AI-Based System for Automatic Detection and Analysis of Two-Wheeled Vehicle Violations at Crosswalks and Linkage to Administrative Enforcement
An AI-based system for detecting motorcycle violations and helmet compliance on crosswalks addresses the limitations of existing systems, ensuring efficient and effective traffic enforcement and public safety management.
Patent Information
- Application Number
- KR1020260025711
- Authority / Receiving Office
- KR · KR
- Patent Type
- Patents
- Current Assignee / Owner
- Filing Date
- 2026-02-09
- Publication Date
- 2026-11-06
- Estimated Expiration
- 2046-02-09
AI Technical Summary
Existing traffic enforcement systems struggle to accurately identify motorcycle violations on crosswalks and manage additional public safety information like helmet compliance and wanted vehicle status, leading to inefficiencies and safety risks.
An AI-based system that integrates image analysis, deep learning object recognition, and administrative linkage to automatically detect motorcycle violations, analyze helmet usage, and determine if a vehicle is wanted or fleeing, storing data for administrative sanctions.
Enables accurate, real-time enforcement of motorcycle violations and helmet compliance, improving traffic safety and public security management by integrating data management and administrative processes.
Smart Images

Figure 112026016937031-PAT00003_ABST
Abstract
Description
Technology Field
[0001] The present invention relates to an AI-based automatic detection and analysis of two-wheeled vehicle driving violations on crosswalks and a system linked to administrative sanctions.
[0002] More specifically, the present invention relates to the field of traffic enforcement and public safety technology, and more specifically, to an AI-based automatic enforcement system that analyzes video data of a crosswalk captured by AI on a local server at the site to automatically determine whether a motorcycle is driving on the crosswalk, stores the violation in a database, and transmits it to the relevant department of a government agency to carry out administrative sanctions.
[0003] The present invention corresponds to a smart traffic enforcement technology that integrates intelligent image analysis, deep learning object recognition, license plate recognition, and administrative system linkage technology. Background Technology
[0005] Recently, violations involving motorcycles driving on crosswalks have been increasing, posing a serious threat to pedestrian safety.
[0006] However, existing traffic enforcement systems are primarily designed around automobiles, so they have limitations in accurately identifying small vehicles with different driving characteristics, such as motorcycles.
[0007] Furthermore, most current pedestrian crossing enforcement is conducted by reviewing CCTV footage retrospectively, which consumes a significant amount of enforcement manpower and makes real-time response difficult. In particular, there is a problem in that additional public safety information, such as helmet non-compliance or wanted vehicle status, is managed in a separate system, making integrated utilization difficult.
[0008] Therefore, a system is required that automatically detects motorcycle violations at crosswalks while simultaneously comprehensively analyzing the status of the motorcycle and its rider.
[0009] Korean Patent Publication No. 10-2023-0036699 addresses the unfortunate reality that most vehicle accidents involving children at crosswalks near elementary schools are caused by minor driver negligence. Despite these areas surrounding elementary schools being specific locations subject to systematic speed limits and special management, such accidents are not decreasing and are actually on the rise. Considering the persistent conditions for accidents caused by driver negligence, this invention was devised with the need for stronger institutional safeguards and safety facilities to alert drivers, at least in areas around elementary schools. It implements a new safety device that unfolds toward the road in accordance with the pedestrian signal of the crosswalk traffic light to block vehicles from entering the crosswalk, thereby preventing vehicles from entering the crosswalk during the pedestrian signal and allowing pedestrians to cross safely.
[0010] Korean Patent Publication No. 10-2023-0120369 was devised in consideration of the fact that although there have been recent cases utilizing LEDs or warning alarms near crosswalks to prevent safety accidents caused by pedestrian inattention due to smartphone use, these measures are applied only to visual or auditory safety, making practical prevention difficult. By illuminating the crosswalk, drivers can easily identify pedestrians, and through the pedestrian information unit, drivers can recognize that a pedestrian is crossing their driving path, thereby enabling safer driving.
[0011] However, the aforementioned conventional literature had a problem in that it could not prevent safety accidents caused by dangerous driving, even though two-wheeled vehicles such as motorcycles should cross crosswalks on foot. Prior art literature
[0013] Republic of Korea Patent Publication No. 10-2023-0036699 (Published March 15, 2023, Title: Pedestrian Crossing Safety Device) Republic of Korea Patent Publication No. 10-2023-0120369 (Published August 17, 2023, Title: Pedestrian Crossing Safety Accident Prevention System) The problem to be solved
[0014] The objective of the present invention is to provide an AI-based automatic detection and analysis of motorcycle driving violations on crosswalks and a system for linking administrative sanctions, which automatically and accurately determines whether a motorcycle is driving on a crosswalk from video footage of a crosswalk, and automatically transmits the violation to the relevant government department to carry out administrative sanction procedures.
[0015] Another objective of the present invention is to provide an AI-based automatic detection and analysis system for pedestrian crossing motorcycle driving violations and a system linked to administrative sanctions, which determines whether a motorcycle rider is wearing a helmet and whether the vehicle is a specific vehicle (wanted vehicle, hit-and-run vehicle, etc.) through image analysis.
[0017] The technical problems of the present invention are not limited to those mentioned above, and other unmentioned technical problems will be clearly understood by those skilled in the art from the description below. means of solving the problem
[0019] The AI-based automatic detection analysis and administrative disposition linkage system of the present invention for achieving the above objective
[0020] A local on-site server that stores and manages various video data and processed data collected at the crosswalk site;
[0021] A government agency linkage communication unit that transmits and receives data between the above-mentioned local server and the server system of an external government agency via wired or wireless communication;
[0022] An image capturing unit that generates image data by capturing images of the monitored area of a crosswalk in real-time or periodically;
[0023] An alarm output unit that outputs at least one of a warning sound, a warning light, and a voice guidance according to image data or analysis results obtained from the above image capturing unit;
[0024] A database storing image data captured by the above-mentioned image capturing unit, analysis result data, and administrative processing history data;
[0025] An AI video analysis unit that performs motorcycle object recognition and driving analysis at a crosswalk based on video data input from the above-mentioned video capturing unit;
[0026] A helmet wearing analysis unit that analyzes whether a motorcycle rider is wearing a helmet based on the AI video analysis unit recognized by the above-mentioned unit; and
[0027] It comprises a specific vehicle determination unit that analyzes a moving object recognized by the above AI image analysis unit and determines whether it is a vehicle that meets specific conditions;
[0028] It is characterized by generating violation information based on the judgment results of the above-mentioned video capturing unit, helmet wearing analysis unit, and specific vehicle judgment unit.
[0029] In the present invention, the AI image analysis unit
[0030] It is characterized by being configured to distinguish and recognize people, vehicles, and protective equipment using a deep learning-based object recognition model.
[0031] In the present invention, the helmet wear analysis unit
[0032] It is characterized by extracting the head region of a motorcycle rider and determining whether a helmet is worn by analyzing at least one of the shape, color, or reflection pattern of the helmet in the area.
[0033] In the present invention, the specific vehicle determination unit
[0034] It is characterized by determining whether a specific vehicle is in a wanted or fleeing situation by using at least one of the vehicle license plate recognition result, vehicle type information, and pre-registered vehicle data. Effects of the invention
[0036] According to the AI-based automatic detection and analysis of pedestrian crossing violations by two-wheeled vehicles and the linkage of administrative sanctions of the present invention, pedestrian crossing violations by two-wheeled vehicles can be automatically and accurately enforced.
[0037] In addition, by analyzing helmet usage and whether the vehicle is wanted or fleeing, traffic safety and public security management effects can be achieved simultaneously.
[0039] In addition to the above, the specific effects of the present invention are described together with the specific details for implementing the invention below. Brief explanation of the drawing
[0041] FIG. 1 is a network configuration diagram of an AI-based automatic detection analysis and administrative disposition linkage system for pedestrian crossing two-wheeled vehicle driving violations according to an embodiment of the present invention. FIG. 2 is a crosswalk configuration diagram to which the present invention is applied. FIG. 3 is a control configuration diagram of the local server or control server of FIG. 1. FIG. 4 is a flowchart of the operation for detecting violations of two-wheeled vehicle driving on a crosswalk according to an embodiment of the present invention. FIG. 5 is a flowchart of the operation for analyzing whether a two-wheeled vehicle rider is not wearing a helmet according to an embodiment of the present invention. FIG. 6 is a flowchart of a specific vehicle determination operation according to an embodiment of the present invention. FIG. 7 is a flowchart of intermittent operation in night and adverse weather environments according to an embodiment of the present invention. Clear embodiments of the present invention are illustrated in the drawings above and will be described in more detail below. These drawings are not intended to limit the scope of the spirit of the present invention in any way, but are intended to enable those skilled in the art to understand the concept of the present invention by referring to specific embodiments. Specific details for implementing the invention
[0042] The present invention is capable of various modifications and may take various forms, and specific embodiments are illustrated in the drawings and described in detail in the text.
[0043] However, this is not intended to limit the invention to the specific disclosed forms, and it should be understood that it includes all modifications, equivalents, and substitutions that fall within the spirit and scope of the invention.
[0044] The terms used in this application are used merely to describe specific embodiments and are not intended to limit the invention. The singular expression includes the plural expression unless the context clearly indicates otherwise.
[0045] Unless otherwise defined, all terms used herein, including technical or scientific terms, have the same meaning as generally understood by those skilled in the art to which this invention pertains.
[0046] Terms such as those defined in commonly used dictionaries should be interpreted as having meanings consistent with their meanings in the context of the relevant technology, and should not be interpreted in an ideal or overly formal sense unless explicitly defined in this application.
[0048] Hereinafter, preferred embodiments of the present invention will be described with reference to the attached drawings.
[0049] FIG. 1 is a network configuration diagram of an AI-based automatic detection analysis of motorcycle driving violations on crosswalks and a system linked to administrative sanctions according to an embodiment of the present invention. FIG. 2 is a crosswalk configuration diagram to which the present invention is applied. FIG. 3 is a control configuration diagram of the local server or control server of FIG. 1. FIG. 4 is a flowchart of the operation for detecting motorcycle driving violations on crosswalks according to an embodiment of the present invention. FIG. 5 is a flowchart of the operation for analyzing whether a motorcycle rider is not wearing a helmet according to an embodiment of the present invention. FIG. 6 is a flowchart of the operation for determining a specific vehicle according to an embodiment of the present invention. FIG. 7 is a flowchart of the operation for enforcement in night and adverse weather environments according to an embodiment of the present invention.
[0051] As illustrated in FIGS. 1 and 2, the AI-based automatic detection analysis and administrative disposition linkage system for pedestrian crossing two-wheeled vehicle driving violations according to an embodiment of the present invention may be composed of a local server (100) installed near a pedestrian crossing, a control server (200) that receives and processes pedestrian crossing violation facts transmitted from a plurality of local servers (100), and a storage medium (300) that stores data processed by the local servers (100). The storage medium (300) may be a USB.
[0052] A video recording device (CCTV or AI camera) (120) installed near an intersection or crosswalk captures road conditions in real time and simultaneously collects traffic signal information, such as pedestrian signals, vehicle signals, and the timing of signal state changes, through a signal controller or a signal information receiving module. The collected video data and signal information are transmitted to a local server (100) installed at the site via wired or wireless communication.
[0053] The local server (100) performs a predefined image analysis algorithm on the received image. Specifically, it uses an artificial intelligence-based object recognition model to identify a motorcycle in the image and determines whether the motorcycle object enters the area.
[0054] The local server (100) determines whether a violation has occurred by combining the motorcycle's location information and traffic signal information. For example, if a motorcycle enters a crosswalk area or crosses the stop line while the pedestrian signal is green, it is determined to be a violation. Alternatively, a violation can be determined based on whether the motorcycle encroaches on a crosswalk within a specific time interval set for pedestrian protection. These determination criteria are applied according to enforcement policies stored in the local server or setting values transmitted from the control server (200). When the violation condition is met, the local server (100) confirms the event as a violation incident.
[0055] When a violation is confirmed, the local server (100) extracts video frames from before and after the time of the violation to generate evidence video or still images. Along with this, it generates violation record data including metadata such as the time of violation, location information, signal status, and equipment ID.
[0056] The control server (200) is a server configured in an administrative agency that distinguishes, identifies, and stores data received from multiple local servers (100). Through this, enforcement data at the national or specific regional level can be managed in an integrated manner.
[0057] The control server (200) performs additional verification procedures based on the received violation data or undergoes verification by an administrator. The verified violation information is linked to a traffic enforcement system or a fine imposition system and used for administrative disposition.
[0058] Additionally, the control server (200) performs analysis of violation frequency, statistics by time period, and risk levels by region, which can be used to establish future enforcement policies or adjust enforcement standards. If necessary, the control server (200) transmits the updated enforcement standards or AI models back to the local server (1).
[0059] The control server (200) has the configuration shown in FIG. 3 and can perform the same control operations as the field local server (100). That is, the local server (100) in FIG. 3 can be replaced by the control server (200).
[0060] According to the above configuration, all administrative actions corresponding to the violation, such as notices of violation or bills, can be processed by the local government based on the records registered in the computer system.
[0062] As illustrated in FIG. 3, the AI-based automatic detection analysis and administrative disposition linkage system for pedestrian crossing two-wheeled vehicle driving violations according to an embodiment of the present invention is composed of a local server (100), a government office linkage communication unit (110), a video capturing unit (120), an alarm output unit (130), a database (140), an AI video analysis unit (151), a helmet wearing analysis unit (152), and a specific vehicle determination unit (153).
[0063] The local server (100) at the site receives a crosswalk video transmitted from the video capturing unit (120) and performs video analysis and data processing, and has a pre-trained deep learning-based object recognition model stored therein.
[0064] The above object recognition model may be a model pre-trained using a large-scale traffic video dataset to distinguish various road moving objects such as motorcycles, pedestrians, passenger cars, and trucks.
[0065] The local server (100) performs object detection and object classification for each received video frame to identify the location, speed, size, and type of an object moving within the video in the form of a bounding box. In this process, if the detected object corresponds to a moving two-wheeled vehicle, a unique identification ID is assigned to the object to enable tracking between frames.
[0066] The local server (100) predefines a specific coordinate area within the image as a crosswalk area (ROI, Region of Interest) at the time of camera installation or during the initial system setup phase.
[0067] The above crosswalk area can be set based on crosswalk painting patterns, lane information, or polygonal coordinates directly designated by the administrator. Subsequently, the local server (100) analyzes the position change of the two-wheeled vehicle object between frames to generate a movement trajectory and determines whether the movement trajectory overlaps with or penetrates the crosswalk area for a certain period of time or longer. Through this, it is possible to distinguish between cases where the vehicle simply drives near the crosswalk and cases where it actually encroaches upon the crosswalk.
[0068] As a result of the analysis, if it is determined that the movement trajectory of a specific two-wheeled vehicle object passes through the crosswalk area or satisfies a preset intrusion condition, the local server (100) determines this as a crosswalk two-wheeled vehicle driving violation event. At this time, the local server (100) automatically collects image data including multiple frames before and after the time of violation, video clips of a certain length, information on the time of violation occurrence, camera installation location information, and device identification information. The collected data is stored in a database (140) along with hash values or electronic signature information for preventing forgery and alteration.
[0069] A local server (100) is installed near a crosswalk to minimize network delay, and the received video is transmitted to the AI video analysis unit (151) of the local server (100).
[0070] The local server (100) acts as a hub for transmitting analysis results to the database (140) and the government office department in charge.
[0071] The government office linkage communication department (110) transmits data stored in the database (140) to the government office in charge using a wired or wireless communication network to perform administrative disposition.
[0072] The data transmitted to the relevant department of the aforementioned government office includes video evidence of the violation and analysis results.
[0073] The video capturing unit (120) includes a camera installed to capture the crosswalk area and provides video data for determining whether a two-wheeled vehicle is driving within the crosswalk, and may include a high-resolution camera and an infrared camera to enable daytime and nighttime shooting.
[0074] The video recording camera (120) may include a high-resolution image sensor and a wide-angle lens to enable stable shooting in both daytime and nighttime environments, and may generate continuous video data at a rate of 30 frames or more, preferably 60 frames or more.
[0075] The video capturing unit (120) is installed, for example, in front of, on the side or above the crosswalk to continuously capture video including the crosswalk area, and the captured video data is transmitted to a local server (100) in real time or at regular intervals.
[0076] Meanwhile, the video recording unit (120) may be configured to be linked with a traffic light under the control of a local server (100) to record video of the crosswalk situation during the pedestrian traffic light illumination time.
[0077] Additionally, the video capturing unit (120) may be configured to capture license plates attached to the front and rear of a specific vehicle that is wanted or fleeing and passing through a crosswalk at times other than when the pedestrian signal light is on, under the control of the local server (100). To this end, the video capturing unit (120) may be configured to rotate so that the license plate is captured accurately.
[0078] In addition, the video recording unit (120) may also record a video of a motorcycle or vehicle moving in reverse.
[0079] The alarm output unit (130) outputs an alarm when a motorcycle traveling on the crosswalk is detected by the AI video analysis unit (151) and when a specific vehicle passing through the crosswalk is detected by the specific vehicle judgment unit (153), and may be provided in a local server (100) or a central control server (2).
[0080] The above alarm is provided in the form of a visual notification, an audible alarm, or a control system pop-up, allowing the person in charge to take immediate response measures.
[0081] A display composed of LEDs, etc., is configured together with the above alarm output unit (130) so that a person in charge can recognize when a motorcycle traveling on the crosswalk is detected and when a specific vehicle passing through the crosswalk is detected by the specific vehicle judgment unit (153).
[0082] The database (140) is a storage unit that stores facts regarding violations of two-wheeled vehicle crossing violations, and stores video, images, time information, location information, helmet wearing status, and specific vehicle determination results at the time of the violation. The data is retained for a certain period for post-verification and administrative disposition.
[0083] The database (140) can be built on a local server or a central control server.
[0084] The AI video analysis unit (151) configured in the violation analysis and administrative processing unit (150) includes a deep learning-based object recognition and tracking algorithm and is an artificial intelligence module equipped in the local server (100) to analyze video data, detecting objects in the video and distinguishing between motorcycles and pedestrians.
[0085] In addition, it determines whether a motorcycle passes through the crosswalk based on a predefined crosswalk area, and generates a violation signal if it is determined that the motorcycle has traveled across the crosswalk. At this time, the speed of the motorcycle is analyzed to distinguish whether it is crossing the crosswalk with pedestrians or traveling across it.
[0086] The helmet wearing analysis unit (152) configured in the violation analysis and administrative processing unit (150) is linked with the AI image analysis unit (151) to analyze the condition of the two-wheeled vehicle rider. When the two-wheeled vehicle is detected, the rider's head area is extracted, and the determination of whether a helmet is worn is performed based on image features or a learned classification model.
[0087] The analysis results are linked with the above violation data and stored in the database (140).
[0088] This allows for the simultaneous management of driving violations on crosswalks and failure to wear a helmet.
[0089] The specific vehicle determination unit (153) configured in the violation analysis and administrative processing unit (150) analyzes whether a motorcycle is a wanted vehicle or a fleeing vehicle, and extracts vehicle identification information of the wanted or fleeing vehicle including a license plate recognition module.
[0090] The extracted identification information is compared with a pre-stored database of specific vehicle lists, and if the comparison result determines that it is a specific vehicle, a specific vehicle detection signal is generated.
[0091] In this way, the specific vehicle judgment unit (153) simultaneously performs traffic enforcement and public security management functions.
[0092] The computerized registration unit (154) configured in the violation analysis and administrative processing unit (150) computerizes the data determined by the specific vehicle judgment unit (153).
[0094] The operation of the present invention configured as described above is explained by case-by-case as follows.
[0095] First, I will explain the automatic detection and storage operation for motorcycle driving violations on crosswalks.
[0096] The video recording camera (120) transmits video data in real time to a local server (100) installed at the site via a communication network.
[0097] The local server (100) determines whether the two-wheeled vehicle object is driving on the crosswalk by determining whether the vehicle's movement path enters the crosswalk ROI, stays for a certain period of time or moves through the ROI.
[0098] Meanwhile, the aforementioned crosswalk area may be set as an area expanded by a certain width compared to the marked crosswalk. This is because motorcycles can cross the road while driving around crosswalks used by pedestrians.
[0099] In this process, a multi-frame-based reliability verification algorithm may be applied to prevent misrecognition caused by temporary overlap of pedestrians, stationary motorcycles, or camera shake. Additionally, it may be possible to further determine whether a motorcycle is in motion during a pedestrian signal by linking with traffic light status information.
[0100] When the analysis results determine that the motorcycle has driven in the crosswalk area, the local server (100) automatically generates a core image frame corresponding to the time of violation and a video clip of a certain length of time before and after the violation.
[0101] In addition, information such as the time of filming, camera installation location, crosswalk identification number, and object tracking ID are generated together in the form of metadata. This violation-related data is automatically stored in a database (140) and can be used later for administrative enforcement, statistical analysis, or integration with a remote control system.
[0102] Thus, the present invention provides the effect of automatically detecting violations of pedestrian crossing driving by motorcycles and systematically storing violation evidence data by combining image-based artificial intelligence analysis and pedestrian crossing area recognition technology.
[0104] Next, the analysis of helmet wearing status is performed as follows.
[0105] The helmet wearing status analysis operation is performed to automatically determine the driver's helmet wearing status for two-wheeled vehicles detected to be driving on a crosswalk.
[0106] The AI video analysis unit (151) expands the analysis target area based on the bounding box of the object as soon as a two-wheeled vehicle object is detected in the video data received from the video recording camera (120).
[0107] Specifically, a region of interest (ROI) is established that extends upward by more than a certain proportion to include not only the body area of the two-wheeled vehicle but also the upper body, shoulders, neck, and head area of the rider located on top of the two-wheeled vehicle.
[0108] The AI video analysis unit (151) may additionally apply a human body pose estimation model or a face and head detection model to precisely separate the occupant's head area from the expanded region of interest.
[0109] In this process, object tracking algorithms can be applied in parallel to stably track the occupant's head position across multiple frames. This enables more accurate analysis of helmet usage even when shaking, changes in shooting angles, or partial occlusion occur during driving.
[0110] For the extracted head region, a pre-trained deep learning-based classification model is applied to determine whether a helmet is being worn.
[0111] The above classification model may be designed to distinguish between a helmet-worn state, a helmet-unworn state, or an indistinguishable state, and may be a model trained using a training dataset that includes data on helmets, hats, hoods, bare heads, etc., having various colors, shapes, and reflective characteristics.
[0112] The AI video analysis unit (151) assigns a confidence score to the classification result for each frame and determines the final helmet wearing status by combining the analysis results of multiple frames.
[0113] If, as a result of the analysis, it is determined that the helmet is not worn beyond a set threshold confidence level, the local server (100) generates a helmet non-wearing violation event for the corresponding motorcycle object. At this time, the image frame at the time of determining that the helmet is not worn, the enlarged image of the helmet analysis area, the determination result, and the confidence information are logically linked with the pedestrian crossing driving violation information. The linked violation information is stored together in the database (140) based on the same event ID or vehicle (motorcycle) object ID.
[0114] In this way, by integrating and storing information on driving violations on crosswalks and violations regarding failure to wear a helmet, it becomes possible to conduct not only enforcement for individual violations but also automated enforcement and statistical analysis of complex traffic law violations.
[0115] Furthermore, administrators can simultaneously verify pedestrian crossing violations and helmet non-compliance through stored data, significantly improving the efficiency and reliability of enforcement operations. This configuration strengthens the safety of motorcycle riders and maximizes the effectiveness of traffic accident prevention.
[0117] Next, the detection and computer registration of specific vehicles (wanted or fleeing vehicles) is carried out as follows.
[0118] Specific vehicle detection and computerized registration operations are performed to rapidly identify wanted or fleeing vehicles and enable an immediate response.
[0119] The local server (100) automatically searches for and extracts a license plate area located at the rear or front of the vehicle body by the AI image analysis unit (151). The extraction of the license plate area is performed by combining a license plate location detection model and an image preprocessing algorithm, and is configured to enable stable recognition even in poor shooting environments such as lighting changes, tilt, contamination, and partial obscuration.
[0120] A license plate recognition module included in a local server (100) performs preprocessing steps such as distortion correction, contrast enhancement, and noise removal on the extracted license plate area, and then applies an optical character recognition (OCR)-based or deep learning-based character recognition model to extract character and number information of the license plate.
[0121] The confirmed license plate information is compared and analyzed in real time with a specific vehicle list database that is stored in a local server (100) or periodically synchronized from a central control server (200).
[0122] The aforementioned specific vehicle list database may include license plate information of vehicles managed for law enforcement or administrative purposes, such as wanted vehicles, getaway vehicles, and vehicles subject to seizure.
[0123] The local server (100) can analyze not only whether the license plate string is a perfect match, but also a similarity-based comparison method that takes into account some character recognition errors.
[0124] As a result of the analysis, if the detected license plate information matches the specific vehicle list mentioned above, or if it is determined to correspond to a wanted vehicle or a fleeing vehicle with a similarity level greater than or equal to the set similarity standard, the local server (100) registers the event occurrence data (date / time / photo / vehicle number) in the computerized register (154). At this time, an alarm may also be output.
[0125] The above alarm event may be displayed as a visual warning on the monitoring screen of the local server (100) itself or provided as an audible alarm through a speaker. Additionally, it may be transmitted in real time to the central control server (200) via a network and displayed on the control system screen as a pop-up notification, color highlight, or message.
[0126] When such an alarm is output, the vehicle's license plate information, captured image or video clip, detection time, installation location information, and violation type information are provided together, allowing the person in charge to immediately grasp the situation. Through this, field agents or control center personnel can rapidly track wanted or fleeing vehicles, request on-site dispatch, or take additional enforcement measures.
[0127] Therefore, according to this operation, the crime vehicle detection function can be performed in an integrated manner, which can significantly improve the efficiency of the public safety assurance and response system.
[0129] Next, the coordination with the relevant government department and the sending of bills are carried out as follows.
[0130] The operation of linking with the relevant government department and sending the notice is performed to quickly and accurately notify the violator of traffic violation information automatically detected and stored by the local server (100) in conjunction with administrative procedures.
[0131] Violation data stored in the database (140) of the local server (100) can be transmitted in batches at a preset interval, for example, at a set time interval or daily interval, or can be transmitted to the local server alarm and control server in real time when a violation event occurs.
[0132] The transmitted violation data includes the results of a complex violation determination in a structured data format, such as the type of violation, captured images and video clips, time of violation, installation location information, license plate information, and whether a helmet was worn.
[0133] The server of the relevant government department stores the received violation data and performs an automated verification procedure to determine whether the data is missing, forged, or altered, and whether legal requirements are met. During this process, the resolution of the violation image, the reliability of license plate recognition, and whether violation judgment criteria are met may be reviewed.
[0135] Next, the operation for performing stable enforcement even in night and adverse weather conditions is as follows.
[0136] The operation for performing stable enforcement even in the above-mentioned night and adverse weather environments is performed to maintain the reliability of motorcycle detection and violation judgment even in situations where image quality deteriorates due to reduced illumination or adverse weather conditions.
[0137] To this end, video cameras installed near crosswalks are configured to include not only a visible light shooting mode but also an infrared (IR) shooting function, or to include a high-sensitivity image sensor that minimizes noise in low-light environments.
[0138] In addition, the camera includes automatic exposure control, automatic white balance, and backlight compensation functions, enabling stable image acquisition even with variations in light intensity caused by nighttime streetlights or vehicle headlights.
[0139] The local server (100) applies an image quality correction algorithm as a preprocessing step to image data received from the camera. The image quality correction algorithm may include brightness and contrast adjustment, noise removal, shake correction, and blur removal, and is designed to mitigate image degradation that occurs in nighttime footage or in situations involving rain or snow. This correction processing is performed in real time to provide image data of a quality suitable for the AI analysis step.
[0140] The AI video analysis unit (151) uses a pre-trained deep learning-based analysis model by utilizing traffic video data containing various environmental noises such as changes in illumination, raindrops, snowflakes, and fog as training data.
[0141] The above model can be configured to detect objects based on the morphological characteristics, movement patterns, and contour information of a two-wheeled vehicle, even if pixel distortion or partial occlusion of objects occurs due to bad weather.
[0142] Additionally, the local server (100) can dynamically adjust analysis parameters according to the shooting environment. For example, an environment-adaptive analysis can be performed by selecting an object recognition model optimized for infrared images at night, or by increasing the noise removal intensity during rainfall. Through this, stable performance can be secured under various environmental conditions compared to relying on a single analysis model.
[0143] Based on this configuration and operation, the accuracy of motorcycle detection and the reliability of violation judgments are maintained above a certain level even in adverse environments such as night, rain, or snow. Consequently, consistent enforcement performance can be secured regardless of the time of day or weather conditions, thereby significantly enhancing the effectiveness and reliability of the traffic enforcement system.
[0145] As seen above, the detailed description of the present invention has described preferred embodiments of the invention; however, this is merely an illustrative description of the best embodiments of the invention and is not intended to limit the invention. Furthermore, it is obvious that anyone skilled in the art to which the present invention pertains can make various modifications and imitations within the scope of the technical concept of the present invention without departing from it. Explanation of the symbols
[0147] 100 : Local Server 110 : Communications Unit 120: Video recording unit 130: Alarm output unit 140: Database 150: Violation Analysis and Administrative Processing Department 151 : AI Video Analysis Unit 152 : Helmet Wear Analysis Unit 153 : Specific Vehicle Identification Unit 200 : Control Server
Claims
Claim 1 A local server at the site that stores and manages various video data and processed data collected at the crosswalk; the local server calculates reliability using object tracking results for multiple frames, confirms a violation only if it exceeds a set threshold, determines whether a violation has occurred by analyzing the pedestrian signal status and the movement trajectory within the crosswalk area together, generates violation record data including multiple frames before and after the time of violation, the time of violation occurrence, the installation location, and the object tracking ID, and stores the violation data in a database along with hash values or electronic signature information; a government agency linkage communication unit that transmits and receives data between the local server and the server system of an external government agency via wired or wireless communication; a video capturing unit that generates video data by capturing images of the monitored area of the crosswalk in real-time or periodically; an alarm output unit that outputs at least one of a warning sound, a warning light, or voice guidance according to the video data or analysis results obtained from the video capturing unit; a database that stores video data captured by the video capturing unit, analysis result data, and administrative processing history data, and stores crosswalk driving violation information and helmet wearing analysis results linked to the same incident ID or object ID; and video data input from the video capturing unit An AI video analysis unit that performs motorcycle object recognition and driving analysis in a crosswalk area set as a surveillance area extended outward beyond the actual crosswalk marking, configured to distinguish and recognize people, vehicles, and protective equipment respectively using a deep learning-based object recognition model, generates a movement trajectory of a two-wheeled vehicle object for a predefined crosswalk area (ROI), and determines a violation of driving on a crosswalk if the movement trajectory overlaps with or penetrates the crosswalk area for a certain period of time or longer;An AI-based automatic detection and analysis of motorcycle driving violations at crosswalks and linked administrative disposition system, characterized by generating violation information based on the judgment results of the image capturing unit, the helmet wearing analysis unit, and the specific vehicle determination unit, which analyzes whether a motorcycle rider recognized by the AI image analysis unit is wearing a helmet, extracts the head region of the motorcycle rider, analyzes at least one of the shape, color, or reflection pattern of the helmet in the area to determine whether a helmet is worn, and generates an ROI including the rider's head region by expanding the bounding box of the two-wheeled vehicle object upward; and a specific vehicle determination unit that analyzes a moving object recognized by the AI image analysis unit to determine whether it is a vehicle that meets specific conditions, and determines whether it is a specific vehicle in a wanted or fleeing situation using at least one of a vehicle license plate recognition result, vehicle type information, and pre-registered vehicle data. Claim 2 delete Claim 3 delete Claim 4 delete
Citation Information
Patent Citations
Unmanned traffic enforcement system for enhancing accuracy of enforcement
KR102656252B1
Smart information detection system mounted on the vehicle and smart information detection method using the same
KR101066081B1
Deep learning-based risk factor object detection system and traffic safety guidance method for pedestrian crossing safety
KR1020250022950A
Traffic violation enforcement system for motor bicycle
KR102368127B1
