Ordinary road slope deformation time sequence monitoring method based on TL-1 satellite

Abstract

The invention relates to the technical field of image analysis, in particular to a common road slope deformation time sequence monitoring method based on a TL-1 satellite, and the method comprises the steps: constructing a three-dimensional model of a target region based on a historical image and a coordinate system; inputting each satellite image in the time sequence image into an image recognition model to obtain a corresponding recognition image, extracting a crack area in the target area based on each recognition image, obtaining crack information in each satellite image, and obtaining first deformation time sequence information according to crack change; updating each satellite image into the three-dimensional model according to a time sequence, acquiring displacement change information and morphological change information of the target area in a preset time period, and acquiring second deformation time sequence information based on the displacement change information and the morphological change information; and obtaining a slope deformation prediction result according to the first deformation time sequence information, the second deformation time sequence information and a deformation prediction model. According to the invention, the problem of poor identification precision of slope deformation is solved, and the monitoring precision of slope deformation is improved.

Description

Technical Field

[0001] The present application relates to the field of image analysis technology, and in particular to a method for time-series monitoring of slope deformation of ordinary highways based on a TL-1 satellite. Background Art

[0002] Ordinary highway slopes are subject to deformation and landslide risks due to natural erosion, climate change, human activities and other factors. Traditional slope monitoring methods rely on ground sensors and manual inspections, and have problems such as high cost, untimely data collection, and limited coverage. With the rapid development of transportation infrastructure, the stability problem of ordinary highway slopes has become increasingly prominent. Traditional slope monitoring methods have problems such as high cost, low efficiency, and poor real-time performance. Therefore, it is urgent to develop an efficient and real-time slope deformation monitoring method. Similar prior art includes a Chinese patent application with publication number CN118552848A, which proposes a slope deformation monitoring method, device, electronic device and storage medium, the method comprising: collecting a monitoring image of a slope monitoring area through an image acquisition device; determining the current illumination type of the slope monitoring area based on the monitoring image; determining a template image of the slope monitoring area under the illumination type based on the acquisition time of the monitoring image and the illumination type; comparing the monitoring image with the template image to determine whether the slope monitoring area is deformed. Through the method and device, shadow changes caused by different lighting directions and lighting intensities are avoided, false alarms caused by lighting changes are reduced, and the accuracy of slope deformation monitoring is improved. In addition, similar prior art also includes a Japanese patent application with publication number JP2023140677A, which provides a slope instability position extraction system, including: an acquisition unit for acquiring a visible light image of a collapse extraction area on the surface and a vegetation index map representing the distribution of vegetation indexes in the collapse extraction area; a detection process for detecting low indices; an area in the vegetation index map where the vegetation index is less than a threshold, and an extraction process for extracting the slope collapse position by pattern matching the judgment target area including the low index area in the visible light image, which can extract the slope instability position at a low cost and improve the extraction accuracy at the same time. Although the above two patent applications both solve the problem of slope monitoring and identification, they do not obtain the deformation characteristics within a preset time period based on the acquisition time series, nor do they obtain the specific deformation category and the corresponding deformation trend based on the deformation characteristics. Therefore, the development trend of the slope deformation cannot be obtained. The purpose of the present invention is to provide a time-series monitoring method for ordinary highway slope deformation based on the LT-1 satellite, which can achieve high-precision prediction of slope deformation and improve the reliability and safety of monitoring. Summary of the invention

[0003] In order to better solve the above problems, the present invention provides a method for monitoring the deformation of ordinary highway slopes based on LT-1 satellite. The method: Acquire a historical image of the target area, preprocess the historical image, select a fixed object outside the target area as a reference object, use the reference object as the origin of a coordinate system, and construct a three-dimensional model of the target area based on the historical image and the origin of the coordinate system; Input each satellite image in the time-series images acquired within a preset time period into an image recognition model to obtain a recognition image corresponding to the satellite image, extract a crack region in the target region based on each recognition image, and obtain crack information corresponding to the crack region in each satellite image, obtain crack changes through all the crack information corresponding to the time-series images, and obtain first deformation time-series information according to the crack changes; updating each of the satellite images in the time series images to the three-dimensional model in time sequence, and obtaining the position information and morphological information of the slope in the target area in each of the satellite images, obtaining the displacement change information and morphological change information of the target area within the preset time period according to the position information and morphological information of the slope corresponding to each of the satellite images, and obtaining the second deformation time series information based on the displacement change information and the morphological change information; A slope deformation prediction result is obtained according to the first deformation time series information, the second deformation time series information and a deformation prediction model.

[0004] As a preferred technical solution of the present invention, the acquisition of the crack information includes: Preprocess each satellite image in the time series images acquired within the preset time period and input the preprocessed satellite image into the image recognition model to acquire the recognition image corresponding to each satellite image; wherein the recognition image is a binary image, and is an image that distinguishes the crack area and the slope area in the corresponding target area by pixel values, and the satellite image is a three-dimensional image; A crack area is extracted based on the crack probability value corresponding to the crack in the identified image, and each of the crack areas is divided into a plurality of sub-segments according to a set distance in the vertical direction, and the first tangent and the second tangent of both sides of each of the sub-segments are obtained, and the first angle and the second angle between each adjacent sub-segment and the first tangent and the second tangent on the same side of the sub-segment are calculated, wherein the first angle and the second angle are the angles on one side inside the crack area, and when at least one of the first angle and the second angle corresponding to at least one of the adjacent sub-segments and the sub-segment is less than a set angle, the sub-segment is regarded as a crack segment, and when at least one of the adjacent sub-segments and the sub-segment corresponding to ... When the first angle and the second angle are both greater than or equal to the set angle, at least one of the adjacent sub-segments is used as the target sub-segment, and the target sub-segment and the sub-segment are merged into a new sub-segment, and the two sides of the first angle are connected with the two intersection points of the corresponding sub-segment and used as the third tangent of the new sub-segment. Similarly, the fourth tangent on the other side of the new sub-segment is obtained, and this step is repeated to obtain multiple crack segments in the crack area, wherein the crack angle between adjacent crack segments is greater than or equal to the set angle, and the crack area in the identified image, the crack segment corresponding to each crack area, and the size information and position information of each crack segment are used as the crack information.

[0005] As a preferred technical solution of the present invention, the acquisition of the first deformation time series information of the target area includes: The crack segment in the crack information corresponding to each satellite image in the time series image is compared with the crack information of the crack segment at the same position in the crack information corresponding to an adjacent satellite image in the time series in chronological order to obtain crack change information, wherein the acquisition time of the adjacent satellite image is later than the acquisition time of the satellite image, and the first deformation time series information of the target area within the preset time period corresponding to the time series image is obtained based on the crack change information corresponding to each satellite image and the adjacent satellite image, wherein the crack change information includes size change and depth change.

[0006] As a preferred technical solution of the present invention, the training of the image recognition model includes: The image recognition model is created by a model creation unit, and a sample image is read from a storage unit, wherein the sample image includes first label information, and the first label information is a pre-labeled label. The sample image and the first label information are used as first training data to train the image recognition model, and a moving weight mean of a prediction result corresponding to each pixel in the sample image is periodically calculated during a training iteration process, and the prediction result obtained in this iteration is replaced by the moving weight mean, and second label information is set for a second pixel whose prediction result is greater than a set threshold, wherein the second pixel is a pixel not labeled with the pre-labeled label, and a pixel including the pre-labeled label is used as a first pixel; Acquire the second pixel adjacent to at least two of the first pixels, and use at least two of the first pixels as reference pixels, wherein the first label information corresponding to the reference pixels is the same and is located on both sides of the second pixel, compare the second label information corresponding to the second pixel with the first label information corresponding to the reference pixel, and when the second label information is the same as the first label information corresponding to the reference pixel, add the second pixel and the corresponding second label information as the first learning sample pixel to the second training data, otherwise, re-extract the difference feature between the second pixel and the reference pixel and enhance it, and re-train the image recognition model through the second training data.

[0007] As a preferred technical solution of the present invention, the acquisition of the second deformation time series information includes: Each pre-processed satellite image in the time series images acquired within the preset time period is sequentially updated to the three-dimensional model, and the position information and morphological information of the slope in the satellite images adjacent in time are compared to obtain displacement change information and slope morphological information change, and the displacement change information and the slope morphological change information are used as the second deformation time series information.

[0008] As a preferred technical solution of the present invention, obtaining the slope deformation prediction result includes: The first deformation time series information and the second deformation time series information are fused in time series to obtain slope deformation time series information, and the slope deformation time series information is input into a deformation prediction model to obtain the slope deformation prediction result, which includes the deformation development trend and slope deformation type of the target area in the future.

[0009] As a preferred technical solution of the present invention, the deformation prediction model is a model trained with historical deformation data, wherein the historical deformation data includes displacement deformation, crack deformation and corresponding deformation types, and also includes the entire development process of the displacement deformation and the crack deformation during the entire deformation process of the deformation type.

[0010] As a preferred technical solution of the present invention, the deformation development trend of the target area includes the displacement change trend, morphology change trend and crack change trend of the slope over time.

[0011] The present invention also provides a common highway slope deformation time-series monitoring system based on TL-1, the system is used to implement the above method, the system comprises: A model creation unit, configured to obtain a historical image of a target area, preprocess the historical image, select a fixed object outside the target area as a reference object, and use the reference object as the origin of a coordinate system, and construct a three-dimensional model of the target area based on the historical image and the origin of the coordinate system; The calculation unit is configured to: input each satellite image in the time-series images acquired within the preset time period into an image recognition model to acquire an identification image corresponding to the satellite image, extract a crack area in the target area based on each identification image, and acquire crack information corresponding to the crack area in each satellite image, acquire crack changes through all the crack information corresponding to the time-series images, and acquire first deformation time-series information according to the crack changes; update each satellite image in the time-series images to the three-dimensional model in chronological order, and acquire position information and morphological information of the slope in the target area in each satellite image, acquire displacement change information and morphological change information of the target area within the preset time period according to the position information and morphological information of the slope corresponding to each satellite image, and acquire second deformation time-series information based on the displacement change information and the morphological change information; A prediction unit is used to obtain a slope deformation prediction result according to the first deformation time series information, the second deformation time series information and a deformation prediction model.

[0012] The present invention provides a computer storage medium, characterized in that the storage medium stores program instructions, wherein when the program instructions are executed, the device where the storage medium is located is controlled to execute the above method.

[0013] Compared with the prior art, the beneficial effects of the present invention are at least as follows: The present invention combines the historical images and time-series images of the LT-1 satellite to construct a three-dimensional model of the target area and extract the crack area and its change information, thereby realizing high-precision monitoring of the deformation of ordinary highway slopes. This method can provide more accurate deformation data, and improves the accuracy and reliability of monitoring compared to traditional monitoring methods. Using the time-series images of the LT-1 satellite, the present invention can update the displacement change and morphological change information of the target area in real time and obtain the latest deformation time-series information. This real-time monitoring capability enables the management department to respond to the deformation of the slope in a timely manner, improves the timeliness of responding to geological disasters, automatically extracts crack information through the image recognition model, and can ensure the accuracy of the above image recognition model without the need for precise labeling of sample images, reduces labor intensity and human errors, and improves the efficiency and accuracy of crack analysis. Combined with the first deformation time series information and the second deformation time series information, through the mutual cooperation of the above technical solutions, it is not only possible to predict the future deformation development trend and deformation type of the slope, providing a scientific basis for preventing and mitigating geological disasters, but also reduces dependence on ground sensors and manual inspections, reduces monitoring costs, and improves monitoring efficiency, especially in remote or inaccessible areas. The present invention not only focuses on the deformation of cracks, but also combines displacement changes and morphological changes. Information, through data fusion technology, provides a more comprehensive deformation analysis. The method of the present invention is suitable for slope deformation monitoring under various environmental conditions, is not limited by natural conditions such as weather and terrain, and has strong environmental adaptability. BRIEF DESCRIPTION OF THE DRAWINGS

[0014] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the accompanying drawings required for use in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other accompanying drawings can be obtained based on these accompanying drawings without paying creative work.

[0015] Figure 1 This is a flow chart of a method for time-series monitoring of slope deformation of ordinary highways based on LT-1 satellite in an embodiment of the present application; Figure 2 It is a schematic diagram of the structure of the ordinary highway slope deformation time-series monitoring system based on the LT-1 satellite in the embodiment of the present application. DETAILED DESCRIPTION

[0016] The embodiments of the present application provide a method, device, equipment and medium for customizing feed formulas based on artificial intelligence. The terms "first", "second", "third", "fourth", etc. (if any) in the specification and claims of the present application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the data used in this way can be interchangeable where appropriate, so that the embodiments described here can be implemented in an order other than that illustrated or described here. In addition, the terms "including" or "having" and any variations thereof are intended to cover non-exclusive inclusions, for example, a process, method, system, product or device that includes a series of steps or units is not necessarily limited to those steps or units that are clearly listed, but may include other steps or units that are not clearly listed or inherent to these processes, methods, products or devices.

[0017] For ease of understanding, the specific process of the embodiment of the present application is described below. Figure 1 As shown, the time-series monitoring method for slope deformation of ordinary highways based on LT-1 satellite in the embodiment of the present application includes: Step S1: acquiring a historical image of a target area, preprocessing the historical image, selecting a fixed object outside the target area as a reference object, and taking the reference object as the origin of a coordinate system, and constructing a three-dimensional model of the target area based on the historical image and the origin of the coordinate system; Specifically, by preprocessing the historical images of the target area, the preprocessing includes angle adjustment and denoising, etc., and selecting the above-mentioned fixed object outside the above-mentioned target area and maintaining a certain distance from the above-mentioned target area as a reference object, and establishing a coordinate system with the above-mentioned reference object as the origin of the coordinate system, it can be ensured that even if the above-mentioned target area is deformed, the position of the coordinate origin remains unchanged, and the slope deformation can be accurately measured. The above-mentioned three-dimensional model is established with the above-mentioned historical images and the above-mentioned coordinate system with the reference object as the origin. Through the above-mentioned technical solution, the above-mentioned three-dimensional model of the above-mentioned target area can be obtained, thereby laying a foundation for accurately calculating the crack deformation information, displacement changes and morphological changes of the above-mentioned target area.

[0018] Step S2: inputting each satellite image in the time-series images acquired within the preset time period into an image recognition model to acquire a recognition image corresponding to the satellite image, extracting a crack region in the target region based on each recognition image, and acquiring crack information corresponding to the crack region in each satellite image, acquiring crack changes through each crack information corresponding to the time-series image, and acquiring first deformation time-series information according to the crack changes; Specifically, by inputting each satellite image in the above-mentioned time series image into the above-mentioned image recognition model, the corresponding above-mentioned recognition image is obtained, and the slope area and the crack area in the above-mentioned satellite image are obtained according to the above-mentioned recognition image, and the crack information of the above-mentioned crack area is obtained. The crack information corresponding to the above-mentioned satellite image is compared with the crack information corresponding to the adjacent satellite image in time series to obtain crack change information, and the crack change information between each of the above-mentioned satellite images and the above-mentioned adjacent satellite images within the above-mentioned preset time period is used as the above-mentioned first deformation time series information of the above-mentioned target area within the above-mentioned preset time period. Through the above-mentioned technical scheme, the crack change information of the crack section corresponding to each crack area in the above-mentioned target area can be accurately obtained, which lays a foundation for further combining the second deformation time series information, i.e., the displacement change information and the morphological change information of the slope to obtain accurate slope deformation time series information.

[0019] Step S3: updating each of the satellite images in the time series images to the three-dimensional model in time sequence, and obtaining the position information and morphological information of the slope in the target area in each of the satellite images, obtaining the displacement change information and morphological change information of the target area within the preset time period according to the position information and morphological information of the slope corresponding to each of the satellite images, and obtaining the second deformation time series information based on the displacement change information and the morphological change information; Specifically, by updating each of the satellite images preprocessed in the time series images acquired within the preset time period into the three-dimensional model, and recording the corresponding slope position and shape in the three-dimensional model in sequence, the position change and shape change of the slope in the three-dimensional model corresponding to each two adjacent satellite images are calculated, and the displacement change information and shape change information of the slope within the preset time period are obtained based on the position change and shape change, and the displacement change information and shape change information of the slope are used as the second deformation time series information of the target area. Through the above technical solution, the displacement change and shape change of the slope in the target area within the preset time period can be accurately obtained, laying a foundation for further obtaining the slope deformation of the target area.

[0020] Step S4: obtaining a slope deformation prediction result according to the first deformation time series information, the second deformation time series information and a deformation prediction model.

[0021] Specifically, by fusing the first deformation time series information and the second deformation time series information in time series, the complete time series information of the slope deformation that changes with time can be obtained, and based on the slope deformation time series information and the deformation prediction model, the deformation development trend of the target area in the future and the deformation type of the slope can be obtained. Through the above technical solution, the deformation development trend and deformation type of the target area can be understood in time, providing a basis for taking timely intervention measures to prevent deformation deterioration. Furthermore, the acquisition of the crack information includes: Preprocess each satellite image in the time series images acquired within the preset time period and input the preprocessed satellite image into the image recognition model to acquire the recognition image corresponding to each satellite image; wherein the recognition image is a binary image, and is an image that distinguishes the crack area and the slope area in the corresponding target area by pixel values, and the satellite image is a three-dimensional image; Specifically, the preset time period is a time period in which the LT-1 satellite can capture images of the target area. The time series images are all satellite images captured in chronological order at a set period within the preset time period. Each of the satellite images in the time series images is pre-processed and input into the image recognition model. The pre-processing includes enhancement, denoising, angle adjustment, and light correction. The angle correction can be used to obtain a front image of the target area, i.e., the direction of the slope where the crack is located, thereby facilitating the identification of the crack. The image recognition model can be used to obtain the probability of each pixel being divided into a slope or a crack, i.e., the identification image. Since the identification image is a binary image, the slope and the crack can be more clearly distinguished. The technical solution can obtain an accurate identification image, laying a foundation for further extraction of crack information.

[0022] A crack area is extracted based on the crack probability value corresponding to the crack in the identified image, and each of the crack areas is divided into a plurality of sub-segments according to a set distance in the vertical direction, and the first tangent and the second tangent of both sides of each of the sub-segments are obtained, and the first angle and the second angle between each adjacent sub-segment and the first tangent and the second tangent on the same side of the sub-segment are calculated, wherein the first angle and the second angle are the angles on one side inside the crack area, and when at least one of the first angle and the second angle corresponding to at least one of the adjacent sub-segments and the sub-segment is less than a set angle, the sub-segment is regarded as a crack segment, and when at least one of the adjacent sub-segments and the sub-segment corresponding to ... When the first angle and the second angle are both greater than or equal to the set angle, at least one of the adjacent sub-segments is used as the target sub-segment, and the target sub-segment and the sub-segment are merged into a new sub-segment, and the two sides of the first angle are connected with the two intersection points of the corresponding sub-segment and used as the third tangent of the new sub-segment. Similarly, the fourth tangent on the other side of the new sub-segment is obtained, and this step is repeated to obtain multiple crack segments in the crack area, wherein the crack angle between adjacent crack segments is greater than or equal to the set angle, and the crack area in the identified image, the crack segment corresponding to each crack area, and the size information and position information of each crack segment are used as the crack information.

[0023] Specifically, a pixel area in the above-mentioned recognition image having a crack probability value greater than a first set threshold is taken as a crack area, and a pixel area in the above-mentioned recognition image having a slope probability value greater than a second set threshold is taken as a slope area, wherein the first set threshold and the second set threshold are both greater than or equal to 0.8, and the maximum values ​​of the above-mentioned crack probability value and the above-mentioned slope probability value are 1, and each of the above-mentioned crack areas in the above-mentioned recognition image is divided into a plurality of the above-mentioned sub-segments according to a set distance in the vertical direction, the above-mentioned set distance is the length of k pixel rows, and the value of k is greater than or equal to 3 and less than or equal to 5, and the first tangent and the second tangent of the two side edges of each of the above-mentioned sub-segments are obtained, and the above-mentioned first tangent and the above-mentioned second tangent are respectively line segments that are tangent to the pixels on the corresponding side edge close to the other side edge and intersect with the corresponding side edge, and the above-mentioned first angle between the first tangents on the same side of the two adjacent above-mentioned sub-segments and the above-mentioned second angle between the above-mentioned second tangents on the same side are calculated, wherein the above-mentioned first angle and the above-mentioned second angle are both angles pointing to the above-mentioned crack area, and in the two adjacent sub-segments of the above-mentioned sub-segment, the first angle and the second angle corresponding to each adjacent sub-segment are at least 1. When at least one of the angles is less than the set angle, that is, when the crack trends of the above sub-segment and the two adjacent sub-segments are inconsistent, the above sub-segment is taken as a crack segment; when the first angle and the second angle between at least one adjacent sub-segment of the above sub-segment and the above sub-segment are both greater than the set angle, at least one adjacent sub-segment is merged with the above sub-segment into a new sub-segment, and the two sides of the above first angle are connected with the intersection of the corresponding two sub-segments to obtain the third tangent of the above new sub-segment, and the fourth tangent of the above new sub-segment corresponding to the second angle is obtained in the same way, and the method of obtaining the above crack segment is repeated to obtain the crack segment in each of the above crack areas, and each of the above crack areas in the above recognition image, the crack segment corresponding to each crack area, and the scale information and position information of each of the above crack segments are taken as the above crack information. Through the above technical solution, not only the recognition image of each satellite image in the above time series image can be obtained, but also the crack area in the above recognition image and the crack segment of each crack area can be obtained, so as to further obtain the change trend of the crack corresponding to the above time series image, thereby laying a foundation for obtaining the above first deformation time series information.

[0024] Furthermore, acquiring the first deformation time series information of the target area includes: The crack segment in the crack information corresponding to each satellite image in the time series image is compared with the crack information of the crack segment at the same position in the crack information corresponding to an adjacent satellite image in the time series in chronological order to obtain crack change information, wherein the acquisition time of the adjacent satellite image is later than the acquisition time of the satellite image, and the first deformation time series information of the target area within the preset time period corresponding to the time series image is obtained based on the crack change information corresponding to each satellite image and the adjacent satellite image, wherein the crack change information includes size change and depth change.

[0025] Specifically, by comparing the crack information corresponding to the above satellite image with the crack information corresponding to the adjacent satellite image according to the acquisition time sequence of each satellite image in the above time series image, the crack change information is obtained, and based on the crack change information between each of the above satellite images and the above adjacent satellite images within the above preset time period, the above first deformation time series information of the above target area within the above preset time period is obtained. Through the above technical scheme, the crack change information of the crack section corresponding to each crack area in the above target area can be accurately obtained, laying a foundation for further combining the second deformation time series information, i.e., the displacement change of the slope to obtain accurate slope deformation time series information.

[0026] Furthermore, the training of the image recognition model includes: The image recognition model is created by a model creation unit, and a sample image is read from a storage unit, wherein the sample image includes first label information, and the first label information is a pre-labeled label. The sample image and the first label information are used as first training data to train the image recognition model, and a moving weight mean of a prediction result corresponding to each pixel in the sample image is periodically calculated during a training iteration process, and the prediction result obtained in this iteration is replaced by the moving weight mean, and second label information is set for a second pixel whose prediction result is greater than a set threshold, wherein the second pixel is a pixel not labeled with the pre-labeled label, and a pixel including the pre-labeled label is used as a first pixel; Specifically, when annotating the sample image, if pixel-level annotation is performed, the workload is relatively large and requires a lot of manpower. Therefore, a pre-annotation method, i.e., partial annotation, is adopted to add pre-annotated labels to the sample image, i.e., the first label information. The first label information includes slope pixel label information corresponding to a part of the slope area and crack pixel label information corresponding to a part of the crack area in the sample image. The sample image and the corresponding first label information are used as the first training data for the image recognition model. During the training process, the moving weighted average of the prediction results corresponding to each pixel in the sample image is periodically calculated, i.e., the weighted sum of the prediction value of each pixel this time and the moving weighted average of the previous time, wherein the weight of the current prediction result is greater than the weight of the previous time. The weight of the moving weight average is calculated, and the moving weight average replaces the prediction result, and the second label information is set for the second pixel whose prediction result is greater than the set threshold, the prediction result is the maximum value of the first probability that the pixel is a slope pixel and the second probability that the pixel is a crack pixel, and the second label information is set for the second pixel whose prediction result is greater than the set threshold, for example: the probability that the second pixel is a slope is 0.85, then the second label information is that the second pixel is a slope pixel, wherein the second pixel is a pixel that is not labeled with the pre-labeled label, that is, a pixel of unsupervised learning, and through the technical solution, the predicted label of the unlabeled pixel, that is, the second label information, is obtained, laying a foundation for further learning based on the second label information and the corresponding second pixel.

[0027] Acquire the second pixel adjacent to at least two of the first pixels, and use at least two of the first pixels as reference pixels, wherein the first label information corresponding to the reference pixels is the same and is located on both sides of the second pixel, compare the second label information corresponding to the second pixel with the first label information corresponding to the reference pixel, and when the second label information is the same as the first label information corresponding to the reference pixel, add the second pixel and the corresponding second label information as the first learning sample pixel to the second training data, otherwise, re-extract the difference feature between the second pixel and the reference pixel and enhance it, and re-train the image recognition model through the second training data.

[0028] Specifically, since the above-mentioned first pixel is supervised training, the prediction result of the first pixel is more accurate than the prediction result of the second pixel. Since the above-mentioned second pixel adjacent to a single first pixel may be located at the boundary between the slope and the crack, it is impossible to accurately judge whether the second label information of the above-mentioned second pixel is the same as the first label information of the above-mentioned first pixel. Therefore, by obtaining the above-mentioned second pixels adjacent to at least two of the above-mentioned first pixels, and taking the above-mentioned first pixels as reference pixels, the first label information corresponding to the above-mentioned reference pixels is compared with the above-mentioned second label information, wherein the above-mentioned reference pixels are located on both sides of the above-mentioned second pixel. When the second label information of the above-mentioned second pixel is the same as the first label information of the above-mentioned reference pixel, it means that the above-mentioned second pixel is adjacent to the first pixel. The prediction result of the pixel, that is, the second label information, is accurate. Therefore, the second pixel and the corresponding second label information are used as the second training data. On the contrary, when the second label information of the second pixel is different from the first label information of the reference pixel, it means that the label information of the second pixel is not accurate enough. The difference feature between the second pixel and the reference pixel is re-extracted, and the difference feature is enhanced to improve the recognition between the second pixel and the reference pixel. At the same time, the image recognition model is re-trained with the second training data. Through the technical solution, the pre-annotated satellite image can be accurately recognized at the pixel level, and the slope area and the crack area can be accurately identified, thereby improving the recognition accuracy of slope deformation.

[0029] Furthermore, the obtaining of the second deformation time series information includes: Each pre-processed satellite image in the time series images acquired within the preset time period is sequentially updated to the three-dimensional model, and the position information and morphological information of the slope in the satellite images adjacent in time are compared to obtain displacement change information and slope morphological information change, and the displacement change information and the slope morphological change information are used as the second deformation time series information.

[0030] Specifically, by updating each of the satellite images preprocessed in the time series images obtained within the preset time period into the three-dimensional model, and recording the corresponding slope position information and morphological information in the three-dimensional model in sequence, the position change and morphological change of the slope in the three-dimensional model corresponding to each two adjacent satellite images are calculated, and the displacement change information and morphological change information of the slope within the preset time period are obtained based on the position change and morphological change, and the displacement change information and morphological change information of the slope are used as the second deformation time series information of the target area. Through the above technical solution, the displacement change information and morphological change information of the slope in the target area within the preset time period can be accurately obtained, laying a foundation for further obtaining the slope deformation of the target area.

[0031] Furthermore, obtaining the slope deformation prediction result includes: The first deformation time series information and the second deformation time series information are fused in time series to obtain slope deformation time series information, and the slope deformation time series information is input into a deformation prediction model to obtain the slope deformation prediction result, which includes the deformation development trend and slope deformation type of the target area in the future.

[0032] Specifically, by fusing the above-mentioned first deformation time series information and the above-mentioned second deformation time series information in time series, the complete time series information of the slope deformation that changes with time can be obtained, and based on the above-mentioned slope deformation time series information and the above-mentioned deformation prediction model, the future deformation development trend of the above-mentioned target area and the above-mentioned edge deformation type can be obtained. Through the above-mentioned technical solution, the deformation development trend and deformation type of the above-mentioned target area can be understood in time, providing a basis for taking timely intervention measures to prevent deformation deterioration.

[0033] Furthermore, the deformation prediction model is a model trained with historical deformation data, and the historical deformation data includes displacement deformation, slope morphology change, crack deformation and corresponding deformation types, and also includes the entire development process of the displacement deformation, slope morphology change and crack deformation during the entire deformation process of the deformation type.

[0034] Specifically, the above-mentioned displacement deformation, slope morphology change and crack deformation are the deformation information of the corresponding deformation type changing over time. Through the above-mentioned technical scheme, the above-mentioned deformation prediction model with higher accuracy can be obtained, thereby providing a basis for obtaining the corresponding deformation type and the subsequent slope deformation trend through the displacement changes and crack changes in the above-mentioned target area.

[0035] Furthermore, the deformation development trend of the target area includes the displacement change trend, morphology change trend and crack change trend of the slope over time.

[0036] The present invention also provides a time-series monitoring system for slope deformation of ordinary highways based on TL-1, and the system is used to implement the above method, such as Figure 2 As shown, the system comprises: A model creation unit, configured to obtain a historical image of a target area, preprocess the historical image, select a fixed object outside the target area as a reference object, and use the reference object as the origin of a coordinate system, and construct a three-dimensional model of the target area based on the historical image and the origin of the coordinate system; The calculation unit is configured to: input each satellite image in the time-series images acquired within the preset time period into an image recognition model to acquire an identification image corresponding to the satellite image, extract a crack region in the target region based on each identification image, and acquire crack information corresponding to the crack region in each satellite image, acquire a crack change trend through all the crack information corresponding to the time-series images, and acquire first deformation time-series information according to the crack change trend; update each satellite image in the time-series images to the three-dimensional model in chronological order, and acquire position information and morphological information of the slope in the target region in each satellite image, acquire displacement change information and morphological change information of the target region within the preset time period according to the position information and morphological information of the slope corresponding to each satellite image, and acquire second deformation time-series information based on the displacement change information and the morphological change information; A prediction unit is used to obtain a slope deformation prediction result according to the first deformation time series information, the second deformation time series information and a deformation prediction model.

[0037] The present invention also provides a computer storage medium, wherein the storage medium stores program instructions, wherein when the program instructions are executed, the device where the storage medium is located is controlled to execute the above method.

[0038] In summary, the present invention combines the historical images and time-series images of the LT-1 satellite to construct a three-dimensional model of the target area and extract the crack area and its change information, thereby realizing high-precision monitoring of the deformation of ordinary highway slopes. This method can provide more accurate deformation data, and improves the accuracy and reliability of monitoring compared to traditional monitoring methods. Using the time-series images of the LT-1 satellite, the present invention can update the displacement change and morphological change information of the target area in real time and obtain the latest deformation time-series information. This real-time monitoring capability enables the management department to respond to the deformation of the slope in a timely manner, improves the timeliness of responding to geological disasters, automatically extracts crack information through the image recognition model, and can ensure the accuracy of the above image recognition model without the need for precise labeling of sample images, reduces labor intensity and human errors, and improves the efficiency and accuracy of crack analysis. Combined with the first deformation time series information and the second deformation time series information, through the mutual cooperation of the above technical solutions, it is not only possible to predict the future deformation development trend and deformation type of the slope, providing a scientific basis for preventing and mitigating geological disasters, but also reduces dependence on ground sensors and manual inspections, reduces monitoring costs, and improves monitoring efficiency, especially in remote or inaccessible areas. The present invention not only focuses on the deformation of cracks, but also combines displacement changes and morphological changes. Information, through data fusion technology, provides a more comprehensive deformation analysis. The method of the present invention is suitable for slope deformation monitoring under various environmental conditions, is not limited by natural conditions such as weather and terrain, and has strong environmental adaptability.

[0039] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the above-described systems, systems and units can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0040] If the integrated unit is implemented in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product. The computer software product is stored in a storage medium, including several instructions to enable a computer device (which can be a personal computer, server, or network device, etc.) to execute all or part of the steps of the method described in each embodiment of the present application. The aforementioned storage medium includes: U disk, mobile hard disk, read-only memory (ROM), random access memory (RAM), disk or optical disk and other media that can store program codes.

[0041] As described above, the above embodiments are only used to illustrate the technical solutions of the present application, rather than to limit them. Although the present application has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A time-series monitoring method for slope deformation of ordinary highways based on LT-1 satellite, characterized in that: The method: Acquire a historical image of the target area, preprocess the historical image, select a fixed object outside the target area as a reference object, use the reference object as the origin of a coordinate system, and construct a three-dimensional model of the target area based on the historical image and the origin of the coordinate system; Input each satellite image in the time-series images acquired within a preset time period into an image recognition model to obtain a recognition image corresponding to the satellite image, extract a crack region in the target region based on each recognition image, and obtain crack information corresponding to the crack region in each satellite image, obtain crack changes through all the crack information corresponding to the time-series images, and obtain first deformation time-series information according to the crack changes; updating each of the satellite images in the time series images to the three-dimensional model in time sequence, and obtaining the position information and morphological information of the slope in the target area in each of the satellite images, obtaining the displacement change information and morphological change information of the target area within the preset time period according to the position information and morphological information of the slope corresponding to each of the satellite images, and obtaining the second deformation time series information based on the displacement change information and the morphological change information; A slope deformation prediction result is obtained according to the first deformation time series information, the second deformation time series information and a deformation prediction model.

2. The method according to claim 1, characterized in that: The acquisition of the crack information includes: Preprocess each satellite image in the time series images acquired within the preset time period and input the preprocessed satellite image into the image recognition model to acquire the recognition image corresponding to each satellite image; wherein the recognition image is a binary image, and is an image that distinguishes the crack area and the slope area in the corresponding target area by pixel values, and the satellite image is a three-dimensional image; A crack area is extracted based on the crack probability value corresponding to the crack in the identified image, and each of the crack areas is divided into a plurality of sub-segments according to a set distance in the vertical direction, and the first tangent and the second tangent of both sides of each of the sub-segments are obtained, and the first angle and the second angle between each adjacent sub-segment and the first tangent and the second tangent on the same side of the sub-segment are calculated, wherein the first angle and the second angle are the angles on one side inside the crack area, and when at least one of the first angle and the second angle corresponding to at least one of the adjacent sub-segments and the sub-segment is less than a set angle, the sub-segment is regarded as a crack segment, and when at least one of the adjacent sub-segments and the sub-segment corresponding to ... When the first angle and the second angle are both greater than or equal to the set angle, at least one of the adjacent sub-segments is used as the target sub-segment, and the target sub-segment and the sub-segment are merged into a new sub-segment, and the two sides of the first angle are connected with the two intersection points of the corresponding sub-segment and used as the third tangent of the new sub-segment. Similarly, the fourth tangent on the other side of the new sub-segment is obtained, and this step is repeated to obtain multiple crack segments in the crack area, wherein the crack angle between adjacent crack segments is greater than or equal to the set angle, and the crack area in the identified image, the crack segment corresponding to each crack area, and the size information and position information of each crack segment are used as the crack information.

3. The method according to claim 1, characterized in that The acquiring of the first deformation time series information of the target area includes: The crack segment in the crack information corresponding to each satellite image in the time series image is compared with the crack information of the crack segment at the same position in the crack information corresponding to an adjacent satellite image in the time series in chronological order to obtain crack change information, wherein the acquisition time of the adjacent satellite image is later than the acquisition time of the satellite image, and the first deformation time series information of the target area within the preset time period corresponding to the time series image is obtained based on the crack change information corresponding to each satellite image and the adjacent satellite image, wherein the crack change information includes size change and depth change.

4. The method according to claim 1, characterized in that The training of the image recognition model includes: The image recognition model is created by a model creation unit, and a sample image is read from a storage unit, wherein the sample image includes first label information, and the first label information is a pre-labeled label. The sample image and the first label information are used as first training data to train the image recognition model, and a moving weight mean of a prediction result corresponding to each pixel in the sample image is periodically calculated during a training iteration process, and the prediction result obtained in this iteration is replaced by the moving weight mean, and second label information is set for a second pixel whose prediction result is greater than a set threshold, wherein the second pixel is a pixel not labeled with the pre-labeled label, and a pixel including the pre-labeled label is used as a first pixel; Acquire the second pixel adjacent to at least two of the first pixels, and use at least two of the first pixels as reference pixels, wherein the first label information corresponding to the reference pixels is the same and is located on both sides of the second pixel, compare the second label information corresponding to the second pixel with the first label information corresponding to the reference pixel, and when the second label information is the same as the first label information corresponding to the reference pixel, add the second pixel and the corresponding second label information as the first learning sample pixel to the second training data, otherwise, re-extract the difference feature between the second pixel and the reference pixel and enhance it, and re-train the image recognition model through the second training data.

5. The method according to claim 1, characterized in that The acquisition of the second deformation time series information includes: Each pre-processed satellite image in the time series images acquired within the preset time period is sequentially updated to the three-dimensional model, and the position information and morphological information of the slope in the satellite images adjacent in time are compared to obtain displacement change information and slope morphological information change, and the displacement change information and the slope morphological change information are used as the second deformation time series information.

6. The method according to claim 1, characterized in that Obtaining the slope deformation prediction result includes: The first deformation time series information and the second deformation time series information are fused in time series to obtain slope deformation time series information, and the slope deformation time series information is input into a deformation prediction model to obtain the slope deformation prediction result, which includes the deformation development trend and slope deformation type of the target area in the future.

7. The method according to claim 1, characterized in that The deformation prediction model is a model trained with historical deformation data, wherein the historical deformation data includes displacement deformation, slope morphology change, crack deformation and corresponding deformation types, and also includes the entire development process of the displacement deformation, slope morphology change and crack deformation during the entire deformation process of the deformation type.

8. The method according to claim 6, characterized in that The deformation development trend of the target area includes the displacement change trend, morphology change trend and crack change trend of the slope over time.

9. A time-series monitoring system for slope deformation of ordinary highways based on TL-1, the system being used to implement the method according to any one of claims 1 to 8, characterized in that: The system comprises: A model creation unit, configured to obtain a historical image of a target area, preprocess the historical image, select a fixed object outside the target area as a reference object, and use the reference object as the origin of a coordinate system, and construct a three-dimensional model of the target area based on the historical image and the origin of the coordinate system; The calculation unit is configured to: input each satellite image in the time-series images acquired within the preset time period into an image recognition model to acquire an identification image corresponding to the satellite image, extract a crack area in the target area based on each identification image, and acquire crack information corresponding to the crack area in each satellite image, acquire crack changes through all the crack information corresponding to the time-series images, and acquire first deformation time-series information according to the crack changes; update each satellite image in the time-series images to the three-dimensional model in chronological order, and acquire position information and morphological information of the slope in the target area in each satellite image, acquire displacement change information and morphological change information of the target area within the preset time period according to the position information and morphological information of the slope corresponding to each satellite image, and acquire second deformation time-series information based on the displacement change information and the morphological change information; A prediction unit is used to obtain a slope deformation prediction result according to the first deformation time series information, the second deformation time series information and a deformation prediction model.

10. A computer storage medium, characterized in that: The storage medium stores program instructions, wherein when the program instructions are executed, the device where the storage medium is located is controlled to execute the method according to any one of claims 1 to 8.

Images