Vehicle backup radar off-line detection system and method

Abstract

This application provides a vehicle reversing radar off-line detection system and method, including: a multi-bus data acquisition device, which synchronously acquires CAN, LIN, and Ethernet data streams during a standardized vehicle reversing process and configures time markers for each data stream based on the same hardware clock source; a spatiotemporal alignment determination module, which identifies the response signals of the same physical event in different buses, calculates the maximum timing deviation between time markers, and determines whether each data stream meets timing consistency requirements; a multimodal joint verification module, which, when timing consistency is met, performs joint determination on acoustic response, optical response, and image response characteristics based on the aligned time axis and outputs the detection result; and a response handling module, which triggers a vehicle interception command and maintenance scheduling process when the detection result is a failure. This application solves core problems such as the inability to synchronize cross-protocol data, the lack of acoustic-optical-image collaborative verification, the fact that faults are only located at the system level, and the lack of an automatic handling mechanism.

Description

Technical Field

[0001] This application relates to the field of vehicle off-line inspection technology, and in particular to a vehicle reversing radar off-line inspection system and method. Background Technology

[0002] Existing reversing radar EOL detection relies on manual operation or a single-protocol industrial control computer, which has three technical limitations: First, it only parses CAN bus messages and cannot simultaneously acquire LIN light signals and Ethernet video streams, making the timing relationship between the sound, light, and image systems unverifiable, and the true delay between brake light illumination and braking signal cannot be accurately measured; Second, the alarm sound only determines whether it is present or not, lacking a means to quantify the nonlinear gradual change characteristics of frequency with distance, and cannot identify abnormal pitch change rates; Third, visual indicators such as the curvature of reversing image auxiliary lines and obstacle mark recognition are out of sync with the actual vehicle movement state, and a dynamic consistency verification mechanism based on the actual reversing trajectory has not been established. Summary of the Invention

[0003] The purpose of this application is to provide a vehicle reversing radar off-line detection system and method to alleviate the aforementioned technical problems existing in the prior art.

[0004] In a first aspect, the present invention provides a vehicle reversing radar decommissioning detection system, comprising a multi-bus data acquisition device, a spatiotemporal alignment determination module, a multimodal joint verification module, and a response processing module; wherein... The multi-bus data acquisition device is used to synchronously acquire data streams from different communication buses during the standardized reversing process of the vehicle, and to configure time identifiers for each data stream based on the same hardware clock source. The spatiotemporal alignment determination module is used to identify the response signals of the same physical event in different data streams, calculate the maximum timing deviation between the time markers, and determine whether the timing consistency between each bus data stream is satisfied according to the preset allowable conditions. The multimodal joint verification module is used to perform feature joint determination on the features of acoustic response, optical response and image response respectively based on the aligned time identifier when the temporal consistency determination result is satisfied, so as to obtain the detection determination result. The response and handling module is used to trigger a vehicle interception command and maintenance scheduling process when the detection and judgment result is not passed.

[0005] In an optional implementation, the multi-bus data acquisition device is provided with multiple bus interfaces, each of which is connected to a different protocol bus in the vehicle network. The multi-bus data acquisition device integrates a high-precision hardware clock circuit. The hardware clock circuit provides a synchronization clock signal to each bus interface through a physical synchronization path. Each bus interface generates a time identifier based on the synchronization clock signal when acquiring data.

[0006] In an optional implementation, when the multimodal joint verification module performs joint determination of acoustic response features, it constructs an acoustic relationship model between obstacle distance and alarm sound characteristic frequency; the acoustic relationship model is used to characterize the variation law of alarm sound characteristic frequency increasing as obstacle distance decreases and approaching a stable value as obstacle distance increases; The multimodal joint verification module generates an ideal frequency sequence at the corresponding distance based on the obstacle distance data acquired in real time in the vehicle network and the acoustic relationship model.

[0007] In an optional implementation, the multimodal joint verification module performs frequency domain transformation on the measured alarm sound, extracts the dominant frequency component in each time segment, and forms a measured frequency sequence. The measured frequency sequence is dynamically matched with the ideal frequency sequence to obtain the matching distance value between the two. When the matching distance value is lower than the preset matching threshold, the overall shape of the sound response is deemed to be qualified.

[0008] In an optional implementation, the multimodal joint verification module obtains a measured gradient sequence by differentiating the measured frequency sequence and an ideal gradient sequence by differentiating the ideal frequency sequence. Calculate the correlation coefficient between the measured gradient sequence and the ideal gradient sequence; When the correlation coefficient is greater than a preset threshold, the dynamic change trend of the sound is deemed acceptable.

[0009] In an optional implementation, when the multimodal joint verification module performs joint determination of optical response features, it uses the state of a specific bit in the braking command message as a command validity flag and the state of a specific bit in the lighting status message as a lighting flag; the time interval between the transition of the command validity flag and the transition of the lighting flag is used as a response delay, and the duration for which the lighting flag remains in a valid state is recorded.

[0010] In an optional implementation, when the multimodal joint verification module performs joint determination of image response features, it performs edge detection and Hough linear transformation on the reversing image video stream, fits an auxiliary guide line, and calculates its curvature value. The target recognition model is invoked to identify preset obstacle markers in the video frame, and the confidence score of the recognition result is output. The absolute deviation of the curvature value from the standard curvature value, the confidence level value, and the preset threshold are all used as the criteria for judging the qualification of the imaging system.

[0011] In an optional implementation, the spatiotemporal alignment determination module includes: The hardware synchronization unit is used to distribute the same high-precision clock reference to each bus interface through a physical synchronization path, so that the time identifier of each bus data stream corresponds to a common hardware time source. An event-driven compensation unit is used to identify physical alignment events that generate deterministic response signals on multiple buses, and to calculate the time offset between buses based on the response time of the physical alignment events on each bus. The deviation monitoring and judgment unit is used to continuously track the changing trend of the time offset during the detection process, generate a health index characterizing the consistency of multi-bus timing, and make a pass / fail judgment on the health index according to preset allowable conditions.

[0012] In an optional implementation, the response handling module is further configured to: Receive the structured detection results output by the multimodal joint verification module, and parse the judgment status of each response dimension of the structured detection results; When the judgment status of any response dimension is "failed", a maintenance handling instruction uniquely corresponding to the failed dimension is automatically generated, and the maintenance handling instruction is sent to the factory production execution system through a preset communication path. The maintenance handling instruction includes at least: vehicle identification, fault location level, recommended maintenance action and designated handling station. The fault location level is determined by the multimodal joint verification module during the collaborative judgment process based on the time alignment relationship and signal transmission path of each bus data stream.

[0013] Secondly, the present invention provides a method for detecting the decommissioning of a vehicle reversing radar, comprising: During the standardized reversing operation of the vehicle, data streams from different communication buses are synchronously acquired through a multi-bus data acquisition device, and each data stream is configured with a time identifier based on the same hardware clock source. Based on the alignment event identification, the response signals of each data stream to the same physical event are identified, the maximum timing deviation between the time markers is calculated, and the timing consistency between each bus data stream is determined according to the preset allowable conditions. When the timing consistency determination result is satisfied, based on the aligned time identifier, the features characterizing the acoustic response, optical response and image response are jointly analyzed and collaboratively determined, and the detection determination result is output. If any response dimension fails in the detection and judgment results, a vehicle interception command and maintenance scheduling process are triggered.

[0014] The vehicle reversing radar offline detection system and method provided in this application introduces the same hardware clock source through a multi-bus data acquisition device, and synchronously configures time stamps for CAN, LIN and Ethernet bus messages, fundamentally eliminating the problem of time heterogeneity of cross-protocol data, and providing a unified and reliable time reference for the collaborative analysis of sound, light and shadow systems. Based on this, the spatiotemporal alignment judgment module can accurately identify the response signals of the same physical event (such as braking operation) in different buses and quantify the maximum timing deviation, thereby effectively distinguishing between network transmission delay and inherent module response lag, and supporting module-level fault location. The multimodal joint verification module, based on the aligned unified time axis, performs synchronous correlation analysis on features such as acoustic frequency gradient, optical illumination timing, and curvature of image auxiliary lines, making up for the blind spots of single-dimensional detection: on the one hand, by quantifying the nonlinear gradual change characteristics of sound frequency with distance, it identifies abnormal pitch change rate; on the other hand, it performs dynamic consistency verification between the reversing image auxiliary lines and the actual reversing trajectory, ensuring that the visual prompts match the actual movement state. The response and handling module automatically triggers interception and maintenance scheduling according to the judgment results, realizing a quality closed-loop process and significantly shortening the problem response cycle. In summary, this system effectively solves the core problems in existing technologies, such as the inability to synchronize cross-protocol data, the lack of collaborative verification of sound, light and shadow, the inability to locate faults only at the system level, and the lack of automatic handling mechanisms. Attached Figure Description

[0015] To more clearly illustrate the technical solutions in the specific embodiments of this application or the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of this application. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0016] Figure 1 A structural diagram of a vehicle reversing radar off-line detection system provided in this application embodiment; Figure 2 A standardized testing system architecture for vehicle reversing scenarios is provided in this application embodiment; Figure 3 This application provides a schematic diagram of the physical layout of a detection area. Figure 4 This application provides a specific implementation process for a vehicle reversing detection process. Figure 5 The workflow of a multi-protocol parsing engine provided in this application embodiment; Figure 6 A schematic diagram of lighting timing matching provided in an embodiment of this application; Figure 7This is a flowchart of a method for detecting the decommissioning of a vehicle reversing radar, provided in an embodiment of this application. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of the embodiments of this application clearer, the technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. The components of the embodiments of this application described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0018] Therefore, the following detailed description of the embodiments of this application provided in the accompanying drawings is not intended to limit the scope of the claimed application, but merely to illustrate selected embodiments of the application. All other embodiments obtained by those skilled in the art based on the embodiments of this application without inventive effort are within the scope of protection of this application.

[0019] It should be noted that similar labels and letters in the following figures indicate similar items. Therefore, once an item is defined in one figure, it does not need to be further defined and explained in subsequent figures.

[0020] This application provides a vehicle reversing radar decommissioning detection system, see [link]. Figure 1 As shown, the system includes a multi-bus data acquisition device, a spatiotemporal alignment determination module, a multimodal joint verification module, and a response processing module; among which, The multi-bus data acquisition device is used to synchronously acquire data streams from different communication buses during a standardized reversing process of a vehicle, and configures each data stream with a time stamp based on the same hardware clock source. The multi-bus data acquisition device is an on-board OBD data acquisition instrument integrating CANFD, LIN, and Ethernet interfaces. It has an internal hardware time synchronization unit, which acts as a slave clock for a precise time protocol, achieving high-precision time synchronization with the master clock deployed in the detection area. All bus messages, including radar distance signals on the CANFD bus, brake light status signals on the LIN bus, and reversing image video streams transmitted via Ethernet, are timestamped by the same hardware clock at the moment of acquisition, ensuring that data from different protocols have a unified, reliable, and comparable time reference.

[0021] The spatiotemporal alignment determination module is used to identify the response signals of the same physical event in different data streams, calculate the maximum timing deviation between time stamps, and determine whether the data streams of each bus meet the timing consistency requirements based on preset tolerance conditions. This module uses an emergency braking event as a typical alignment event: when the ESP module in the CAN bus sends a braking pressure message, the BCM in the LIN bus sends a brake light illumination signal, and a brake light illumination frame appears in the Ethernet video stream, the timestamps of the corresponding messages or frames are extracted, the absolute time deviations between each pair are calculated, and the maximum value is taken as the maximum observation time deviation for this acquisition. If this maximum deviation is less than a preset threshold (e.g., 10 milliseconds), the three-bus data are determined to meet the timing consistency requirements, allowing subsequent joint verification; otherwise, the detection data is marked as invalid to avoid misjudgments due to timing inaccuracies.

[0022] In one specific implementation, the aforementioned spatiotemporal alignment determination module can employ a cross-protocol message spatiotemporal alignment model based on hardware clock synchronization and software offset compensation to achieve a unified time base and reliable association for CAN, LIN, and Ethernet bus data. The core objective of this model is to control the systematic deviation and random jitter of timestamps from different bus protocol messages within a preset allowable range (e.g., the maximum observation time deviation is less than 10 milliseconds), ensuring that signals generated by the same physical event on different buses can be reconstructed to the same absolute time axis for collaborative analysis.

[0023] The model's inputs include: raw message streams from the onboard data acquisition unit, each message accompanied by a local timestamp under its respective bus clock domain; and preset alignment event identifiers, which are physical events capable of generating deterministic messages on multiple buses, such as test start signals, synchronization pulses, or emergency braking moments. The model's outputs include: global timestamps after all messages have been unified to a reference time (e.g., PTP master clock time); the current time offset and clock drift rate of each bus relative to the time reference; and the maximum observation time deviation used to determine data timing accuracy.

[0024] The model employs a hierarchical architecture. The first layer is hardware-level clock synchronization, implemented based on the IEEE 1588v2 protocol: the data acquisition unit acts as a PTP slave clock, synchronizing with the high-precision PTP master clock deployed in the detection area via an Ethernet interface. The CAN and LIN bus controllers integrated within the acquisition unit are synchronized with the PTP slave clock through hardware design or have a known fixed delay, thereby assigning a high-precision hardware timestamp based on the same clock source to all captured messages. The second layer is software-level message alignment, used to compensate for and verify residual time deviations caused by factors such as bus protocol network latency, message scheduling, and processing interruptions.

[0025] The core of software-layer alignment lies in using alignment events for offset estimation and compensation. Taking a typical alignment event—emergency braking—as an example: when the ESP module in the CAN bus sends a braking pressure message, the BCM in the LIN bus feeds back a brake light illumination signal, and a brake light illumination frame appears in the Ethernet video stream, the hardware timestamps t_CAN, t_LIN, and t_Eth are extracted respectively. The pairwise absolute deviations are calculated, and the maximum value is taken as the observation time deviation Δt_observed for this event. Throughout the detection cycle, a series of observation deviations are obtained by monitoring multiple alignment events, and then a drift model is fitted or a conservative maximum value is taken as the global maximum observation time deviation Δt_max for the current detection batch. For the globally reliable timestamp of any message, its uncertainty window is defined as hardware timestamp ± Δt_max / 2. When the timing accuracy meets a preset threshold (e.g., Δt_max < 10ms), this uncertainty can be ignored, and the hardware timestamp can be directly used for subsequent analysis.

[0026] The qualification criterion for this model is whether Δt_max is less than a preset engineering threshold. When the requirement is met, the multi-protocol data collected is considered to be time-aligned and reliable, and can be used for subsequent sound, light, and shadow collaborative verification; otherwise, the data is marked as time-unreliable, and the detection results are invalid. This model plays the following core roles in the present invention: First, it provides a time-reliable basis for multimodal joint verification, making the delay calculation between braking signals and light feedback comparable; second, it supports module-level fault attribution, distinguishing between network transmission delay and inherent module response lag through accurate time alignment, and locating the fault to a specific ECU or its software algorithm or hardware driver; third, the maximum observation time deviation itself serves as a system health indicator, which can be used to evaluate the vehicle network status and the performance of the acquisition system.

[0027] Furthermore, the aforementioned multimodal joint verification module is used to perform feature joint judgment on the acoustic response, optical response, and image response based on the aligned time markers when the timing consistency judgment result is satisfied, to obtain the detection judgment result. Under the premise that the maximum observation time deviation meets the preset requirements, this module performs three collaborative analyses: First, acoustic response verification, which involves performing spectrum analysis on the alarm audio signal collected by the microphone, extracting the instantaneous main frequency change sequence over time, matching it with the theoretical sequence generated by the ideal frequency change model constructed based on the obstacle distance, and calculating the overall similarity and correlation of the two sequences to double-judge whether the standard is met; Second, optical response verification, which involves extracting the effective time of the braking signal in the CAN bus and the effective time of the brake light illumination state in the LIN bus, calculating the time difference between the two, and requiring that both the delay and the illumination duration meet the preset indicators; Third, image response verification, which involves performing image recognition on the reversing image video stream, extracting auxiliary calibration lines and fitting their curvature, while simultaneously detecting preset obstacle markers in the image, and judging whether the curvature deviation and recognition confidence are within the allowable range. If all three verifications pass, the module outputs a "detection passed" conclusion; if any one of them fails, the module outputs "detection failed".

[0028] The response and handling module is used to trigger vehicle interception commands and maintenance scheduling processes when a detection judgment result is received as failing. This module receives the structured judgment results output by the multimodal joint verification module, parses the fault type and location information (e.g., brake light delay exceeding a preset threshold, indicating BCM response timeout), and automatically generates a structured work order containing the VIN code, fault code, and suggested maintenance actions. On one hand, it sends an interception command to the AGV scheduling system through the central control APP to guide the vehicle to the designated maintenance bay; on the other hand, it pushes the work order to the workshop MES system to start a closed-loop maintenance process, ensuring timely response and handling.

[0029] In summary, this system ensures the uniformity of spatiotemporal benchmarks for multi-source data through hardware-level time synchronization, achieves reliable timing determination through event-driven deviation quantization, conducts joint verification of three-dimensional features of sound, light, and shadow based on aligned time axes, and directly connects the determination results to the production execution system. It constructs a complete closed loop for EOL detection from data acquisition, collaborative analysis, fault attribution to automatic handling, and solves key problems in existing technologies such as protocol fragmentation, isolated verification, coarse positioning, and delayed response.

[0030] For ease of understanding, the specific components of the system will be described in detail below.

[0031] The aforementioned multi-bus data acquisition device is equipped with multiple bus interfaces, each connecting to different protocol buses in the vehicle network. Specifically, the device features three physically isolated bus interfaces: a CANFD interface (compliant with ISO11898-1:2015 standard, supporting 5 Mbps), a LIN interface (compliant with LIN 2.2A protocol, baud rate 19.2 kbps), and an Ethernet interface (compliant with IEEE 802.3bw 100BASE-T1 standard, bandwidth 100 Mbps). These interfaces are respectively connected to the power domain CANFD bus (carrying key signals such as radar distance and brake pressure), the body domain LIN bus (carrying actuator status signals such as lights and buzzers), and the intelligent driving domain vehicle Ethernet (carrying 1280×720@30fps reversing camera video stream). Each interface is connected to the main control unit via an independent transceiver to avoid electrical crosstalk and protocol conflicts.

[0032] The multi-bus data acquisition device integrates a high-precision hardware clock circuit. This circuit provides a synchronization clock signal to each bus interface via a physical synchronization path. Each bus interface generates a time stamp based on the synchronization clock signal when acquiring data. This hardware clock circuit is implemented based on the IEEE 1588v2 protocol, and the timestamps of all messages and video frames are generated by this unified hardware clock with an accuracy of 1 μs. During standardized reversing operations, each bus interface directly latches the current hardware clock count as a time stamp at the moment the hardware interrupt for data acquisition is triggered, ensuring that data from different protocols have a consistent, reliable, and comparable time reference.

[0033] Furthermore, when performing joint determination of acoustic response features, the aforementioned multimodal joint verification module constructs an acoustic relationship model between obstacle distance and alarm sound characteristic frequencies. This acoustic relationship model is an exponentially decaying mathematical expression: f(d) = k1·e -k ² d +b Where d is the obstacle distance in meters, and its value is derived from the radar ranging data parsed in real time from the CANFD bus message with ID 0x305; f(d) is the ideal characteristic frequency of the alarm sound at the corresponding distance in Hertz, representing the main frequency or the rhythmic frequency of the "beep" sound; k1 is the initial amplitude coefficient, reflecting the frequency increase at the closest distance; k2 is the attenuation coefficient, which determines the gradient of frequency change with distance; b is the reference frequency, representing the steady-state frequency at long distances; k1, k2, and b are all fixed parameters determined by vehicle model calibration and stored in the model configuration library of the cloud analysis layer.

[0034] The aforementioned acoustic relationship model characterizes the variation of the alarm tone's characteristic frequency as the distance to the obstacle decreases, and then approaches a stable value as the distance to the obstacle increases. That is, when d decreases (the vehicle approaches the obstacle), e... -k ² d As the term increases, f(d) increases significantly, simulating the auditory characteristics of the alarm sound changing from slow to fast; when d increases (the vehicle moves away from the obstacle), e -k ² d As the value approaches 0, f(d) approaches b, resulting in a smooth, gradual prompting effect; the derivative of this model is f′(d) = k1k2e -k ² d The absolute value of is the frequency gradient, which itself increases exponentially as d decreases, accurately depicting the nonlinear gradual change law of "the closer, the steeper", which conforms to the human-machine interaction design specifications of reversing radar.

[0035] The multimodal joint verification module generates ideal frequency sequences for corresponding distances based on real-time obstacle distance data and acoustic relationship models acquired from the vehicle network. This module continuously parses the distance value d from the message with ID 0x305 on the CANFD bus, forming a measured distance sequence {d1, d2, …, d...} in chronological order. n}; Put each d i Substituting into f(d)=k1·e -k ² di +b, calculate the ideal frequency sequence {f(d1), f(d2), …, f(d)} n This sequence serves as the benchmark for acoustic verification. Subsequently, it is compared with the measured frequency sequence extracted after the sound is collected by the microphone and processed by Mel spectrum transformation. The overall matching degree is compared using the dynamic time warping algorithm, and the gradient sequence correlation coefficient of the two sequences is calculated respectively. If both criteria are met, the acoustic response is deemed qualified.

[0036] Furthermore, the aforementioned multimodal joint verification module performs frequency domain transformation on the measured alarm sound, extracting the dominant frequency components within each time segment to form a measured frequency sequence. In specific implementation, the raw audio signal collected by the vehicle-mounted microphone is received at a sampling rate of 48 kHz and a bit depth of 16 bits. The audio signal is divided into frames with a frame length of 20 ms and a frame shift of 10 ms. After adding a Hanning window to each frame, a short-time Fourier transform (STFT) is performed to obtain the time spectrum. The time spectrum is then mapped to a Mel frequency scale to generate a Mel spectrum diagram. Subsequently, on each frame of the Mel spectrum, the location of the energy peak is searched along the frequency axis, and its corresponding actual physical frequency is taken as the dominant frequency component of that frame, thereby generating a measured frequency sequence {f1, f2, …, f} arranged in chronological order. n}

[0037] The measured frequency sequence is dynamically matched with the ideal frequency sequence to obtain the matching distance value between them. In one implementation, the dynamic matching process can employ the Dynamic Time Warping (DTW) algorithm, using the measured frequency sequence as the query sequence and the acoustic relationship model f(d) = k1·e -k ² d +b combines the ideal frequency sequence generated by the real-time distance sequence as a reference sequence; the DTW algorithm finds the optimal path to minimize the overall deformation cost of the two sequences by nonlinearly stretching or compressing the time axis, and outputs the normalized matching distance value dist, which is used to correct the deviation between the measured sound and the ideal sound in the overall trend of change.

[0038] When the matching distance value is lower than the preset matching threshold, the overall shape of the sound response is deemed acceptable. This preset matching threshold is determined based on vehicle model calibration data, with a value range of 0.15 to 0.25 (normalized DTW distance unit). This threshold is configured in the cloud analysis layer and updated with the vehicle model version. If the calculated normalized matching distance value is less than this threshold, it indicates that the overall frequency change pattern of the measured alarm sound meets the standard requirements and satisfies the "overall similarity" judgment condition. This judgment result only indicates that the shape is acceptable; it still needs to be combined with the gradient consistency rule (i.e., the gradient correlation coefficient between the two sequences is >0.95) to form a dual judgment conclusion.

[0039] Furthermore, the aforementioned multimodal joint verification module obtains the measured gradient sequence by differentiating the measured frequency sequence and the ideal gradient sequence by differentiating the ideal frequency sequence. In practical applications, the measured frequency sequence is a discrete-time sequence {f1, f2, …, f} extracted from the Mel spectrum. n The time interval is 10 ms (corresponding to frame shift); a first-order forward difference is performed on it, i.e., Δf i =f i+1 f i The measured gradient sequence {Δf1, Δf2, …, Δf} is obtained. n-1 The ideal frequency sequence is derived from the acoustic relational model f(d) = k1·e -k ² d +b combines the distance sequence {d1, d2, …, d} obtained in real-time from the CANFD bus ID 0x305 message. n The ideal gradient sequence {Δ} is generated, sampled at a time step of 10 ms, and subjected to the same forward differencing to obtain the desired gradient sequence. 1, Δ 2, …,Δ n-1}

[0040] Calculate the correlation coefficient between the measured gradient sequence and the ideal gradient sequence. For example, use the Pearson correlation coefficient formula to calculate the degree of linear correlation between the two gradient sequences: r = cov(Δf, Δf). ) / [σ(Δf)·σ(Δ ]], where cov represents the covariance and σ represents the standard deviation; the range of this coefficient is [ [1, 1] is used to characterize the synchronicity and directional consistency of the changing trends of two sequences.

[0041] When the correlation coefficient is greater than a preset threshold, the dynamic change trend of the sound is deemed acceptable. In a specific example, the preset threshold can be set to 0.95. This value is fixed in the cloud analysis layer and originates from the statistical analysis results of the gradient response characteristics of a large number of qualified sample vehicles' alarm sounds during the vehicle calibration phase. If the calculated r > 0.95, it indicates that the frequency change rate of the measured alarm sound (i.e., the nonlinear process of increased abruptness) is highly consistent with the exponential decay law described by the ideal model, thus satisfying the gradient consistency judgment condition. This judgment, together with the aforementioned DTW matching distance judgment, constitutes a dual rule, and both must be satisfied simultaneously for the sound response to be deemed acceptable.

[0042] When the aforementioned multimodal joint verification module performs joint determination of optical response characteristics, it uses the state of a specific bit in the braking command message as a valid command flag and the state of a specific bit in the lighting status message as a lighting flag. The braking command message originates from the CAN bus message with ID 0x112. Bit 7 (corresponding to mask 0x80) of this message is used to indicate whether the brake pedal is depressed: when this bit changes from 0 to 1, it is considered that a braking command has been issued. The lighting status message originates from the LIN bus message with ID 0x2A. Bit 3 (corresponding to mask 0x08) of this message is used to indicate whether the brake light is actually lit: when this bit changes from 0 to 1, it is considered that the light is lit in response.

[0043] The time interval between the instruction valid flag transition and the light-on flag transition is used as the response delay, and the duration for which the light-on flag remains valid is recorded. In practical applications, after detecting a 0x112 message bit 7 transition from 0 to 1 at time t1, the status of 0x2A message bit 3 is continuously monitored; when the first 0x2A message bit 3 transition from 0 to 1 is detected, its corresponding timestamp t2 is recorded, and the response delay Δt = t2 is calculated. t1; then continue tracking the continuous period when bit3 remains 1, recording the duration from its start time t2 to the moment it first falls back to 0 t3, i.e., the duration T = t3. t2.

[0044] Furthermore, when the aforementioned multimodal joint verification module performs joint determination of image response features, it performs edge detection and Hough line transform on the reversing image video stream, fits the auxiliary guide line, and calculates its curvature value. In specific implementation, the reversing image video stream transmitted via Ethernet is received, with a resolution of 1280×720 and a frame rate of 30 fps; each frame of the image is first preprocessed by grayscale conversion and Gaussian filtering, and then edge detection is performed using the Canny algorithm to extract high-contrast contours in the image; within the preset ROI (region of interest), the Hough line transform is applied to identify multiple straight line segments corresponding to the reversing auxiliary calibration line, and a cubic spline curve is fitted based on the least squares method; finally, the curvature value is calculated based on this curve.

[0045] The system uses a target recognition model to identify preset obstacle markers in video frames and outputs the confidence score of the recognition result. It utilizes a pre-trained YOLOv5 target detection model to identify preset standard obstacle markers (such as test cones, targets, etc.) in video frames. The model outputs the position coordinates of each detection box and the corresponding category confidence score, taking the highest confidence score as the confidence score of the recognition result. This model is deployed in a cloud-based analysis layer, with the input being a raw RGB image without compression or downsampling to ensure recognition accuracy.

[0046] The absolute deviation of the curvature value from the standard curvature value, the confidence level, and the preset threshold are all used as criteria for judging the quality of the imaging system. The standard curvature value is the theoretical value determined by the calibration of this vehicle model. The absolute deviation |Δκ| between the measured curvature and the standard curvature is calculated, and this deviation is less than 5%. At the same time, it is judged whether the confidence level of the YOLOv5 output is higher than 0.9. The imaging response is judged to be qualified only when both conditions are met: |Δκ| < 5% and confidence level > 0.9.

[0047] Furthermore, the aforementioned spatiotemporal alignment determination module includes: The hardware synchronization unit distributes the same high-precision clock reference to each bus interface via a physical synchronization path, ensuring that the time signatures of each bus data stream correspond to a common hardware time source. This hardware synchronization unit is implemented based on the IEEE 1588v2 protocol, with its core being a PTP slave clock circuit. It performs hardware-level time synchronization with the PTP master clock deployed in the detection area via an Ethernet interface. After synchronization, the unit uses low-jitter PCB routing to deliver a unified 10 MHz reference clock and a 1PPS pulse signal to the local clock inputs of the CANFD interface, LIN interface, and Ethernet video acquisition path, respectively. Each bus interface directly latches the current PTP count value as a time signature at the moment a hardware interrupt is triggered when a message or video frame is captured.

[0048] The event-driven compensation unit identifies physical alignment events that generate deterministic response signals on multiple buses and calculates the time offset between buses based on the response times of the physical alignment events on each bus. The event-driven compensation unit uses an emergency braking event as a typical physical alignment event: when the vehicle performs the third stage of emergency braking in a standardized reversing procedure, a brake signal bit 7 transition is detected in the CANFD bus ID 0x112 message, a light illumination bit 3 transition is detected in the LIN bus ID 0x2A message, and a brake light illumination frame is detected in the Ethernet video stream (confirmed by brightness abrupt changes and area identification). The corresponding timestamps t_CAN, t_LIN, and t_Eth are extracted respectively, and the pairwise difference Δt_CAN_LIN = |t_CAN t_LIN|, Δt_LIN_Eth=|t_LIN t_Eth|, Δt_Eth_CAN=|t_Eth t_CAN|; The deviation monitoring and judgment unit continuously tracks the changing trend of time offset during the detection process, generates a health index characterizing the timing consistency of the multi-bus system, and judges the pass / fail status of this health index according to preset tolerance conditions. Within a single detection cycle (e.g., 30 seconds), this unit continuously monitors no fewer than 5 alignment events, records the three time differences calculated each time, and takes the maximum value as the observation time deviation Δt for that detection; Δt is defined as the health index of the multi-bus data for this detection; the preset tolerance condition can be set to Δt < 10 ms. If the Δt corresponding to all alignment events meets this condition, the acquired multi-bus data is judged to have spatiotemporal consistency and is allowed to proceed to subsequent joint verification; otherwise, the data is marked as invalid.

[0049] Furthermore, the aforementioned response handling module is further used for: The system receives the structured detection results output by the multimodal joint verification module and parses the judgment status of each response dimension of the structured detection results. The structured detection results include three dimensions: radar sound gradation, brake light response, and reversing image. Each dimension is encapsulated in JSON format, with fields including dimension name, judgment status, measurement value, threshold, unit, confidence level, and global PTP timestamp to ensure unambiguity during the parsing process.

[0050] When any response dimension's status is deemed "failed," a repair instruction uniquely corresponding to that failed dimension is automatically generated and sent to the factory's production execution system via a preset communication path. The "failed" determination here employs a mechanism where a single-dimensional failure triggers an interception. Each fault dimension is bound to a specific handling logic, with the following mapping relationships: sound gradation failure is associated with a software algorithm defect in the radar control module; brake light delay is associated with a body control module (BCM) response timeout; and image curvature deviation is associated with camera calibration parameters or the image processing unit. The preset communication path involves the central control APP directly connecting to the AGV scheduling system via an encrypted channel. The communication protocol complies with vehicle network security standards, and the end-to-end transmission delay meets the engineering constraints of the repair response.

[0051] The maintenance and handling instructions include at least: vehicle identification number (VIN), fault location level, recommended maintenance action, and designated handling station. The vehicle identification number uses a VIN code, which is collected in real-time by the OBD system and encrypted to embed the instruction. The fault location level is determined by the multimodal joint verification module based on the time alignment of the bus data stream and the signal transmission path: for example, under the premise of timing consistency, if the delay between the braking signal and the brake light illumination exceeds a threshold, it is attributed to the internal response of the BCM module; if the radar distance and audio spectrum comparison are abnormal, it is located to the audio generation algorithm of the radar control module. The recommended maintenance action is generated based on the fault location level and references the maintenance knowledge base code to call the standard operating procedure (SOP); the designated handling station is preset according to the AGV scheduling system topology, and the AGV automatically plans the path after receiving the instruction.

[0052] In summary, this system performs cross-protocol message spatiotemporal alignment based on hardware time synchronization, achieving a unified time base and reliable association for CAN, LIN, and Ethernet data, providing a foundation for collaborative verification of multiple systems including sound, light, and image. Through acoustic modeling and dynamic spectrum analysis, it quantifies the frequency gradient characteristics of radar alarm sounds as they change with distance. A two-dimensional timing matching mechanism of control signals and execution states is employed to achieve module-level fault attribution for lighting responses. Combining computer vision and target detection technologies, it simultaneously verifies the geometric accuracy of reversing image guide lines and the reliability of obstacle recognition. The detection logic has evolved from single-dimensional judgment to multi-modal joint judgment, and fault location has extended from the system level to the ECU software algorithm or hardware driver. The system connects cloud-side analysis, vehicle-side execution, and production line handling links, automatically generating structured maintenance instructions and directly connecting to the production execution system, improving detection coverage, diagnostic accuracy, and maintenance response efficiency, forming a data-driven, process-controlled intelligent detection mode.

[0053] In another embodiment, see Figure 2 As shown, a standardized testing system for vehicle reversing scenarios is provided. The system specifically includes a vehicle-side execution layer, a cloud-based analysis layer, and a central control APP layer.

[0054] The vehicle-side execution layer is configured to execute a preset standardized reversing procedure. Specifically, the vehicle starts from a distance of 3 meters from the obstacle, with its initial position determined based on UWB (Ultra-Wideband) positioning technology, which provides extremely high accuracy of up to ±1 cm. Throughout the reversing process, the vehicle speed is controlled to not exceed 5 km / h, and speed stability is ensured by a PID controller with a control accuracy error within ±0.2 km / h. This standardized reversing procedure includes three consecutive execution phases: the first phase is straight-line reversing, where the vehicle travels backward in a straight line for 2 meters; the second phase is offset reversing, where the vehicle reverses at a fixed 15° deflection angle, causing the trajectory to deviate from the initial straight line; and the third phase is emergency braking, where the system triggers emergency braking to bring the vehicle to a complete stop. In addition, the vehicle-side execution layer is also equipped with a multi-protocol data acquisition device. This acquisition device has an OBD interface to support CAN FD, Ethernet and LIN protocols, can achieve 100% throughput message capture and timestamp accuracy of 1μs, and supports dual-mode redundant wireless transmission of Wi-Fi 6 and 5G. At the same time, it has built-in AES-256 and TLS 1.x encryption modules to ensure data security.

[0055] The cloud-based analysis layer includes a multi-protocol parsing engine and a sound gradient detection algorithm module. The sound gradient detection algorithm employs a Mel-spectrum-based Dynamic Time Warping (DTW) method to convert the acquired actual alarm sound signal into a Mel-spectrum. The DTW algorithm then calculates the "distance" between the measured sound's Mel-spectrum and a preset ideal alarm sound gradient curve. This distance quantifies the similarity between the measured and ideal sounds in terms of overall pattern and trend. The algorithm further employs a dual-judgment rule: Rule 1 is an overall similarity threshold; the calculated DTW distance must be less than a preset threshold to ensure the overall alarm sound pattern is basically correct. Rule 2 is gradient change consistency; the correlation coefficient between the gradient sequence of the measured sound spectrum and the gradient sequence of the ideal sound must be calculated, and this coefficient must be greater than 0.95 to ensure the dynamic change process of the alarm sound is highly consistent with the ideal standard. The detection result is considered passed only when both conditions of overall shape similarity and highly consistent trend are met simultaneously.

[0056] The central control APP is configured to dynamically display the test results and generate comprehensive decisions. Specifically, the displayed content includes details of each test item: the radar sound gradient status is "passed," indicating that the system detected that the frequency change of the alarm sound meets the standard with a confidence level of 98.2%; the brake light response status is "failed," indicating a critical fault—a 320-millisecond delay in the brake signal, which exceeds the safety threshold; the reversing camera status is "passed," indicating that the imaging system is working normally, and the deviation between its auxiliary calibration line and the standard position is extremely small, only 0.3 pixels. Based on the above test results, especially due to the failure of the "brake light response" function, the system automatically generates an instruction requiring the vehicle to enter the interception zone for repairs. The road test can only continue or the vehicle can be released after the fault has been resolved.

[0057] Therefore, this invention executes a standardized reversing procedure through high-precision positioning and speed control on the vehicle side, combines cloud-based multi-protocol parsing and sound gradation dual detection algorithms, and generates a complete vehicle reversing function testing and fault interception system through comprehensive judgment and decision-making by the central control APP layer.

[0058] Based on the above system architecture, this application also provides a specific detection process, which will be described in detail below.

[0059] To achieve the aforementioned standardized reversing procedure and multi-dimensional functional testing, this invention has set up specialized hardware facilities in the testing area. For example... Figure 3 As shown, a simulated obstacle wall is placed at the core of the test area. This obstacle wall serves as a virtual obstacle and detection benchmark during the vehicle's reversing process. Its surface or perimeter is equipped with precise distance scales to calibrate the vehicle's starting position and measure distances, ensuring that the vehicle can start accurately from a distance of 3 meters from the obstacle.

[0060] The simulated obstacle wall integrates multiple high-precision sensors to automate the determination of vehicle responses. Specifically, these include: Sound barometer array: This array uses an A-weighted network, achieving a measurement accuracy of up to ±0.5dB, to accurately capture the alarm sound signal emitted by the vehicle during reversing. The collected data will be transmitted to the cloud analysis layer as the raw input for the sound gradient detection algorithm to verify whether the alarm sound conforms to the preset gradient curve.

[0061] High frame rate camera: This camera is configured to capture the brake light status at the rear of the vehicle at a rate of 120 frames per second (fps) to accurately measure the response time from when the vehicle issues a braking command to when the brake lights actually illuminate. This configuration provides the hardware basis for detecting abnormal latency (such as an over-threshold latency of 320 milliseconds) in the "Brake Light Response" item.

[0062] In the test of the moving line planning, the vehicle first enters the designated radar detection area to start the reverse procedure. Subsequently, the vehicle reverses towards the simulated obstacle wall strictly according to the preset three-stage actions (straight reverse, offset reverse, emergency braking). After the entire test process is completed, the vehicle drives out of the moving area to complete this detection.

[0063] As Figure 4 shown, the specific implementation process of this system during the vehicle reverse detection is as follows: After the vehicle enters the detection area in autonomous driving, the multi-protocol data acquisition instrument synchronously collects the CAN, LIN, and Ethernet messages of the whole vehicle in real time and uploads them to the cloud in real time; at the same time, the sound pressure meter array and the high-speed camera built in the detection area synchronously collect the radar alarm sound and the reverse image video stream. After receiving the multi-source data, the cloud performs collaborative analysis based on the message spatio-temporal alignment model, and successively completes the joint verification of the radar ranging function, the radar sound gradual change characteristic, the brake light response timing sequence, and the curvature of the reverse image auxiliary line and obstacle recognition. The cloud outputs the detection result and determines whether it is qualified: If all the detection items pass, the vehicle automatically enters the next process; if any item is unqualified, the vehicle is automatically diverted to the repair area to complete the detection closed-loop. The background system supports the viewing and downloading of the whole vehicle messages, and the messages are all attached with the accurate acquisition time for easy traceability and analysis.

[0064] Figure 5 shows the working process of a multi-protocol parsing engine. First, the CAN message, LIN message, and Ethernet message are accessed in parallel; among them, the CAN message is used for the analysis of the radar sound frequency to extract the frequency gradual change characteristics related to the distance; the LIN message is used for the decoding of the light state to obtain the real-time state of the execution components such as the brake light; the Ethernet message is used for the extraction of the video stream characteristics to identify the curvature of the reverse image auxiliary line and the obstacle identification. The above three analysis results are synchronously input into the function coordination judgment module, which performs multi-modal joint verification on the radar sound, light response, and reverse image based on the time reference provided by the message spatio-temporal alignment model, and finally outputs the coordination judgment results of each response dimension to support the subsequent fault attribution and maintenance decision-making.

[0065] Further, when performing the light timing matching, as Figure 6 shown, first parse the brake pedal opening signal from the CAN message to extract the effective moment of the brake signal; parse the light state from the LIN message to extract the lighting moment of the brake light. Based on the unified time reference provided by the message spatio-temporal alignment model, calculate the time delay between the two signals and determine whether the delay is less than the preset threshold (for example, 100 milliseconds), and at the same time verify whether the lighting duration of the brake light meets the requirements (for example, not less than 500 milliseconds). If both conditions are met, it is determined that the light response timing is qualified; otherwise, it is determined as unqualified. This algorithm is the core part of the optical response verification in the multi-modal joint verification, and realizes the module-level fault attribution of the light system through accurate timing matching.

[0066] Based on the above system description, this embodiment provides a method for detecting the offline status of a vehicle reversing radar. (See [link]). Figure 7 As shown, the method mainly includes the following steps: During the standardized reversing operation of the vehicle, the S710 synchronously acquires data streams from different communication buses through a multi-bus data acquisition device and configures a time identifier for each data stream based on the same hardware clock source.

[0067] The S720 identifies the response signals of the same physical event in each data stream based on the alignment event, calculates the maximum timing deviation between time markers, and determines whether the timing consistency between each bus data stream is satisfied according to preset tolerance conditions.

[0068] When the timing consistency determination result is satisfied, S730 performs joint analysis and collaborative determination on the features representing acoustic response, optical response and image response based on the aligned time markers, and outputs the detection determination result.

[0069] S740: When any response dimension fails in the detection and judgment results, a vehicle interception command and maintenance scheduling process are triggered.

[0070] Based on the above scheme, this embodiment provides a specific implementation example. Taking the detection of a pure electric vehicle as an example, the vehicle automatically drives to the starting position of the detection area (3.0 m from the obstacle wall) and reverses at a speed of 4.8 km / h, simultaneously collecting CAN FD messages (ID 0x305, radar distance signal), LIN messages (ID 0x2B, light status), and Ethernet video streams (1280×720@30fps). After the multimodal joint verification module completes data analysis in the cloud, it detects two anomalies: First, the radar alarm sound gradually fails, and at a distance of 1.2 m from the obstacle, the measured frequency gradient deviates from the ideal model by 38%; second, the braking signal and brake light illumination delay reaches 320 ms, exceeding the allowable threshold of 100 ms. After verifying the reliability of the timing based on the message spatiotemporal alignment model, the diagnosis is that the radar control module algorithm is faulty and the body control module (BCM) response timeout. Based on this, the central control APP generates a handling instruction, including the vehicle's VIN code (SN2035X7), the final conclusion "unqualified," the designated interception station (station 3 in zone B), and repair suggestions: refresh the radar control module software and check the BCM power line impedance. This instruction is sent to the AGV scheduling system and MES system through a preset communication path, guiding the vehicle into the rework area and initiating the closed-loop repair process.

[0071] Unless otherwise specifically stated, the relative steps, numerical expressions, and values ​​of the components and steps described in these embodiments do not limit the scope of this application.

[0072] If the aforementioned functions are implemented as software functional units and sold or used as independent products, they can be stored in a processor-executable, non-volatile, computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or a portion of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as USB flash drives, portable hard drives, read-only memory (ROM), random access memory (RAM), magnetic disks, or optical disks.

[0073] In the description of this application, it should also be noted that, unless otherwise expressly specified and limited, the terms "set up," "install," "connect," and "link" should be interpreted broadly. For example, they can refer to a fixed connection, a detachable connection, or an integral connection; they can refer to a mechanical connection or an electrical connection; they can refer to a direct connection or an indirect connection through an intermediate medium; and they can refer to the internal connection of two components. Those skilled in the art can understand the specific meaning of the above terms in this application based on the specific circumstances.

[0074] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of this application, and are not intended to limit them. Although this application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that modifications can still be made to the technical solutions described in the foregoing embodiments, or equivalent substitutions can be made to some or all of the technical features therein. Such modifications or substitutions do not cause the essence of the corresponding technical solutions to deviate from the scope of the technical solutions of the embodiments of this application.

Claims

1. A vehicle reversing radar off-line detection system, characterized in that, It includes a multi-bus data acquisition device, a spatiotemporal alignment determination module, a multimodal joint verification module, and a response processing module; among which, The multi-bus data acquisition device is used to synchronously acquire data streams from different communication buses during the standardized reversing process of the vehicle, and to configure time identifiers for each data stream based on the same hardware clock source. The spatiotemporal alignment determination module is used to identify the response signals of the same physical event in different data streams, calculate the maximum timing deviation between the time markers, and determine whether the timing consistency between each bus data stream is satisfied according to the preset allowable conditions. The multimodal joint verification module is used to perform feature joint determination on the features of acoustic response, optical response and image response respectively based on the aligned time identifier when the temporal consistency determination result is satisfied, so as to obtain the detection determination result. The response and handling module is used to trigger a vehicle interception command and maintenance scheduling process when the detection and judgment result is not passed.

2. The system according to claim 1, characterized in that, The multi-bus data acquisition device is equipped with multiple bus interfaces, each of which is connected to a different protocol bus in the vehicle network. The multi-bus data acquisition device integrates a high-precision hardware clock circuit. The hardware clock circuit provides a synchronization clock signal to each bus interface through a physical synchronization path. Each bus interface generates a time identifier based on the synchronization clock signal when acquiring data.

3. The system according to claim 1, characterized in that, When the multimodal joint verification module performs joint determination of acoustic response features, it constructs an acoustic relationship model between obstacle distance and alarm sound characteristic frequency; the acoustic relationship model is used to characterize the change law of alarm sound characteristic frequency increasing as obstacle distance decreases and approaching a stable value as obstacle distance increases; The multimodal joint verification module generates an ideal frequency sequence at the corresponding distance based on the obstacle distance data acquired in real time in the vehicle network and the acoustic relationship model.

4. The system according to claim 3, characterized in that, The multimodal joint verification module performs frequency domain transformation on the measured alarm sound, extracts the main frequency component in each time segment, and forms the measured frequency sequence. The measured frequency sequence is dynamically matched with the ideal frequency sequence to obtain the matching distance value between the two. When the matching distance value is lower than the preset matching threshold, the overall shape of the sound response is deemed to be qualified.

5. The system according to claim 4, characterized in that, The multimodal joint verification module obtains the measured gradient sequence by differentiating the measured frequency sequence and the ideal gradient sequence by differentiating the ideal frequency sequence. Calculate the correlation coefficient between the measured gradient sequence and the ideal gradient sequence; When the correlation coefficient is greater than a preset threshold, the dynamic change trend of the sound is deemed acceptable.

6. The system according to claim 1, characterized in that, When the multimodal joint verification module performs joint determination of optical response features, it uses the state of a specific bit in the braking command message as the command validity flag and the state of a specific bit in the light status message as the light illumination flag; the time interval between the transition of the command validity flag and the transition of the light illumination flag is used as the response delay, and the duration for which the light illumination flag remains in a valid state is recorded.

7. The system according to claim 1, characterized in that, When the multimodal joint verification module performs joint determination of image response features, it performs edge detection and Hough linear transformation on the reversing image video stream, fits the auxiliary guide line and calculates its curvature value. The target recognition model is invoked to identify preset obstacle markers in the video frame, and the confidence score of the recognition result is output. The absolute deviation of the curvature value from the standard curvature value, the confidence level value, and the preset threshold are all used as the criteria for judging the qualification of the imaging system.

8. The system according to claim 1, characterized in that, The spatiotemporal alignment determination module includes: The hardware synchronization unit is used to distribute the same high-precision clock reference to each bus interface through a physical synchronization path, so that the time identifier of each bus data stream corresponds to a common hardware time source. An event-driven compensation unit is used to identify physical alignment events that generate deterministic response signals on multiple buses, and to calculate the time offset between buses based on the response time of the physical alignment events on each bus. The deviation monitoring and judgment unit is used to continuously track the changing trend of the time offset during the detection process, generate a health index characterizing the consistency of multi-bus timing, and make a pass / fail judgment on the health index according to preset allowable conditions.

9. The system according to claim 1, characterized in that, The response handling module is further used for: Receive the structured detection results output by the multimodal joint verification module, and parse the judgment status of each response dimension of the structured detection results; When the judgment status of any response dimension is "failed", a maintenance handling instruction uniquely corresponding to the failed dimension is automatically generated, and the maintenance handling instruction is sent to the factory production execution system through a preset communication path; The maintenance and handling instructions include at least: vehicle identification, fault location level, recommended maintenance action, and designated handling station; wherein, the fault location level is determined by the multimodal joint verification module during the collaborative judgment process based on the time alignment relationship and signal transmission path of each bus data stream.

10. A method for detecting the decommissioning of a vehicle reversing radar, characterized in that, include: During the standardized reversing operation of the vehicle, data streams from different communication buses are synchronously acquired through a multi-bus data acquisition device, and each data stream is configured with a time identifier based on the same hardware clock source. Based on the alignment event identification, the response signals of each data stream to the same physical event are identified, the maximum timing deviation between the time markers is calculated, and the timing consistency between each bus data stream is determined according to the preset allowable conditions. When the timing consistency determination result is satisfied, based on the aligned time identifier, the features characterizing the acoustic response, optical response and image response are jointly analyzed and collaboratively determined, and the detection determination result is output. If any response dimension fails in the detection and judgment results, a vehicle interception command and maintenance scheduling process are triggered.

Images