CN-122024479-A - Traffic road network supervision method, equipment and medium

CN122024479ACN 122024479 ACN122024479 ACN 122024479ACN-122024479-A

Abstract

The invention discloses a traffic road network supervision method, equipment and a medium, which relate to the technical field of intelligent traffic information processing and data quality control and comprise the steps of S1, acquiring initial image data from a road network environment through multi-source sensor fusion, calibrating the installation position and shooting angle of acquisition equipment by adopting a standardized protocol to obtain a road network image set with uniform parameters, S2, preprocessing according to the obtained road network image set, removing noise by adopting Gaussian filtering, and applying histogram equalization to enhance contrast ratio and determining an enhanced image after definition improvement if the image brightness is lower than a preset threshold value.

Inventors

YIN JIANBIN
SU HANG
HAO LANXIA
Lv Zhouxing
DONG YAPING
LIU SHUNGANG

Assignees

山东北网数据科技有限公司

Dates

Publication Date: 20260512
Application Date: 20260211

Claims (10)

1. A traffic road network supervision method, comprising: S1, acquiring initial image data from a road network environment through multi-source sensor fusion, and calibrating the installation position and shooting angle of acquisition equipment by adopting a standardized protocol to obtain a road network image set with uniform parameters; S2, preprocessing according to the obtained road network image set, removing noise by adopting Gaussian filtering, and if the brightness of the image is lower than a preset threshold value, applying histogram equalization to enhance contrast, and determining an enhanced image with enhanced definition; S3, extracting edge features of the enhanced image, detecting contours by adopting a Canny operator, judging that the contours are fuzzy abnormal if the number of the detected contours is less than an expected range, and obtaining a feature image marked with potential fuzzy marks; s4, acquiring coordinate information from the feature image marked with the potential fuzzy mark, fitting a road geometric model by using a least square method, and if the fitting residual exceeds a preset threshold, marking the fitting residual as a coordinate offset error to obtain a correction image with error type classification; S5, inputting the corrected images classified according to the error types into a convolutional neural network model, analyzing image textures and semantic features by a pre-trained model, judging whether distortion caused by insufficient light exists or not, and obtaining a verification image with quality scores; S6, grouping the verification images with respect to quality scores by using a clustering algorithm, clustering similar deviation images by using a K-means method, and if the deviation between a clustering center and a standard template is greater than a threshold value, performing iterative correction to determine a final high-quality road network data set; and S7, extracting real-time state indexes from the final high-quality road network data set, adopting time sequence analysis to track the change trend, and triggering an alarm mechanism if the index fluctuation exceeds a normal interval to obtain a reliable data stream for supervision decision.
2. The traffic road network supervision method according to claim 1, wherein S1 comprises: acquiring multi-mode data containing images and environmental information from a road network environment through a multi-source sensor, and aligning the sensor data by adopting a time stamp synchronization technology to obtain a multi-source data set with consistent time; Calibrating the sensor installation position and the shooting angle in the time-consistent multi-source data set by adopting a standardized protocol, and adjusting parameters by a geometric transformation algorithm to obtain an image set with uniform parameters; Denoising and enhancing the image set with unified parameters, removing noise and improving image definition by adopting a median filtering algorithm to obtain a clear image set; And comparing the image parameters with a preset threshold value by a consistency verification algorithm aiming at the clear image set, and if the parameter deviation is smaller than the preset threshold value, judging that the image parameters are consistent, so as to obtain a road network image set passing verification.
3. The traffic road network supervision method according to claim 1, wherein S2 comprises: acquiring a first image from an image set, smoothing the first image by adopting Gaussian filtering, and carrying out weighted average on pixel points through convolution operation to obtain a denoised second image; Calculating a brightness average value according to the second image, and if the brightness average value is lower than a preset threshold value, carrying out pixel value redistribution on the second image by adopting histogram equalization to obtain an enhanced third image; Extracting edge features from the third image, and processing the third image through an edge detection algorithm to obtain a fourth image containing road network features; and carrying out sharpening processing on the fourth image, and enhancing the contrast of edge pixels of the fourth image through Laplacian filtering to obtain a fifth image with improved definition.
4. The traffic road network supervision method according to claim 1, wherein S3 comprises: Carrying out contour detection on the enhanced image by adopting a Canny operator, and extracting a strong edge and a weak edge through double-threshold processing to obtain a first feature image containing contour features; Traversing pixel points according to the first characteristic image, and calculating the total number of continuous edges to obtain a contour quantity value; if the contour quantity value is lower than a preset threshold value, applying fuzzy judgment logic to the first characteristic image to obtain a second characteristic image marked with fuzzy marks; and carrying out contrast enhancement on the outline pixels of the second characteristic image through Laplace filtering to obtain an optimized third characteristic image.
5. The traffic road network supervision method according to claim 1, wherein S4 comprises: extracting feature points from the original image marked with the potential fuzzy marks, acquiring key point coordinates by adopting an edge detection algorithm, and generating the first coordinate set; Fitting the first coordinate set by adopting a least square method to generate parameters of a road geometric model, and obtaining the first fitting model; if the residual error of the first fitting model exceeds a preset threshold value, judging that the coordinate deviation is wrong by comparing the deviation of the first coordinate set and the first fitting model, and generating the error type label; And adjusting the first coordinate set according to the error type label, and generating a corrected image by adopting a data point registration method to obtain the corrected image.
6. The traffic road network supervision method according to claim 1, wherein S5 comprises: obtaining pixel values from the corrected image, enhancing image textures by adopting a gray histogram equalization method, and generating the first texture enhanced image; If the brightness value of the first texture enhanced image is lower than a preset threshold value, extracting semantic features through a pre-training model of a convolutional neural network to obtain the first feature set; According to the first feature set, analyzing the light shortage degree of the image area by adopting an optical flow algorithm, judging whether distortion exists or not, and generating the first distortion annotation set; and comparing the first distortion annotation set with a preset quality scoring standard, and adjusting the pixel distribution of the image to obtain a verification image.
7. The traffic road network supervision method according to claim 1, wherein S6 comprises: acquiring pixel distribution from the verification image, extracting texture features by adopting a gray level co-occurrence matrix method, and generating the first feature set; clustering and grouping the image features by adopting a K-means method according to the first feature set, and determining a clustering center based on similar deviation to obtain the first clustering group; Extracting a clustering center from the first clustering group, performing deviation comparison with a standard template by adopting an Euclidean distance method, and performing iterative correction through pixel value adjustment if the deviation value is greater than a preset threshold value to obtain the second clustering group; and integrating the image areas by adopting a space adjacency analysis method according to the second cluster group to generate a high-quality road network data set, and determining final output.
8. The traffic road network supervision method according to claim 1, wherein S7 comprises: Acquiring a real-time state index from the final high-quality road network data set, extracting time sequence data by adopting a sliding window method, and grouping based on a time stamp to obtain the dynamic state data set; Calculating a change trend by adopting a time sequence analysis method aiming at the dynamic state data set, and generating a trend curve by an autoregressive moving average algorithm to obtain the index change sequence; If the fluctuation value of the index change sequence exceeds a preset normal interval threshold value, triggering an alarm mechanism through a logic comparison method to generate the abnormal event record; And merging the real-time state indexes with alarm information by adopting a data integration method according to the abnormal event record, generating a reliable data stream for supervision decision, and determining final output.
9. A traffic road network supervision device comprising a memory and a processor, the memory and the processor being communicatively connected to each other, the memory having stored therein computer instructions, the processor executing the computer instructions to perform the traffic road network supervision method according to any one of claims 1-8.
10. A computer readable storage medium having stored thereon computer instructions for causing a computer to perform the traffic road network supervision method according to any one of claims 1 to 8.

Description

Traffic road network supervision method, equipment and medium Technical Field The invention relates to the technical field of intelligent traffic information processing and data quality control, in particular to a traffic road network supervision method, equipment and medium. Background The road network information is used as a core support for traffic management and city planning, and the accuracy of the road network information is directly related to the scientificity of supervision decision-making and the safety of public travel. The high-quality road network data not only needs to reflect the real-time state of the road, but also needs to ensure the reliability of the data in the whole process of collection, processing and application. However, the problem of data quality of the current road network information frequently occurs, which severely restricts the supervision efficiency and the intelligent development of the traffic system. How to construct a dynamic, accurate and reliable road network data quality control system becomes a major subject to be overcome in the field of traffic management. The existing road network data management method has obvious defects in technical realization. Many systems rely on a single device to collect data, neglecting the normalization of hardware parameters such as device installation positions, shooting angles and the like, and causing deviation of the data at the source. For example, some monitoring devices cannot clearly reflect road conditions due to improper installation locations, or data distortion due to angular offset. In addition, the data processing link lacks an automatic detection means, and the problems of blurred images, wrong timestamps or coordinate offset and the like can be usually found only manually, so that the efficiency is low and omission is easy. These problems make it difficult for the data quality to meet high precision regulatory requirements. A further technical difficulty is how to ensure standardization of data acquisition and real-time identification of data errors. The lack of unified specifications for the installation positions and parameter settings of the acquisition devices results in large differences in format and accuracy of data acquired by different devices. For example, different cameras on the same road section may have different resolutions, and thus may have uneven image quality, and it is difficult to directly use the image quality in road network analysis. Another difficulty that arises is that automatic detection techniques for data quality have not yet matured. The existing system is difficult to quickly identify abnormal data such as fuzzy images or coordinate offsets, and the hysteresis can lead the error data to enter a decision process especially in a dynamic road network environment of high-frequency acquisition. For example, a road condition image blurred due to insufficient light may be mistaken for a clear road, thereby affecting the accuracy of traffic signal regulation. Disclosure of Invention The invention aims to provide a traffic road network supervision method, equipment and medium, which solve the problems existing in the prior art. In order to achieve the purpose, the invention provides a traffic road network supervision method, which is characterized by comprising the following steps: S1, acquiring initial image data from a road network environment through multi-source sensor fusion, and calibrating the installation position and shooting angle of acquisition equipment by adopting a standardized protocol to obtain a road network image set with uniform parameters; S2, preprocessing according to the obtained road network image set, removing noise by adopting Gaussian filtering, and if the brightness of the image is lower than a preset threshold value, applying histogram equalization to enhance contrast, and determining an enhanced image with enhanced definition; S3, extracting edge features of the enhanced image, detecting contours by adopting a Canny operator, judging that the contours are fuzzy abnormal if the number of the detected contours is less than an expected range, and obtaining a feature image marked with potential fuzzy marks; s4, acquiring coordinate information from the feature image marked with the potential fuzzy mark, fitting a road geometric model by using a least square method, and if the fitting residual exceeds a preset threshold, marking the fitting residual as a coordinate offset error to obtain a correction image with error type classification; S5, inputting the corrected images classified according to the error types into a convolutional neural network model, analyzing image textures and semantic features by a pre-trained model, judging whether distortion caused by insufficient light exists or not, and obtaining a verification image with quality scores; S6, grouping the verification images with respect to quality scores by using a clustering algorithm, clustering similar devia