A vehicle-mounted intelligent illegal parking detection and evidence collection device and method thereof

Abstract

The vehicle-mounted intelligent illegal parking capture and evidence collection device and method include a main body, a quick-release mounting bracket, a capture button, an enforcement mobile terminal, and integrated dual-mode imaging system, AI intelligent recognition engine, positioning and time synchronization module, communication and storage module, audio-visual linkage unit, and power management module. The quick-release bracket uses a multi-spacing positioning hole + spring clip + hand-tightening bolt design to adapt to various motorcycle models for quick assembly and disassembly without special tools. The dual-mode imaging system integrates dual lenses, a gyroscope, and an automatic fill light array to achieve image capture with image stabilization and high definition. The AI ​​intelligent recognition engine uses dual violation judgment logic based on parking space area and effective distance filtering to accurately identify illegally parked vehicles. The audio-visual linkage unit connects to the motorcycle's original horn to achieve flexible voice persuasion. The communication and storage module encrypts and uploads the evidence data with superimposed time and location information and performs local backup. This addresses the shortcomings of poor adaptability, low enforcement efficiency, insufficient judgment accuracy, and weak linkage.

Description

Technical Field

[0001] This invention relates to the field of intelligent traffic enforcement equipment technology, and in particular to a quick-release vehicle-mounted intelligent illegal parking detection and evidence collection device and method adapted to police motorcycles. It is especially suitable for the automatic identification, evidence collection, and flexible dissuasion of illegal parking in urban road mobile patrol scenarios, belonging to the field of technological innovation in mobile traffic enforcement equipment. Background Technology

[0002] Currently, enforcement of illegal parking on urban roads mainly relies on two methods: fixed-point electronic police equipment and on-site manual evidence collection by law enforcement personnel. A search revealed that current vehicle-mounted illegal parking detection equipment still has many shortcomings. For example, CN201510564839.2, a device and method for collecting evidence of illegal oblique lateral parking, applied for by Beijing Xinluwei Technology Co., Ltd., while achieving mobile monitoring, is not adapted to the specific characteristics of police motorcycle frames due to its fixed structure installed externally on the vehicle. Other devices use fixed installation structures, which are cumbersome to install and remove and lack an interface for linkage with the vehicle's original system.

[0003] Existing technologies generally suffer from the following core defects:

[0004] Under the manual evidence collection mode, law enforcement officers need to stop the vehicle, take photos on site, and enter vehicle information and enforcement location. The average time for collecting evidence of a single illegal parking violation is no less than 5 minutes. Moreover, it is limited by environmental factors such as morning and evening rush hours and inclement weather, and cannot achieve full-time and full-area coverage. Even with the existing vehicle-mounted equipment, most violations still require manual assistance to judge, and the automated closed-loop enforcement has not been achieved, which still requires a lot of law enforcement resources.

[0005] In addition, traditional vehicle-mounted illegal parking detection equipment usually relies solely on single image recognition to determine violations, without combining a legal parking space area filtering mechanism based on real-time video image recognition, nor introducing an effective capture distance filtering mechanism based on the conversion between the target outline and the bottom pixel distance of the image. Therefore, it is easy to misjudge parking in legal parking spaces as illegal, or the license plate image may be blurred or fail to be recognized due to excessive capture distance. The overall misjudgment rate is generally higher than 5%, which is difficult to meet the accuracy and rigor requirements of traffic law enforcement.

[0006] Furthermore, most of the existing equipment consists of independent working units and lacks an interface for integration with the original systems on police motorcycles. It also lacks a design for integration with the original audio-visual systems, making it impossible to use the original loudspeakers for gentle persuasion, which could easily lead to law enforcement conflicts.

[0007] In view of the shortcomings of the existing technologies, such as low enforcement efficiency, poor installation compatibility, insufficient accuracy in judging violations, and weak system linkage, there is an urgent need in this field for a vehicle-mounted intelligent illegal parking capture and evidence collection device that can be quickly disassembled and adapted, accurately judge violations, and efficiently link up to meet the needs of mobile law enforcement on urban roads. Summary of the Invention

[0008] The present invention aims to overcome the shortcomings of existing vehicle-mounted illegal parking capture and evidence collection devices, such as poor installation adaptability, low enforcement efficiency, insufficient accuracy in judging violations, and weak system linkage, and provides a quick-release vehicle-mounted intelligent illegal parking capture and evidence collection device and corresponding method.

[0009] In other words, the object of this invention is to solve at least one of the following specific problems:

[0010] A quick-release mounting structure is provided that is compatible with various models of police motorcycles, enabling rapid disassembly and assembly without special tools, keeping the disassembly and assembly time within 30 seconds, thereby improving the efficiency of equipment turnover among multiple motorcycles.

[0011] A dual-logic approach to illegal parking detection is implemented: parking area filtering and effective capture distance filtering. Parking area filtering uses real-time AI recognition of parking space markings or signs in video images to generate a dynamic mask of legal parking areas. Effective capture distance filtering calculates the actual physical distance by extracting the vertical pixel distance from the lowest point of the vehicle's outline to the bottom edge of the image and combining this with the camera's installation height and viewing angle parameters. This reduces the error rate of illegal parking detection to below 0.1%, ensuring the legality, accuracy, and enforcement effectiveness of the generated evidence data.

[0012] This enables deep integration between the equipment and the original systems of police motorcycles. By connecting to the original motorcycle horn through a dedicated interface, the warning messages can be clearly heard in a driving environment, thus improving the flexibility of law enforcement.

[0013] To solve the above-mentioned technical problems, this application provides the following technical solution:

[0014] The present invention discloses a vehicle-mounted intelligent illegal parking capture and evidence collection device, comprising a main body, a quick-release mounting bracket, a capture button, a dedicated law enforcement mobile terminal, and functional modules integrated within the main body. The specific structure and connection relationship of each component are as follows:

[0015] The main body of the equipment uses a high-strength sealed outer shell;

[0016] Quick-release mounting bracket, which detachably connects the main body of the equipment to the frame of the police motorcycle;

[0017] A dedicated mobile law enforcement terminal, which communicates with the main unit of the device;

[0018] Functional modules, integrated within the main body of the device, include:

[0019] Dual-mode imaging system for acquiring road video and capturing images;

[0020] The AI ​​intelligent recognition engine is used to perform dual illegal parking judgments, including vehicle detection, filtering of legal parking space areas, and filtering of effective distance.

[0021] The sound and light linkage unit is compatible with the original horn interface of police motorcycles and is used to play warning messages.

[0022] The communication and storage module is used for encrypted data upload and local backup.

[0023] The AI ​​intelligent recognition engine is connected to the dual-mode imaging system, the audio-visual linkage unit, and the communication and storage module. The main body of the device is adapted to multiple vehicle models through a quick-release mounting bracket.

[0024] This invention also provides a method for capturing and collecting evidence of illegal parking by an onboard intelligent camera, using the aforementioned equipment, and including the following steps:

[0025] S1: The main body of the device is mounted on the police motorcycle using a quick-release mounting bracket, and the capture button is connected.

[0026] S2: The dual-mode imaging system collects road video in real time and transmits it to the AI ​​intelligent recognition engine to perform dual illegal parking judgment;

[0027] S3: If the parking violation is determined, the dual-mode imaging system is triggered to capture the image and overlay time and location information;

[0028] S4: Controls the motorcycle's original horn to play a warning message via the sound and light linkage unit;

[0029] S5: The communication and storage module uploads the encrypted evidence data to the law enforcement platform and performs local backup.

[0030] Furthermore, the quick-release mounting bracket requires no special tools and can be disassembled and assembled by snap-fit ​​and bolt locking, with a disassembly and assembly time of no more than 30 seconds, and is compatible with at least 3 models of police motorcycle frames.

[0031] Furthermore, the effective distance filtering range is 0.5 meters to 3 meters, and this distance is determined by the following method: based on the image captured by the camera, the vertical pixel distance from the lowest point of the target outline of the illegally parked vehicle to the bottom edge of the image is extracted, and combined with the installation height of the camera and the lens angle parameters, the pixel distance is converted into the actual physical distance from the camera to the vehicle.

[0032] Furthermore, the filtering of legal parking space areas is achieved through video image analysis: real-time parking space recognition is performed on the current scene, parking markings or parking space signs on the road surface are detected, a dynamic legal parking space area mask is generated, and it is determined whether the target vehicle is within the range of the mask, thereby excluding legal parking situations. Furthermore, the audio-visual linkage unit includes a voice synthesis module, pre-stores at least 5 warning voice messages, and plays them through the motorcycle's original speaker, with the volume adapted to the ambient noise of the driving environment.

[0033] To ensure the validity and accuracy of the captured evidence, the system employs a dual filtering mechanism: effective distance filtering and legal parking space area filtering. Specifically:

[0034] 1. Implementation method of effective distance filtering

[0035] The effective distance filter is used to limit the target vehicle to a physical distance of 0.5 meters to 3 meters in order to trigger high-definition capture. This distance is accurately calculated based on the principle of monocular vision geometric ranging.

[0036] The system first uses the main camera in the dual-mode imaging system to acquire images of the road ahead; the AI ​​intelligent recognition engine performs target detection on the image and locates the outline boundary of the illegally parked vehicle; then, it extracts the lowest point of the vehicle outline in the image coordinate system (i.e., the pixel row position closest to the bottom of the image) and calculates the vertical pixel distance between it and the bottom edge of the image (denoted as d_pix).

[0037] Combining known camera installation parameters—including camera installation height H (vertical height relative to the ground, typically 1.2 meters), lens vertical field of view θ (e.g., 60 degrees), and total image height h_pix (e.g., 1080 pixels)—the system converts pixel distance to actual physical distance D using the following formula:

[0038]

[0039] in, , representing the initial angle between the camera's optical axis and the ground (assuming the camera is installed tilted downwards). When the calculated D satisfies 0.5m≤D≤3m, the system determines that the target is within the effective recognition distance and allows it to proceed to the subsequent capture process; otherwise, it is ignored, thus effectively avoiding false detection at long distances or recognition errors caused by distortion at close distances.

[0040] H (installation height) needs to be determined during equipment initialization through a calibration process—using a calibration board placed 1m directly in front of the motorcycle, the height is calculated by image recognition, with an error tolerance of ≤0.05m; θ is an inherent parameter of the lens, pre-stored in the system.

[0041] 2. Implementation method of filtering legal parking space areas

[0042] To distinguish between "illegal parking" and "legal parking," this invention abandons the traditional method of relying on pre-stored high-precision electronic maps and instead adopts real-time parking space recognition technology based on video images to achieve dynamic and adaptive filtering of legal areas.

[0043] That is, while the AI ​​intelligent recognition engine identifies illegally parked vehicles, it performs semantic segmentation or key point detection operations on the same frame image to specifically identify parking space markings on the road surface, including but not limited to white squares, T-shaped lines, numbered markings, or dedicated parking space indicator signs. This process is completed in real time on the edge AI chip through a lightweight deep learning model. The model can be an improved YOLOv8-Parking or MobileNetV3 combined with the U-Net architecture.

[0044] The system generates a binary mask based on the recognition results, where legal parking areas are marked as "1" and the remaining areas as "0". Then, the bottom center point or bounding box of the illegally parked vehicle is projected onto this mask to determine whether it is completely or primarily located within a legal parking area. If the vehicle does not fall within any legal parking area mask, it is considered illegally parked; otherwise, it is considered legally parked and no capture is triggered.

[0045] This method does not rely on external map data updates and is applicable to complex urban scenarios such as unmarked roads, temporary parking spaces, and construction changes, significantly improving the robustness and versatility of the system.

[0046] 3. Closed-loop enforcement process

[0047] When both of the above filtering conditions are met simultaneously, the system automatically triggers dual-view high-definition capture, generating an evidence package containing a panoramic view of the front and rear of the vehicle, and overlaying Beidou coordinates provided by the positioning and time synchronization module, a timestamp generated by the timing unit, and the road name obtained by matching the electronic map; at the same time, it can optionally play directional voice persuasion prompts through the sound and light linkage unit; finally, the encrypted data is uploaded to the traffic police integrated command platform through the 4G / 5G network, forming a complete "non-contact flexible law enforcement closed loop".

[0048] Furthermore, the dual-mode imaging system includes dual-lens cameras, image stabilization components, and an automatic illumination array, which automatically activates based on ambient light levels.

[0049] Furthermore, the communication and storage module supports 4G / 5G network communication, uses AES-256 encryption before data is uploaded, and the captured images comply with the "GA / T 832-2014" evidence collection standard.

[0050] In addition, in step S1, the quick-release mounting bracket can be disassembled and assembled using only two operations: snap-fit ​​and bolt locking, without the need for special tools.

[0051] In step S2, the double illegal parking determination is a joint determination of "outside the legal parking space area + within an effective distance of 0.5 - 3 meters".

[0052] Furthermore, it also includes: law enforcement officers triggering the capture by manually pressing the capture button, directly executing steps S3-S5.

[0053] Compared with the prior art, the vehicle-mounted intelligent illegal parking capture and evidence collection device and method of the present invention have at least the following beneficial effects:

[0054] Highly adaptable to various installation types and significantly improved assembly / disassembly efficiency: This invention's quick-release mounting bracket, through a combination design of "multi-spacing positioning holes + spring clips + hand-tightening bolts," can be adapted to at least three mainstream police motorcycle models. Compared to the existing fixed mounting brackets' "single model adaptation" mode, the compatibility range is greatly expanded. Simultaneously, the equipment can be assembled and disassembled without the need for special tools, and the assembly / disassembly time can be controlled within 30 seconds. Compared to existing equipment that takes more than 30 minutes to assemble and disassemble, efficiency is increased by up to 60 times, effectively reducing time and labor costs during equipment turnover.

[0055] This invention achieves high accuracy in determining illegal parking, and the acquired evidence data is legal and valid. It constructs a dual illegal parking determination logic of "parking space area filtering + effective distance filtering," adding two key dimensions compared to existing single-image recognition methods. Parking space area filtering is based on real-time recognition of road parking space markings or signs in video images. Through semantic segmentation or object detection, a dynamic mask of legal parking spaces is generated, accurately distinguishing between legal and illegal parking without relying on pre-stored electronic maps. Effective distance filtering is achieved through monocular visual geometric ranging: the system extracts the vertical pixel distance from the lowest point of the illegally parked vehicle's outline to the bottom edge of the image. Combined with the camera's installation height and lens angle parameters, the actual physical distance from the camera to the vehicle is calculated, triggering capture only within a range of 0.5 to 3 meters, ensuring that key information such as license plates is clearly identifiable. Actual road test data shows that the error rate of this invention can be reduced to below 0.1%, far lower than the generally higher than 5% error rate of existing equipment. The generated evidence data (including dual-view images, timestamps, geographic coordinates, and structured license plate information) fully complies with the requirements of the "GA / T832–2014 Technical Specification for Image Evidence Collection of Road Traffic Safety Violations" and can be directly used as law enforcement evidence.

[0056] The system of this invention exhibits excellent interconnectivity, enabling flexible law enforcement. Its audio-visual linkage unit connects to the original motorcycle horn via a dedicated waterproof interface, and, combined with a power amplification module and automatic volume adjustment function, ensures that the warning messages are clearly audible even while the vehicle is in motion. Compared to the independent small speakers of existing equipment, the voice transmission distance is increased to over 50 meters, significantly enhancing the effectiveness of the warnings. Furthermore, replacing direct punishment with voice warnings effectively reduces law enforcement conflicts, aligns with the current policy direction of "flexible law enforcement," and enhances the humane aspect of law enforcement.

[0057] The following description, in conjunction with the accompanying drawings, further elaborates on a vehicle-mounted intelligent illegal parking detection and evidence collection device and method of the present invention. Attached Figure Description

[0058] Figure 1 This is a schematic diagram of the installation of a vehicle-mounted intelligent illegal parking capture and evidence collection device according to the present invention;

[0059] Figure 2 This is a block diagram showing the internal structure modules of an intelligent vehicle-mounted illegal parking capture and evidence collection device according to the present invention.

[0060] Figure 3 This is a schematic diagram of the process for a vehicle-mounted intelligent illegal parking capture and evidence collection method according to the present invention.

[0061] in:

[0062] 1-Main body of the device (high-strength aluminum alloy shell); 2-Quick-release mounting bracket; 3-Motorcycle body; 4-Snapshot button (handlebar control terminal); 6-Dual-mode imaging system, including 6.1-Camera, 6.2-Gyroscope, 6.3-Fill-in light array; 7-AI intelligent recognition engine; 8-Positioning and time synchronization module; 9-Communication and storage module; 10-Audio-optical linkage unit; 11-Power management module.

[0063] The following is in conjunction with the appendix Figure 1-3 This document provides a detailed description of the specific component parameters and control logic of the present invention, enabling those skilled in the art to fully and accurately implement the technical solution of the present invention. It should be noted that the component models and parameters used in this embodiment are merely preferred options and are not intended to limit the scope of protection of the present invention. Those skilled in the art can select other equivalent components to replace them according to actual needs; any components that can achieve the technical effects of the present invention should fall within the scope of protection of the present invention. Detailed Implementation

[0064] like Figure 1-3 As shown, this invention relates to a vehicle-mounted intelligent illegal parking capture and evidence collection device, comprising a main body, a quick-release mounting bracket, a capture button, a dedicated law enforcement mobile terminal, and functional modules integrated within the main body. The specific structure and connection relationships of each component are as follows:

[0065] Figure 1 This is an installation diagram of the vehicle-mounted intelligent illegal parking capture and evidence collection device of the present invention. Figure 2 This is a block diagram showing the connection of the internal structural modules of the device, which clearly illustrates the composition of each functional module, the signal flow and the logical coordination mechanism, and intuitively presents the system architecture of the technical solution.

[0066] Equipment body 1 (i.e. Figure 1 (Mid-tail housing structure): Utilizing a unibody 6061-T6 high-strength aluminum alloy shell, this alloy features high strength (tensile strength ≥310MPa), strong corrosion resistance, and light weight, enabling it to withstand motorcycle vibrations and harsh outdoor environments. An internal Q235 steel metal frame supports and secures the functional modules, preventing displacement during vibration. Fluororubber seals with an IP67 rating are installed between the shell and the frame, effectively preventing rainwater and dust from entering the equipment. The main body dimensions are designed to be 400mm × 350mm × 400mm, precisely fitting the installation space of mainstream police motorcycles' rear ends and avoiding any impact on the motorcycle's stability after installation.

[0067] Quick-release mounting bracket 2: As the connecting component between the main body of the equipment and the motorcycle frame 3, it is made of high-strength aluminum alloy in one piece and includes three parts: a fixed base, a movable buckle, and a locking bolt. The fixed base connects to the pre-set mounting holes at the rear of the motorcycle frame via two sets of M8 bolts. The base has three sets of positioning holes with different spacing, which can adapt to the frame structures of at least three mainstream police motorcycle models (such as the CFMoto 650, Qianjiang Benelli 502, etc.). The movable buckle matches the slot on the bottom of the main body of the equipment and achieves quick engagement through spring drive. After engagement, it is further secured with a set of hand-tightening locking bolts. Disassembly and assembly can be completed without special tools, and the entire process takes no more than 30 seconds. The connection surface between the bracket and the main body of the equipment is equipped with anti-slip rubber pads, which can increase friction and buffer vibrations during motorcycle operation, protecting the main body of the equipment.

[0068] Detection button 4: Features a compact design with an IP65 waterproof rating, fixed to the left or right side of the motorcycle handlebars. Its position can be flexibly adjusted according to the operating habits of law enforcement officers. The detection button includes a manual detection button and a status indicator light.

[0069] The manual capture button employs a self-locking mechanical structure with a bistable switching function: upon first press, the button locks in the "on" position, triggering the device to enter continuous capture mode, and the status indicator light illuminates; upon subsequent press, the button resets to the "off" position, stopping capture, and the status indicator light turns off. This design avoids frequent pressing operations and provides intuitive feedback on the current capture status through the indicator light, preventing misoperation or status confusion. The capture button connects to the communication interface of the device body via a 2.5-meter shielded cable. The cable shielding layer uses a tinned copper wire braided structure, effectively suppressing electromagnetic interference generated by the motorcycle ignition system, motor, and other onboard electronic devices, ensuring stable and reliable control signal transmission, and adapting to the installation spacing between the handlebars and the rear device body on different motorcycle models.

[0070] (II) Functional modules and their connections (see) Figure 2 )

[0071] The dual-mode imaging system (6) is the core component for image acquisition, integrating three sub-modules: a 6.1 camera, a 6.2 gyroscope, and a 6.3 supplementary lighting array. The camera employs a single-lens panoramic design with a focal length of 2.8mm and a field of view of 120°, used to acquire panoramic images of the roadside parking area. License plate recognition relies on an AI intelligent recognition engine to perform super-resolution enhancement and detail extraction on local areas within the panoramic image. The camera uses a 1 / 2.8-inch CMOS sensor with a resolution of 8 megapixels and a frame rate of 25fps, ensuring both high definition and smoothness during motorcycle operation, guaranteeing clear identification of key information such as vehicle outlines, license plates, and road markings in dynamic scenes. The gyroscope is a MEMS type with a range of ±2000° / s and an accuracy of ±0.01° / s, acquiring motorcycle vibration data in real time and transmitting it to the AI ​​intelligent recognition engine, compensating for the impact of vibration on the image through an image stabilization algorithm. The supplementary lighting array consists of four high-brightness LEDs arranged in a strip below the camera. Each LED has a power of 3W and a color temperature of 5500K (close to natural light). The supplementary lighting distance is 0.5-3m. The array uses a light sensor to detect the ambient illuminance in real time and automatically turns on when the illuminance is below 15 lux, ensuring good capture performance in low-light environments. The dual-mode imaging system connects to the AI ​​intelligent recognition engine via an LVDS interface, receiving manual capture trigger signals (low-level trigger, voltage range 3.3V-5V) from the capture button or automatic capture commands from the AI ​​intelligent recognition engine to complete image acquisition.

[0072] The AI ​​intelligent recognition engine (7) is the core control component of the device. It adopts an embedded AI computing module based on the NVIDIA Jetson XavierNX architecture, integrating a 6-core ARM Cortex-A57 processor and 48 CUDA cores, with a peak computing power of 21 TOPS, which can meet the requirements of real-time image recognition and multi-task parallel processing under high-resolution video streams. The AI ​​intelligent recognition engine is connected to the dual-mode imaging system (6) through the PCIe 3.0 interface, receives the 25 frames per second (fps) panoramic video stream with 8 million pixels transmitted by it, and executes a three-step illegal parking judgment logic through a pre-loaded deep learning model:

[0073] The first step is vehicle detection: The YOLOv8 target detection algorithm is used to identify motor vehicles in each frame of the image, and the bounding box and confidence score are output; when the confidence score is not lower than 99.5%, it is determined to be a valid vehicle target;

[0074] The second step is parking space area filtering: semantic segmentation or key point detection is performed on the current video frame to identify parking space markings (such as white squares, T-shaped lines) or dedicated parking space signs on the road surface in real time, and a dynamic legal parking space area mask is generated; the bounding box of the effective vehicle target is projected onto the mask, and if more than 70% of its area is outside the legal parking space area, it is judged as a suspected illegal parking.

[0075] The third step is effective distance filtering: Extract the lowest point of the suspected illegally parked vehicle's outline in the image (i.e., the pixel row position closest to the bottom edge of the image), and calculate the vertical pixel distance d_pix from this point to the bottom edge of the image; combine this with the known camera installation height H (e.g., 1.4 meters), lens vertical field of view θ (e.g., 60°), and total image height h_pix (e.g., 1520 pixels), and calculate the actual physical distance D through geometric relationships:

[0076] ,

[0077] in The initial angle between the camera's optical axis and the ground; when the calculated distance D satisfies 0.5 meters ≤ D ≤ 3 meters, it is determined to be a valid illegally parked target.

[0078] When a vehicle simultaneously meets both conditions of being "outside the legal parking area" and "within the effective capture distance," the AI ​​intelligent recognition engine (7) outputs the illegal parking judgment result and generates a capture command (TTL level signal) and a voice persuasion command (UART protocol, baud rate 9600). Positioning and time synchronization module (8): integrates a Beidou positioning chip and a BDS timing unit. The positioning chip uses the Huaxin antenna UM220-IVN module, with a positioning accuracy of 1-3 meters, and supports satellite-based augmentation system (SBAS) differential positioning, which can improve the positioning accuracy in complex urban environments; the timing unit uses BDS dual-mode timing, with a timing accuracy of no more than 1 microsecond, ensuring accurate timestamps. The positioning and time synchronization module is connected to the AI ​​intelligent recognition engine through the UART interface, responds to the AI ​​intelligent recognition engine's requests, and outputs latitude and longitude coordinates (format WGS-84), time data (format UTC), and positioning status information in real time. The module has a built-in backup battery, which can ensure that the positioning and timing functions operate normally for no less than 2 hours when the motorcycle power is disconnected, ensuring continuous time data.

[0079] Communication and storage module (9): This is the core module for data transmission and storage, integrating a 4G / 5G communication unit, an encrypted storage card interface, and a data processing unit. The communication unit uses a Huawei ME909S-8214G module, supporting TD-LTE / FDD-LTE full frequency bands with a downlink speed of no less than 150Mbps, ensuring rapid uploading of evidence data. The encrypted storage card interface supports MicroSD cards with a maximum capacity of 128GB, using the AES-256 encryption algorithm to encrypt and protect stored data, preventing data tampering. The data processing unit uses an STM32H743 microcontroller, receiving latitude, longitude, and time data transmitted from the positioning and time synchronization module. It then uses a character overlay algorithm to overlay this information onto a specified location in the captured image (timestamp in the upper left corner, latitude and longitude in the upper right corner), generating evidence data that conforms to the "GA / T832-2014 Technical Specification for Image Evidence Collection of Road Traffic Safety Violations". The communication and storage module connects to the AI ​​intelligent recognition engine via the SPI interface, receives upload commands from the AI ​​intelligent recognition engine, uploads the encrypted evidence data to the backend law enforcement platform via the 4G network, and simultaneously stores it on an encrypted storage card for local backup.

[0080] Sound and light linkage unit (10): includes a speech synthesis module, a power amplifier module, and a motorcycle original horn interface. The speech synthesis module uses the iFlytek XFS5152 chip, supports offline speech synthesis, and pre-stores 5 different scenario warning voices (such as "Parking is prohibited here, please leave as soon as possible" and "Violation of parking regulations will be punished according to law"), with a speech synthesis accuracy of no less than 98%; the power amplifier module uses the TDA7850 chip, with an output power of no less than 45W, which can drive the motorcycle original horn (usually 4Ω / 20W); the motorcycle original horn interface uses a 3-pin waterproof aviation plug (pin 1: voice signal input, 0-5V analog voltage; pin 2: power amplifier output, matching 4Ω impedance; pin 3: ground). Volume adaptive logic: the base volume is 75dB, which is dynamically adjusted according to the vehicle speed sensor signal - it remains unchanged when the vehicle speed is <30km / h, and increases by 20% when the speed is ≥30km / h, which is achieved through the STM32 microcontroller PID algorithm. The power amplifier module uses a TDA7850 chip, with an output power of no less than 45W. It is compatible with the horn interface in the motorcycle wiring harness and controls the voice signal output via a relay. The sound and light linkage unit connects to the AI ​​intelligent recognition engine via a UART interface, receives voice warning commands from the AI ​​intelligent recognition engine, selects the corresponding warning voice, amplifies it through the power amplifier module, and drives the motorcycle's original horn to play it. The playback volume can be automatically adjusted according to the motorcycle's speed (volume increases by 20% when the speed is not lower than 30km / h). The volume can also be adaptively adjusted by connecting the motorcycle's speed sensor signal through the ACC interface of the power management module, or by calculating the speed using Beidou positioning data from the positioning and time synchronization module.

[0081] Power Management Module (11): This is the core of the equipment's power supply. It uses the LM2596-ADJ switching power supply chip, supports a wide voltage input range of 12V-24V (adapting to the power systems of different models of police motorcycles), and outputs two sets of stable voltages, 5V / 3A and 12V / 2A, to power each functional module. This module integrates an ACC power detection circuit. The equipment automatically turns on when the motorcycle starts (ACC is powered on) and shuts down after a 30-second delay when the motorcycle is turned off (ACC is powered off), ensuring that unuploaded data is stored normally. The module also has overvoltage protection (threshold 28V), overcurrent protection (threshold 5A), and short-circuit protection circuits to prevent abnormal power supply from damaging the equipment. The power management module is connected to the motorcycle's ACC power supply interface through a power harness. The harness is made of flame-retardant material to ensure safe use.

[0082] See Figure 3 The usage process of the device of this invention is as follows:

[0083] Step 1: Equipment Installation (see...) Figure 1 )

[0084] The quick-release mounting bracket 2 is connected to the pre-set mounting holes at the rear of the police motorcycle frame using two sets of M8×20 bolts. The bolt tightening torque is controlled at 15 N·m to ensure the mounting base is securely installed. The slot at the bottom of the device body 1 is aligned with the movable buckle on the mounting base, and 50 N of pressure is applied to make the buckle fully engage with the slot, completing the initial fixation. The hand-tightening locking bolt on the side of the mounting base is rotated until the bolt head is tightly fitted to the side of the device body, and a tightening torque of 5 N·m is applied to complete the secondary fixation. The capture button 4 is fixed to the left side of the motorcycle handlebars using a special clamp. The clamp adjustment range is 22-32 mm to accommodate handlebars of different diameters. One end of the connecting cable is inserted into the communication interface of the device body, and the other end is inserted into the interface of the capture button. The cable is fixed by the cable clip on the motorcycle frame to prevent the cable from being worn due to shaking during driving. The entire installation process does not require the use of special tools such as wrenches and screwdrivers and takes about 25 seconds, which meets the design requirements.

[0085] Step 2: System Initialization

[0086] When the police motorcycle is started, the ACC power supply is turned on. The ACC detection circuit of the power management module 11 detects a 12V voltage signal and automatically starts the power supply program. It outputs 5V / 3A voltage to power the core modules such as the AI ​​intelligent recognition engine 7 and the positioning and time synchronization module 8, and outputs 12V / 2A voltage to power the supplementary light array of the dual-mode imaging system 6 and the power amplification module of the audio-visual linkage unit 10. After the positioning and time synchronization module (8) is started, it searches for Beidou satellite signals and completes the positioning lock after about 10 seconds. The positioning accuracy is 2m. The time synchronization unit completes the time calibration simultaneously and sets the current UTC time (format YYYY-MM-DD HH:MM:SS) and latitude and longitude coordinates (format N39°54′30″). E116°23′15″) is transmitted to the AI ​​intelligent recognition engine (7) and the communication and storage module (9); after the AI ​​intelligent recognition engine (7) is started, it loads the pre-stored YOLOv8 vehicle detection model and the lightweight semantic segmentation model (such as U-Net or YOLO-Parking variant) for road parking space recognition. Without relying on the pre-stored electronic map, it can detect parking lines or parking space signs in the video stream in real time, generate a dynamic legal parking space area mask, and complete the system initialization.

[0087] Step 3: Patrol data collection and AI-based violation detection

[0088] When law enforcement officers patrol on police motorcycles, the camera (6.1) of the dual-mode imaging system (6) is activated and simultaneously collects road video. The video stream is transmitted to the AI ​​intelligent recognition engine (7) at a rate of 30fps through the LVDS interface. The gyroscope (6.2) collects motorcycle vibration data in real time with a sampling rate of 100Hz and transmits the vibration angular velocity data to the AI ​​intelligent recognition engine (7). The AI ​​intelligent recognition engine performs anti-shake processing on the video image through the Kalman filter algorithm to compensate for the image shift caused by vibration. The light sensor detects the ambient illuminance in real time. When the illuminance is 12 lux (below the 15 lux threshold), the supplementary light array (6.3) is automatically turned on, and four high-brightness LED beads (subject to the detailed technical parameters of the specific implementation) are lit simultaneously to supplement the shooting area.

[0089] The AI ​​intelligent recognition engine (7) receives the video stream after image stabilization and executes the following three-step illegal parking judgment logic:

[0090] The first step is vehicle detection: The YOLOv8 algorithm is used to detect targets in each frame of video image and identify motor vehicle targets in the image; when a motor vehicle is detected, its bounding box coordinates and confidence score are output; if the confidence score is ≥0.95, it is determined to be a valid vehicle target and proceeds to the next step of judgment.

[0091] The second step is parking space area filtering: semantic segmentation or key point detection is performed on the current video frame to identify parking space markings (such as white squares or T-shaped lines) or dedicated parking space signs on the road surface in real time, and a dynamic legal parking space area mask is generated; the bounding box of the valid vehicle target is projected onto the mask. If more than 70% of its area is outside the legal parking space area defined by the mask, it is judged as suspected illegal parking and proceeds to the third step; otherwise, it is considered as legal parking and the judgment process is terminated.

[0092] The third step is effective distance filtering: Extract the lowest point of the suspected illegally parked vehicle's outline in the image (i.e., the pixel row position closest to the bottom edge of the image), and calculate the vertical pixel distance d_pix from this point to the bottom edge of the image; combine this with the known camera installation height H (e.g., 1.4 meters), lens vertical field of view θ (e.g., 60°), and total image height h_pix (e.g., 1520 pixels), and calculate the actual physical distance D from the camera to the vehicle through geometric relationships: ,in Let D be the initial angle between the camera's optical axis and the ground. When the calculated distance D satisfies 0.5 meters ≤ D ≤ 3 meters, it is considered a valid illegally parked target.

[0093] When a vehicle meets both the conditions of "suspected illegal parking" and "distance within the effective range", the AI ​​intelligent recognition engine (7) outputs the final illegal parking judgment result and records the current time.

[0094] Step 4: Capture and Overlay Information

[0095] While outputting the illegal parking judgment result, the AI ​​intelligent recognition engine (7) sends a capture command (TTL high-level signal, lasting 100ms) to the dual-mode imaging system (6). After receiving the command, the dual-mode imaging system (6) controls the camera (6.1) to capture an image with a resolution of 3840x2160 and transmits it to the communication and storage module (9) through the LVDS interface. At the same time, the AI ​​intelligent recognition engine (7) sends a data request command to the positioning and time synchronization module (8), and the positioning and time synchronization module (8) transmits the current timestamp (accurate to the second) and latitude and longitude coordinate data to the communication and storage module (9) through the UART interface. After receiving the panoramic image, close-up image, and time and location data, the STM32H743 microcontroller of the communication and storage module (9) overlays the timestamp (format: "2024-05-2014:30:00") onto the upper left corner of the image and the latitude and longitude coordinates (format: "N39°54′30″ E116°23′15″") onto the upper right corner of the image using a character overlay algorithm, generating evidence data that conforms to the GA / T 832-2014 standard. After overlaying, the microcontroller performs CRC32 verification on the entire evidence image data, generates a verification code, and stores or uploads it along with the image to ensure the integrity and tamper-proof nature of the data during transmission and storage.

[0096] Step 5: Audible and visual warnings and data upload

[0097] While the AI ​​intelligent recognition engine (7) sends the capture command, it also sends a voice warning command (UART protocol, baud rate 9600) to the sound and light linkage unit (10). The command includes the warning voice number of the current scene (e.g., number 1 corresponds to "Parking is prohibited here, please leave as soon as possible"). After receiving the command, the voice synthesis module (XFS5152 chip) of the sound and light linkage unit (10) calls the voice data corresponding to number 1 and synthesizes a voice signal (analog signal, amplitude 1Vpp). The voice signal is transmitted to the power amplifier module (TDA7850 chip), and after power amplification, it outputs a 45W drive signal, which controls the original horn circuit of the motorcycle through the relay to drive the original horn to play the warning voice. At this time, the motorcycle speed is 35km / h, and the system automatically increases the volume by 20% to ensure that people within 50 meters can hear it clearly. The warning voice plays for 8 seconds. After the playback is completed, the relay is disconnected and the horn returns to normal working status.

[0098] After completing the overlay and CRC32 verification of the evidence data, the communication and storage module (9) encrypts the evidence data using the AES-256 encryption algorithm. The encryption key is stored in the internal security chip of the device and cannot be read from the outside. After encryption, the communication unit (Huawei ME909S-821 module) establishes a 4G network connection with the backend law enforcement platform and uploads the encrypted evidence data to the platform via the HTTPS protocol. At the same time, the encrypted evidence data is written to the encrypted storage card for local backup. The storage path is " / SD card / evidence data / 20240520 / 143000 / ", which facilitates subsequent data traceability. After the data upload is completed, the communication and storage module sends an upload completion signal to the AI ​​intelligent recognition engine.

[0099] Once the AI ​​intelligent recognition engine identifies the license plate number of a illegally parked vehicle, the voice synthesis module automatically reads the license plate information and fills in placeholders from a preset script template (e.g., 'Hello, owner of vehicle number Zhejiang A12345, parking is prohibited here. Please move your vehicle as soon as possible.'), then synthesizes a personalized warning message and plays it.

[0100] Step 6: Manually trigger the snapshot (optional)

[0101] In actual law enforcement, if the vehicle-mounted intelligent illegal parking capture and evidence collection equipment fails to automatically identify illegal parking behavior due to factors such as vehicle obstruction, extreme lighting conditions, or special vehicle models, law enforcement officers can activate the manual supplementary shooting mode: using the dedicated law enforcement mobile terminal configured with this system, the built-in camera is used to take on-site photos of the illegally parked vehicle and its surrounding environment, obtaining images that include a panoramic view of the vehicle, a clear license plate, and road reference objects.

[0102] Step 6.1: Law enforcement officers open the pre-installed APP on the terminal, click the "Manual Capture" button, and the terminal camera starts to capture a panoramic image of the illegally parked vehicle (which must include the license plate and road references).

[0103] Step 6.2: The APP automatically calls the terminal's GNSS module to obtain latitude and longitude coordinates (WGS-84 format) and synchronizes the UTC timestamp (accuracy ±1 second) through network time synchronization.

[0104] Step 6.3: The terminal establishes a connection with the communication and storage module (9) of the main body of the device via Bluetooth 5.0 or 4G network, and uses UART protocol (baud rate 9600) to synchronize positioning / time data to ensure consistency with automatic capture data.

[0105] Step 6.4: The APP encapsulates the image, coordinates, time, and law enforcement officer ID into JSON format and uploads it to the traffic police platform via HTTPS.

[0106] This process serves as a fault-tolerant mechanism for automatic image capture, and the data conforms to the GA / T 832-2014 standard. It ensures that the timestamps and latitude / longitude coordinates of the manually captured data are consistent with those of the automatically captured data.

[0107] The platform uniformly archives and manages both manually captured data and vehicle-mounted automatic capture data. Both types of data meet the requirements of the "GA / T 832–2014 Technical Specification for Image Evidence Collection of Road Traffic Safety Violations" and have the same legal effect, and can be used as valid law enforcement evidence.

[0108] This manual re-photographing mechanism, as an effective supplement to the vehicle-mounted automatic identification system, significantly improves the system's coverage and fault tolerance in complex urban scenarios, achieving the enforcement goal of "photographing everything that should be photographed and monitoring without blind spots." The embodiments described above are merely preferred embodiments of the present invention and are not intended to limit the scope of the invention. Various modifications and improvements made to the technical solutions of the present invention by those skilled in the art without departing from the spirit of the invention should fall within the protection scope defined by the claims of the present invention.

Claims

1. A vehicle-mounted intelligent illegal parking snapshot evidence taking device, characterized in that, include: The main body of the equipment is made of a high-strength aluminum alloy sealed shell; A quick-release mounting bracket can detachably connect the main body of the equipment to the frame of a police motorcycle; A capture button is connected to the main body of the device via a cable. It is configured to generate a capture trigger signal in response to a press operation and has a built-in LED indicator to display the capture status in real time. Functional modules, integrated within the main body of the device, include: Dual-mode imaging system for acquiring road video and capturing images; The AI ​​intelligent recognition engine is used to perform dual illegal parking judgments, including vehicle detection, legal parking space area filtering, and effective distance filtering. The sound and light linkage unit is compatible with the original horn interface of police motorcycles and is used to play warning messages. The communication and storage module is used for encrypted data upload and local backup. The AI ​​intelligent recognition engine is connected to the dual-mode imaging system, the audio-visual linkage unit, and the communication and storage module. The main body of the device is adapted to multiple vehicle models through a quick-release mounting bracket. The vehicle-mounted intelligent illegal parking capture and evidence collection device is installed at the rear of the police motorcycle. The power management module is electrically connected to the motorcycle's ACC power supply and operates automatically when the vehicle starts. The dual-mode imaging system continuously collects video streams from the roadside ahead and transmits them to the AI ​​intelligent recognition engine. The AI ​​intelligent recognition engine has a built-in dual illegal parking filtering mechanism, namely, effective distance filtering function and legal parking space area filtering function, which analyzes the video stream in real time to achieve non-contact illegal parking enforcement. The dual filtering mechanism consists of effective distance filtering and legal parking space area filtering. The effective distance filtering function is implemented in the following way: The AI ​​intelligent recognition engine receives road images captured by the camera in the dual-mode imaging system, locates the outline boundary of illegally parked vehicles, and extracts the vertical pixel distance d_pix from the lowest point of the outline to the bottom edge of the image; combined with the preset camera installation height H, lens vertical field of view θ, and total image height h_pix, the effective distance filtering function is implemented using the formula... Convert the pixel distance to the actual physical distance D, where The distance is considered valid only if D satisfies 0.5m ≤ D ≤ 3m. The legal parking space area filtering function is implemented in the following way: The AI ​​intelligent recognition engine performs semantic segmentation or key point detection on a single frame of the video stream, identifies parking space markings and dedicated parking space indicator signs on the road surface, and generates a binary mask with "1" for legal parking space areas and "0" for other areas; the bottom center point or bounding box of the illegally parked vehicle is projected onto the binary mask, and if the vehicle does not fall into any legal parking space mask area, it is determined to be illegally parked.

2. The vehicle-mounted intelligent illegal parking detection and evidence collection device according to claim 1, characterized in that, The quick-release mounting bracket includes a fixed base, movable buckles, and locking bolts. It adopts a detachable connection method of buckle snap-fit ​​and bolt locking, which can be disassembled and assembled without special tools, and is compatible with at least three mainstream models of police motorcycle frames.

3. The vehicle-mounted intelligent illegal parking detection and evidence collection device according to claim 1, characterized in that, The audio-visual linkage unit includes a speech synthesis module, which uses a text-to-speech (TTS) engine and has multiple pre-set persuasion script templates containing license plate placeholders. After the system identifies the license plate number of the illegally parked vehicle, it selects a pre-set template from the multiple persuasion script templates, fills the identified license plate number into the license plate placeholder, generates a complete broadcast text, synthesizes the speech through the TTS engine, and plays it through the motorcycle's original speaker after power amplification.

4. The vehicle-mounted intelligent illegal parking detection and evidence collection device according to claim 1, characterized in that, The dual-mode imaging system includes a single-lens panoramic camera, an image stabilization component, and an automatic fill light array, which automatically turns on according to a predetermined value based on the ambient illumination.

5. A vehicle-mounted intelligent illegal parking detection and evidence collection device according to any one of claims 1-4, characterized in that, The main body of the device adopts a high-strength sealed shell, which is made of 6061-T6 aluminum alloy with a tensile strength ≥310MPa; the communication and storage module supports 4G / 5G networks, uses AES-256 encryption algorithm to encrypt the evidence data before uploading, and simultaneously achieves local backup.

6. A method for capturing and collecting evidence of illegally parked vehicles using an intelligent onboard camera, characterized in that, Using the device according to any one of claims 1-5, the method includes the following steps: S1: Equipment installation: Fix the main body of the equipment to the rear of the police motorcycle using a quick-release mounting bracket, complete the communication connection between the handlebar operation terminal and the main body of the equipment, and connect the power management module to the motorcycle's ACC power supply. S2: Video capture and dual illegal parking determination. The dual-mode imaging system continuously captures video streams from the roadside ahead and transmits them to the AI ​​intelligent recognition engine. The AI ​​intelligent recognition engine completes dual illegal parking determination through effective distance filtering and legal parking space area filtering. S3: Capture and information overlay. If both illegal parking criteria are met, the dual-mode imaging system is triggered to capture high-definition images and generate an evidence package. The communication and storage module receives BeiDou coordinates and timestamp data provided by the positioning and time synchronization module, overlays them onto the evidence package and matches the road name. S4: Audio-visual dissuasion; the audio-visual linkage unit plays dissuasion voice messages according to the instructions of the AI ​​intelligent recognition engine. S5: Data Upload. The communication and storage module uploads the encrypted evidence package to the traffic police integrated command platform.

7. The method for capturing and collecting evidence of illegal parking by an onboard vehicle according to claim 6, characterized in that, In step S1, the quick-release mounting bracket can be disassembled and assembled without special tools, and is completed by only two operations: snap-fit ​​and bolt locking.

8. A method for capturing and collecting evidence of illegal parking by an onboard vehicle according to claim 6, characterized in that, In step S2, the logic for the dual illegal parking determination is as follows: the vehicle is determined to be illegally parked only when it is outside the legal parking space area and the actual physical distance D between it and the equipment satisfies 0.5m≤D≤3m.

9. A method for capturing and collecting evidence of illegal parking by an onboard vehicle according to claim 6, characterized in that, When both illegal parking criteria in step S2 are met, the system automatically triggers the dual-mode imaging system to capture high-definition images and generate an evidence package. The communication and storage module receives BeiDou coordinates and timestamp data provided by the positioning and time synchronization module, overlays them onto the evidence package, and matches the road name. Simultaneously, the audio-visual linkage unit plays a warning message. Finally, the communication and storage module uploads the encrypted evidence package to the traffic police integrated command platform, forming a non-contact flexible law enforcement closed loop. Law enforcement officers also directly execute the subsequent image overlay, voice warning, and data upload processes in this step through a dedicated law enforcement mobile terminal.

Images