Hierarchical scanning, automatic identification and processing method based on traffic violation snapshot pictures

Abstract

The invention relates to a hierarchical scanning, automatic identification and processing method based on traffic violation snapshot pictures, and belongs to the field of image processing and traffic management. The technical method aims at improving the scanning efficiency, accurately recognizing traffic violation information and achieving automatic processing and uploading so as to improve the processing efficiency and accuracy of traffic violation snapshot data. Comprising the steps of hierarchical scanning, automatic identification, data verification and uploading, and exception handling and feedback. According to the method, the stored snapshot pictures are quickly positioned and extracted by presetting the folder hierarchy depth and being compatible with various folder hierarchy structures, and each picture is marked, so that repeated scanning and missing detection are avoided, and the scanning efficiency is remarkably improved. The OCR technology is combined with the deep learning algorithm, so that the recognition accuracy is remarkably improved, and particularly, error recognition can be effectively reduced in a complex environment. After identification is completed, the system verifies the accuracy of an identification result through an automatic verification mechanism, seamless connection and processing of data are completed, manual intervention is reduced, and law enforcement efficiency is improved.

Description

Technical Field

[0001] The invention relates to the field of image processing and traffic management, and in particular to a layered scanning, automatic identification and processing method based on traffic violation captured pictures. Background Art

[0002] With the development of intelligent transportation systems, traffic violation capture technology has been widely used in urban traffic management. By deploying illegal capture equipment on key roads, pictures of traffic violations can be automatically captured and recorded. These pictures are usually stored in the server in a hierarchical structure based on time or geographic location, such as "year / month / day" or "jurisdiction / area / road section". However, the existing technology has the following problems when processing these captured pictures: Low scanning efficiency: Due to the complex image storage structure and large number, it is difficult for the existing method to efficiently traverse the folders at all levels, resulting in a slow scanning process and easy omission or repeated scanning. Insufficient recognition accuracy: Due to factors such as ambient lighting and overlapping of multiple vehicles, the recognition accuracy of the camera may be limited, resulting in errors or missing information such as license plate number, violation time and location. Subsequent processing lags: The existing technology lags behind in data processing and uploading after recognition, and cannot achieve automated data verification, formatting and uploading, affecting law enforcement efficiency.

[0003] Therefore, there is an urgent need for a technical method that can improve scanning efficiency, accurately identify traffic violation information, and realize automated processing and uploading, so as to improve the processing efficiency and accuracy of traffic violation capture data. Summary of the invention

[0004] The present invention provides a layered scanning, automatic recognition and processing method based on traffic violation captured images, aiming to solve the problems of low scanning efficiency, poor recognition accuracy and imperfect subsequent processing in the prior art, thereby significantly improving the processing efficiency, accuracy and intelligence level of traffic violation captured data.

[0005] The technical solution of the present invention includes the following main steps: hierarchical scanning: by presetting the folder hierarchy depth, compatible with multiple folder hierarchy structures, hierarchical scanning is performed on the captured images in the server specified folder. The system can automatically traverse the folders of each level, quickly locate and extract the stored captured images, and mark each image to avoid repeated scanning and missed detection, and significantly improve the scanning efficiency. Automatic recognition: OCR technology is combined with deep learning algorithm to automatically extract information such as license plate number, violation time, violation location, camera IP, violation behavior, vehicle type, etc. in the image, optimize the recognition effect in complex scenes, such as uneven lighting, multiple vehicles overlapping, etc., improve the recognition accuracy, and compare with the preset template or database to automatically generate a data format that meets the background requirements. Significantly improve the recognition accuracy, especially in complex environments, it can effectively reduce misidentification. Data verification and upload: After the recognition is completed, the system verifies the accuracy of the recognition result through an automatic verification mechanism to ensure that the license plate number format, time range, location legality, etc. meet the standards. After the verification is passed, the system automatically uploads the recognition data to the traffic violation management system to complete the seamless connection and processing of the data, reduce manual intervention, and improve law enforcement efficiency. Data is uploaded to the traffic violation management system through an encrypted transmission protocol to ensure data security and real-time synchronization. Exception handling and feedback mechanism: For exceptions that occur during the recognition process (such as recognition errors, poor image quality, etc.), the system will automatically mark and generate detailed feedback reports for manual verification and correction to ensure efficient and accurate operation of the system and optimize the entire processing flow. The system has a self-learning function, which optimizes the recognition algorithm based on historical data and manual correction results to continuously improve accuracy.

[0006] Preferably, in the hierarchical scanning step, the system uses multi-threaded parallel scanning technology to greatly improve scanning efficiency while ensuring accuracy. This technology can process images in parallel in multiple folder levels, shorten the scanning time, and reduce the performance bottleneck caused by single-threaded processing. It supports flexible configuration, and users can set parameters such as scanning path, layer depth, file format, and scanning cycle.

[0007] Preferably, in the automatic recognition step, the image recognition technology combines the deep learning model with the traditional OCR (optical character recognition) technology, which can automatically recognize information such as the license plate number, the time of violation, the location of violation, and optimize the complex scenes such as multiple vehicles overlapping, blurring, and uneven lighting in the image. The deep learning model not only improves the recognition accuracy through large-scale data training, but also continuously optimizes the recognition algorithm to adapt to different traffic environments and capture scenes.

[0008] Preferably, in the data verification and uploading steps, the system automatically converts the identified data into a unified standard format through a data formatting module, and uploads it through an encryption protocol to ensure the security and confidentiality of the data. The uploaded data will be synchronized to the cloud database in real time for subsequent query and analysis by the traffic management department. During the data uploading process, the system will also perform real-time status monitoring to ensure the stability and reliability of the data uploading process.

[0009] Preferably, in the exception handling and feedback mechanism, the system will automatically generate a detailed log report based on the exception type (such as license plate recognition failure, substandard image quality, etc.), and notify relevant staff through backend station messages, etc. Staff can manually check the exception based on the feedback report and correct the abnormal data through the system interface to ensure the efficient operation of the entire system.

[0010] Preferably, the system also has a self-learning function, which can automatically optimize the recognition algorithm based on historical data and manual correction results. After each manual intervention, the system will use the corrected data as a training set to further improve recognition accuracy and reduce the need for future manual intervention.

[0011] Preferably, the hierarchical scanning system supports flexible configuration, and can flexibly adjust the depth of the folder hierarchy and the scanning strategy according to different traffic management requirements and server structures. Users can choose to scan snapshots of a specified time period or a specific area according to specific circumstances, thereby improving the flexibility and pertinence of the scan.

[0012] Preferably, the system also includes an abnormal alarm module. When abnormal data is identified or upload fails, the system will automatically trigger an alarm mechanism to promptly notify the administrator for processing. The alarm information includes the abnormal type, occurrence time, related pictures, etc., to ensure that the problem can be quickly resolved.

[0013] Preferably, the system supports seamless integration with the existing traffic violation management system, can directly transmit the recognition results and processed data to the traffic violation management platform, supports conversion of multiple data formats, and ensures the compatibility and fluency of data between different platforms.

[0014] Preferably, the system can also integrate the recognition function of various types of traffic violations, including but not limited to illegal parking, running red lights, speeding, driving in the wrong direction, vehicle restrictions, time-limited traffic, etc. The system can automatically extract relevant violation information according to different violation types and classify them for subsequent analysis and management.

[0015] Preferably, the system can also integrate the recognition function of various types of traffic violations, including but not limited to illegal parking, running red lights, speeding, driving in the wrong direction, vehicle restrictions, time-limited traffic, etc. The system can automatically extract relevant violation information according to different violation types and classify them for subsequent analysis and management.

[0016] Optionally, the system supports processing of multiple image formats, including common formats such as JPEG and PNG, and can pre-process images, such as denoising, contrast enhancement, etc., to improve image quality and further improve recognition accuracy.

[0017] Optionally, the system has adaptive capabilities and can automatically adjust the parameters of the image recognition algorithm according to different cameras and capture environments. For example, in a poorly lit environment, the system can automatically enhance the brightness and contrast of the image to optimize the recognition effect; in the case of heavy traffic, the system can automatically adjust the processing strategy for multiple overlapping vehicles to reduce recognition errors.

[0018] The present invention provides an efficient, accurate and intelligent method for processing traffic violation snapshot images, which has significant technical advantages. Through layered scanning, automatic identification, data verification and uploading and other technical means, it not only improves the processing efficiency of traffic violation data, but also improves the accuracy and automation level of identification, can effectively optimize the workflow of the traffic management system, improve law enforcement efficiency, and provide strong support for the further development of intelligent transportation systems. BRIEF DESCRIPTION OF THE DRAWINGS

[0019] Figure 1 :The present invention is a flow chart of a layered scanning, automatic identification and processing method based on traffic violation captured pictures.

[0020] Figure 2 : Hierarchical scanning flowchart.

[0021] Figure 3 : Automatic recognition flow chart. DETAILED DESCRIPTION

[0022] To make the content, technical solutions and advantages of the embodiments of the present application clearer, the specific implementation methods of the embodiments of the present application are described in detail below in conjunction with the accompanying drawings. The described embodiments are part of the embodiments of the present application, not all of the embodiments.

[0023] Figure 1 The present invention is a flowchart of a hierarchical scanning and automatic identification and processing method based on traffic violation snapshot images. In the present invention, the process of the method and system may include: reading configuration files, hierarchical scanning folders, identifying anomalies, automatically identifying image content, data verification, data formatting and encryption, and data uploading.

[0024] In the first embodiment, the illegal pictures captured by the front-end illegal capture device are divided into: illegal event pictures, illegal evidence pictures, checkpoint passing pictures, etc. These pictures will be uploaded to multiple servers such as server A, server B, server C, etc., and will be stored in different naming formats (the naming format and rules of each capture device may be different). The stored folder levels are not necessarily exactly the same. There are date formats (year / month / day / picture) and address formats (jurisdiction / area / section / picture); the software will run in multiple servers. After the software is turned on, it will automatically perform multi-threaded fast scanning, identification, and upload to the background server according to the contents set in the configuration file.

[0025] Figure 2 The hierarchical scanning process is one of the core parts of the present invention, which mainly realizes the efficient classification and batch processing of traffic violation captured pictures to improve the processing efficiency and resource utilization of the system.

[0026] Specifically, initializing the scanning parameters is the starting point of the entire process. The system reads the configuration file to complete the loading and initialization of the relevant parameters, providing a basis for subsequent hierarchical scanning. Configuration file content: Scan path, specify the root directory path to be scanned (for example, / illegal parking capture / event picture / ); file type, specify the file format to be processed (such as .jpg, .png, etc.); scan cycle, set the time interval for the system to perform the scan (such as scanning every 5 seconds); log path, define the storage path of the scan log, and record detailed information during the scanning process; background server address, abnormal files will be directly uploaded to the background server for manual review; level depth, recursive loop scanning will be performed according to the level depth, and no further scanning will be performed after exceeding the level depth; default number of threads, the number of multi-threaded scans that are enabled by default when running automatically.

[0027] The initialization operation will read the configuration file content and perform parameter verification to ensure that the path is valid and the file type is legal; initialize system resources, such as memory allocation, log file creation, scan queue initialization, etc.; record the initialization success information in the system log for subsequent tracking.

[0028] Specifically, hierarchical scanning of folders is the core step of the present invention. It mainly scans the folders under the root directory layer by layer through hierarchical recursive scanning to ensure the comprehensive collection of image files. Hierarchical division rules: According to the directory structure, the scanning path is divided into different levels, which may exist: First-level directory: year classification, such as "2024". Second-level directory: month classification, such as "12". Third-level directory: date classification, such as "01". Fourth-level directory: road section classification, such as "Ciqikou Business District Strictly Controlled Section 1". Hierarchical scanning logic: recursive scanning, the system starts from the root directory, traverses the subdirectories layer by layer, and finds all qualified image files. Multi-threaded optimization, for large batches of files, the system can use multi-threaded parallel scanning to improve scanning efficiency. Scanning process records: record statistical information such as the total number of scanned files, the number of valid files, and the number of abnormal files. The scanning status of each scanning path is recorded in the system log to facilitate troubleshooting.

[0029] Specifically, the main goal of file screening and classification is to remove invalid or abnormal files and classify valid files to provide data support for the automatic recognition process. File integrity check: File size check, skip files with a size of 0 or below the threshold. File readability check, try to open the file to ensure that the file is not damaged. Metadata check, check whether the EXIF ​​information of the image is complete, such as shooting time, device number, etc.

[0030] Specifically, abnormal file processing: During the scanning and screening process, abnormal files may be encountered, such as duplicate files, damaged files, or files that do not comply with the naming rules. The system will mark and process these files to ensure the integrity and accuracy of the data. Abnormal file types: Duplicate files, duplicate images are found by comparing the file name or hash value. Damaged files, files that cannot be opened or read. Abnormally named files, files that do not comply with the preset naming rules. Processing method: Move the abnormal file to the designated "abnormal folder". Record the file path, abnormal type, and time in the abnormal log. Provide a manual review interface for administrators to check and repair abnormal files.

[0031] Specifically, after scanning and screening, the system will output the scanning results and generate a picture file information list with the classification results of valid files, providing input data for subsequent automatic recognition.

[0032] Figure 3 The automatic recognition process is another core part of the present invention, which mainly realizes the extraction of key information and data standardization in the captured pictures.

[0033] Specifically, image reading and preprocessing: the system reads valid image files in the scan results one by one. Denoise, remove noise points, and improve image quality. Grayscale, simplify image data, and reduce the amount of calculation. Binarization, extract key information areas such as license plate, time, and location. Image enhancement, deal with problems such as uneven lighting and blur, to ensure recognition accuracy. The system marks the image before recognition to prevent repeated recognition by multiple threads. After recognition is completed, the mark status is updated to "processed" to ensure efficient execution of the recognition task.

[0034] Specifically, license plate recognition: Use the target detection algorithm (hyperlpr3) to locate the license plate area, crop the license plate area and convert it to RGB mode, load the cropped license plate image, define the HSV color range, and use HSV to identify the overall license plate color. Perform OCR recognition on the located license plate area and extract the license plate number. Use regular expressions to verify the license plate format to ensure validity.

[0035] Specifically, the identification of illegal behaviors, vehicle types, illegal time, illegal address, etc.: use OCR to extract time information from the image and format the time into a standard format (YYYY-MM-DD HH:MM:SS); OCR extracts location information and compares it with the preset location database for standardized output. Extract the camera registration number or IP address from the image; use a deep learning target detection model to identify illegal behaviors in the image (such as running a red light, crossing the line, etc.); use an image classification model and license plate color to comprehensively identify vehicle types (such as small cars, large cars, new energy small cars, new energy large cars, motorcycles, etc.).

[0036] Specifically, for data verification and formatting, the integrity and legality of the extracted data are verified to ensure that all fields meet the requirements. The extracted data is organized into a standard format and the pictures are renamed again: 10355_202401217_182924_728_Import_Teaching_Vehicle_Limit_50.22.36.211_Yu AQ39C9_42_40_Blue.jpg. The naming content includes the illegal act code, year, month, day, hour, minute, second, three-digit milliseconds, import / export, illegal module name bound to the illegal code, IP address of the抓拍 device, illegal license plate number, vehicle speed, section speed limit, and license plate color. And JSON format data for transmission is generated: {'data_type': '10355', 'ip': '50.22.36.211', 'car_id': 'Yu AQ39C9', 'car_type': '02', 'wf_time': '202401217182924','speed': '42', 'lim_speed': '40', 'img_type': 'Import','model_name': 'Teaching Vehicle Limit'}.

[0037] Specifically, for data encryption and upload: For data encryption, the recognition results are encrypted to ensure the security of data transmission. For data upload, the encrypted data is uploaded to the background system or database through an interface. For result feedback, the system returns the upload result and records the success or failure status.

[0038] Specifically, for abnormal data processing and self-learning: For abnormal data marking, the data with failed recognition or incomplete data is marked for manual review. For self-learning optimization, the data corrected manually is used as a training sample to optimize the recognition model and improve the system accuracy.

[0039] The above are only the embodiments of this application. The present invention is not limited by the above embodiments. What is described in the above embodiments and the specification is only the preferred examples of the present invention and is not used to limit the present invention. Without departing from the spirit and scope of the present invention, the present invention will have various changes and improvements, and these changes and improvements all fall within the scope of the present invention claimed; the scope of protection claimed by the present invention is defined by the appended claims and their equivalents.

Claims

1. A layered scanning, automatic identification and processing method based on traffic violation snapshot images, characterized in that: The following steps are involved: Layered scanning: Through the preset folder level depth, multi-threaded layered scanning is performed on the traffic violation captured pictures stored in the server, automatically traversing each level of folders, locating and extracting qualified pictures, and marking the pictures to avoid repeated scanning and missed detection; Automatic identification: Pre-process the extracted images, including denoising, grayscale, binarization and image enhancement, and use OCR technology combined with deep learning algorithms to automatically identify information such as license plate number, violation time, violation location, camera IP, violation behavior and vehicle type in the image; Data verification and upload: Verify the recognition results, verify the format of the license plate number, time range and location legitimacy. After verification, format the data and upload it to the traffic violation management system through encryption protocol to achieve data synchronization; Exception handling and feedback: Abnormal files that appear during the recognition process are marked and feedback reports are generated for manual review and correction. The system has a self-learning function that optimizes the recognition algorithm based on historical data and manual correction results to improve recognition accuracy.

2. The method according to claim 1, characterized in that In the hierarchical scanning step, multi-threaded parallel scanning technology is used to process folders at each level in parallel. Users can configure the scanning path, hierarchical depth, file format and scanning cycle according to their needs.

3. The method according to claim 1, characterized in that In the automatic recognition step, the image recognition technology is combined with the deep learning model and OCR technology to optimize the processing of complex scenes such as uneven lighting, overlapping vehicles, blurred images, etc., and automatically extract and recognize key information in the picture.

4. The method according to claim 1, characterized in that: During the data verification and uploading steps, the system will automatically format the identified data into a unified standard and upload it to the background traffic violation management system through an encrypted transmission protocol to ensure data security and real-time synchronization.

5. The method according to claim 1, characterized in that In the exception handling and feedback step, the system marks abnormal files with recognition failure, poor image quality or irregular naming, generates an exception log report, and provides a manual review interface for staff to check and correct.

6. The method according to claim 1, characterized in that The system has a self-learning function and can automatically optimize the recognition algorithm based on historical recognition data and manual correction results to continuously improve the recognition accuracy.

7. The method according to claim 1, characterized in that The system supports multiple image formats, including but not limited to JPEG and PNG formats, and can perform pre-processing on images, such as denoising, contrast enhancement and brightness adjustment, to improve recognition quality.

8. The method according to claim 1, characterized in that: The system supports seamless integration with the existing traffic violation management system, realizes conversion and transmission of multiple data formats, and ensures the compatibility and fluency of data between different platforms.

9. The method according to claim 1, characterized in that: The system is equipped with an abnormal alarm module, which automatically triggers the alarm mechanism when abnormal data is identified or upload fails, and notifies the administrator to handle the problem.

10. The method according to claim 1, characterized in that The system can automatically extract relevant violation information and classify it according to different violation types, including illegal parking, running red lights, speeding, driving in the wrong direction, vehicle restrictions, time-limited traffic, etc.

Images