Vehicle-mounted motor vehicle road violation identification instrument

Abstract

The invention relates to a vehicle-mounted motor vehicle road violation recognition instrument, and relates to the technical field of intelligent traffic. The device comprises a camera module integrated with a wide-angle lens and an infrared light supplement unit, wherein the camera module supports 1080P and 30fps high-definition video acquisition and dynamic exposure adjustment; the multi-band GPS / Beidou positioning module and the 9-axis IMU cooperate to realize continuous positioning; a deep learning algorithm module based on a CNN-RNN hybrid model accurately judges violation behaviors such as retrograde driving, lane changing without turning on a steering lamp, solid line lane changing and the like through YOLO target detection, FastSAM lane line segmentation and CRNN license plate recognition technologies; and a 4G / 5G and block chain encrypted data uploading module is arranged, so that reliable transmission of violation evidence packets is realized. According to the invention, a multi-modal data fusion mechanism is innovatively adopted, and a vehicle-mounted power supply and standby battery intelligent switching system is combined, so that all-weather violation monitoring can be realized in a complex road environment, high-confidence structured evidence is provided for traffic law enforcement, and violation identification accuracy and evidence chain integrity are effectively improved.

Description

Technical Field

[0001] The present invention relates to the field of intelligent transportation technology, in particular to a vehicle-mounted road violation identification device for motor vehicles. Background Art

[0002] Current traffic violation monitoring relies mainly on fixed cameras and manual inspections, which have problems such as large blind spots and delayed responses. Existing on-board recording equipment mostly focuses on accident evidence collection and lacks the ability to actively identify violations. In particular, the misjudgment rate is high in complex scenarios (such as changing lanes without using a turn signal, changing lanes on a solid line), and it is difficult to associate violations with accurate spatiotemporal information. Traditional algorithms are not adaptable enough to low light conditions, license plate obstructions, etc., resulting in an incomplete chain of evidence and restricting the efficiency of traffic law enforcement. In addition, the power supply and data transmission of on-board equipment are not stable enough, affecting the real-time and credibility of violation evidence. Therefore, there is an urgent need for an active on-board violation recognition system that integrates multi-source perception, high-precision positioning and intelligent analysis.

[0003] Therefore, a vehicle-mounted motor vehicle road violation identification device is needed to solve the above problems. Summary of the Invention

[0004] In order to solve the problems of the prior art, the present invention provides a vehicle-mounted motor vehicle road violation identification device.

[0005] In order to solve the above technical problems, the present invention is implemented through the following technical solutions: In a first aspect, a vehicle-mounted motor vehicle road violation identification device comprises:

[0006] The camera module is used to collect high-definition video images of the vehicle's driving environment in real time, supporting image capture in high resolution and low light conditions;

[0007] GPS positioning module, used to obtain the vehicle's real-time location, driving trajectory and speed information;

[0008] A deep learning algorithm module, based on a model combining a convolutional neural network (CNN) and a recurrent neural network (RNN), processes video data collected by the camera module to identify traffic violations and license plate information of target vehicles. Violations include at least one of driving against traffic, changing lanes without using a turn signal, and changing lanes on a solid line.

[0009] Memory module, used to store video data, violation determination results and GPS location information, and supports cyclic overwriting storage;

[0010] The power supply module, including the vehicle power adapter and backup battery, automatically switches the power source according to the vehicle ignition status to ensure continuous operation;

[0011] The data upload module uploads violation evidence (including pictures, video clips), license plate information and GPS location data to the cloud server or traffic management platform via the wireless network.

[0012] In this aspect, the specific configuration is:

[0013] A camera module, comprising at least one high dynamic range (HDR) image sensor, supporting a resolution of 1920×1080P or higher, a frame rate of 30-60fps, a wide-angle lens (field of view ≥140°) and an infrared LED fill light unit (wavelength 850nm), for continuously capturing video streams of the road environment in front of and to the sides of the vehicle, both day and night;

[0014] GPS positioning module, integrated with a multi-band satellite signal receiver (supporting GPS, Beidou, and GLONASS), with a positioning accuracy of ≤1.5 meters, real-time recording of vehicle latitude and longitude, speed, heading angle, and timestamp, and synchronized with the camera module's clock;

[0015] The deep learning algorithm module is deployed on an embedded processor (such as the ARM Cortex-A72 or an AI acceleration chip with equivalent computing power), including:

[0016] The traffic violation recognition unit, based on the deep learning model of Yolo and FastSAM network architecture, detects and tracks vehicles in the video stream and identifies violations such as lane changes without using the turn signal, lane changes on a solid line, driving against traffic, running a red light, and speeding.

[0017] The license plate recognition unit uses a CRNN (convolutional recurrent neural network) combined with a CTC loss OCR model to extract license plate characters from the video (supporting Chinese characters, letters, numbers, and province abbreviations);

[0018] The multimodal data fusion module associates the violation's timestamp, video clips (5 seconds before and 10 seconds after the violation), license plate recognition results, GPS location, and vehicle speed data into a structured evidence package.

[0019] Memory module, using eMMC 5.1 or SSD storage media, with a capacity of ≥128GB, supporting H.265 encoded video compression storage and cyclic overwriting strategy (overwriting the earliest data in chronological order);

[0020] The power supply module includes an onboard 12V / 24V DC-DC voltage regulator circuit and a removable lithium polymer battery (capacity ≥ 5000mAh). The battery provides power for at least 8 hours of standby after the vehicle is turned off.

[0021] Data upload module, integrated with 4G / 5G modem (compatible with Cat.12 and above) and Wi-Fi 6 wireless communication unit, uploads evidence packages to the cloud server via HTTPS protocol encryption, supports breakpoint resuming and multi-threaded concurrent transmission

[0022] In a specific embodiment of the first aspect, the camera module includes a wide-angle lens and an infrared fill light unit, supports a resolution of 1080P or higher, has a frame rate of not less than 30fps, and has a dynamic exposure adjustment function to adapt to lighting changes;

[0023] The camera module further includes:

[0024] Polarizing filter to eliminate windshield reflections that interfere with image quality;

[0025] Fisheye lens distortion correction algorithm, which converts 140° field of view video into perspective projection view through bilinear interpolation;

[0026] Dynamic exposure control logic automatically adjusts sensor gain and shutter speed based on ambient light intensity (0.1 lux to 100,000 lux).

[0027] In a specific embodiment of the first aspect, the GPS positioning module has a positioning accuracy of less than 3 meters and is integrated with a speed sensor to synchronously record vehicle speed data at the time of the violation;

[0028] GPS positioning module:

[0029] Integrated 9-axis MEMS inertial measurement unit (IMU), including a 3-axis gyroscope, accelerometer, and magnetometer, for maintaining positioning continuity through dead reckoning when satellite signals are lost;

[0030] Built-in high-precision differential positioning (RTK) interface, which can connect to external base station data to improve positioning accuracy to centimeter level;

[0031] Real-time matching of vehicle position and lane data on electronic maps to determine the lane attributes where the violation occurred (such as whether it was on a solid line or in an emergency lane).

[0032] In a specific implementation of the first aspect, the deep learning algorithm module further includes:

[0033] License plate recognition submodule, which extracts license plate characters through OCR technology;

[0034] The behavior analysis submodule determines whether the vehicle violates traffic regulations based on video time series data, and verifies the compliance of the lane and road section where the violation occurred in combination with GPS location information.

[0035] In a specific implementation of the first aspect, the data upload module supports multiple network protocols (4G / 5G, Wi-Fi), and temporarily stores data in the memory module when the network connection is interrupted, and automatically resumes the data after the network is restored;

[0036] Data upload module:

[0037] After a violation is detected, metadata in JSON format is automatically generated, and the evidence package is hashed and encrypted using blockchain technology to ensure that the data cannot be tampered with. Priority upload strategies based on traffic management areas are supported (e.g., violations in school areas are transmitted first).

[0038] In a specific embodiment of the first aspect, the system further includes a user interaction module for providing real-time feedback of violation alerts to the driver and providing a local preview of violation evidence and a complaint portal; specifically, the system includes a touch screen (7 inches, capacitive) for displaying camera images, violation alerts (audio and visual prompts), and historical evidence packages in real time;

[0039] Voice interaction unit, which allows drivers to query violation records through natural language (e.g., "retrieve the most recent lane change violation");

[0040] The appeal interface allows drivers to add voice or text comments on misjudged violations and submit review requests through the data upload module.

[0041] In a specific embodiment of the first aspect, the device is integrated into the vehicle in the form of a vehicle-mounted rearview mirror or a standalone device, and is powered by an OBD interface or a cigarette lighter power supply; a detachable bracket is fixed to the inside of the vehicle windshield, and the bracket has a built-in cooling fan (speed ≤ 25dB);

[0042] Supports external radar or LiDAR modules via the USB-C interface to enhance the detection accuracy of vehicle distance and speed;

[0043] Provides API interfaces for third-party applications to call, including real-time violation data streams, electronic fence warnings, and driving behavior scores.

[0044] In a second aspect, a method for monitoring motor vehicle violations using a vehicle-mounted motor vehicle road violation identification device comprises the following steps:

[0045] S1: Real-time acquisition of vehicle-surrounding video streams and GPS location data; hardware clock alignment between the GPS module and the camera module ensures that each image frame is associated with millisecond-accurate timestamps and location data.

[0046] S2: Identify traffic violations and license plates of other vehicles or the vehicle itself through a deep learning model; segment the video stream into an image sequence at 30 frames per second and input it into the deep learning model; perform target detection (output vehicle bounding box and turn signal status), lane line segmentation (output solid line / dashed line classification), and license plate recognition on each frame; based on the detection results of multiple consecutive frames, determine whether the vehicle has completed the entire illegal action (such as starting and ending a lane change on a solid line).

[0047] S3: Associate the violation with the GPS location and timestamp to generate a structured evidence package. For violations, extract the video clip containing the complete action process (from the first 5 seconds to the last 10 seconds) and overlay the GPS trajectory layer and violation type label.

[0048] S4: Automatically upload evidence packages to the cloud platform for review by traffic management departments; dynamically adjust the resolution (720P to 4K) and upload frequency of evidence packages based on the severity of the violation (e.g., running a red light is high-risk, while changing lanes without turning on the signal is ordinary) and network bandwidth status. BRIEF DESCRIPTION OF THE DRAWINGS

[0049] Figure 1 It is a schematic diagram of the overall system architecture of the present invention.

[0050] Figure 2 It is a schematic diagram of the hardware composition of the present invention.

[0051] Figure 3 It is a data processing flow diagram of the present invention.

[0052] Figure 4 It is a schematic diagram of the deep learning algorithm processing of the present invention.

[0053] Figure 5 Schematic diagram of user-cloud interaction according to the present invention.

[0054] Figure 6 It is a schematic diagram of the power management logic of the present invention. DETAILED DESCRIPTION

[0055] The following will be combined with the embodiments of the present invention to clearly and completely describe the technical solutions in the embodiments of the present invention. Obviously, the embodiments described are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0056] like Figures 1 to 4 The figure shows a vehicle-mounted motor vehicle road violation identification device.

[0057] 1. Camera Module

[0058] Hardware configuration:

[0059] Sensor: Sony IMX415 stacked CMOS sensor (1 / 2.8-inch, 8.3M effective pixels), supports HDR mode (dynamic range ≥ 100dB) and STARVIS2 low-light technology (minimum illumination 0.001lux);

[0060] Optics: Wide-angle lens (140° diagonal field of view, f / 1.6 aperture) composed of 6 glass elements, equipped with a polarizing filter to eliminate windshield reflections;

[0061] Processing unit: Built-in ISP image signal processor (such as Ambarella CV25), supporting real-time distortion correction (fisheye to perspective projection) and 3D noise reduction algorithm.

[0062] Workflow:

[0063] After the vehicle is started, the camera continuously captures RAW format images at 30fps;

[0064] The ISP processor performs the following real-time processing:

[0065] Demosaic conversion to YUV 4:2:2 format;

[0066] Dynamic range compression (adjusting the gamma curve based on scene brightness distribution);

[0067] Multi-frame synthesis (merging 3 consecutive frames to improve signal-to-noise ratio in low light conditions);

[0068] Output 1080P@30fps H.264 encoded video stream to the memory module, and cache the original frame to RAM for the algorithm module to call.

[0069] 2. GPS positioning module

[0070] Hardware configuration:

[0071] Chipset: U-blox ZED-F9P multi-frequency GNSS receiver (supports GPS L1 / L2, Beidou B1 / B2, GLONASSL1 / L2);

[0072] Auxiliary positioning: integrated LSM6DSO 6-axis IMU (accelerometer + gyroscope) and QMC5883L magnetometer;

[0073] Interface: Communicates with the main control chip via UART, with a baud rate of 115200bps.

[0074] Positioning process:

[0075] The GNSS chip outputs positioning data (latitude and longitude, speed, and heading angle) once per second, with a positioning accuracy of ≤1.5 meters (in an open environment);

[0076] When satellite signals are lost (e.g. in a tunnel), the IMU calculates the position based on dead reckoning (DR):

[0077] Displacement is calculated based on the accelerometer integration, and heading change is calculated based on the gyroscope integration;

[0078] The magnetometer corrects heading drift, with an error accumulation of ≤3 meters / minute;

[0079] The positioning data and camera clock are synchronized via the PTP protocol, with a timestamp error of ≤10ms.

[0080] 3. Deep Learning Algorithm Module

[0081] Model Architecture:

[0082] Object detection: Use the YOLO model (input resolution 640×640, anchor boxes optimized for vehicle size) to output vehicle bounding boxes and classification labels (regular vehicle, truck, motorcycle, etc.);

[0083] Lane segmentation: Based on the FastSAM model (MobileNetV3 as the backbone), it outputs pixel-level solid / dashed line classification results;

[0084] License plate recognition: Using the CRNN model (CNN feature extraction + BiLSTM sequence modeling + CTC decoding), it supports the recognition of blue license plates, yellow license plates, and new energy vehicle license plates;

[0085] Behavioral analysis: Based on the LSTM network, the detection results of 10 consecutive frames (about 0.33 seconds) are modeled in a time series to determine whether behavioral conditions such as changing lanes without turning on the indicator light or crossing the solid line are met.

[0086] Reasoning process:

[0087] Extract the current frame and the previous 9 frames from the camera RAM;

[0088] The target detection model outputs the vehicle's position and turn signal status (by detecting changes in the brightness of the taillight area);

[0089] The lane segmentation model outputs the current lane type (solid line / dashed line) and the relative position of the vehicle;

[0090] The LSTM network uses time series data to determine traffic violations:

[0091] Lane change without signaling: The vehicle crosses the lane line and the turn signal on the corresponding side is not activated within 0.5 seconds before or after the lane change;

[0092] Solid line lane change: When the vehicle crosses the lane line, the lane line type is solid and the continuous crossing time ≥ 0.2 seconds;

[0093] If a violation is determined, trigger the evidence fixation process.

[0094] IV. Data Upload Module

[0095] Hardware and Protocol:

[0096] Communication chip: Quectel EC25-E 4G module (supporting Cat 4, downlink 150Mbps), built-in eSIM card;

[0097] Transmission protocol: Communicate with the Alibaba Cloud IoT platform based on the MQTT protocol, and the data packet is encrypted with AES-256.

[0098] Upload process:

[0099] When a violation is detected, the system generates the following data packets:

[0100] Video evidence: H.264 video clip from 5 seconds before to 10 seconds after the violation (with GPS track overlay);

[0101] Metadata: Violation type, license plate number, location, timestamp, confidence level in JSON format (such as {"violation": "solid line lane change", "license_plate": "Beijing A12345", "confidence": 0.91});

[0102] The data packet is compressed and divided into multiple 1MB chunks, and uploaded in slices through the 4G network;

[0103] If the transmission is interrupted, record the breakpoint location and resume transmission from the breakpoint when the network resumes.

[0104] V. Power Supply Module

[0105] Circuit design:

[0106] Vehicle power supply: Draw power through the OBD-II interface (12V DC), and convert it to a stable output of 5V / 3A through the MPQ4475 buck-boost chip;

[0107] Backup battery: Use 18650 lithium-ion battery cells (two in parallel, total capacity 6000mAh), support PD fast charging (fully charged in 1 hour); <00002​​​​​​​

[0111] Switching logic:

[0112] When the vehicle is ignited, the vehicle power supply is used first to supply power and charge the battery;

[0113] After the vehicle is turned off, it automatically switches to battery power. If the camera detects continued inactivity (no change in IMU data), it enters standby mode after 10 minutes.

[0114] When the battery level drops below 10%, a low battery alarm is triggered and data uploading stops.

[0115] 6. Installation and debugging

[0116] Physical Installation:

[0117] Use 3M VHB double-sided tape to fix the device body behind the windshield rearview mirror;

[0118] The centerline of the camera lens is aligned with the vehicle's forward direction, and the pitch angle is automatically calibrated using a built-in gravity sensor;

[0119] The GPS antenna is placed on the roof or edge of the windshield to improve signal quality.

[0120] Software Configuration:

[0121] When starting up for the first time, complete the following settings via the mobile app (Bluetooth connection):

[0122] Bind vehicle VIN code;

[0123] Select the traffic rules template for your region (e.g. China GB / T 20234-2022);

[0124] Calibrate the spatiotemporal alignment parameters of the camera and GPS;

[0125] The system automatically updates the algorithm model and electronic map data via OTA every 24 hours.

[0126] 7. Implementation Verification Example

[0127] Scenario: A vehicle changes lanes illegally on a solid road

[0128] Data collection:

[0129] The camera captures the target vehicle moving from the left lane to the right lane;

[0130] GPS positioning shows that the vehicle is on a certain city expressway (this section is marked as a solid line on the electronic map);

[0131] Algorithm judgment:

[0132] The target detection model identifies the vehicle's bounding box and the status of the turn signal (the right turn signal is not activated).

[0133] The lane line segmentation model outputs that the current lane line is a solid white line.

[0134] The LSTM network confirms that the lane change action crosses a solid line and lasts for 0.3 seconds.

[0135] Evidence generation:

[0136] Capture the video of the lane change process (timestamp synchronized and superimposed).

[0137] Extract the license plate number "沪B123#5" and associate it with the GPS coordinates (31.2304°N, 121.4737°E).

[0138] Data upload:

[0139] The encrypted evidence package is uploaded to the traffic management platform via the 4G network, and the system returns a receipt of successful reception.

[0140] Although the embodiments of the present invention have been shown and described, those of ordinary skill in the art can understand that various changes, modifications, substitutions, and variations can be made to these embodiments without departing from the principles and spirit of the present invention. The scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A vehicle-mounted motor vehicle road violation identification device, characterized in that: include: The camera module is used to collect high-definition video images of the vehicle's driving environment in real time, supporting image capture in high resolution and low light conditions; GPS positioning module, used to obtain the vehicle's real-time location, driving trajectory and speed information; A deep learning algorithm module, based on a model combining a convolutional neural network (CNN) and a recurrent neural network (RNN), processes video data collected by the camera module to identify traffic violations and license plate information of target vehicles. Violations include at least one of driving against traffic, changing lanes without using a turn signal, and changing lanes on a solid line. Memory module, used to store video data, violation determination results and GPS location information, and supports cyclic overwriting storage; The power supply module, including the vehicle power adapter and backup battery, automatically switches the power source according to the vehicle ignition status to ensure continuous operation; The data upload module uploads violation evidence (including pictures, video clips), license plate information and GPS location data to the cloud server or traffic management platform via the wireless network.

2. The vehicle-mounted motor vehicle road violation identification device according to claim 1, characterized in that: The camera module includes a wide-angle lens and an infrared fill light unit, supports 1080P and above resolution, a frame rate of not less than 30fps, and has a dynamic exposure adjustment function to adapt to lighting changes.

3. The vehicle-mounted motor vehicle road violation identification device according to claim 1, characterized in that: The GPS positioning module has a positioning accuracy of less than 3 meters and is integrated with a speed sensor to synchronously record vehicle speed data during violations.

4. The vehicle-mounted motor vehicle road violation identification device according to claim 1, characterized in that: The deep learning algorithm module further includes: License plate recognition submodule, which extracts license plate characters through OCR technology; The behavior analysis submodule determines whether the vehicle violates traffic regulations based on video time series data, and verifies the compliance of the lane and road section where the violation occurred in combination with GPS location information.

5. The vehicle-mounted motor vehicle road violation identification device according to claim 1, characterized in that: The data upload module supports multiple network protocols (4G / 5G, Wi-Fi), and temporarily stores data in the memory module when the network connection is interrupted, and automatically resumes the upload after the network is restored.

6. The vehicle-mounted motor vehicle road violation identification device according to claim 1, characterized in that: It also includes a user interaction module for providing real-time feedback of violation alerts to drivers and providing a local preview of violation evidence and a complaint portal.

7. The vehicle-mounted motor vehicle road violation identification device according to claim 1, characterized in that: The device is integrated into the vehicle in the form of a vehicle-mounted rearview mirror or an independent device, and is powered by an OBD interface or a cigarette lighter power supply.

8. A method for monitoring motor vehicle violations using a vehicle-mounted motor vehicle road violation identification device, characterized in that: The following steps are involved: S1: Real-time collection of vehicle surrounding video stream and GPS location data; S2: Identify traffic violations and license plates of other vehicles or the vehicle itself through deep learning models; S3: Associate the violation behavior with the GPS location and timestamp to generate a structured evidence package; S4: Automatically upload the evidence package to the cloud platform for review by the traffic management department.

Images