A method, apparatus and device for determining the actual size of a lesion

Abstract

The application provides a method, device and equipment for determining actual size of disease, the method comprises the following steps: acquiring actual disease radar atlas of test road based on the result of radar detection of the test road; setting forward parameter based on disease position and size in the test road, and acquiring disease forward atlas; drawing disease color pixel map by taking the actual disease radar atlas as reference; constructing evolution model from the actual disease radar atlas, through the disease forward atlas to the disease color pixel map; inputting actual disease radar atlas of road to be tested into the evolution model, and acquiring target disease color pixel map corresponding to the road to be tested. The application belongs to the field of road exploration, through radar detection of the road, the color pixel map of road disease is acquired, the actual size of invisible disease in the road can be observed directly, and the maintenance personnel can select appropriate maintenance measures to repair the road disease.

Description

Technical Field

[0001] This invention belongs to the field of road exploration, and specifically relates to a method, apparatus and equipment for determining the actual size of road defects. Background Technology

[0002] Ground-penetrating radar (GPR) can transmit and receive high-frequency broadband electromagnetic waves in the microwave band. Since electromagnetic waves are reflected at the interface of underground media, analyzing the waveform characteristics of the reflected electromagnetic waves can reveal the spatial location, material composition, and other features of underground targets. Therefore, using GPR to collect road defect images is a common method. However, the actual road defect images collected by GPR are affected by objective factors such as road conditions, operator skill, and limited sample size. For example, traditional methods involve first collecting data on locations where road defects might occur, and then excavating the road to verify the data. This approach makes it difficult to clearly investigate the actual internal environment of the road, thus failing to establish a specific connection between the images and the defects. Summary of the Invention

[0003] In view of the above problems, embodiments of the present invention provide a method, apparatus and equipment for determining the actual size of a disease, so as to overcome the above problems or at least partially solve the above problems.

[0004] In a first aspect, the present invention provides a method for determining the actual size of a disease, the method comprising:

[0005] Based on the results of radar detection of the test road, the actual damage radar map of the test road is obtained;

[0006] Based on the location and size of the defects in the test road, forward modeling parameters were set, and forward modeling maps of the defects were obtained.

[0007] Using the actual disease radar map as a reference, a color pixel map of the disease is drawn based on the location and size of the disease in the test road; wherein, the location and size of the disease in the color pixel map correspond one-to-one with the actual disease radar map;

[0008] An evolutionary model is constructed from the actual disease radar map, through the disease forward model, to the disease color pixel map;

[0009] The actual damage radar map of the road to be tested is input into the evolution model to obtain the target damage color pixel map corresponding to the road to be tested. The target damage color pixel map is used to describe the actual size of the damage in the road to be tested.

[0010] Furthermore, the method for constructing the evolutionary model from the actual disease radar map, through the disease forward model, to the disease color pixel map includes:

[0011] Deep learning is performed on the actual disease radar map and the disease forward model map to construct a first inversion model; wherein, the first inversion model is used to output the disease forward model simulation map corresponding to the actual disease radar map;

[0012] Deep learning is performed on the forward modeling atlas of the disease and the color pixel image of the disease to construct a second inversion model; wherein, the second inversion model is used to output the color pixel image of the target disease corresponding to the forward modeling atlas of the disease;

[0013] The evolution model is obtained based on the first inversion model and the second inversion model.

[0014] Furthermore, the method for obtaining a forward modeling map of the disease by setting forward modeling parameters based on the location and size of the disease in the test road includes:

[0015] Obtain the road structure of the test road;

[0016] Based on the forward modeling parameters, the location and size of the defects in the road structure and the test road are modeled in forward modeling to obtain a defect forward modeling map;

[0017] The forward modeling parameters are consistent with the parameters used for radar detection of the test path.

[0018] Furthermore, before obtaining the actual damage radar map of the test road based on the results of radar detection of the test road, the method further includes:

[0019] Obtain the radar map of the detected disease corresponding to the results of the radar detection;

[0020] Based on the radar map of the detected disease, the depth of the disease location is obtained;

[0021] Based on the depth of the location of the disease, the radar map of the disease detection is subjected to target processing, which includes at least one of DC drift, static correction, gain, bandpass filtering and background removal.

[0022] The radar map of the detected defects after the target has been processed is determined as the actual radar map of the defects of the test road.

[0023] Furthermore, the target processing of the radar map of the detected disease based on the depth of the disease location includes:

[0024] When the depth is at the first depth, the radar map of the detected defects is processed once, and the radar map of the detected defects after the first signal processing is determined as the actual radar map of the defects of the test road.

[0025] When the depth of the disease location is at the second depth, the radar map of the detected disease after the first signal processing is subjected to a second signal processing, and the radar map of the detected disease after the second signal processing is determined as the actual radar map of the disease of the test road.

[0026] The primary signal processing includes: DC drift processing, static correction processing, gain processing, bandpass filtering processing, and background removal processing; the secondary signal processing includes the gain processing.

[0027] The first depth is less than the second depth.

[0028] Furthermore, the step of drawing a color pixel map of the defects based on the location and size of the defects in the test road, using the actual defect radar map as a reference, includes:

[0029] Obtain the location, width, and dielectric constant of the defects in the test path;

[0030] Obtain the length of the disease in the actual disease radar map;

[0031] Obtain the dielectric constant of the road structure of the test road;

[0032] Based on the length, location, width, dielectric constant of the lesion, and dielectric constant of the road structure, a color pixel map of the lesion is drawn.

[0033] Furthermore, the method for obtaining the disease length in the actual disease radar map includes:

[0034] Obtain the waveform of the actual disease radar map;

[0035] Based on the waveform diagram, the first arrival radar wave of the disease is picked up; wherein, the first arrival radar wave of the disease includes multiple first arrival radar waves;

[0036] Based on the multiple first-arrival radar waves, the starting and ending positions of the disease are obtained by judging the multiple first-arrival radar waves according to the disease length index.

[0037] Based on the difference between the starting and ending positions of the disease, the length of the disease in the actual disease radar map is obtained.

[0038] Further, the step of determining the starting and ending positions of the disease based on the plurality of first-arrival radar waves according to the disease length index includes:

[0039] Obtain the valley information of the plurality of first-arrival radar waves; wherein, the valley information is the ratio of valley amplitude to valley height;

[0040] The disease length index is constructed based on the ratio of the trough amplitude to the trough height.

[0041] The two first-arrival radar waves that satisfy the disease length index and are located at the middle end of the plurality of first-arrival radars are determined as the starting first-arrival radar wave and the ending first-arrival radar wave.

[0042] Based on the trough position of the initial arrival radar wave at the starting point and the trough position of the initial arrival radar wave at the ending point, the starting point position and the ending point position of the disease are obtained.

[0043] A second aspect of the present invention provides an apparatus for determining the actual size of a disease, the apparatus comprising:

[0044] First acquisition module: used to acquire the actual damage radar map of the test road based on the results of radar detection of the test road;

[0045] The second acquisition module is used to set forward modeling parameters and acquire forward modeling maps of the diseases based on the location and size of the diseases in the test road.

[0046] The drawing module is used to draw a color pixel map of the disease based on the location and size of the disease in the test road, with the actual disease radar map as a reference; wherein the location and size of the disease in the color pixel map correspond one-to-one with the actual disease radar map.

[0047] The third acquisition module is used to construct an evolution model from the actual disease radar map, through the disease forward model map, to the disease color pixel map;

[0048] The fourth acquisition module is used to input the actual damage radar map of the road to be tested into the evolution model to obtain the target damage color pixel map corresponding to the road to be tested. The target damage color pixel map is used to describe the actual size of the damage in the road to be tested.

[0049] A third aspect of the present invention provides an electronic device, including a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the method described in the first aspect of the present invention.

[0050] The method provided in this invention includes: a method for determining the actual size of a disease, the method comprising: firstly, obtaining an actual disease radar map of the test road based on the results of radar detection of the test road; and secondly, obtaining a disease forward modeling map based on the location and size of the disease in the test road by setting forward modeling parameters; then, using the actual disease radar map as a reference, drawing a disease color pixel map based on the location and size of the disease in the test road; making the location and size of the disease in the disease color pixel map correspond one-to-one with the actual disease radar map, establishing the relationship between the disease pixel map and the actual radar map; thereby realizing the construction of an evolutionary model from the actual disease radar map, through the disease forward modeling map to the disease color pixel map, through which the actual radar map can be linked to the actual disease size, achieving the purpose of quantitatively determining the actual size of the disease.

[0051] Finally, the actual radar image of the road under test is input into the evolutionary model to obtain a color pixel image of the target defect corresponding to the road under test. This color pixel image describes the actual size of the defect in the road under test. The method provided by this invention allows for the direct determination of the size and location of latent road defects from actual radar images, facilitating maintenance personnel to select appropriate maintenance measures for road repair. It also provides theoretical analysis methods and technical support for the identification and inversion of abnormal areas within road structures, further enriching research in related fields. Attached Figure Description

[0052] To more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings used in the description of the embodiments of the present invention will be briefly introduced below. Obviously, the drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0053] Figure 1 This is a flowchart illustrating the steps of a method for determining the actual size of a disease according to an embodiment of the present invention;

[0054] Figure 2 This is a schematic diagram of an actual disease radar map provided in an embodiment of the present invention;

[0055] Figure 3 This is a schematic diagram of a disease forward modeling atlas provided in an embodiment of the present invention;

[0056] Figure 4 This is a schematic diagram of a disease forward modeling simulation atlas output by an SZ inversion model provided in an embodiment of the present invention;

[0057] Figure 5 This is a schematic diagram of a color pixel image of a disease provided in an embodiment of the present invention;

[0058] Figure 6 This is a schematic diagram of a target lesion color pixel image output by a ZB inversion model provided in an embodiment of the present invention;

[0059] Figure 7 This is provided by the embodiments of the present invention. Figure 2 A schematic diagram of the corresponding disease atlas A-scan;

[0060] Figure 8 This is provided by the embodiments of the present invention. Figure 7 Corresponding A-scan trough diagram;

[0061] Figure 9 This is a schematic diagram of a device for determining the actual size of a disease according to an embodiment of the present invention. Detailed Implementation

[0062] Exemplary embodiments of the present invention will now be described in more detail with reference to the accompanying drawings. While exemplary embodiments of the invention are shown in the drawings, it should be understood that the invention may be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that this invention will be thorough and complete, and will fully convey the scope of the invention to those skilled in the art.

[0063] The detection and intelligent interpretation of roadbed defects have become urgent problems in my country's highway construction and maintenance management. Current research focuses more on automatic identification techniques for certain latent defects rather than determining their size, resulting in radar images failing to fully reflect the true condition of the road interior. Furthermore, data collectors can only obtain a general description of latent road defects from radar images, such as their depth, but cannot determine their actual size, hindering in-depth research and interpretation of defect identification characteristics.

[0064] Therefore, this invention provides a method for determining the actual size of road defects, in order to optimize the current problem of failing to reflect the size of road defects inside the road.

[0065] Figure 1 This is a flowchart illustrating the steps of a method for determining the actual size of a disease according to an embodiment of the present invention, referred to below. Figure 1 The steps include:

[0066] Step S101: Based on the results of radar detection of the test road, obtain the actual damage radar map of the test road.

[0067] In this embodiment, a full-scale standardized road structural damage test road is constructed based on road defects and structures present in daily life. Defects may include common hidden road defects such as cracks, potholes, loosening, and subsidence. Road structures may include cement pavement or asphalt pavement. This embodiment does not limit the specific road defects and structures; it is merely an example. After the test road is constructed, ground-penetrating radar (GPR) antennas of different frequencies are selected to detect the test road, thus obtaining actual defect maps from different GPR antennas. The different frequency GPR antennas may include: a 500MHz radar shielded antenna, an 800MHz radar shielded antenna, and a 300MHz radar shielded antenna. The appropriate operating frequency radar shielded antenna is selected based on the requirements of the test road for on-site detection. This invention is not limited to these frequencies; the choice depends on the available equipment on site.

[0068] For example, the test road is a five-kilometer asphalt road with multiple hidden defects. The location and size of the defects, as well as the road structure, are known. A 300MHz radar shielded antenna is selected, and the test road is probed according to the preset initial position and step size of the radar antenna. The detection result will be a complete five-kilometer raw radar detection map. This raw radar detection map corresponds to the 300MHz radar shielded antenna. Then, this raw radar detection map is input into radar processing software. The radar software processes it to obtain a processed raw radar detection map. The defects in the raw radar map are then cropped according to preset size parameters to obtain actual defect radar maps of the test road. For example, cropping the five-kilometer raw radar detection map to a size of 3m × 40ns will obtain multiple actual defect radar maps with preset size parameters.

[0069] Step S102: Based on the location and size of the defects in the test road, set forward modeling parameters and obtain the forward modeling map of the defects.

[0070] In this embodiment, due to environmental signal interference, the actual damage radar map cannot reflect the damage independently, and the map is relatively messy, which is not conducive to deep learning for inversion. Therefore, it is necessary to construct a damage forward model map. When constructing a full-scale standardized road structure damage test road, the corresponding detection data (i.e., forward modeling parameters) of the ground penetrating radar detection of the test road are input based on the known damage size and location, as well as the road structure of the test road. This data is then fed into the forward model radar map generation software, which can achieve the purpose of obtaining the damage forward model map by setting the forward modeling parameters.

[0071] For example, based on the radar detection parameters of the test road in step S101, forward modeling parameters are set. These parameters include setting a time window based on the frequency corresponding to the radar antenna frequency (which can be understood as the travel time of electromagnetic waves), obtaining the dielectric constant and conductivity of the asphalt pavement and defects, the initial position and step size of the radar antenna, and constructing a defect forward model using the determined parameters. The road structure of the test road, as well as the size and depth of each embedded defect, are set with preset size parameters identical to those in the actual radar map. The complete defect forward model map is then cropped to obtain a defect forward model map corresponding to the actual radar map. In this embodiment, the defect forward model map corresponding to the actual defect radar map of the test road can be obtained by inputting the forward modeling parameters and size parameters corresponding to the actual defect radar map of the test road into the radar forward model map generation software. This lays the foundation for the subsequent deep learning of the SZ inversion model and ensures accuracy.

[0072] Step S103: Using the actual disease radar map as a reference, draw a color pixel map of the disease based on the location and size of the disease in the test road; wherein the location and size of the disease in the color pixel map correspond one-to-one with the actual disease radar map.

[0073] In this embodiment, the length of the disease is determined by the disease constant index, and the width and depth of the disease in the test road are determined according to the location and size of the disease in the test road. The disease color pixel map is then drawn quantitatively to ensure that the location and size of the disease in the disease color pixel map correspond one-to-one with the actual disease radar map.

[0074] For example, taking an asphalt road structure as a test road with multiple defects as an example, the dielectric constant of the cracks and the dielectric constant of the asphalt are obtained. In the drawing software, the corresponding dielectric constants are assigned different colors to draw a color pixel map. During the drawing process, the length of the defect is based on the length of the defect in the actual radar image, while the width and location of the defect are based on the test road. This ensures that the size and location of the defect in the pixel map correspond one-to-one with the actual radar image. By observing the different colors of the defects in the defect pixel map and the road structure, the specific size and location of the defects can be determined, laying the foundation for the deep learning of the subsequent ZB inversion model and ensuring accuracy.

[0075] Step S104: Construct an evolution model from the actual disease radar map, through the disease forward model map, to the disease color pixel map.

[0076] In this embodiment, an evolutionary model is constructed, including constructing an inversion model from the actual disease radar map to the disease forward model, finding the relationship between the actual disease map and the disease forward model, constructing an inversion model from the disease forward model to the disease pixel map, finding the relationship between the disease forward model and the disease pixel map, and finally finding the relationship between the actual disease forward model and the disease color pixel map through the relationship between the actual disease map and the disease forward model, and the relationship between the disease forward model and the disease color pixel map. The evolutionary model is constructed so that when an actual disease radar map is input into the evolutionary model, the evolutionary model outputs a disease color pixel map corresponding to the actual radar map.

[0077] Step S105: Input the actual damage radar map of the road to be tested into the evolution model to obtain the target damage color pixel map corresponding to the road to be tested. The target damage color pixel map is used to describe the actual size of the damage in the road to be tested.

[0078] In this embodiment, the actual radar map of the road to be tested can be input into the evolutionary model obtained in step S104 to obtain the target defect color pixel map corresponding to the road to be tested. The target defect color pixel map is used to describe the actual size of the defects in the road to be tested. The road to be tested refers to a real road that can be used in real life, not the test road provided in this embodiment. The target defect color pixel map refers to the defect color pixel map of the real road. Through this defect color pixel map, road maintenance personnel can easily and intuitively determine the size and location of road defects and carry out precise treatment of road defects.

[0079] In one embodiment, the method for constructing the evolutionary model from the actual disease radar map, through the disease forward model, to the disease color pixel map includes: performing deep learning on the actual disease radar map and the disease forward model to construct a first inversion model; wherein the first inversion model is used to output a disease forward model simulation map corresponding to the actual disease radar map; performing deep learning on the disease forward model simulation map and the disease color pixel map to construct a second inversion model; wherein the second inversion model is used to output the target disease color pixel map corresponding to the disease forward model simulation map; and obtaining the evolutionary model based on the first inversion model and the second inversion model.

[0080] In this embodiment, deep learning is performed on the actual disease radar map and the disease forward model map. Based on the relationship between the two disease forward model maps, a first inversion model is constructed. The first inversion model is used to output the disease forward model simulation map corresponding to the actual disease radar map. Deep learning is then performed on the disease forward model simulation map and the disease color pixel map. Based on the relationship between the disease forward model map and the disease color pixel map, a second inversion model is constructed. The second inversion model is used to output the target disease color pixel map corresponding to the disease forward model simulation map. Finally, an evolutionary model is constructed by combining the first inversion model and the second inversion model. The evolutionary model is used to output the target disease color pixel map corresponding to the actual disease radar map.

[0081] For example, the construction of the first inversion model and the second inversion model are explained below:

[0082] First, the first inversion model, namely the SZ inversion model, is constructed. The U-NET network is used to construct and optimize the inversion model via gradient descent. All maps used are adjusted to a size of 512×512 pixels. The training set should contain more than 100 maps, and the number of maps used for training should be greater than 50% of the total number of maps. Taking a total of 1000 maps as an example, 800 maps are selected for model training. After training, the remaining forward simulation maps are used as the test set to verify the model's accuracy. Each training iteration uses 35 maps, and the model parameters are set as follows:

[0083] (1)niter=400, nither_decay=400.

[0084] (2) batch_size = 35.

[0085] The training set of the SZ inversion model is divided into two parts: set a and set b. Set a stores the actual radar images used for training, referencing... Figure 2 , Figure 2 This is a schematic diagram of an actual disease radar map provided in an embodiment of the present invention. Set b stores the forward modeling maps of diseases corresponding to the maps in set a. (Refer to...) Figure 3 , Figure 3 This is a schematic diagram of a disease forward modeling map provided in an embodiment of the present invention. As can be seen from the diagram, the disease forward modeling map eliminates interference from environmental and other signals, displaying only the radar waves reflecting the disease, laying the foundation for subsequent quantitative determination of disease size from waveforms. The actual radar map and the disease forward modeling map together form the training dataset for the SZ inversion model. During deep learning, the correspondence between the two is used to train the SZ inversion model's ability to identify radar waves reflecting the disease in the radar map, filtering out the radar waves reflecting the disease from the actual radar map and outputting the disease forward modeling simulation map corresponding to the actual disease radar map. Figure 4 , Figure 4This is a schematic diagram of a disease forward simulation map output by an SZ inversion model provided in an embodiment of the present invention.

[0086] Next, a second inversion model, namely the ZB inversion model, is constructed. Similarly, based on the same principle as constructing the first inversion model, taking a total of 1000 maps as an example, 700 maps are selected for model training. After training, the remaining disease forward maps are used as the test set to verify the accuracy of the model. Each training session uses 35 maps, and the model parameters are set as follows:

[0087] (1)niter=350, nither_decay=350.

[0088] (2) batch_size = 35.

[0089] The training set of the ZB inversion model is divided into two parts: set c and set d. Set c stores the forward simulation maps of the diseases being trained, such as... Figure 4 As shown, set d stores the color pixel images of diseases corresponding to the atlas in set c, refer to... Figure 5 , Figure 5 This is a schematic diagram of a disease color pixel map provided in an embodiment of the present invention. During deep learning, the ZB inversion model is trained to identify diseases in the forward modeling simulation map by establishing a correspondence between the two. The disease forward modeling simulation map is used to link the actual radar map with the disease pixel map, and the target disease color pixel map corresponding to the actual disease forward modeling map is output. Figure 6 , Figure 6 This is a schematic diagram of a target lesion color pixel image output by a ZB inversion model provided in an embodiment of the present invention. By comparison... Figure 6 and Figure 5 It can be seen that the ZB inversion model has high accuracy. Figure 6 Color pixel images of the disease can intuitively reflect the size and depth of the disease.

[0090] Finally, by constructing an evolutionary model using the SZ inversion model and the ZB inversion model, the evolutionary model can output a color pixel image of the target disease corresponding to the actual radar image when an actual disease radar image is input into the evolutionary model. The color pixel image of the target disease can accurately reflect the actual size of the disease.

[0091] In one embodiment, the method of setting forward modeling parameters based on the location and size of defects in the test road and obtaining a forward modeling map of defects includes: obtaining the road structure of the test road; performing forward modeling on the road structure, the location and size of defects in the test road based on the forward modeling parameters to obtain a forward modeling map of defects; wherein the forward modeling parameters are consistent with the parameters for radar detection of the test road.

[0092] In this embodiment, the dielectric constant and conductivity of the test road structure are obtained, and forward modeling parameters consistent with those used for radar detection of the test road are set. These parameters include the time window, the dielectric constant and conductivity of the materials required for the model, the initial position and step size of the radar antenna, and, using the determined forward modeling parameters, a forward modeling map of the defects is constructed based on the test road data. Figure 2 and Figure 3 , Figure 3 It is based on Figure 2 The forward model of the disease was constructed using the parameters. From the figure, it can be seen that... Figure 3 The forward model of the disease compared to Figure 2 The actual disease forward modeling map filters out the radar waves reflecting the disease from the radar map, making the disease display clearer.

[0093] In one embodiment, before obtaining the actual damage radar map of the test road based on the results of radar detection of the test road, the method further includes: obtaining a detected damage radar map corresponding to the results of the radar detection; obtaining the depth of the damage location based on the detected damage radar map; performing target processing on the detected damage radar map based on the depth of the damage location, the target processing including at least one of DC drift, static correction, gain, bandpass filtering and background removal; and determining the target-processed detected damage radar map as the actual damage radar map of the test road.

[0094] In this embodiment, a radar map of detected defects is obtained for radar detection of the test road. The depth of the defect locations in the radar map is determined. Target processing is performed on the detected defect radar map, which includes at least one of the following: DC drift, static correction, gain adjustment, bandpass filtering, and background removal. The target-processed detected defect radar map is then determined as the actual defect radar map of the test road. Specifically, DC drift is used to zero the DC drift of the radar signal; static correction is used to remove the image above the first peak of the radar signal; gain adjustment is used to vertically increase the radar signal, enhancing it; bandpass filtering is used to filter out unwanted clutter; and background removal is used to remove background noise from the radar map.

[0095] In one embodiment, the target processing of the radar map of the detected disease based on the depth of the disease location includes: when the depth is at a first depth, performing a first signal processing on the radar map of the detected disease, and determining the radar map of the detected disease after the first signal processing as the actual radar map of the disease on the test road; when the depth of the disease location is at a second depth, performing a second signal processing on the radar map of the detected disease after the first signal processing, and determining the radar map of the detected disease after the second signal processing as the actual radar map of the disease on the test road; wherein, the first signal processing includes: DC drift, static correction, gain, bandpass filtering and background removal; the second signal processing includes the gain processing; the first depth is less than the second depth.

[0096] In this embodiment, the effective signal depth of the radar spectrum is in the range of 0–2m. The processing flow for the defect spectrum in the 0–2m depth range is as follows: DC drift removal, static correction, gain adjustment, bandpass filtering, and background removal. Since the 1–2m depth range is affected by reflected radar waves, a secondary gain processing is required for the defect spectrum in this depth range to highlight the radar waves within this range. Therefore, it is necessary to analyze the depth of the defect location. When the depth of the defect location is in the range of 0–1m, a first signal processing step is performed, which includes: DC drift removal, static correction, gain adjustment, bandpass filtering, and background removal. The radar spectrum of the detected defect after the first signal processing step is determined as the actual defect radar spectrum for the test path.

[0097] When the depth of the lesion location is within the range of 1 to 2 meters, a second signal processing is performed on the basis of the first signal processing. The second signal processing includes gain processing. At this time, the radar map of the lesion detected after the second signal processing is determined as the actual radar map of the lesion on the test road.

[0098] In one embodiment, the step of drawing a color pixel map of defects based on the location and size of defects in the test road, with reference to the actual defect radar map, includes: obtaining the location, width, and dielectric constant of defects in the test road; obtaining the length of defects in the actual defect radar map; obtaining the dielectric constant of the road structure of the test road; and drawing the color pixel map of defects based on the length, location, width, dielectric constant of the defects, and dielectric constant of the road structure.

[0099] In this embodiment, with Figure 2 For reference, obtain Figure 2The actual length of the defects in the radar map is obtained, followed by the location, width, dielectric constant of the defects, and dielectric constant of the road structure in the test road. Using graphics software, different colors are assigned to the dielectric constants of the defects and the road structure, with black for the dielectric constant and gray for the dielectric constant of the road structure, thus creating a color pixel map of the defects. Figure 5 As shown, Figure 5 Therefore Figure 2 Color pixel map of the disease drawn for reference.

[0100] In one embodiment, the method for obtaining the disease length in the actual disease radar map includes: obtaining a waveform of the actual disease radar map; picking the first arrival radar waves of the disease based on the waveform; wherein the first arrival radar waves of the disease include multiple first arrival radar waves; judging the multiple first arrival radar waves according to a disease length index based on the multiple first arrival radar waves to obtain the starting position and ending position of the disease; and obtaining the disease length in the actual disease radar map based on the difference between the starting position and the ending position of the disease.

[0101] In this embodiment, since radar waves undergo abrupt changes and form peaks when entering another medium, the thickness of structural layers of different materials can be quantitatively determined based on the distance between adjacent peaks. Based on this principle, radar processing software converts the actual radar image of the defect into a waveform diagram, i.e., an A-scan single-channel waveform diagram, as shown below. Figure 7 , Figure 7 This is provided by the embodiments of the present invention. Figure 2 The corresponding A-scan diagram of the disease map is used to analyze the waveform. The disease contains multiple first-arrival radar waves. By using the disease length index to judge the multiple first-arrival radar waves, the starting and ending positions of the disease can be determined. The difference between the starting and ending positions of the disease in the horizontal direction of the road surface is the actual length of the disease.

[0102] In one embodiment, the step of determining the starting and ending positions of the disease based on the plurality of first-arrival radar waves according to a disease length index includes: obtaining trough information of the plurality of first-arrival radar waves; wherein the trough information is the ratio of trough amplitude to trough height; constructing the disease length index based on the ratio of trough amplitude to trough height; quantitatively determining the trough information of the plurality of first-arrival radar waves based on the disease length index, selecting two first-arrival radar waves whose trough information satisfies the disease length index and are located at the middle end of the plurality of first-arrival radar waves, and determining them as the starting and ending first-arrival radar waves; obtaining the starting and ending positions of the disease based on the trough positions of the starting and ending first-arrival radar waves.

[0103] In this embodiment, extensive research on disease patterns shows that the most obvious location of the disease is at the trough below the wave crest. Therefore, this invention selects the trough below the wave crest of the radar first arrival wave for research, obtaining trough information from multiple first arrival radar waves, and referring to... Figure 8 , Figure 8 This is provided by the embodiments of the present invention. Figure 7 The corresponding A-scan (single-channel waveform) trough diagram shows that the black shaded area represents the trough below the peak. By using the trough information below the most prominent peak in the A-scan waveform, the start and end points of the disease can be determined, thus accurately obtaining the length of the disease in the radar image. The trough information is represented by the ratio of trough amplitude l to trough width h, constructing a disease length index BC. The expression for the disease length index is as follows:

[0104]

[0105] Where a represents the i-th A-scan trough information, b represents the difference between the two adjacent A-scan trough information of the i-th A-scan, and the length of the defect is the road surface length between the two A-scans at the start and end points.

[0106] By setting specific constraints on the length index of the disease, the conditions for meeting the length index can be determined. In this invention, the trough below the most obvious peak at the disease location should have an amplitude-to-width ratio greater than or equal to 0.4, and the change at this point between the two preceding and following A-scans should not exceed 0.2. The trough information that meets the disease length index and is located at the two first-arrival radar waves at the ends of multiple first-arrival radars is selected as the starting and ending first-arrival radar waves. The disease length can then be obtained as the road surface length corresponding to the two A-scans at the starting and ending points. By measuring the road surface length, the actual length of the disease can be determined.

[0107] Furthermore, it can be set that when the disease is obvious, the difference between the disease length obtained by the disease length index BC and the actual disease length can be up to 7%; when the disease is not obvious, the difference of BC can be up to 27%, which is highly accurate. This satisfies the need for quantitatively reflecting the disease length and also meets the requirements for drawing the disease color pixel map required by the inversion model. Through the disease length index provided by this invention, the disease length can be quantitatively determined based on the ratio of the trough amplitude to the trough width at the disease location, linking the radar spectrum with the disease size. This achieves the goal of obtaining the disease length from the radar spectrum, reducing the difficulty of manually interpreting the disease radar spectrum.

[0108] In a second aspect, the present invention provides an apparatus for determining the actual size of a disease, referring to... Figure 9 , Figure 9 This is a schematic diagram of a device for determining the actual size of a disease according to an embodiment of the present invention. The device includes: a first acquisition module 901, a second acquisition module 902, a drawing module 903, a third acquisition module 904, and a fourth acquisition module 905.

[0109] First acquisition module 901: used to acquire the actual damage radar map of the test road based on the results of radar detection of the test road;

[0110] The second acquisition module 902 is used to set forward modeling parameters and acquire forward modeling maps of diseases based on the location and size of the diseases in the test road.

[0111] Drawing module 903: used to draw a color pixel map of the disease based on the location and size of the disease in the test road, with the actual disease radar map as a reference; wherein the location and size of the disease in the color pixel map correspond one-to-one with the actual disease radar map;

[0112] The third acquisition module 904 is used to construct an evolution model from the actual disease radar map, through the disease forward model map to the disease color pixel map;

[0113] The fourth acquisition module 905 is used to input the actual damage radar map of the road to be tested into the evolution model to obtain the target damage color pixel map corresponding to the road to be tested. The target damage color pixel map is used to describe the actual size of the damage in the road to be tested.

[0114] The device provided by this invention allows for the training of a road defect detection system using actual radar and forward modeling maps of road defects. This process filters background noise from the actual maps, highlighting the defects and obtaining an SZ inversion model, which outputs a forward modeling map of the defects. Then, the forward modeling map is used to train a ZB inversion model, linking the radar map to the size and location of the defects. The forward modeling map serves as a bridge, connecting the actual radar map to the pixel image of the defects.

[0115] Based on these two inversion models, the invention achieves the goal of directly obtaining information about road defects from radar images. This allows for the direct determination of the size and location of hidden road defects from actual radar images, facilitating appropriate responses from maintenance personnel. This enhances the application capabilities of ground-penetrating radar in actual road detection and provides technical support for the identification and inversion of abnormal areas within structures. Furthermore, this invention is simple to operate, does not require advanced radar signal interpretation skills, and can be used by operators with or without radar knowledge.

[0116] In this embodiment, the above-mentioned device is applied to road radar detection to obtain the actual radar map of the road. The size and location of hidden road defects can be directly obtained from the actual radar map, so that maintenance personnel can make reasonable responses. It can also provide certain theoretical analysis methods and technical support for the identification and inversion of abnormal areas inside the road structure, and further enrich the research in related fields.

[0117] A third aspect of the present invention provides an electronic device, including a memory, a processor, and a computer program stored in the memory, wherein the processor executes the computer program to implement the steps of the method described in the first aspect of the present invention.

[0118] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on the differences from other embodiments. The same or similar parts between the various embodiments can be referred to each other.

[0119] Embodiments of the present invention are described with reference to flowchart illustrations and / or block diagrams of methods and apparatus according to embodiments of the present invention. It should be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing terminal device to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing terminal device, generate instructions for implementing the flowchart illustrations. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0120] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing terminal device to operate in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0121] These computer program instructions can also be loaded onto a computer or other programmable data processing terminal equipment, causing a series of operational steps to be performed on the computer or other programmable terminal equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable terminal equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0122] Although preferred embodiments of the present invention have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of the embodiments of the present invention.

[0123] Finally, it should be noted that in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or terminal device that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or terminal device. Without further limitations, an element defined by the phrase "comprising one..." does not exclude the presence of other identical elements in the process, method, article, or terminal device that includes said element.

[0124] The present invention has provided a detailed description of a method, apparatus, and device for determining the actual size of a disease. Specific examples have been used to illustrate the principles and implementation methods of the present invention. The descriptions of the above embodiments are only for the purpose of helping to understand the method and core ideas of the present invention. At the same time, those skilled in the art will recognize that there will be changes in the specific implementation methods and application scope based on the ideas of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A method for determining the actual size of a disease, characterized in that, The method includes: Based on the results of radar detection of the test road, the actual damage radar map of the test road is obtained; Based on the location and size of defects in the test road, forward modeling parameters are set to obtain a forward modeling map of defects, including: obtaining the road structure of the test road; and performing forward modeling on the road structure, the location and size of defects in the test road, and the defects based on the forward modeling parameters to obtain a forward modeling map of defects; wherein the forward modeling parameters are consistent with the parameters used for radar detection of the test road; Using the actual disease radar map as a reference, a color pixel map of the disease is drawn based on the location and size of the disease in the test road; wherein, the location and size of the disease in the color pixel map correspond one-to-one with the actual disease radar map; An evolutionary model is constructed from the actual disease radar map, through the disease forward model, to the disease color pixel map, including: performing deep learning on the actual disease radar map and the disease forward model to construct a first inversion model and find the relationship between the actual disease map and the disease forward model; wherein, the first inversion model is used to output the disease forward model simulation map corresponding to the actual disease radar map; performing deep learning on the disease forward model simulation map and the disease color pixel map to construct a second inversion model and find the relationship between the disease forward model and the disease color pixel map; wherein, the second inversion model is used to output the target disease color pixel map corresponding to the disease forward model simulation map; based on the first inversion model and the second inversion model, the evolutionary model is obtained, and the relationship between the actual disease forward model and the disease forward model, and the relationship between the disease forward model and the disease color pixel map, is found; The actual damage radar map of the road to be tested is input into the evolution model to obtain the target damage color pixel map corresponding to the road to be tested. The target damage color pixel map is used to describe the actual size of the damage in the road to be tested.

2. The method according to claim 1, characterized in that, Before obtaining the actual damage radar map of the test road based on the radar detection results, the method further includes: Obtain the radar map of the detected disease corresponding to the results of the radar detection; Based on the radar map of the detected disease, the depth of the disease location is obtained; Based on the depth of the location of the disease, the radar map of the disease detection is subjected to target processing, which includes at least one of DC drift, static correction, gain, bandpass filtering and background removal. The radar map of the detected defects after the target has been processed is determined as the actual radar map of the defects of the test road.

3. The method according to claim 2, characterized in that, The target processing of the radar map of the detected disease based on the depth of the disease location includes: When the depth is at the first depth, the radar map of the detected defects is processed once, and the radar map of the detected defects after the first signal processing is determined as the actual radar map of the defects of the test road. When the depth of the disease location is at the second depth, the radar map of the detected disease after the first signal processing is subjected to a second signal processing, and the radar map of the detected disease after the second signal processing is determined as the actual radar map of the disease of the test road. The primary signal processing includes: DC drift processing, static correction processing, gain processing, bandpass filtering processing, and background removal processing; the secondary signal processing includes the gain processing. The first depth is less than the second depth.

4. The method according to claim 1, characterized in that, The step of drawing a color pixel map of the defects based on the actual radar map of the defects and the location and size of the defects in the test road, with reference to the actual defect radar map, includes: Obtain the location, width, and dielectric constant of the defects in the test path; Obtain the length of the disease in the actual disease radar map; Obtain the dielectric constant of the road structure of the test road; Based on the length, location, width, dielectric constant of the lesion, and dielectric constant of the road structure, a color pixel map of the lesion is drawn.

5. The method according to claim 4, characterized in that, The method for obtaining the disease length in the actual disease radar map includes: Obtain the waveform of the actual disease radar map; Based on the waveform diagram, the first arrival radar wave of the disease is picked up; wherein, the first arrival radar wave of the disease includes multiple first arrival radar waves; Based on the multiple first-arrival radar waves, the starting and ending positions of the disease are obtained by judging the multiple first-arrival radar waves according to the disease length index. Based on the difference between the starting and ending positions of the disease, the length of the disease in the actual disease radar map is obtained.

6. The method according to claim 5, characterized in that, The step of determining the starting and ending positions of the disease based on the multiple first-arrival radar waves according to the disease length index includes: Obtain the valley information of the plurality of first-arrival radar waves; wherein, the valley information is the ratio of valley amplitude to valley height; The disease length index is constructed based on the ratio of the trough amplitude to the trough height. The two first-arrival radar waves that satisfy the disease length index and are located at the middle end of the plurality of first-arrival radars are determined as the starting first-arrival radar wave and the ending first-arrival radar wave. Based on the trough position of the initial arrival radar wave at the starting point and the trough position of the initial arrival radar wave at the ending point, the starting point position and the ending point position of the disease are obtained.

7. A device for determining the actual size of a disease, characterized in that, The device includes: First acquisition module: used to acquire the actual damage radar map of the test road based on the results of radar detection of the test road; The second acquisition module is used to set forward modeling parameters and acquire a forward modeling map of the defects based on the location and size of the defects in the test road. This includes: acquiring the road structure of the test road; performing forward modeling on the road structure, the location and size of the defects in the test road based on the forward modeling parameters, and acquiring a forward modeling map of the defects; wherein the forward modeling parameters are consistent with the parameters used for radar detection of the test road. The drawing module is used to draw a color pixel map of the disease based on the location and size of the disease in the test road, with the actual disease radar map as a reference; wherein the location and size of the disease in the color pixel map correspond one-to-one with the actual disease radar map. The third acquisition module is used to construct an evolutionary model from the actual disease radar map, through the disease forward model, to the disease color pixel map, including: performing deep learning on the actual disease radar map and the disease forward model to construct a first inversion model and find the relationship between the actual disease map and the disease forward model; wherein, the first inversion model is used to output the disease forward model simulation map corresponding to the actual disease radar map; performing deep learning on the disease forward model simulation map and the disease color pixel map to construct a second inversion model and find the relationship between the disease forward model and the disease color pixel map; wherein, the second inversion model is used to output the target disease color pixel map corresponding to the disease forward model simulation map; based on the first inversion model and the second inversion model, the evolutionary model is acquired, and the relationship between the actual disease forward model and the disease forward model, and the relationship between the disease forward model and the disease color pixel map are found; The fourth acquisition module is used to input the actual damage radar map of the road to be tested into the evolution model to obtain the target damage color pixel map corresponding to the road to be tested. The target damage color pixel map is used to describe the actual size of the damage in the road to be tested.

8. An electronic device comprising a memory, a processor, and a computer program stored in the memory, characterized in that, The processor executes the computer program to implement the method for determining the actual size of the lesion as described in any one of claims 1 to 6.

Images