CN-122024172-A - Rural highway construction progress monitoring method and device, electronic equipment and storage medium

CN122024172ACN 122024172 ACN122024172 ACN 122024172ACN-122024172-A

Abstract

The invention provides a rural highway construction progress monitoring method, a device, electronic equipment and a storage medium. The method can monitor progress of rural highway construction projects based on multi-source visual perception and deep learning, is suitable for automatic, high-precision and real-time progress tracking, quality assessment and risk early warning of rural highway construction processes, and realizes accurate, automatic and low-cost monitoring of rural highway construction progress, engineering quality compliance and potential risks.

Inventors

CUI YINGSHOU
FAN WENTAO
MA JINGYU
Wang Yaoze
ZHANG SHUZHEN
LU ZILIN

Assignees

交通运输部科学研究院

Dates

Publication Date: 20260512
Application Date: 20260407

Claims (10)

1. A rural highway construction progress monitoring method, the method comprising: the method comprises the steps that a fixed shooting module acquires multi-state construction image data, wherein the multi-state construction image data comprises construction area high-definition visual images, geographic position information and time stamp information acquired under different time nodes; the road identification module performs edge detection on the multi-time-state construction image data and extracts road boundary information, wherein the road boundary information comprises a real boundary of a road, a construction area range and structural characteristics; The fusion recognition module is used for modeling the association relation between the visual characteristics and the numerical characteristics by utilizing a self-attention mechanism through a twin neural network to generate standardized construction progress characteristic data; And the progress analysis module combines the road boundary information output by the road recognition module and the construction feature recognition result output by the fusion recognition module to execute construction progress deviation assessment, engineering quality compliance judgment and potential construction risk prediction.
2. The method of claim 1, wherein the fixed shooting module comprises a fixed camera shooting sub-module and a data preprocessing sub-module, and the step of acquiring the multi-temporal construction image data by the fixed shooting module comprises the following steps: The fixed-machine-position shooting submodule is deployed at a key node of a construction site, original image data are automatically collected every day according to a preset time period, and geographic position information, timestamp information and ambient illumination intensity are synchronously recorded; The data preprocessing sub-module receives the original image data acquired by the fixed-camera shooting sub-module, sequentially adopts a Gaussian filter algorithm to remove noise and a polynomial correction method to carry out geometric correction, carries out normalization processing on the image data based on a Z-Score method, and outputs a standardized image dataset as multi-temporal construction image data.
3. The method of claim 2, wherein the road recognition module comprises an edge detection sub-module, and wherein the road recognition module performs edge detection on the multi-temporal construction image data to extract road boundary information, and the method comprises the steps of: the edge detection sub-module performs the following processing steps S1-S7 on the normalized image dataset: S1, performing graying, size normalization and gray stretching treatment on image data in the standardized image data set, and setting a concerned area by suppressing non-key texture information through Gaussian blur, wherein the concerned area is set as the lower half part of the image data; s2, extracting edge features of the image data by adopting a Canny algorithm, and converting the edge features into an edge line set by means of random Hough transformation; S3, calculating slope and intercept of each line in the edge line set, and determining whether any 2 lines in the edge line set are redundant lines or not based on the slope and the intercept, if so, reserving 1 line to obtain an effective edge line set; S4, extracting all pixel points of the effective edge line set to form a set in the obtained attention area aiming at each line in the effective edge line set; s5, carrying out matching processing on the effective edge line set based on an edge classification result; s6, carrying out gray scale correction on the area between the matched line pairs; S7, executing S1-S3 again on the corrected gray level image, and finally selecting two effective edge lines at the leftmost side and the rightmost side as road boundary lines.
4. The method of claim 3, wherein the road identification module further comprises a rural highway extraction sub-module, the method further comprising: And the rural highway extraction submodule carries out enhancement processing by taking the area between the two road boundary lines as a target rural highway based on the road boundary lines output by the edge detection submodule, and suppresses the area outside the two road boundary lines as a background to generate rural highway image data.
5. The method of claim 3, wherein the fusion recognition module comprises a feature extraction submodule, a feature extraction submodule and a feature analysis unit, wherein the feature extraction submodule adopts a twin neural network based on a transducer architecture; The twin neural network unit comprises two Transformer subunits with the same structure and shared weight, wherein the Transformer subunits respectively input the rural highway image data in the front and rear stages, output characteristic diagrams through four downsampling stages after being embedded in sequence, and fuse the characteristic diagrams output by the two Transformer subunits in a channel splicing mode; The feature fusion unit adopts a gradient sensing multi-scale feature fusion mechanism, and adds the multi-scale feature map output by the cavity space pyramid subunit and the enhanced gradient map output by the gradient change feature subunit element by element to generate a multi-scale fusion feature map; The feature analysis unit adopts a mode of sensing the attention of the strip by a scale to obtain a multi-scale fusion feature map from the context as a construction feature recognition result.
6. The method according to claim 5, wherein after the step of the feature analysis unit obtaining the multi-scale fusion feature map from the context as the construction feature recognition result by using the scale-aware band attention, the method further comprises: The fusion recognition module performs weighted summation based on the learnable parameters and the multi-scale fusion feature map, and outputs a final construction progress map, wherein the final construction progress map is used for representing the difference between the image data of the rural highways in the front period and the rear period so as to reflect the change of the rural highways in the construction range, the pavement material, the identification line and the elements of the guideboard.
7. The method of claim 6, wherein the progress analysis module comprises a progress assessment sub-module, and wherein the progress analysis module performs construction progress deviation assessment, engineering quality compliance determination, and potential construction risk prediction steps based on the final construction progress map, comprising: the progress evaluation submodule receives the road boundary information output by the road identification module and the final construction progress chart output by the fusion identification module, and executes the following processing steps: determining the actual geometric range of the current construction area based on the road boundary information and the final construction progress map, and calculating the area of the actual construction area by a pixel area integration method; Comparing the actual construction area with the planned construction area, and calculating the progress deviation rate; And judging the construction state according to the progress deviation rate and a preset threshold value.
8. The method of claim 7, wherein the progress analysis module further comprises a change statistics sub-module, wherein the progress analysis module performs construction progress deviation assessment, engineering quality compliance determination, and potential construction risk prediction steps based on the final construction progress map, further comprising: the change statistics submodule receives the road boundary information output by the road identification module and the final construction progress chart output by the fusion identification module; counting the change condition of the road width and the marking line according to the road boundary information output by the road identification module; And marking a change area in the original data of the rural highway image according to the rural highway image data and the final construction progress chart output by the fusion identification module so as to count the actual change condition of the highway.
9. A rural highway construction progress monitoring system, the system comprising: the fixed shooting module is used for acquiring multi-state construction image data, wherein the multi-state construction image data comprises construction area high-definition visual images, geographical position information and time stamp information acquired under different time nodes; The road identification module is used for carrying out edge detection on the multi-time-state construction image data and extracting road boundary information, wherein the road boundary information comprises a real boundary of a road, a construction area range and structural characteristics; The fusion recognition module is used for modeling the association relation between the visual characteristics and the numerical characteristics by using a self-attention mechanism through a twin neural network to generate standardized construction progress characteristic data; And the progress analysis module is used for combining the road boundary information output by the road recognition module and the construction characteristic recognition result output by the fusion recognition module to execute construction progress deviation assessment, engineering quality compliance judgment and potential construction risk prediction.
10. An electronic device comprising a memory, a processor, the memory having stored thereon a computer program executable on the processor, wherein the processor, when executing the computer program, implements the rural highway construction progress monitoring method of any one of claims 1-8.

Description

Rural highway construction progress monitoring method and device, electronic equipment and storage medium Technical Field The invention relates to the technical field of intelligent traffic and engineering management, in particular to a rural highway construction progress monitoring method, device, electronic equipment and storage medium. Background At present, with the deep advancement of the country plain, the rural highway construction scale is continuously enlarged. However, due to the characteristics of complex geographical environment, insufficient supervision manpower, construction dispersion and the like, the traditional progress management mode relying on manual inspection or regular reporting has the problems of low efficiency, strong hysteresis, strong subjectivity, difficulty in quantification and the like, and cannot meet the requirements of fine engineering management. Therefore, an intelligent monitoring method integrating high-precision edge detection, time sequence feature matching and multi-scale attention mechanisms is needed to realize objective, automatic and quantifiable dynamic supervision of the whole rural highway construction process. Disclosure of Invention In view of the above, the invention aims to provide a rural highway construction progress monitoring method, a device, electronic equipment and a storage medium, which aim to solve the problems of construction progress evaluation lag, strong subjectivity in quality judgment, untimely risk early warning and the like caused by relying on manual inspection or general remote sensing means in the prior art, and realize accurate, automatic and low-cost monitoring on rural highway construction progress, engineering quality compliance and potential risk. The method comprises the steps of obtaining multi-state construction image data by a fixed shooting module, enabling the multi-state construction image data to comprise high-definition visual images, geographical position information and time stamp information of construction areas collected under different time nodes, conducting edge detection on the multi-state construction image data by a road identification module, extracting road boundary information, enabling the road boundary information to comprise real boundaries, construction area ranges and structural features of roads, enabling a fusion identification module to model association relations between visual features and numerical features through a self-attention mechanism through a twin neural network to generate standardized construction progress feature data, enabling the local construction features in the multi-state construction image data in the front period and the rear period to be matched through the twin neural network to output construction feature identification results, and enabling a progress analysis module to execute construction progress deviation assessment, engineering quality compliance judgment and potential construction risk prediction by combining the road boundary information output by the road identification module and the construction feature identification results output by the fusion identification module. The fixed shooting module comprises a fixed machine position shooting sub-module and a data preprocessing sub-module, wherein the fixed shooting module is used for acquiring multi-temporal construction image data and comprises the steps of deploying the fixed machine position shooting sub-module at a key node of a construction site, automatically acquiring original image data every day according to a preset period, synchronously recording geographic position information, timestamp information and ambient illumination intensity, and the data preprocessing sub-module is used for receiving the original image data acquired by the fixed machine position shooting sub-module, sequentially adopting a Gaussian filtering algorithm to remove noise and a polynomial correction method to carry out geometric correction, carrying out normalization processing on the image data based on a Z-Score method and outputting a standardized image data set as multi-temporal construction image data. In an alternative embodiment of the present application, the road recognition module includes an edge detection sub-module; the road recognition module performs edge detection on multi-state construction image data and extracts road boundary information, and the method comprises the following processing steps of S1-S7, wherein S1 is performed on the standardized image data set, gray scale, size normalization and gray scale stretching processing are performed on the image data in the standardized image data set, non-critical texture information is suppressed through Gaussian blur, a region of interest is set, the region of interest is set to be the lower half part of an image of the image data, S2 is used for extracting edge features of the image data by means of a Canny algorithm and converting the edge features into an edge line