CN-121978644-A - Intelligent highway disease detection method and system based on three-dimensional ground penetrating radar

CN121978644ACN 121978644 ACN121978644 ACN 121978644ACN-121978644-A

Abstract

The invention relates to the technical field of road detection, in particular to an intelligent road disease detection method and system based on a three-dimensional ground penetrating radar, comprising the steps of acquiring three-dimensional radar data and processing the three-dimensional radar data to obtain a horizontal slice image along the depth direction and vertical section images of a plurality of channels along the survey line direction; the method comprises the steps of determining the boundary position of an underground structure layer, selecting a horizontal slice image with corresponding depth according to the boundary position, inputting a first disease detection model to obtain a preliminary disease candidate region, mapping the preliminary candidate region to a vertical section image, intercepting a local vertical section image of each channel, inputting a second disease detection model to obtain disease category information, carrying out fusion decision based on the transverse position of the candidate region and the disease category information of each channel, determining the final disease type, associating the final disease type to the candidate region, and realizing consistency of disease positioning and type judgment by cooperatively utilizing the horizontal slice image and the multi-channel vertical section image, thereby improving detection accuracy and stability.

Inventors

JIN GUANGLAI
CAI WENLONG
WANG SHUANGPING

Assignees

江苏中路工程技术研究院有限公司

Dates

Publication Date: 20260505
Application Date: 20251223

Claims (10)

1. The intelligent road disease detection method based on the three-dimensional ground penetrating radar is characterized by comprising the following steps of: Acquiring and processing three-dimensional radar data to obtain horizontal slice images distributed along the depth direction and vertical section images of a plurality of channels distributed along the survey line direction; determining a boundary line position of the subsurface structure layer based on the plurality of vertical section images; selecting a horizontal slice image with corresponding depth according to the boundary position, and inputting the horizontal slice image into a first disease detection model to obtain a preliminary disease candidate region on a horizontal slice; Mapping the preliminary disease candidate region to the vertical section image, and intercepting local vertical section images corresponding to the candidate region in each channel; inputting the local vertical section images of all the channels into a second disease detection model, and obtaining disease category information of all the channels; And carrying out fusion decision based on the transverse position of the preliminary disease candidate region and the disease category information of each channel, determining the final disease type, and associating the final disease type to the preliminary disease candidate region.
2. The intelligent detection method for highway diseases based on three-dimensional ground penetrating radar according to claim 1, wherein the fusion decision comprises: counting the detected disease category information in each channel and the coverage width of the detected disease category information in the transverse range of the preliminary disease candidate area; Calculating the comprehensive score of each disease type according to the preset disease type severity weight and the coverage width; And determining a primary disease type and/or a secondary disease type as the final disease type according to the comprehensive score.
3. The intelligent detection method for highway diseases based on three-dimensional ground penetrating radar according to claim 2, wherein the composite score is calculated by the following formula: ; In the formula, A comprehensive score for disease category k; weights set according to disease severity; the number of channels being adjacent channels and the disease type being the same.
4. The intelligent detection method for highway diseases based on three-dimensional ground penetrating radar according to claim 1, wherein the first disease detection model comprises a C3WC module, and the C3WC module is used for enhancing the identification capability of weak-edge and large-scale diseases by embedding wavelet transform into convolution operation to extract high-frequency edge features and low-frequency texture features in a horizontal slice image at the same time.
5. The intelligent detection method for highway diseases based on three-dimensional ground penetrating radar according to claim 1, wherein the first disease detection model or the second disease detection model comprises an OEM module, and the OEM module generates an attention weight graph by the following steps: performing edge detection on the input image to generate an initial edge mask; normalizing the initial edge mask and mapping the initial edge mask with Gaussian weights to generate an attention weight map with high center weight and gradually low edge weight; multiplying the attention weight graph with the characteristic graph of the first disease detection model or the second disease detection model to strengthen the characteristic response conforming to the interlayer disease prior morphology.
6. The intelligent detection method for highway diseases based on three-dimensional ground penetrating radar according to claim 1, wherein the obtaining the preliminary disease candidate region on the horizontal slice comprises: performing sliding window detection on the horizontal slice image to generate a plurality of detection windows; inputting the detection window into the first disease detection model to obtain a plurality of initial disease candidate frames; And carrying out fusion and filtering treatment on the plurality of initial disease candidate frames to form the initial disease candidate region set.
7. The intelligent detection method for highway diseases based on three-dimensional ground penetrating radar according to claim 1, wherein mapping the preliminary disease candidate region to the vertical section image, and intercepting local vertical section images corresponding to the candidate region in each channel, comprises: Determining a corresponding transverse position interval of the preliminary disease candidate region in each channel vertical section image based on a transverse coordinate range and a channel number of the preliminary disease candidate region in the horizontal section image; and according to the preset longitudinal analysis depth, image blocks are cut from the vertical section images of all channels along the transverse position interval, so that the local vertical section images are formed.
8. The method for intelligently detecting highway diseases based on three-dimensional ground penetrating radar according to claim 1, wherein the step of obtaining disease category information of each channel comprises the following steps: Inputting the local vertical section image of each channel into the second disease detection model; Obtaining disease categories and related attribute information corresponding to each local area, which are output by the second disease detection model; and storing the disease category and the related attribute information according to channels based on the transverse coordinates of the preliminary disease candidate region.
9. The intelligent detection method for highway diseases based on three-dimensional ground penetrating radar according to claim 1, wherein determining the boundary position of the underground structure layer comprises: performing pixel-level superposition on a plurality of vertical section images to obtain a superposition image; smoothing the superimposed image; Calculating the vertical gradient of the smoothed image to obtain gradient information; And detecting local peaks based on the gradient information, and screening out effective peaks according to preset conditions to serve as boundary positions of the underground structure layer.
10. Highway disease intelligent detection system based on three-dimensional ground penetrating radar, characterized in that the system includes: the data acquisition and preprocessing module is used for acquiring and processing three-dimensional radar data to obtain horizontal slice images distributed along the depth direction and vertical section images of a plurality of channels distributed along the measuring line direction; A subsurface horizon determination module for determining a boundary line position of a subsurface structural layer based on a plurality of the vertical section images; The initial positioning module of the horizontal slice disease is used for selecting a horizontal slice image with a corresponding depth according to the boundary position, and inputting the horizontal slice image into the first disease detection model to obtain a preliminary disease candidate area on the horizontal slice; a vertical section image intercepting module, configured to map the preliminary defect candidate area to the vertical section image, and intercept local vertical section images corresponding to the candidate area in each channel; The vertical section disease fine classification module is used for inputting the local vertical section images of all the channels into a second disease detection model to obtain disease category information of all the channels; and the multidimensional information fusion decision module is used for carrying out fusion decision based on the transverse position of the preliminary disease candidate region and the disease category information of each channel, determining the final disease type and relating to the preliminary disease candidate region.

Description

Intelligent highway disease detection method and system based on three-dimensional ground penetrating radar Technical Field The invention relates to the technical field of road detection, in particular to an intelligent road disease detection method and system based on a three-dimensional ground penetrating radar. Background In the long-term service process of a highway, hidden diseases such as interlayer void, looseness, cracks, water seepage and the like are easily generated in the internal structure of the highway under the action of vehicle load and environmental factors, the diseases are difficult to perceive from the surface, but the road bearing capacity is obviously weakened, even collapse and other safety accidents are caused, the ground penetrating radar is used as an efficient, lossless and high-resolution underground detection technology, the ground penetrating radar becomes an important means for detecting the highway diseases, the traditional method mainly relies on a single-channel vertical section image acquired by the two-dimensional ground penetrating radar to carry out manual interpretation or simple analysis, however, the image information of a single dimension is limited, the disease characteristics are easy to be confused with medium interface reflection and random noise, and the detection result has the inherent limitations of high misjudgment rate, multiple missed detection, difficulty in three-dimensional positioning, quantitative evaluation and the like. With the development of the three-dimensional ground penetrating radar technology, a horizontal slice image reflecting horizontal distribution and a multi-channel vertical section image reflecting a vertical structure can be synchronously acquired, a data basis is provided for three-dimensional identification of diseases, however, the conventional detection method cannot fully utilize the cooperative value of the multi-dimensional data, namely, although the horizontal slice image can clearly show the planar spread of the diseases, the characteristics of weak texture and blurred edges exist, a general target detection model is difficult to directly adapt and cannot judge the specific type of the diseases, while the detection based on the vertical section can provide category information, but a single channel analysis result is greatly interfered by noise, has poor spatial consistency and lacks effective association with horizontal positioning, so that the three-dimensional identification process of the diseases has key bottlenecks such as feature confusion, positioning deviation, type misjudgment and the like, and the problems seriously restrict the automation level and reliability of highway disease detection. The information disclosed in this background section is only for enhancement of understanding of the general background of the invention and should not be taken as an acknowledgement or any form of suggestion that this information forms the prior art already known to a person skilled in the art. Disclosure of Invention The invention provides a highway disease intelligent detection method and system based on a three-dimensional ground penetrating radar, so that the problems in the background technology are effectively solved. In order to achieve the purpose, the technical scheme adopted by the invention is that the intelligent detection method for the highway diseases based on the three-dimensional ground penetrating radar comprises the following steps: Acquiring and processing three-dimensional radar data to obtain horizontal slice images distributed along the depth direction and vertical section images of a plurality of channels distributed along the survey line direction; determining a boundary line position of the subsurface structure layer based on the plurality of vertical section images; selecting a horizontal slice image with corresponding depth according to the boundary position, and inputting the horizontal slice image into a first disease detection model to obtain a preliminary disease candidate region on a horizontal slice; Mapping the preliminary disease candidate region to the vertical section image, and intercepting local vertical section images corresponding to the candidate region in each channel; inputting the local vertical section images of all the channels into a second disease detection model, and obtaining disease category information of all the channels; And carrying out fusion decision based on the transverse position of the preliminary disease candidate region and the disease category information of each channel, determining the final disease type, and associating the final disease type to the preliminary disease candidate region. Further, making a fusion decision includes: counting the detected disease category information in each channel and the coverage width of the detected disease category information in the transverse range of the preliminary disease candidate area; Calculating the comprehensive