Self-calibration portable speedometer and zero debugging layout method thereof

Abstract

The present application relates to a kind of self-calibration portable speedometer and its zero debugging layout method, the speedometer includes: radar-video space-time synchronization module, for establishing the time synchronization of hardware level of millimeter wave radar and image sensor;Self-positioning-self-ranging module, by projecting parallel laser strip to road surface and combining perspective transformation, the installation pose parameter is solved;Joint external parameter self-calibration module, utilize the radar speed and image displacement data of natural passing vehicle, six-degree-of-freedom external parameter matrix is solved by least square method;Adaptive light supplement decision module, based on license plate gray closed-loop PID control adjusts light intensity;Edge QoS feedback module, real-time evaluation multidimensional quality index and control equipment state.The present application realizes the rapid deployment without manual measurement and physical calibration object, not only shortens the deployment time, improves speed measurement precision and night license plate recognition rate, and has edge real-time quality evaluation function, eliminates "sick law enforcement".

Description

Technical Field

[0001] This invention belongs to the field of intelligent traffic monitoring technology, specifically relating to a self-calibrating portable speed measuring instrument and its zero-adjustment deployment method. Background Technology

[0002] Portable speed measuring devices are core equipment in traffic enforcement and temporary speed enforcement scenarios. Their technological development has gone through three stages:

[0003] The first generation of mechanical calibration equipment relied on manual measurement of the equipment's installation position and angle using tools such as tape measures and levels. This resulted in large measurement errors, long setup times, and difficulty in use in environments without lane markings or at night.

[0004] The second-generation semi-automatic calibration model introduces laser rangefinders to assist in ranging, but still requires manual input of external parameters. Furthermore, radar calibration and image calibration are processed separately, and a unified spatiotemporal benchmark has not been established. The supplementary lighting intensity is fixed, resulting in severe overexposure or underexposure at night. There is a lack of equipment-side quality assessment mechanisms, posing a risk of "enforcing regulations with defects."

[0005] The third-generation edge intelligence uses a pre-trained model for license plate recognition, but still relies on calibration boards or closed roads for external parameter calibration. It has low multi-sensor time synchronization accuracy (>1ms) and weak adaptive capability under complex lighting conditions.

[0006] Patent CN109597037A discloses a radar calibration method and apparatus. This method and apparatus require setting a 15cm × 15cm physical calibration plate on the radar calibration line and calculating the offset by detecting the coordinates of the calibration points. This technology relies on a physical calibration plate, making it unsuitable for the dynamic deployment scenarios of portable speedometers, and it does not address the time synchronization and joint calibration of video sensors.

[0007] Patent CN112083387A discloses a radar calibration method and apparatus. This method and apparatus require multiple corner reflectors to be arranged around the radar, and GPS data is used to verify the radar data. This technology requires the deployment of multiple corner reflectors, making on-site operation complex and unable to achieve rapid deployment. It also does not involve the joint calibration of radar and video.

[0008] In summary, existing technologies generally suffer from problems such as high reliance on manual labor, inconsistent spatiotemporal benchmarks, poor environmental adaptability, and uncontrollable quality, and cannot meet the needs for fast, accurate, and reliable speed measurement in complex and dynamic environments such as factory areas, industrial park construction sections, and temporary law enforcement points. Summary of the Invention

[0009] The purpose of this invention is to overcome the shortcomings of the prior art and provide a self-calibrating portable speed measuring instrument and its zero-adjustment deployment method. This invention can achieve rapid deployment and high-precision speed measurement without manual measurement or physical calibration, and can evaluate the quality of the entire process in real time at the device end.

[0010] To achieve the above objectives, the present invention adopts the following technical solution: a self-calibrating portable speed measuring instrument, comprising:

[0011] The radar-video time-space synchronization module is used to establish hardware-level time synchronization between the millimeter-wave radar and the image sensor after the device is started, so that the time deviation between the two does not exceed a preset first threshold.

[0012] The self-positioning-self-range measurement module includes a laser projection unit, which is used to project a preset active structured light onto the road surface. Based on the image analysis of the active structured light and combined with the perspective transformation principle, the installation pose parameters of the speed measuring instrument relative to the road surface are calculated.

[0013] The joint extrinsic parameter self-calibration module is used to solve the six-degree-of-freedom extrinsic parameter matrix between the millimeter-wave radar and the image sensor, based on the working relationship established by the radar-video spatiotemporal synchronization module and the self-localization-self-range measurement module, using radar detection data and video detection data from at least three naturally passing vehicles.

[0014] An adaptive supplementary lighting decision module is used to detect the license plate area in a video image in real time, and to adjust the luminous intensity of the supplementary lighting unit in a closed loop with the image grayscale value of the license plate area as the control target, so that the grayscale value of the license plate area is stabilized within a preset target range.

[0015] The edge QoS feedback module is used to evaluate multiple preset quality dimensions of the speed measuring device in real time during deployment and law enforcement, and control the working status of the speed measuring device or issue prompt information based on the evaluation results.

[0016] Furthermore, the radar-video spatiotemporal synchronization module includes:

[0017] The PTP master clock unit is used to receive the second pulse signal from GNSS and use it to calibrate the local oscillator to generate a high-precision reference clock.

[0018] The hardware timestamp unit is used to stamp the sampling time of the millimeter-wave radar and the frame exposure time of the image sensor with hardware timestamps based on the high-precision reference clock, respectively.

[0019] A phase-locked loop unit is used to lock the modulation signal period of the millimeter-wave radar to an integer multiple of the frame period of the image sensor.

[0020] Furthermore, the self-localization-self-range measurement module also includes a GNSS-RTK / IMU tightly coupled unit, used to acquire the geodetic coordinates and attitude information of the velocimeter;

[0021] The laser projection unit is a structured light projector used to project at least two parallel laser strips with a known spacing onto the road surface;

[0022] The self-localization-self-rangement module calculates the installation pose parameters by analyzing the ratio between the pixel spacing of the parallel laser stripe in the image and the known spacing, and combining the attitude information output by the GNSS-RTK / IMU tightly coupled unit; the installation pose parameters include the equipment installation height, pitch angle and lateral distance of the equipment from the lane line.

[0023] Furthermore, the parallel laser strip is generated by a 940nm VCSEL array line laser, and its power meets human eye safety standards.

[0024] Furthermore, the joint extrinsic parameter self-calibration module is specifically used for:

[0025] Collect radar speed sequences and pixel displacement sequences from images of naturally passing vehicles during their passage.

[0026] Based on the radar velocity sequence and pixel displacement sequence, a system of linear overdetermined equations is constructed regarding the six-degree-of-freedom extrinsic parameter matrix;

[0027] The overall least squares method is used to solve the linear overdetermined equations to simultaneously optimize radar measurement noise and video tracking noise, thereby obtaining the initial values ​​of the six-degree-of-freedom extrinsic parameter matrix.

[0028] Furthermore, the joint extrinsic parameter self-calibration module is also used for:

[0029] The point cloud detected by the millimeter-wave radar is projected onto the image plane using the solved six-degree-of-freedom extrinsic parameter matrix;

[0030] Calculate the reprojection error between the projected point and the corresponding video detection target;

[0031] When the reprojection error exceeds a preset second threshold, new natural traffic vehicle data is automatically collected and recalibrated.

[0032] Furthermore, the adaptive supplemental lighting decision module includes:

[0033] A deep learning detection unit is used to detect and locate license plate areas in video images in real time;

[0034] A PID controller is used to calculate the error between the actual grayscale value and the target grayscale value of the license plate area, and output a control quantity based on the error.

[0035] The PWM drive unit is used to adjust the pulse width modulation duty cycle of the supplementary light unit according to the control quantity output by the PID controller, thereby changing its luminous intensity.

[0036] Furthermore, the adaptive supplemental lighting decision module also includes a polarizing filter, which is disposed in front of the image sensor to suppress secondary reflection glare generated by the road surface or vehicle body.

[0037] Furthermore, the edge QoS feedback module performs real-time evaluation of several preset quality dimensions, including: GNSS positioning status, laser projection detection confidence, time synchronization deviation, joint calibration accuracy, and supplementary lighting control stability.

[0038] The operating status includes an enforcement-ready status indicating successful deployment, an early warning status indicating the presence of non-fatal anomalies, and an alarm status indicating deployment failure requiring manual intervention.

[0039] The present invention also provides a zero-adjustment deployment method for the above-mentioned self-calibrating portable speedometer, comprising the following steps:

[0040] Step S1: After the device is powered on, it projects active structured light onto the road surface through the self-positioning-self-range measurement module, calculates the installation pose parameters of the speed measuring instrument relative to the road surface, and establishes hardware-level time synchronization between the millimeter-wave radar and the image sensor through the radar-video spatiotemporal synchronization module.

[0041] Step S2: Based on the spatiotemporal reference and pose parameters established in Step S1, the six-degree-of-freedom extrinsic parameter matrix between the millimeter-wave radar and the image sensor is automatically solved and verified using data from at least three naturally passing vehicles through the joint extrinsic parameter self-calibration module.

[0042] Step S3: After successful calibration, the adaptive supplementary lighting decision module is activated. Through closed-loop control, the grayscale value of the detected license plate area is stabilized within the preset target range, and the system enters the law enforcement ready state.

[0043] Step S4: During deployment and enforcement, the edge QoS feedback module performs real-time evaluation of multiple preset quality dimensions and dynamically adjusts the working status based on the evaluation results.

[0044] Compared with the prior art, the present invention has the following beneficial effects:

[0045] 1. Zero manual measurement and rapid deployment: This invention generates virtual lane lines through active laser projection and combines perspective transformation inversion to completely eliminate manual measurement and physical calibration objects, reducing deployment time from more than 20 minutes in the prior art to within 30 seconds.

[0046] 2. High-precision spatiotemporal synchronization: This invention uses PTP hardware timestamp + PLL phase-locked loop to achieve sub-millisecond synchronization (≤0.1ms) between radar and video, laying the foundation for high-precision fusion speed measurement.

[0047] 3. Adaptive ambient lighting: This invention is based on closed-loop PID control of license plate grayscale to solve the problem of overexposure / underexposure in nighttime lighting, and improves the nighttime license plate recognition rate to over 99%.

[0048] 4. Joint calibration without calibration board: This invention uses naturally passing vehicles as mobile calibration objects and solves the 6-DOF extrinsic parameters through the TLS algorithm to achieve automatic joint calibration of radar and video, reducing the speed measurement error from ±8km / h to ±2km / h.

[0049] 5. Real-time edge quality assessment: This invention constructs a five-dimensional QoS assessment system at the device end. If the quality is not up to standard, it will be proactively reported, thus preventing "enforcement with defects" from the source. Attached Figure Description

[0050] Figure 1 This is a system architecture block diagram of a self-calibrating portable speedometer according to an embodiment of the present invention. Detailed Implementation

[0051] The present invention will be further described below with reference to the accompanying drawings and embodiments.

[0052] It should be noted that the following detailed descriptions are exemplary and intended to provide further explanation of this application. Unless otherwise specified, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains.

[0053] It should be noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to limit the exemplary embodiments according to this application. As used herein, the singular form is intended to include the plural form as well, unless the context clearly indicates otherwise. Furthermore, it should be understood that when the terms "comprising" and / or "including" are used in this specification, they indicate the presence of features, steps, operations, devices, components, and / or combinations thereof.

[0054] like Figure 1 As shown, this embodiment provides a self-calibrating portable speed measuring device, including: a radar-video spatiotemporal synchronization module, a self-localization-self-range measurement module, a joint extrinsic parameter self-calibration module, an adaptive supplementary lighting decision module, and an edge QoS feedback module. These modules cooperate to achieve full automation from device power-on to law enforcement readiness.

[0055] The radar-video spatiotemporal synchronization module is used to establish hardware-level time synchronization between the millimeter-wave radar and the image sensor after the device is started, so that the time deviation between the two does not exceed a preset first threshold (e.g., set to 0.1ms).

[0056] Specifically, the radar-video spatiotemporal synchronization module includes:

[0057] 1) The PTP master clock unit is used to receive the second pulse signal from GNSS and use it to calibrate the local oscillator (OCXO) to generate a high-precision reference clock;

[0058] 2) Hardware timestamp unit, used to stamp hardware timestamps (timestamp resolution up to 10ns) at the sampling time of the millimeter-wave radar and the frame exposure time of the image sensor, respectively, based on the high-precision reference clock.

[0059] 3) A phase-locked loop unit is used to lock the modulation signal period of the millimeter-wave radar (e.g., chirp period of 100μs) and the frame period of the image sensor (e.g., 33.3ms@30fps) into an integer multiple relationship (e.g., 333:1).

[0060] The self-positioning-self-range measurement module includes a laser projection unit and a GNSS-RTK / IMU tightly coupled unit. The laser projection unit projects a preset active structured light (dual-beam parallel laser strip) onto the road surface. Based on image analysis of the active structured light and combined with perspective transformation principles, the installation attitude parameters of the speedometer relative to the road surface are calculated. The GNSS-RTK / IMU tightly coupled unit acquires the geodetic coordinates and attitude information of the speedometer.

[0061] The self-localization-self-rangement module calculates the installation pose parameters by analyzing the ratio between the pixel spacing of the parallel laser stripe in the image and the known spacing, and by combining the attitude information output by the GNSS-RTK / IMU tightly coupled unit. The installation pose parameters include the device installation height, pitch angle, and lateral distance between the device and the lane line.

[0062] In this embodiment, the laser projection unit is specifically a structured light projector (such as a 940nm VCSEL array line laser with a divergence angle of 15°×1° and a power of 3W, which is used to project at least two parallel laser strips with a known spacing (such as a factory-calibrated spacing of 50cm) onto the road surface.

[0063] The joint extrinsic parameter self-calibration module is used to solve the six-degree-of-freedom extrinsic parameter matrix between the millimeter-wave radar and the image sensor based on the working foundation established by the radar-video spatiotemporal synchronization module and the self-localization-self-range measurement module, using radar detection data (radial velocity) and video detection data (pixel displacement) from at least three naturally passing vehicles, and through joint calibration algorithms such as the total least squares method (TLS).

[0064] Specifically, the joint extrinsic parameter self-calibration module is used for:

[0065] 1) Collect radar velocity sequences and pixel displacement sequences in images of naturally passing vehicles during their passage;

[0066] 2) Based on the radar velocity sequence and pixel displacement sequence, construct a system of linear overdetermined equations about the six-degree-of-freedom extrinsic parameter matrix;

[0067] 3) The overall least squares method is used to solve the linear overdetermined equations to simultaneously optimize radar measurement noise and video tracking noise, and to obtain the initial value of the six-degree-of-freedom extrinsic parameter matrix.

[0068] The joint extrinsic parameter self-calibration module is also used for:

[0069] 1) Project the point cloud detected by the millimeter-wave radar onto the image plane using the solved six-degree-of-freedom extrinsic parameter matrix;

[0070] 2) Calculate the reprojection error between the projected point and the corresponding video detection target;

[0071] 3) When the reprojection error is greater than the preset second threshold (e.g., set to 0.3 pixels), new natural traffic vehicle data will be automatically collected and recalibrated.

[0072] The adaptive fill light decision module is used to detect the license plate area in the video image in real time, and uses the image gray value of the license plate area as the control target to adjust the light intensity of the fill light unit (such as a strobe lamp) in a closed loop so that the gray value of the license plate area is stabilized within a preset target range (such as 110±5).

[0073] Specifically, the adaptive supplemental lighting decision module includes:

[0074] 1) Deep learning detection units (such as the YOLOv7-tiny network) are used to detect and locate license plate areas in video images in real time;

[0075] 2) A PID controller (parameters can be set to Kp=0.5, Ki=0.1, Kd=0.05) is used to calculate the error between the actual gray value and the target gray value of the license plate area, and output a control quantity based on the error;

[0076] 3) PWM drive unit, used to adjust the pulse width modulation duty cycle (e.g., 0-127 level) of the supplementary light unit according to the control quantity output by the PID controller, thereby changing its luminous intensity;

[0077] 4) A polarizing filter is placed in front of the image sensor to suppress secondary reflection glare from the road surface or vehicle body.

[0078] The edge QoS feedback module is used to evaluate multiple preset quality dimensions of the speed measuring device in real time during deployment and law enforcement, and control the working status of the speed measuring device or issue prompt information based on the evaluation results.

[0079] The preset quality dimensions include: GNSS positioning status (weight 20%), laser projection detection confidence (weight 20%), time synchronization deviation (weight 20%), joint calibration accuracy (weight 20%), and supplementary lighting control stability (weight 20%).

[0080] The working states include: a law enforcement ready state indicating successful deployment (green light always on), a warning state indicating a non-fatal anomaly (yellow light flashing), and an alarm state indicating deployment failure requiring manual intervention (red light flashing + NB-IoT reporting).

[0081] This embodiment also provides a zero-adjustment deployment method for the above-mentioned self-calibrating portable speedometer, including the following steps:

[0082] Step S1: After the device is powered on, it projects active structured light onto the road surface through the self-positioning-self-range measurement module, calculates the installation pose parameters of the speed measuring instrument relative to the road surface, and establishes hardware-level time synchronization between the millimeter-wave radar and the image sensor through the radar-video spatiotemporal synchronization module.

[0083] Step S2: Based on the spatiotemporal reference and pose parameters established in Step S1, the six-degree-of-freedom extrinsic parameter matrix between the millimeter-wave radar and the image sensor is automatically solved and verified using data from at least three naturally passing vehicles through the joint extrinsic parameter self-calibration module.

[0084] Step S3: After successful calibration, the adaptive supplementary lighting decision module is activated. Through closed-loop control, the grayscale value of the detected license plate area is stabilized within the preset target range, and the system enters the law enforcement ready state.

[0085] Step S4: During deployment and enforcement, the edge QoS feedback module performs real-time evaluation of multiple preset quality dimensions and dynamically adjusts the working status based on the evaluation results.

[0086] The specific implementation of the present invention will be further described in detail below through specific embodiments.

[0087] 1. Radar-video spatiotemporal synchronization module (sub-millisecond hardware synchronization)

[0088] Unlike existing technologies that use software-based or NTP network time synchronization, this invention employs a three-level synchronization mechanism: IEEE 1588v2 PTP protocol + hardware timestamp + PLL phase-locked loop.

[0089] PTP master clock: The local OCXO is disciplined by the PPS (Pulse Per Second) signal of GNSS-RTK to generate a 10MHz reference clock;

[0090] Hardware timestamp: Both the radar ADC sampling time and the CMOS frame exposure time are timestamped via FPGA, with a timestamp resolution of 10ns;

[0091] PLL (Phase-Locked Loop): Locks the radar chirp period (typically 100μs) to an integer multiple relationship with the video frame period (33.3ms@30fps), ensuring: TradarTvideo = 100μs 33.3ms = 333∈Z+

[0092] Synchronization accuracy verification: For a target of 120km / h (33.3m / s), a time deviation of 0.1ms results in a spatial deviation of 3.3mm, which is far lower than the radar range resolution (0.15m@77GHz), meeting the joint calibration accuracy requirements.

[0093] 2. Self-localization-self-distance measurement module (virtual lane line laser inversion)

[0094] Unlike existing technologies that rely on manual measurement with a measuring tape or fixed calibration objects, this invention proposes an "active structured light projection + perspective transformation inversion" mechanism:

[0095] Hardware configuration:

[0096] Dual-frequency GNSS-RTK (BeiDou B1I / B2a + GPS L1 / L2): Planar accuracy ±1cm, elevation accuracy ±2cm;

[0097] MEMS-IMU (6-axis): angular velocity noise <0.01° / s, acceleration noise <0.1mg;

[0098] 940nm VCSEL line laser: dual-beam parallel light, divergence angle 15°×1°, power 3W, Class 1 eye-safe.

[0099] Algorithm flow:

[0100] Laser projection: Projecting two parallel laser strips with a known spacing onto the road surface (factory-calibrated spacing d0=50cm);

[0101] Image acquisition: The laser stripe image is acquired by CMOS, and the pixel coordinates of the laser stripe are obtained by adaptive threshold segmentation and skeleton extraction;

[0102] Perspective transformation inversion:

[0103] Establish the Homography matrix H from the image coordinate system (u,v) to the road surface coordinate system (X,Y);

[0104] Using the proportional relationship between the actual laser stripe spacing d0 and the image spacing dpixel, the height H of the equipment above the road surface and the pitch angle θ are calculated: H = dpixelf⋅d0⋅cosθ, where f is the camera focal length;

[0105] Lane width calculation: Automatically calculates lane width L and distance D from the device to the nearest lane line by using the geometric relationship between the laser strip and the lane line (if any) or the curb.

[0106] Accuracy verification: Under typical installation height of 1.5m and elevation angle of 25°, the laser inversion distance accuracy is ±1cm and the angle accuracy is ±0.1°, which is 50 times better than manual measurement (±5°). This invention requires no physical calibration objects and creates "virtual lane lines" through active laser projection, completing self-distance measurement within 3 seconds.

[0107] 3. Joint external parameter self-calibration module (license plate grayscale constant closed loop)

[0108] Unlike existing technologies that rely on fixed power or ambient light sensing, this invention constructs a closed loop of "license plate area detection - grayscale evaluation - PWM modulation":

[0109] Perception front-end: YOLOv7-tiny lightweight network, 7ms single-frame inference after NNIE acceleration;

[0110] Detection target: License plate region (ROI), output bounding box (x,y,w,h);

[0111] Control Algorithm:

[0112] Controlled quantity: Average grayscale value of the license plate area (Gactual);

[0113] Setting value: Gtarget=110 (8-bit, corresponding to midtones, preserving license plate character contrast);

[0114] Control quantity: strobe light PWM duty cycle Dpwm∈[0,127] (128 levels, 0.78% increment);

[0115] PID controller: Dpwm(k)=Kp⋅e(k)+Ki⋅∑i=0ke(i)+Kd⋅[e(k)−e(k−1)] where e(k)=Gtarget−Gactual(k), and Kp=0.5, Ki=0.1, Kd=0.05 are tuned.

[0116] Anti-interference design:

[0117] Polarizing filter: A linear polarizer is installed in front of the CMOS sensor, with the angle orthogonal to the polarization direction of the strobe light, to suppress secondary reflection glare from the road surface / vehicle body;

[0118] Time-domain difference: When the grayscale change is greater than 20% for 3 consecutive frames, it is judged as interference, and the PWM value of the previous cycle is maintained.

[0119] Performance metrics: License plate grayscale control accuracy ±5, convergence time <100ms (3 frames @ 30fps), nighttime recognition rate ≥99%.

[0120] This invention enables adaptive control of supplemental lighting, and for the first time uses the grayscale of the license plate as the controlled variable to achieve microsecond-level PWM closed-loop control.

[0121] 4. Adaptive Illumination Decision Module (Natural Vehicle TLS Joint Calibration)

[0122] Unlike existing calibration plate methods or closed road methods, this invention proposes a joint calibration based on Total Least Squares of naturally passing vehicles:

[0123] Mathematical model:

[0124] Radar measurement: Target velocity vradar (radial);

[0125] Video measurement: Displacement of the target in the image Δp (pixels);

[0126] Geometric relationship: vradar = k⋅Δp + b, where k is the proportionality coefficient and b is the system deviation;

[0127] Extrinsic parameter matrix: 6-DOF rigid body transformation T=[R∣t], where R∈SO(3) is the rotation matrix and t∈R3 is the translation vector.

[0128] Calibration process:

[0129] Data collection: Automatically collect data from at least 3 naturally passing vehicles (5 recommended for improved robustness), and record data for each vehicle:

[0130] Radar: Point cloud velocity sequence {vi}i=1N;

[0131] Video: The pixel coordinate sequence of the center point of the license plate {(ui,vi)}i=1N;

[0132] Displacement calculation: The license plate pixel displacement Δp is calculated using optical flow or centroid tracking.

[0133] Construction of the overdetermined equation system: v1v2⋮vM=Δp1Δp2⋮ΔpM11⋮1[kb] where M≥3 is the number of vehicles;

[0134] TLS solution: Considering the noise present in both radar and video measurements, the total least squares method (SVD decomposition) is used to solve for T;

[0135] Reprojection verification: Project the radar point cloud onto the image plane and calculate the reprojection error: ereproj=N1∑i=1N∥pimagei−π(T⋅pradari)∥ where π(⋅) is the projection function, and the convergence condition is ereproj<0.3 pixel.

[0136] This invention requires no calibration plate, no road closures, and no human or computer intervention throughout the entire process, truly achieving "installation and calibration on demand." It utilizes naturally passing vehicles as "mobile calibration objects" to achieve self-calibration of radar-video combined external parameters.

[0137] 5. Edge QoS Feedback Module (Full-Process Confidence Assessment)

[0138] Unlike existing technologies that rely on post-event sampling inspections, this invention constructs a five-dimensional QoS evaluation system at the device level, achieving zero tolerance for "enforcement with defects":

[0139]

[0140] State machine design:

[0141] Idle: Power-on initialization;

[0142] Sensing: Completes spatiotemporal synchronization within 200ms;

[0143] Calibrating: Performs joint calibration;

[0144] Validating: QoS assessment;

[0145] Enforcing (green light always on): All passes, enter enforcement mode;

[0146] Alerting (red light flashing): A critical item has failed and requires manual intervention.

[0147] Communication protocol: NB-IoT reports JSON format status codes with a delay of <2s and supports remote diagnostics.

[0148] 6. System Integration Examples

[0149] Scenario: Temporary speed limit section on a factory road, without lane markings, nighttime lighting <0.1 lux, requiring rapid deployment of speed measuring devices.

[0150] The complete deployment process of this invention is detailed below.

[0151] Step 1: Power on the device (T=0s)

[0152] With a 12V / 20Ah lithium battery, the battery life is >8 hours;

[0153] Press the power button to start the device and automatically initialize all modules;

[0154] Status lights: Yellow light is always on (initializing).

[0155] Step 2: GNSS positioning (T=0-3s)

[0156] Connect an external GNSS antenna (magnetic attachment to the car roof) to receive BeiDou B1I / B2a and GPS L1 / L2 signals;

[0157] RTK base stations broadcast differential correction data via 4G networks;

[0158] Key algorithm: Dual-frequency GNSS-RTK + IMU tightly coupled, utilizing the short-time high-precision characteristics of IMU to assist attitude maintenance when GNSS loses lock;

[0159] Output data:

[0160] Planar coordinates: E=3951234.56m, N=12345678.90m (accuracy ±1cm)

[0161] Elevation: U = 15.2m (accuracy ±2cm)

[0162] Attitude angles: roll=2.1°, pitch=24.5°, yaw=185.3° (accuracy ±0.1°)

[0163] Error handling: If a fixed solution is not obtained within 3 seconds, it is marked as a floating-point solution mode, and the subsequent process continues but the precision requirement is reduced.

[0164] Step 3: Laser ranging (T=3-5s)

[0165] The 940nm VCSEL laser is activated with a drive current of 1.2A, projecting dual parallel laser beams onto the road surface;

[0166] Key Algorithm:

[0167] Adaptive threshold segmentation: The binarization threshold is dynamically adjusted according to the ambient light intensity to ensure robust laser strip extraction;

[0168] Skeleton extraction: The center line of the laser strip is extracted using the Zhang-Suen thinning algorithm, with sub-pixel accuracy;

[0169] Perspective transformation inversion:

[0170] The laser strip spacing is d0=50cm as specified at the factory.

[0171] The installation height is calculated using the pixel spacing dpixel of the laser stripe in the image and the camera focal length f=8mm: H=dpixelf⋅d0=1.52m

[0172] The consistency of laser inversion was verified by combining the IMU elevation angle of 24.5°;

[0173] Calculate the lane width L = 3.75m and the edge distance D = 0.85m.

[0174] Output data: H=1.52m, θ=24.5°, L=3.75m, D=0.85m (accuracy ±1cm);

[0175] Anomaly Handling: If the laser bar is blocked (confidence level < 0.9), the red light will flash, and the NB-IoT will report "laser bar not visible".

[0176] Step 4: Time synchronization (T=5-5.2s)

[0177] Configure the PTP master clock: Discipline the local OCXO with GNSS-PPS signals to generate a 10MHz reference clock;

[0178] Key algorithm: The PLL phase-locked loop locks the radar chirp period of 100μs with the video frame period of 33.3ms in a 333:1 integer multiple relationship;

[0179] Verification metric: timestamp deviation 0.05ms (<0.1ms threshold);

[0180] Error handling: If the synchronization deviation is >1ms, the yellow light will flash as a warning, and after 3 retries, the red light will indicate an error.

[0181] Step 5: Vehicle data collection (T=5.2-18s)

[0182] YOLOv7-tiny network detects vehicles in real time with a confidence threshold of 0.85;

[0183] The first vehicle (truck, speed 45km / h):

[0184] Radar data acquisition: 12 sampling points, velocity sequence {v_i}=[45.2, 45.1, 44.9, ..., 45.0] km / h;

[0185] Video capture: License plate center pixel coordinate sequence {(u_i,v_i)}, frame rate 30fps;

[0186] Displacement calculation: centroid tracking method, Δp = [12.5, 12.3, 12.4, ..., 12.5] pixels;

[0187] The second vehicle (sedan, speed 52km / h): 15 sampling points were collected;

[0188] The third vehicle (SUV, speed 38km / h): collected 10 sampling points;

[0189] Data alignment: Radar velocity is aligned to the video timestamp via linear interpolation to ensure timing consistency.

[0190] Step 6: TLS Configuration (T=18-23s)

[0191] Constructing an overdetermined system of equations:

[0192] A total of 37 valid sampling points were collected from 3 vehicles;

[0193] Matrix A (37×2): [Δp_i, 1];

[0194] Vector b (37×1): [v_i];

[0195] SVD decomposition solution:

[0196] The augmented matrix C=[A|b] is decomposed using SVD: C=UΣVᵀ;

[0197] The solution is the right singular vector corresponding to the minimum singular value σ_min = 0.023.

[0198] The proportionality constant k = 0.0234 and the deviation b = 0.12 are obtained.

[0199] Reconstructing the extrinsic parameter matrix:

[0200] Combine the attitude angle output by the GNSS-IMU with the [H,θ,D] retrieved by laser inversion;

[0201] Optimize the solution for the 6-DOF extrinsic parameter [R|t].

[0202] Reprojection verification:

[0203] The radar point cloud is projected onto the image and compared with the center of the measured license plate.

[0204] Calculate the reprojection error: e_reproj = 0.23 pixels (< 0.3 threshold, pass);

[0205] Error handling: If the error is >0.3 pixels, automatically return to step 5 to add 3 more vehicles for recalibration.

[0206] Step 7: QoS Assessment (T=23-28s)

[0207] Five-dimensional rating:

[0208] GNSS fixed solution: 20 points (obtained by fixed solution);

[0209] Laser inversion: 19 points (confidence level 0.96);

[0210] Time synchronization: 20 minutes (0.05ms < 0.1ms);

[0211] Joint calibration: 20 points (0.23px < 0.3px);

[0212] Fill light control: 18 points (grayscale convergence in progress);

[0213] Overall score: 97 points (>90 threshold, passed);

[0214] Status switch: Green light is always on, entering law enforcement mode;

[0215] Anomaly Handling: If the score is less than 90, the red light flashes, and the NB-IoT reports "Manual intervention required" in JSON format.

[0216] json

[0217] {

[0218] "device_id": "PSD2024XXXX",

[0219] "timestamp": "2024-01-15T09:30:00Z",

[0220] "state": "ALERTING",

[0221] "qos_score": 85,

[0222] "details": {

[0223] "gnss_fix": 1,

[0224] "laser_conf": 0.95,

[0225] "sync_offset_us": 0.5,

[0226] "calib_error_px": 0.35,

[0227] "lighting_stability": 0

[0228] },

[0229] "alert_code": 0x04,

[0230] "alert_desc": "Reprojection error exceeds the limit"

[0231] }

[0232] ```

[0233] Step 8: Enforcement Ready (T=28s)

[0234] - Activate adaptive fill light closed loop:

[0235] - YOLOv7-tiny detects license plate ROI with an inference time of 7ms;

[0236] - Initial PWM duty cycle D_pwm = 64 (50%);

[0237] - The measured grayscale value of the first frame is G_actual=85, with an error e=25;

[0238] - PID calculation: D_pwm=64+0.5×25=76.5→77;

[0239] - The measured grayscale value of the second frame is G_actual=98, with an error e=12;

[0240] - PID calculation: D_pwm=77+0.5×12+0.1×(25+12)×0.033=77+6+0.12=83.12→83;

[0241] - The measured grayscale value of the 3rd frame is G_actual=108, and the error is e=2;

[0242] - Convergence complete, grayscale stabilizes within the range of 110±5;

[0243] - Start speed enforcement:

[0244] - The radar continuously emits FMCW signals to detect vehicle speed;

[0245] - Continuous video capture, license plate recognition rate >99%;

[0246] - Data is stored locally and simultaneously transmitted back in real time via a 4G network.

[0247] Step 9: Continuous monitoring and anomaly handling (T>28s)

[0248] - Normal law enforcement status: Green light is always on, QoS score is consistently >90;

[0249] - 23:15 GNSS briefly lost lock (floating-point solution):

[0250] - Detection: GNSS fixed solution flag = 0;

[0251] - Handling: Yellow light flashing, continue enforcement, log the incident;

[0252] - Recovery: 23:20 GNSS re-fixed, green light remains on;

[0253] - 01:30 Laser overheating:

[0254] - Test: Laser NTC temperature = 62°C;

[0255] - Handling: Automatic power reduction by 50%, supplementary lighting PID adaptive adjustment, no impact;

[0256] - Recovery: When the temperature drops to 55°C, the power is restored to 100%.

[0257] Step 10: Withdrawal (End of Enforcement)

[0258] - One-click shutdown, automatically saving calibration parameters to local Flash;

[0259] - Upload law enforcement data to the cloud, including:

[0260] - Speed ​​measurement records: time, speed, license plate, and image;

[0261] - Device status: QoS score, anomaly logs;

[0262] - Total deployment time: 28 seconds (from power-on to law enforcement readiness);

[0263] - Speed ​​measurement accuracy during law enforcement: ±1.8km / h (actual measurement after calibration);

[0264] - Nighttime license plate recognition rate: 99.4% (27 vehicles were captured, 26 of which were recognized).

[0265] The above description is merely a preferred embodiment of the present invention and is not intended to limit the invention in any other way. Any person skilled in the art may make changes or modifications to the above-disclosed technical content to create equivalent embodiments. However, any simple modifications, equivalent changes, and modifications made to the above embodiments based on the technical essence of the present invention without departing from the scope of the present invention shall still fall within the protection scope of the present invention.

Claims

1. A self-calibrating portable speedometer, characterized in that, include: The radar-video time-space synchronization module is used to establish hardware-level time synchronization between the millimeter-wave radar and the image sensor after the device is started, so that the time deviation between the two does not exceed a preset first threshold. The self-positioning-self-range measurement module includes a laser projection unit, which is used to project a preset active structured light onto the road surface. Based on the image analysis of the active structured light and combined with the perspective transformation principle, the installation pose parameters of the speed measuring instrument relative to the road surface are calculated. The joint extrinsic parameter self-calibration module is used to solve the six-degree-of-freedom extrinsic parameter matrix between the millimeter-wave radar and the image sensor, based on the working relationship established by the radar-video spatiotemporal synchronization module and the self-localization-self-range measurement module, using radar detection data and video detection data from at least three naturally passing vehicles. An adaptive supplementary lighting decision module is used to detect the license plate area in a video image in real time, and to adjust the luminous intensity of the supplementary lighting unit in a closed loop with the image grayscale value of the license plate area as the control target, so that the grayscale value of the license plate area is stabilized within a preset target range. The edge QoS feedback module is used to evaluate multiple preset quality dimensions of the speed measuring device in real time during deployment and law enforcement, and control the working status of the speed measuring device or issue prompt information based on the evaluation results.

2. The self-calibrating portable speedometer according to claim 1, characterized in that, The radar-video spatiotemporal synchronization module includes: The PTP master clock unit is used to receive the second pulse signal from GNSS and use it to calibrate the local oscillator to generate a high-precision reference clock. The hardware timestamp unit is used to stamp the sampling time of the millimeter-wave radar and the frame exposure time of the image sensor with hardware timestamps based on the high-precision reference clock, respectively. A phase-locked loop unit is used to lock the modulation signal period of the millimeter-wave radar to an integer multiple of the frame period of the image sensor.

3. The self-calibrating portable speedometer according to claim 1, characterized in that, The self-positioning-self-range measurement module also includes a GNSS-RTK / IMU tightly coupled unit, used to acquire the geodetic coordinates and attitude information of the velocity measuring instrument; The laser projection unit is a structured light projector used to project at least two parallel laser strips with a known spacing onto the road surface; The self-localization-self-range measurement module calculates the installation pose parameters by analyzing the ratio between the pixel spacing of the parallel laser stripe in the image and the known spacing, and combining the attitude information output by the GNSS-RTK / IMU tightly coupled unit; the installation pose parameters include the equipment installation height, pitch angle and lateral distance of the equipment from the lane line.

4. The self-calibrating portable speedometer according to claim 3, characterized in that, The parallel laser strips are generated by a 940nm VCSEL array line laser, and their power meets human eye safety standards.

5. The self-calibrating portable speedometer according to claim 1, characterized in that, The joint extrinsic parameter self-calibration module is specifically used for: Collect radar speed sequences and pixel displacement sequences from images of naturally passing vehicles during their passage. Based on the radar velocity sequence and pixel displacement sequence, a system of linear overdetermined equations is constructed regarding the six-degree-of-freedom extrinsic parameter matrix; The overall least squares method is used to solve the linear overdetermined equations to simultaneously optimize radar measurement noise and video tracking noise, thereby obtaining the initial values ​​of the six-degree-of-freedom extrinsic parameter matrix.

6. The self-calibrating portable speedometer according to claim 5, characterized in that, The joint extrinsic parameter self-calibration module is also used for: The point cloud detected by the millimeter-wave radar is projected onto the image plane using the solved six-degree-of-freedom extrinsic parameter matrix; Calculate the reprojection error between the projected point and the corresponding video detection target; When the reprojection error exceeds a preset second threshold, new natural traffic vehicle data is automatically collected and recalibrated.

7. The self-calibrating portable speedometer according to claim 1, characterized in that, The adaptive supplemental lighting decision module includes: A deep learning detection unit is used to detect and locate license plate areas in real time from video images; A PID controller is used to calculate the error between the actual grayscale value and the target grayscale value of the license plate area, and output a control quantity based on the error. The PWM drive unit is used to adjust the pulse width modulation duty cycle of the supplementary light unit according to the control quantity output by the PID controller, thereby changing its luminous intensity.

8. The self-calibrating portable speedometer according to claim 7, characterized in that, The adaptive lighting decision module also includes a polarizing filter, which is positioned in front of the image sensor to suppress secondary reflection glare from the road surface or vehicle body.

9. The self-calibrating portable speedometer according to claim 1, characterized in that, The edge QoS feedback module performs real-time evaluation of several preset quality dimensions, including: GNSS positioning status, laser projection detection confidence, time synchronization deviation, joint calibration accuracy, and supplementary lighting control stability. The operating status includes an enforcement-ready status indicating successful deployment, an early warning status indicating the presence of non-fatal anomalies, and an alarm status indicating deployment failure requiring manual intervention.

10. A zero-adjustment deployment method for a self-calibrating portable speedometer according to any one of claims 1-9, characterized in that, Includes the following steps: Step S1: After the device is powered on, it projects active structured light onto the road surface through the self-positioning-self-range measurement module, calculates the installation pose parameters of the speed measuring instrument relative to the road surface, and establishes hardware-level time synchronization between the millimeter-wave radar and the image sensor through the radar-video spatiotemporal synchronization module. Step S2: Based on the spatiotemporal reference and pose parameters established in Step S1, the six-degree-of-freedom extrinsic parameter matrix between the millimeter-wave radar and the image sensor is automatically solved and verified using data from at least three naturally passing vehicles through the joint extrinsic parameter self-calibration module. Step S3: After successful calibration, the adaptive supplementary lighting decision module is activated. Through closed-loop control, the grayscale value of the detected license plate area is stabilized within the preset target range, and the system enters the law enforcement ready state. Step S4: During deployment and enforcement, the edge QoS feedback module performs real-time evaluation of multiple preset quality dimensions and dynamically adjusts the working status based on the evaluation results.

Images