Invention

[0004] In view of the aforementioned existing problems, the present invention is proposed.

[0005] Therefore, this invention provides a method for correcting sensor data in intelligent vehicles based on safety monitoring to solve the problems of insufficient dynamic adaptability and error coupling in sensor networks.

[0006] To solve the above-mentioned technical problems, the present invention provides the following technical solution:

[0007] In a first aspect, the present invention provides a method for correcting sensor data of intelligent vehicles based on safety monitoring, which includes collecting raw monitoring data through intelligent vehicle sensors and preprocessing it, calculating the point cloud density mutation rate, the instantaneous drop in signal-to-noise ratio and the frequency of data continuity interruption, and generating a safety feature code.

[0008] Based on the security rule base, the security feature codes are classified into levels, and combined with the historical fault case database, pattern matching is performed to generate an anomaly diagnosis report of the original monitoring data.

[0009] Construct a physical topology network for intelligent vehicle sensors, dynamically adjust the node states of the physical topology network based on anomaly diagnosis reports, and obtain and execute physical topology network reconstruction instructions.

[0010] Based on the node status of the reconstructed physical topology network, the original monitoring data is corrected twice to generate vehicle safety monitoring data.

[0011] As a preferred embodiment of the intelligent vehicle sensor data correction method based on safety monitoring described in this invention, the original monitoring data includes the spatial position coordinates of the target object, its motion speed value, electromagnetic wave reflection intensity value, three-dimensional point cloud distribution, optical reflectivity value, time synchronization marker, as well as the distance measurement value of nearby obstacles, vehicle acceleration value, and angular velocity value.

[0012] As a preferred embodiment of the intelligent vehicle sensor data correction method based on safety monitoring described in this invention, the steps for calculating the point cloud density abrupt change rate, the instantaneous decrease in signal-to-noise ratio, and the frequency of data continuity interruptions are as follows:

[0013] Based on the 3D point cloud distribution and time synchronization markers, the density abrupt change rate of the 3D point cloud distribution is calculated;

[0014] Based on the density abrupt change rate and electromagnetic wave reflection intensity of the three-dimensional point cloud distribution, the instantaneous drop in signal-to-noise ratio is calculated, and the frequency of continuity interruption is statistically analyzed in conjunction with time synchronization markers.

[0015] As a preferred embodiment of the intelligent vehicle sensor data correction method based on safety monitoring described in this invention, the steps of classifying safety feature codes into levels based on a safety rule base, performing pattern matching in conjunction with a historical fault case database, and generating an anomaly diagnosis report of the original monitoring data are as follows:

[0016] Based on the abnormal status of historical monitoring data, anomaly judgment rules are defined, and the anomaly judgment rules are associated and mapped with the risk level labels in the historical fault case library to generate a safety rule library.

[0017] Based on the security rule base, the security feature codes are classified into levels, and the fault types are matched from the historical fault case library according to the cosine similarity between the security feature codes and the historical fault feature vectors. An abnormal diagnosis report is generated and loaded into the historical fault case library.

[0018] As a preferred embodiment of the intelligent vehicle sensor data correction method based on safety monitoring described in this invention, the steps of constructing a physical topology network of intelligent vehicle sensors and dynamically adjusting the node states of the physical topology network according to anomaly diagnosis reports are as follows:

[0019] Construct a physical topology network for intelligent vehicle sensors and dynamically adjust the node states of the physical topology network based on anomaly diagnosis reports;

[0020] Based on the risk level in the anomaly diagnosis report, perform status marking on the physical topology network nodes, extract and encapsulate the node control parameters, and generate node control instructions.

[0021] The node states of the physical topology network are dynamically adjusted according to node control commands.

[0022] As a preferred embodiment of the intelligent vehicle sensor data correction method based on safety monitoring described in this invention, the step of obtaining and executing the physical topology network reconstruction instruction refers to collecting the state feedback data of the adjusted physical topology network, predicting the health status score of each node in the physical topology network, and generating and executing the physical topology network reconstruction instruction of the intelligent vehicle sensor using the earliest deadline first algorithm.

[0023] As a preferred embodiment of the intelligent vehicle sensor data correction method based on safety monitoring according to the present invention, the steps for correcting the original monitoring data twice based on the node states of the reconstructed physical topology network to generate vehicle safety monitoring data are as follows:

[0024] Based on the node states of the reconstructed physical topology network, the spatial coordinates and weight allocation parameters of the nodes in the normal state are extracted, and the node weight allocation table of the physical topology network is generated by weighted least squares method.

[0025] Based on the node weight allocation table, the weighted residual correction method is used to calculate the deviation between the original monitoring data of each node and the weighted mean, and the original monitoring data is corrected for the first time according to the weight ratio to generate the initial vehicle safety monitoring data.

[0026] Statistical analysis was performed on the confidence weights of each initial vehicle safety monitoring data, and the spatiotemporal consistency of the confidence weights was verified. Failed confidence weights were marked.

[0027] The iterative reweighted least squares method is used to perform secondary correction on the initial vehicle safety monitoring data where the confidence weights have failed, thereby generating vehicle safety monitoring data.

[0028] As a preferred embodiment of the intelligent vehicle sensor data correction method based on safety monitoring described in this invention, the steps for generating the safety feature code are as follows:

[0029] Based on density mutation rate, instantaneous drop in signal-to-noise ratio, and frequency of data continuity interruption, abnormal states of raw monitoring data are identified.

[0030] By using binary bit-field encoding, the abnormal state of the original monitoring data is mapped to a security feature code.

[0031] As a preferred embodiment of the intelligent vehicle sensor data correction method based on safety monitoring described in this invention, the preprocessing includes normalization, filtering and noise reduction, time synchronization calibration, and validity verification.

[0032] Secondly, the present invention provides a smart car sensor data correction system based on safety monitoring, including a data acquisition module, which acquires raw monitoring data through smart car sensors and performs preprocessing, calculates the point cloud density mutation rate, the instantaneous drop in signal-to-noise ratio and the frequency of data continuity interruption, and generates a safety feature code.

[0033] The anomaly diagnosis module classifies security feature codes into levels based on the security rule base, performs pattern matching in combination with the historical fault case database, and generates an anomaly diagnosis report of the original monitoring data.

[0034] The topology network adjustment module constructs the physical topology network of intelligent vehicle sensors, dynamically adjusts the node status of the physical topology network based on anomaly diagnosis reports, and obtains and executes physical topology network reconstruction instructions.

[0035] The data correction module corrects the original monitoring data twice based on the node status of the reconstructed physical topology network to generate vehicle safety monitoring data.

[0036] The beneficial effects of this invention are as follows: It employs multi-level threshold criteria to jointly analyze the point cloud density mutation rate, instantaneous signal-to-noise ratio drop, and data continuity interruption frequency. Furthermore, it generates structured security feature codes through binary bit-field encoding, achieving accurate quantitative characterization of multi-dimensional abnormal states and effectively improving the accuracy and efficiency of fault feature identification under complex operating conditions. By combining weighted residual correction with iterative reweighted least squares, and performing two-level correction on the monitoring data based on the reconstructed node weight allocation table, it achieves progressive error compensation under dynamic performance degradation of the sensor network, improving the reliability and environmental adaptability of the output data. Attached Figure Description

[0037] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the following description of the embodiments will be briefly introduced. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0038] Figure 1 This is a flowchart of a method for correcting sensor data in intelligent vehicles based on safety monitoring.

[0039] Figure 2 This is a schematic diagram of a smart car sensor data correction system based on safety monitoring.

[0040] Figure 3 A flowchart for generating security feature encoding.

[0041] Figure 4 A flowchart for secondary correction of vehicle safety monitoring data. Detailed Implementation

[0042] To make the above-mentioned objects, features and advantages of the present invention more apparent and understandable, the specific embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0043] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0044] Secondly, the term "one embodiment" or "embodiment" as used herein refers to a specific feature, structure, or characteristic that may be included in at least one implementation of the present invention. The phrase "in one embodiment" appearing in different places in this specification does not necessarily refer to the same embodiment, nor is it a single or selective embodiment that is mutually exclusive with other embodiments.

[0045] Reference Figures 1-4 This is one embodiment of the present invention, which provides a method for correcting sensor data of an intelligent vehicle based on safety monitoring, including the following steps:

[0046] S1. Collect raw monitoring data through intelligent vehicle sensors and preprocess it to calculate the point cloud density mutation rate, instantaneous drop in signal-to-noise ratio, and frequency of data continuity interruption, and generate security feature codes.

[0047] The raw monitoring data includes the spatial coordinates of the target object, its velocity, electromagnetic wave reflection intensity, three-dimensional point cloud distribution, optical reflectivity, time synchronization markers, as well as the distance measurement of nearby obstacles, vehicle acceleration, and angular velocity.

[0048] Furthermore, the acquisition of raw monitoring data is achieved through the collaborative work of multiple sensors: the lidar generates the three-dimensional point cloud distribution and spatial coordinates of the target object by emitting laser pulses and receiving reflected signals, while recording the optical reflectivity value; the millimeter-wave radar uses the electromagnetic wave reflection characteristics to measure the motion speed value and reflection intensity value; the ultrasonic sensor is responsible for measuring the distance to nearby obstacles; and the inertial measurement unit (IMU) continuously collects the vehicle's acceleration and angular velocity values.

[0049] Preprocessing includes normalization, filtering and noise reduction, time synchronization calibration, and validity verification.

[0050] Furthermore, normalization is first performed to map the raw monitoring data with different dimensions (such as the reflection intensity values ​​of millimeter-wave radar and the point cloud density of lidar) to a standard range, eliminating the impact of dimensional differences on subsequent analysis. Then, filtering and denoising operations are performed. For lidar data, a statistical outlier filtering algorithm is used to eliminate abnormal point clouds; for millimeter-wave radar signals, Kalman filtering is applied to suppress multipath interference; and for IMU data, sliding window mid-range filtering is applied to eliminate instantaneous noise. The time synchronization calibration stage achieves microsecond-level time alignment of multi-sensor data using a hardware clock method and the PTP protocol (Precision Time Protocol), ensuring timing consistency. Finally, validity verification is performed, filtering the raw monitoring data according to physical rationality rules to remove obvious outliers.

[0051] It should be noted that the physical rationality rule is a verification standard established based on statistical analysis of vehicle historical operation data (such as vehicle speed, acceleration, and position trajectory in actual roads) and environmental perception data (such as lidar point cloud distribution and millimeter-wave radar reflection intensity).

[0052] Based on the 3D point cloud distribution and time-synchronized markers, the density abrupt change rate of the 3D point cloud distribution is calculated through spatial grid density analysis, expressed as:

[0053] ;

[0054] in, express Density abrupt change rate of the 3D point cloud distribution at any given time. express The 3D point cloud density at any given time. express The 3D point cloud density at any given time;

[0055] Based on the density abrupt change rate and electromagnetic wave reflection intensity of the three-dimensional point cloud distribution, the instantaneous decrease in signal-to-noise ratio is calculated using the sliding window standard deviation, and the frequency of continuity interruption is statistically analyzed in conjunction with time synchronization markers.

[0056] Furthermore, the instantaneous signal-to-noise ratio (SNR) drop is calculated based on the density abrupt change rate of the 3D point cloud distribution and the electromagnetic wave reflection intensity value. First, a sliding window of fixed time length is used to segment the electromagnetic wave reflection intensity value. Within the window, the instantaneous SNR fluctuation amplitude is calculated using the standard deviation of the electromagnetic wave reflection intensity value. The density abrupt change rate of the 3D point cloud distribution is used as a dynamic weighting factor to weight and correct the SNR fluctuation amplitude within the window, generating a normalized instantaneous SNR drop value. The time synchronization marker analyzes the timestamp interval of adjacent data packets and counts the number of intervals exceeding the timestamp interval threshold as the data continuity interruption frequency.

[0057] The expression for calculating the instantaneous decrease in signal-to-noise ratio is:

[0058] ;

[0059] in, It is the mean signal-to-noise ratio within the sliding window. express Instantaneous drop in signal-to-noise ratio at any moment express Signal-to-noise ratio at any given moment This represents the standard deviation of the electromagnetic wave reflection intensity values ​​within the current sliding window;

[0060] It should be noted that the time interval threshold is defined based on statistical analysis of network transmission stability requirements (such as maximum allowable delay and average round-trip time), and the value is usually in the range of 0.05 to 0.2 seconds.

[0061] Based on density mutation rate, instantaneous signal-to-noise ratio drop, and data continuity interruption frequency, a multi-level threshold criterion is used to identify abnormal states of the original monitoring data (abnormal point cloud density mutation, abnormal instantaneous signal-to-noise ratio drop, and abnormal data continuity interruption flag).

[0062] Furthermore, when identifying abnormal states based on density mutation rate, instantaneous signal-to-noise ratio (SNR) drop, and data continuity interruption frequency, a multi-level threshold criterion is first established: The first level targets the density mutation rate, setting a density mutation threshold based on statistical analysis of historical monitoring data (typically ranging from 5% to 15%). When the density mutation rate exceeds the threshold, a point cloud density mutation anomaly flag is triggered. The second level uses a sliding window mean comparison method; when the instantaneous SNR drop consistently exceeds three times the standard deviation of the historical mean within the window, an anomaly in the instantaneous SNR drop is marked. The third level uses time series analysis; when the data continuity interruption frequency reaches the reciprocal of the hardware sampling period per unit time, the data continuity interruption anomaly flag is activated. These three anomaly flags are output through a binary bit-field encoding combination.

[0063] By using binary bit-field encoding, the abnormal state of the original monitoring data is mapped to a security feature code.

[0064] Furthermore, based on the density mutation rate, the instantaneous decrease in signal-to-noise ratio, and the frequency of data continuity interruption, an 8-bit binary encoding structure is defined: the lowest three bits correspond to the point cloud density mutation anomaly flag (bit0), the instantaneous decrease in signal-to-noise ratio anomaly flag (bit1), and the data continuity interruption anomaly flag (bit2), respectively; the middle three bits are reserved for extension bits (bit3-bit5); and the highest two bits represent the risk level (bit6-bit7). Finally, an 8-bit security feature code is synthesized through bit operations.

[0065] S2. Based on the security rule base, the security feature codes are classified into levels, and pattern matching is performed in combination with the historical fault case library to generate an anomaly diagnosis report of the original monitoring data.

[0066] Based on the abnormal status of historical monitoring data, support vector machines are used to define anomaly judgment rules, and the K-nearest neighbor algorithm is used to associate and map the anomaly judgment rules with the risk level labels in the historical fault case library to generate a safety rule library.

[0067] Furthermore, based on the abnormal states of historical monitoring data, a support vector machine is used to classify and train the original monitoring data with abnormal states. By maximizing the class margin, the optimal hyperplane is found to determine the boundary conditions that can distinguish different types of anomalies (such as density mutation threshold and three times the standard deviation of the historical mean). This forms a multi-dimensional judgment boundary for point cloud density mutation rate, instantaneous signal-to-noise ratio drop, and frequency of data continuity interruption, resulting in specific anomaly judgment rules. Subsequently, the K-nearest neighbor algorithm is used to calculate the distance between the current security feature code and each historical fault feature vector in the historical fault case library. The K closest historical fault feature vectors are found. Based on the majority category of the risk level labels corresponding to the K historical fault feature vectors, an association mapping relationship is established between the anomaly judgment rules and the risk level labels. Finally, a security rule library containing anomaly judgment rules and corresponding risk level labels is generated.

[0068] Support Vector Machines (SVMs) are trained using supervised learning methods based on historical monitoring data labeled with state categories. It should be noted that the historical fault case library refers to a database that stores historical fault feature vectors and their corresponding risk level labels and fault type information. This database is constructed by collecting various sensor fault cases that occur during the actual operation of intelligent vehicles and storing them as historical fault feature vectors.

[0069] Based on the security rule base, the security feature codes are classified into levels, and the fault types are matched from the historical fault case library according to the cosine similarity between the security feature codes and the historical fault feature vectors. An abnormal diagnosis report is generated and loaded into the historical fault case library.

[0070] Furthermore, the safety feature code is input into the safety rule base. Based on the association mapping relationship between the anomaly judgment rules and risk level labels in the safety rule base, the risk level corresponding to the safety feature code is determined, and the level classification is completed. Then, the cosine similarity between the safety feature code and each historical fault feature vector in the historical fault case library is calculated by weighted dot product. By comparing the cosine similarity values, the fault type corresponding to the historical fault feature vector with the highest cosine similarity is matched from the historical fault case library. Combining the level classification results with the matched fault types, an anomaly diagnosis report is generated, and the generated anomaly diagnosis report is loaded as a new record into the historical fault case library.

[0071] The cosine similarity between the security feature code and the historical fault feature vector is calculated using a weighted dot product, expressed as:

[0072] ;

[0073] in, It is the cosine similarity between the security feature code and the historical fault feature vector. It is the number of bits in the security feature code. It is the security feature coding number The weighting coefficient of the bit (usually ranging from 0.1 to 1.0). It is the first security feature encoding Bit value It is the first of the feature vectors of historical failure cases Bit value;

[0074] Weighting coefficient It is defined based on the importance statistical analysis of historical fault feature vector dimensions;

[0075] It should be noted that, Before participating in the calculation, the dimensions are eliminated by Z-score standardization. S3, Construct the physical topology network of the intelligent vehicle sensors, dynamically adjust the node states of the physical topology network according to the anomaly diagnosis report, and obtain and execute the physical topology network reconstruction instructions;

[0076] Based on the installation location coordinates and communication link relationships of intelligent vehicle sensors, a physical topology network of intelligent vehicle sensors is constructed.

[0077] Furthermore, the three-dimensional installation coordinates of each intelligent vehicle sensor on the vehicle body are obtained. At the same time, the communication connection status (such as normal connection, connection interruption, and signal strength level) and link bandwidth information (such as maximum transmission rate, current available bandwidth, and data packet loss rate) between each sensor are collected. Each intelligent vehicle sensor is used as a node, and the communication link relationship is used as an edge. The spatial layout of the nodes is determined according to the installation coordinates. The connection structure between nodes is established according to the communication link relationship. The nodes and edges are connected topologically using a graph structure representation method, and finally the physical topology network of intelligent vehicle sensors is constructed.

[0078] It should be noted that the communication link relationship refers to the communication connection and its attributes between intelligent vehicle sensors. For example, the connection between sensor A and sensor B is normal and has a maximum transmission rate of 10 Mbps and a packet loss rate of 5%.

[0079] Based on the risk level in the anomaly diagnosis report, the physical topology network nodes are marked with status (isolation / deweighting / normal status), node control parameters are extracted by hash table lookup and encapsulated using DDS, and node control instructions are generated.

[0080] Furthermore, based on the risk level in the anomaly diagnosis report, the risk level is matched with the node status mapping table to determine the status flag that the corresponding physical topology network node should execute. The status flag includes isolation, deweighting, or normal status. Then, using the node identifier as the key, the control parameters corresponding to the current node are extracted from the node control parameter database using a hash table lookup method. These parameters include power outage delay time, weight adjustment value, or recovery command. The obtained node control parameters are serialized and encapsulated according to DDS (Data Distribution Service) to generate a node control command containing the target node identifier, status flag, and control parameters.

[0081] It should be noted that node control commands are hardware-level control protocols used to dynamically adjust the working state of nodes in the physical topology network of intelligent vehicle sensors. They achieve real-time optimization of the network topology by encapsulating node control parameters (such as isolation flags, weighting coefficients, sampling frequencies, etc.).

[0082] Based on node control instructions, dynamically adjust the node status of the physical topology network (such as the power outage delay time of isolated nodes or the weight value of deweighted nodes).

[0083] Furthermore, the target node identifier, status flag, and control parameters contained in the node control command are analyzed using the DDS serialization parsing method to confirm whether the physical topology network node to be adjusted and its corresponding status flag are isolated or deweighted. For example, if the risk level value is greater than or equal to the isolation threshold, the status flag is isolated; if the risk level value is between the deweighting threshold and the isolation threshold, the status flag is deweighted. If the status flag is isolated, the power-off delay time of the isolated node is obtained from the control parameters, and the power supply to the current node is cut off after the power-off delay time is reached. If the status flag is deweighted, the weight value of the deweighted node is obtained from the control parameters, and the original monitoring data of the current node in the physical topology network is reduced.

[0084] It should be noted that the weighting threshold and isolation threshold are set based on statistical analysis of historical failure cases. The weighting threshold usually ranges from 0.5 to 0.8, and the isolation threshold usually ranges from 0.8 to 1.0.

[0085] Collect the state feedback data of the adjusted physical topology network, predict the health status score of each node in the physical topology network using the fuzzy comprehensive evaluation method, and generate and execute the physical topology network reconstruction instructions for the intelligent vehicle sensors using the earliest deadline first algorithm.

[0086] Furthermore, the state feedback data of the adjusted physical topology network is collected. Using the state feedback data as input parameters, combined with preset evaluation factor weights and membership functions, the fuzzy comprehensive evaluation method is used to quantitatively evaluate the operating status of each node, and output the health status score of each node in the physical topology network. Based on the health status score of each node and its priority in the physical topology network, the deadline for the reconstruction task of each node is determined. The reconstruction tasks are arranged in ascending order of deadline, and the earliest deadline first algorithm is used to schedule the reconstruction tasks, generate the physical topology network reconstruction instructions for the intelligent vehicle sensors, and execute them.

[0087] It should be noted that the weights of the evaluation factors are set based on the vehicle's historical operating data, and the values ​​typically range from 0.1 to 1.0.

[0088] Status feedback data includes node online status, communication latency, and data packet loss rate.

[0089] S4. Based on the node status of the reconstructed physical topology network, the original monitoring data is corrected twice to generate vehicle safety monitoring data.

[0090] Based on the node states (isolated / deweighted / normal states) of the reconstructed physical topology network, principal component analysis is used to extract the spatial coordinates and weight allocation parameters of the normal state nodes, and the node weight allocation table of the physical topology network is generated by weighted least squares method.

[0091] Furthermore, based on the node states (isolated / deweighted / normal states) of the reconstructed physical topology network, nodes in the normal state are selected using the state label matching method. The spatial coordinates and weight allocation parameters of the normal state nodes are collected, and a multidimensional data matrix is ​​formed from the spatial coordinates and weight allocation parameters. The multidimensional data matrix is ​​standardized, the covariance matrix is ​​calculated, and the eigenvalues ​​and eigenvectors are solved. Principal components with a cumulative contribution rate greater than the cumulative contribution rate threshold of the principal components are selected. The spatial coordinates and weight allocation parameters of the main feature directions are extracted through principal component analysis, retaining the information that has the greatest impact on the overall distribution. Then, the extracted spatial coordinates and weight allocation parameters are used as input, and corresponding initial weights are assigned according to the spatial distribution differences of each node. The relative reliability between nodes is optimized by weighted least squares method to generate the node weight allocation table of the physical topology network.

[0092] It should be noted that the weighting parameter refers to the value used to adjust the contribution and priority of each node's original monitoring data, which is determined based on the node's position in the physical topology network and its impact on the performance of the physical topology network.

[0093] The threshold for the cumulative contribution rate of principal components is defined by statistical analysis based on the variance distribution characteristics of the original monitoring data under historical normal operating conditions, and its value is usually in the range of 0.85 to 0.95.

[0094] Based on the node weight allocation table, the weighted residual correction method is used to calculate the deviation between the original monitoring data of each node and the weighted mean, and the original monitoring data is corrected for the first time according to the weight ratio to generate the initial vehicle safety monitoring data.

[0095] Furthermore, a weighted mean is calculated based on the original monitoring data of each node and its corresponding weight allocation parameters. Then, a weighted residual correction method is used to calculate the deviation between the monitoring data of each node and the corresponding weighted mean, so as to reflect the degree of deviation of the node's monitoring data from the overall average state. Then, the deviation is compensated according to the weight ratio of the node, that is, the monitoring data of the node is adjusted according to its importance to reduce the difference with the corresponding weighted mean. This completes the first correction of the monitoring data of all nodes, thereby generating the initial vehicle safety monitoring data.

[0096] The confidence weights of each initial vehicle safety monitoring data are statistically analyzed using a sliding window, and the spatiotemporal consistency of the confidence weights is verified, with invalid confidence weights marked.

[0097] Furthermore, by using a sliding window, the initial vehicle safety monitoring data are continuously sampled in the time dimension to form time series data segments. The statistical characteristics of the built-in confidence weights of each time series data segment are obtained, including mean, variance, and rate of change. At the same time, the confidence weights of adjacent sensor nodes at the same moment are compared in the spatial dimension, and the spatial distribution difference method is used to analyze the spatial distribution differences. Based on the preset spatiotemporal consistency threshold, it is determined whether the confidence weights of each initial vehicle safety monitoring data are continuous and stable in time and conform to the proximity relationship in space. If the confidence weights show abrupt changes in the time series or exceed the spatiotemporal consistency threshold range, the confidence weights are marked as invalid confidence weights.

[0098] It should be noted that the spatiotemporal consistency threshold is defined based on road environmental parameters and historical sensor performance data, including a time dimension threshold (typically ranging from 0.1 to 0.3) and a spatial dimension threshold (typically ranging from 0.15 to 0.25).

[0099] The iterative reweighted least squares method is used to perform secondary correction on the initial vehicle safety monitoring data where the confidence weights have failed, thereby generating vehicle safety monitoring data.

[0100] Furthermore, an iterative reweighted least squares method is adopted. The initial vehicle safety monitoring data with invalid confidence weights is used as input, the weights of each data point are initialized to the same value, and the weighted least squares estimate under the current weights is calculated to obtain the preliminary correction result. Based on the absolute value of the residuals between the preliminary correction result and the initial vehicle safety monitoring data obtained by the weighted residual correction method, the Huber weight function is used to update the weight of each data point. Data points with larger residuals are assigned lower weights. The weighted least squares calculation and weight update process is repeated until the weight change is less than the preset convergence threshold or the maximum number of iterations is reached, and finally the corrected vehicle safety monitoring data is output.

[0101] It should be noted that the convergence threshold is defined based on the mean and standard deviation of the L2 norm of the weight differences between two consecutive weights in the historical iteration calculation, and its typical range is: .

[0102] This embodiment also provides an intelligent vehicle sensor data correction system based on safety monitoring, including:

[0103] The data acquisition module collects raw monitoring data through intelligent vehicle sensors and performs preprocessing, calculates the point cloud density mutation rate, instantaneous signal-to-noise ratio drop, and data continuity interruption frequency, and generates security feature codes.

[0104] The anomaly diagnosis module classifies security feature codes into levels based on the security rule base, performs pattern matching in combination with the historical fault case database, and generates an anomaly diagnosis report of the original monitoring data.

[0105] The topology network adjustment module constructs the physical topology network of intelligent vehicle sensors, dynamically adjusts the node status of the physical topology network based on anomaly diagnosis reports, and obtains and executes physical topology network reconstruction instructions.

[0106] The data correction module corrects the original monitoring data twice based on the node status of the reconstructed physical topology network to generate vehicle safety monitoring data.

[0107] This embodiment also provides a computer device applicable to the intelligent vehicle sensor data correction method based on safety monitoring, comprising: a memory and a processor; the memory is used to store computer-executable instructions, and the processor is used to execute the computer-executable instructions to implement the intelligent vehicle sensor data correction method based on safety monitoring as proposed in the above embodiment.

[0108] The computer device can be a terminal, comprising a processor, memory, communication interface, display screen, and input devices connected via a system bus. The processor provides computing and control capabilities. The memory includes non-volatile storage media and internal memory. The non-volatile storage media stores the operating system and computer programs. The internal memory provides an environment for the operation of the operating system and computer programs stored in the non-volatile storage media. The communication interface is used for wired or wireless communication with external terminals; wireless communication can be achieved through Wi-Fi, carrier networks, NFC (Near Field Communication), or other technologies. The display screen can be an LCD screen or an e-ink screen. The input devices can be a touch layer covering the display screen, buttons, a trackball, or a touchpad on the computer device's casing, or an external keyboard, touchpad, or mouse.

[0109] This embodiment also provides a storage medium storing a computer program that, when executed by a processor, implements the intelligent vehicle sensor data correction method based on safety monitoring as proposed in the above embodiments. The storage medium can be implemented by any type of volatile or non-volatile storage device or a combination thereof, such as Static Random Access Memory (SRAM), Electrically Erasable Programmable Read-Only Memory (EEPROM), Erasable Programmable Read Only Memory (EPROM), Programmable Red-Only Memory (PROM), Read-Only Memory (ROM), magnetic storage, flash memory, magnetic disk, or optical disk.

[0110] In summary, this invention achieves accurate quantitative characterization of multi-dimensional abnormal states by employing multi-level threshold criteria to jointly analyze point cloud density mutation rate, instantaneous signal-to-noise ratio drop, and data continuity interruption frequency, and by generating structured security feature codes through binary bit-field encoding. This effectively improves the accuracy and efficiency of fault feature identification under complex operating conditions. Furthermore, by combining weighted residual correction with iterative reweighted least squares, and performing two-level correction on the monitoring data based on the reconstructed node weight allocation table, progressive error compensation under dynamic performance degradation of the sensor network is achieved, improving the reliability and environmental adaptability of the output data.

[0111] It should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not intended to limit it. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the spirit and scope of the technical solutions of the present invention, and all such modifications or substitutions should be covered within the scope of the claims of the present invention.

Claims

1. A method for correcting sensor data in intelligent vehicles based on safety monitoring, characterized in that: include, Raw monitoring data is collected by intelligent vehicle sensors and preprocessed to calculate the point cloud density mutation rate, instantaneous signal-to-noise ratio drop, and data continuity interruption frequency, thereby generating a security feature code. Based on the security rule base, the security feature codes are classified into levels, and combined with the historical fault case database, pattern matching is performed to generate an anomaly diagnosis report of the original monitoring data. Construct a physical topology network for intelligent vehicle sensors, dynamically adjust the node states of the physical topology network based on anomaly diagnosis reports, and obtain and execute physical topology network reconstruction instructions. Based on the node status of the reconstructed physical topology network, the original monitoring data is corrected twice to generate vehicle safety monitoring data.

2. The intelligent vehicle sensor data correction method based on safety monitoring as described in claim 1, characterized in that: The raw monitoring data includes the target object's spatial coordinates, velocity value, electromagnetic wave reflection intensity value, three-dimensional point cloud distribution, optical reflectivity value, time synchronization marker, as well as the distance measurement value of nearby obstacles, vehicle acceleration value, and angular velocity value.

3. The intelligent vehicle sensor data correction method based on safety monitoring as described in claim 2, characterized in that: The steps for calculating the point cloud density abrupt change rate, the instantaneous decrease in signal-to-noise ratio, and the frequency of data continuity interruptions are as follows. Based on the 3D point cloud distribution and time synchronization markers, the density abrupt change rate of the 3D point cloud distribution is calculated; Based on the density abrupt change rate and electromagnetic wave reflection intensity of the three-dimensional point cloud distribution, the instantaneous drop in signal-to-noise ratio is calculated, and the frequency of continuity interruption is statistically analyzed in conjunction with time synchronization markers.

4. The intelligent vehicle sensor data correction method based on safety monitoring as described in claim 1, characterized in that: The steps for classifying security feature codes based on a security rule base, performing pattern matching based on a historical fault case database, and generating an anomaly diagnosis report from the original monitoring data are as follows: Based on the abnormal status of historical monitoring data, anomaly judgment rules are defined, and the anomaly judgment rules are associated and mapped with the risk level labels in the historical fault case library to generate a safety rule library. Based on the security rule base, the security feature codes are classified into levels, and the fault types are matched from the historical fault case library according to the cosine similarity between the security feature codes and the historical fault feature vectors. An abnormal diagnosis report is generated and loaded into the historical fault case library.

5. The intelligent vehicle sensor data correction method based on safety monitoring as described in claim 4, characterized in that: The steps for constructing the physical topology network of intelligent vehicle sensors and dynamically adjusting the node states of the physical topology network based on anomaly diagnosis reports are as follows: Construct a physical topology network for intelligent vehicle sensors and dynamically adjust the node states of the physical topology network based on anomaly diagnosis reports; Based on the risk level in the anomaly diagnosis report, perform status marking on the physical topology network nodes, extract and encapsulate the node control parameters, and generate node control instructions. The node states of the physical topology network are dynamically adjusted according to node control commands.

6. The intelligent vehicle sensor data correction method based on safety monitoring as described in claim 5, characterized in that: The process of obtaining and executing physical topology network reconstruction instructions refers to collecting state feedback data of the adjusted physical topology network, predicting the health status score of each node in the physical topology network, and generating and executing physical topology network reconstruction instructions for intelligent vehicle sensors using the earliest deadline first algorithm.

7. The intelligent vehicle sensor data correction method based on safety monitoring as described in claim 6, characterized in that: The original monitoring data is corrected twice based on the node states of the reconstructed physical topology network to generate vehicle safety monitoring data. The steps are as follows: Based on the node states of the reconstructed physical topology network, the spatial coordinates and weight allocation parameters of the nodes in the normal state are extracted, and the node weight allocation table of the physical topology network is generated by weighted least squares method. Based on the node weight allocation table, the weighted residual correction method is used to calculate the deviation between the original monitoring data of each node and the weighted mean, and the original monitoring data is corrected for the first time according to the weight ratio to generate the initial vehicle safety monitoring data. Statistical analysis was performed on the confidence weights of each initial vehicle safety monitoring data, and the spatiotemporal consistency of the confidence weights was verified. Failed confidence weights were marked. The iterative reweighted least squares method is used to perform secondary correction on the initial vehicle safety monitoring data where the confidence weights have failed, thereby generating vehicle safety monitoring data.

8. The intelligent vehicle sensor data correction method based on safety monitoring as described in claim 1, characterized in that: The steps for generating the security feature code are as follows: Based on density mutation rate, instantaneous drop in signal-to-noise ratio, and frequency of data continuity interruption, abnormal states of raw monitoring data are identified. By using binary bit-field encoding, the abnormal state of the original monitoring data is mapped to a security feature code.

9. The intelligent vehicle sensor data correction method based on safety monitoring as described in claim 1, characterized in that: The preprocessing includes normalization, filtering and noise reduction, time synchronization calibration, and validity verification.

10. A smart car sensor data correction system based on safety monitoring, based on the smart car sensor data correction method based on safety monitoring according to any one of claims 1 to 9, characterized in that: include, The data acquisition module collects raw monitoring data through intelligent vehicle sensors and performs preprocessing, calculates the point cloud density mutation rate, instantaneous signal-to-noise ratio drop, and data continuity interruption frequency, and generates security feature codes. The anomaly diagnosis module classifies security feature codes into levels based on the security rule base, performs pattern matching in combination with the historical fault case database, and generates an anomaly diagnosis report of the original monitoring data. The topology network adjustment module constructs the physical topology network of intelligent vehicle sensors, dynamically adjusts the node status of the physical topology network based on anomaly diagnosis reports, and obtains and executes physical topology network reconstruction instructions. The data correction module corrects the original monitoring data twice based on the node status of the reconstructed physical topology network to generate vehicle safety monitoring data.

Images