Slope displacement monitoring system and method combined with GNSS (Global Navigation Satellite System) and binocular vision

Abstract

The invention provides a GNSS and binocular vision combined slope displacement monitoring system. The GNSS and binocular vision combined slope displacement monitoring system comprises an artificial marker, a binocular camera, a microcomputer, a GNSS positioning system, a solar cell and a data center, the artificial marker is arranged on a target slope, the artificial marker is used for providing mark points for the binocular camera, the binocular camera is used for carrying out image acquisition on the target slope, the GNSS positioning system is used for obtaining absolute coordinates of the slope displacement monitoring system in real time, and the microcomputer is used for controlling, storing and transmitting the slope displacement monitoring system. The solar cell is used for providing electric energy for the slope displacement monitoring system, and the data center is used for processing images transmitted by the microcomputer. The invention further provides a side slope displacement monitoring method combining GNSS and binocular vision, and the side slope displacement monitoring system is adopted. The method is simple in structure, clear in algorithm flow, low in construction cost and beneficial to large-scale popularization and application.

Description

Technical Field

[0001] The present invention belongs to the field of slope monitoring, surveying and mapping remote sensing technology, and in particular relates to a slope displacement monitoring system and method combining GNSS and binocular vision. Background Art

[0002] The current slope displacement monitoring technologies mainly include GNSS displacement monitoring technology, monocular vision displacement monitoring technology, binocular vision displacement monitoring technology, etc. These methods have their own advantages in the field of slope monitoring, but they also have corresponding technical limitations.

[0003] GNSS positioning technology has a wide range of applications in slope displacement, ground subsidence, and bridge deformation monitoring. It can accurately obtain the three-dimensional coordinates of the antenna in a specific coordinate system by receiving continuous signals from GNSS satellites and performing differential calculations. It reflects the movement of the target through changes in the target coordinates, and can achieve centimeter-level or even millimeter-level deformation monitoring. Patent document CN202310115118.8 discloses a GNSS-based slope displacement monitoring system, which combines a satellite positioning module with a rainwater monitoring module, a groundwater monitoring module, a stress monitoring module, and a crack monitoring module, and monitors slope displacement, rainwater, and some other data at the same time to achieve real-time early warning. This method can monitor the slope in real time, but it needs to be used in combination with other monitoring modules. The device installation cost is high, the monitoring station system is complex, and a large number of monitoring devices are required. Li Jingeng from Chang'an University proposed a slope displacement monitoring system based on embedded GNSS. It realizes differential positioning based on an embedded platform and can achieve rapid displacement monitoring. However, the monitoring system needs to be deployed at points of interest on the slope. For slopes with a larger range, more GNSS monitoring systems need to be deployed, and the deployment cost is relatively high. It is difficult to deploy in areas with complex terrain and landforms, and due to the multipath effect, satellite signals are easily blocked in complex terrain areas, which has a certain impact on the accuracy of displacement deformation monitoring.

[0004] Monocular vision technology can realize automated slope displacement monitoring. By taking slope images at regular intervals and analyzing and processing data in real time, continuous monitoring of slope displacement can be achieved. This monitoring technology first sets up several target objects on the slope, such as markers with specific colors, and forms a displacement array. The monocular vision system obtains the slope image, and analyzes and processes the image through image processing technology to identify all target objects in the displacement array, and characterizes the displacement change of the slope with the change in the displacement array. Dalian University of Technology has disclosed a real-time online monitoring system for three-dimensional displacement of slope surface based on monocular vision. The monocular vision monitoring platform is used to collect images of the measuring plate, and the relative displacement of the measuring plate is calculated by analyzing the image to achieve the purpose of monitoring slope deformation. However, monocular vision has the defect of being unable to measure the depth of field and the three-dimensional absolute displacement of the slope, which makes it difficult to warn of slopes with overall sliding or landslide disasters. Xiang Lu from Chongqing Jiaotong University proposed a research on slope displacement monitoring based on monocular vision, which combines a laser rangefinder with a monocular vision camera to perform three-dimensional measurement of the markers placed on the slope, thereby obtaining the three-dimensional coordinates of the markers and monitoring the slope displacement. Although the monocular vision system can perform three-dimensional measurement of the slope, it requires constraints such as laser ranging, and the measurement error in the Z direction is large.

[0005] Binocular vision is a branch of machine vision. Simply put, it uses two cameras to simultaneously capture images of the same scene, and extracts image information through image processing technology to achieve three-dimensional perception of objects in the scene. Compared with a monocular camera, binocular vision can obtain more information such as depth of field, and is thus widely used in slope displacement, ground subsidence, and bridge deformation monitoring. Patent document CN202110597490.8 discloses a binocular vision automated slope displacement monitoring method, which uses a binocular camera to monitor the slope in real time, measures the slope in the image through machine vision technology, and calculates the displacement of the slope. The three-dimensional coordinates obtained by the binocular vision monitoring scheme are calculated using the camera coordinate system of the left camera as the world coordinate system. If the monitoring system and the slope surface slide as a whole, it is difficult for the system to monitor the deformation displacement.

[0006] In short, GNSS displacement monitoring technology needs to be deployed at the points of interest on the slope. It is difficult to deploy in areas with complex terrain and landforms. Due to the influence of multipath effects, the positioning accuracy is affected, which will also have a corresponding impact on the accuracy of displacement monitoring. Although monocular vision monitoring technology has certain advantages in system complexity and deployment cost, due to the limitation of the monocular vision system in obtaining depth of field information, it can only calculate the relative displacement in the XY direction. If it is necessary to measure the displacement in the three-dimensional world coordinate system, other measurement systems are often required to assist in calculating the displacement change, which greatly hinders the development of monocular vision displacement monitoring technology. Compared with monocular vision monitoring technology, the advantage of binocular vision monitoring technology is that it can obtain the depth of field of the object in the photo without the help of other technical means, and reconstruct its three-dimensional coordinates from the target interest points in the photo. However, the reconstruction principle of the binocular vision system is complex, and the software and hardware requirements for the three-dimensional reconstruction system are high. The reconstruction accuracy is affected by the construction parameters and calibration effect of the binocular system. If the monitoring system and the slope surface slide as a whole, it is difficult for the system to monitor the deformation displacement.

[0007] Therefore, it is necessary to design a new slope displacement monitoring technology to carry out slope displacement monitoring. Summary of the invention

[0008] The purpose of the present invention is to provide a slope displacement monitoring system and method combining GNSS and binocular vision, so as to solve the corresponding technical limitations of the current slope displacement monitoring technology mentioned in the background technology.

[0009] To achieve the above-mentioned purpose, the present invention provides a slope displacement monitoring system combining GNSS and binocular vision, including artificial markers, a binocular camera, a microcomputer, a GNSS positioning system, a solar cell and a data center; the artificial markers are set on the target slope, the artificial markers are used to provide marking points for the binocular camera, the binocular camera is used to collect images of the target slope, the GNSS positioning system is used to obtain the absolute coordinates of the slope displacement monitoring system in real time, the microcomputer is used to control the slope displacement monitoring system, store and transmit images, the solar cell is used to provide power to the slope displacement monitoring system, and the data center is used to process images transmitted by the microcomputer.

[0010] In a specific embodiment, the slope displacement monitoring system combining GNSS and binocular vision also includes a bracket, a binocular camera, a microcomputer and a GNSS positioning system are all arranged on the bracket, and the binocular camera, the GNSS positioning system and the solar cell are all connected to the microcomputer; the artificial markers are arranged at the points of interest of the target slope, and a horizontal section line is arranged on the target slope at a horizontal distance of 30 to 50 m, and at least 3 points of interest are arranged on each section line.

[0011] The present invention also provides a slope displacement monitoring method combining GNSS and binocular vision, wherein the slope displacement monitoring method combining GNSS and binocular vision adopts the slope displacement monitoring system combining GNSS and binocular vision as described above, and comprises the following steps:

[0012] Step 1: Set up artificial markers on the target slope, and set up a binocular camera, a microcomputer, a GNSS positioning system and a solar cell in a non-deformed area at a certain distance from the target slope;

[0013] Step 2: The binocular camera takes pictures of the artificial landmarks and performs camera calibration and correction.

[0014] Step 3: The images of artificial marker points taken by the binocular camera are transmitted to the data center through a microcomputer;

[0015] Step 4: The data center identifies and locates the landmarks in the transmitted image and calculates the three-dimensional coordinates;

[0016] Step 5: Analyze the deformation and displacement of the slope through the time-series coordinate changes of the landmark points and formulate early warning rules.

[0017] In a specific implementation, in step 1, a binocular camera, a microcomputer, a GNSS positioning system and a solar cell are arranged in a non-deformation area 20 to 300 m away from the target slope.

[0018] In a specific implementation, in step 2, the Zhang Zhengyou calibration method is used to simultaneously perform monocular camera calibration, camera distortion correction, and binocular camera stereo calibration using a binocular camera; epipolar correction is performed on the binocular camera;

[0019] In step 3, the images captured by the binocular camera are transmitted to the data center via a 4G / 5G mobile communication network;

[0020] In the step 4, the landmark points are extracted by deep learning landmark detection, image preprocessing and corner point detection; the landmark points of the photos taken by the two cameras of the binocular camera are stereo matched; the three-dimensional coordinates of the landmark points are calculated according to the corresponding point coordinates of the landmark points in the photos taken by the two cameras;

[0021] In step 5, the GNSS positioning system is combined with the calculated three-dimensional coordinates of the marker point and calibrated to obtain the Gaussian projection coordinates of the marker point and its geodetic height.

[0022] In a specific implementation, the Bouguet epipolar correction method is used to perform epipolar correction on the binocular camera.

[0023] In a specific implementation, deep learning landmark detection includes: using a deep learning framework to construct a neural network model, constructing a training set and a validation set, and training the neural network model to obtain a landmark detection algorithm model.

[0024] In a specific implementation, the deep learning framework is a YOLOv5 neural network framework.

[0025] In a specific implementation, the image preprocessing includes: image grayscale and image binarization.

[0026] In a specific implementation, the corner point detection uses the Harris corner point detection algorithm to extract the center of the artificial marker point, and uses a sub-pixel corner point positioning method to obtain the precise image plane coordinates of the center of the artificial marker point.

[0027] Compared with the prior art, the present invention has the following beneficial effects:

[0028] The present invention monitors the displacement and deformation of the slope surface based on binocular vision technology, and can realize real-time monitoring of the displacement of the slope surface.

[0029] The present invention realizes non-contact monitoring of the slope surface and reduces safety hazards.

[0030] The slope displacement monitoring system combining GNSS and binocular vision of the present invention has a simple structure, a clear algorithm flow, and a low construction cost, which is conducive to large-scale promotion and application.

[0031] The present invention combines the GNSS satellite positioning system to monitor the absolute displacement of the slope, and can also detect and issue a timely warning when the overall displacement of the slope occurs.

[0032] The present invention uses a deep learning neural network model as a landmark-centric recognition tool in the artificial landmark recognition algorithm. The deep learning method can allow the computer to self-iterate and generate recognition capabilities through training, which saves the process of designing a landmark recognition algorithm, improves work efficiency, and improves the intelligence level of the algorithm.

[0033] The stereo matching algorithm proposed in the present invention is different from the traditional stereo matching method which performs point matching by calculating the similarity degree. It uses the epipolar constraint as the screening condition, which greatly improves the efficiency of stereo matching.

[0034] In addition to the above-described objects, features and advantages, the present invention has other objects, features and advantages. The present invention is further described in detail below. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] The drawings constituting a part of this application are used to provide a further understanding of the present invention. The exemplary embodiments of the present invention and their descriptions are used to explain the present invention and do not constitute an improper limitation of the present invention. In the drawings:

[0036] Figure 1 It is a flow chart of a slope displacement monitoring method combining GNSS and binocular vision of the present invention;

[0037] Figure 2 It is a schematic diagram of a slope displacement monitoring system combining GNSS and binocular vision according to the present invention;

[0038] Among them: 1. Microcomputer; 2. Binocular camera; 3. GNSS positioning system; 4. Artificial markers; 5. Solar cells. DETAILED DESCRIPTION

[0039] The embodiments of the present invention are described in detail below. The specific embodiments described herein are only used to explain the present invention and are not used to limit the present invention.

[0040] Example 1

[0041] The present invention provides a slope displacement monitoring system combining GNSS and binocular vision, including an artificial marker 4, a binocular camera 2, a microcomputer 1, a GNSS positioning system 3, a solar cell 5, a bracket and a data center; the artificial marker is set on the target slope, the artificial marker is used to provide a marker point for the binocular camera, the binocular camera is used to collect images of the target slope, the GNSS positioning system is used to obtain the absolute coordinates of the slope displacement monitoring system in real time, the microcomputer is used to control the slope displacement monitoring system, store and transmit images, the solar cell is used to provide power to the slope displacement monitoring system, and the data center is used to process the images transmitted by the microcomputer. The binocular camera, the microcomputer and the GNSS positioning system are all set on the bracket, and the binocular camera, the GNSS positioning system and the solar cell are all connected to the microcomputer; the artificial marker is set on the point of interest of the target slope, and a horizontal section line is set at a horizontal distance of 30 to 50 meters on the target slope, and at least 3 points of interest are set on each section line.

[0042] The present invention also provides a slope displacement monitoring method combining GNSS and binocular vision, wherein the slope displacement monitoring method combining GNSS and binocular vision adopts the slope displacement monitoring system combining GNSS and binocular vision as described above, and comprises the following steps:

[0043] Step 1: Set up artificial markers on the target slope, and set up a binocular camera, a microcomputer, a GNSS positioning system and a solar cell in a non-deformed area 20 to 300 meters away from the target slope;

[0044] Step 2: The binocular camera takes a photo of the checkerboard calibration plate to perform camera calibration and correction.

[0045] Step 3: The images of artificial marker points taken by the binocular camera are transmitted to the data center through a microcomputer;

[0046] Step 4: The data center identifies and locates the landmarks in the transmitted image and calculates the three-dimensional coordinates;

[0047] Step 5: Analyze the deformation and displacement of the slope through the time-series coordinate changes of the landmark points and formulate early warning rules.

[0048] The step 2 specifically includes:

[0049] The binocular camera is used to collect images of the chessboard calibration plate, and the Zhang Zhengyou calibration method is used to simultaneously perform monocular camera calibration, camera distortion correction, and binocular camera stereo calibration on the binocular camera;

[0050] The Bouguet epipolar correction method is used to correct the epipolar lines of the binocular camera;

[0051] The step 3 specifically includes:

[0052] The images captured by the binocular camera are transmitted to the data center via a microcomputer using a 4G / 5G mobile communication network;

[0053] The step 4 specifically includes:

[0054] The landmark points are extracted through deep learning landmark detection, image preprocessing and corner point detection; the landmark points of the photos taken by the two cameras of the binocular camera are stereo matched; the three-dimensional coordinates of the landmark points are calculated based on the corresponding point coordinates of the landmark points in the photos taken by the two cameras;

[0055] The step 5 specifically includes:

[0056] The GNSS positioning system is combined with the calculated three-dimensional coordinates of the landmark point and calibrated to obtain the Gaussian projection coordinates of the landmark point and its geodetic height.

[0057] Deep learning landmark detection includes: using the YOLOv5 neural network framework to build a neural network model, building a training set and a validation set, and training the neural network model to obtain a landmark detection algorithm model.

[0058] The image preprocessing includes: image grayscale and image binarization.

[0059] The corner point detection adopts Harris corner point detection algorithm to extract the center of the artificial marker point, and uses sub-pixel corner point positioning method to obtain the precise image plane coordinates of the center of the artificial marker point.

[0060] Zhang Zhengyou's calibration method uses a checkerboard calibration plate as a reference. Assuming that the checkerboard calibration plate is located on a specific plane in the world coordinate system, since the size of each checkerboard on the calibration plate is known, the physical coordinates of each corner point in the world coordinate system can be calculated. Then, by taking multiple sets of calibration plate images at different angles and using an image detection algorithm to obtain the coordinates of the corner points in the image. Substituting the pixel coordinates of multiple corner points and the physical coordinates of the corner points in the world coordinate system into the established mapping model, and using an optimization algorithm to calibrate the camera, the camera's internal and external parameter matrix and distortion coefficient can be obtained.

[0061] Since Zhang Zhengyou's calibration method is easy to operate, the reference object only needs a simple chessboard calibration plate, and high-precision calibration results can be obtained by taking multiple pictures. This method has been widely used in the field of visual measurement.

[0062] The YOLOv5 neural network framework includes an input layer (input), a backbone convolution layer (backbone), a multi-scale feature fusion network layer (Neck), and a prediction layer (Prediction). Among them, the input layer resamples the data and plans the image size to a fixed-size image vector as the input data of the neural network. The backbone convolution layer is a feature extraction layer. Through the CSP structure, the image features can be extracted layer by layer to provide more reliable features for subsequent feature fusion. The multi-scale feature fusion network layer fuses the extracted features of different scales to enhance the multi-scale feature recognition ability of the neural network. The prediction layer outputs the recognition probability of the fused large, medium and small scale feature images through an activation function to achieve the function of identifying the target object.

[0063] In this embodiment, a self-built data set is used as a training set and a verification set of the neural network. For the artificial markers used in this embodiment, 3,000 photos taken by mobile phones, cameras, and video cameras in different light and dark environments, with grassland, trees, wasteland, and slope backgrounds, are used as a marker recognition data set, of which 2,500 are used as a training set and 500 are used as a verification set. The trained neural network model is used as a marker detection algorithm model.

[0064] The YOLOv5 neural network model can extract the range of artificial markers in the image for subsequent use by the algorithm. In order to better extract the center of the artificial marker and enable the algorithm to extract the center under various lighting conditions, a series of image preprocessing is required for the marker image.

[0065] This embodiment first performs image preprocessing by graying the image. Since the captured image is easily affected by light, graying the image can improve the contrast of the artificial marker image and the accuracy of the marker center coordinate extraction.

[0066] The present embodiment also adopts the method of image binarization to carry out image preprocessing. The present embodiment adopts Otsu method (Otsu method or maximum inter-class variance method) to carry out the calculation of image binarization. Otsu algorithm is considered to be the best algorithm for threshold selection in image segmentation, and the calculation is simple and is not affected by image brightness and contrast. Otsu algorithm uses the idea of ​​clustering, and the grayscale number of the image is divided into two parts according to the grayscale level, so that the grayscale value difference between the two parts is the largest, and the grayscale difference between each part is the smallest, and a suitable grayscale level is found to divide by calculating the variance.

[0067] The artificial marker point used in this embodiment is a black and white chessboard point, where the center of the marker point is the corner point of the black and white area. This embodiment uses the Harris corner point detection algorithm to extract the center of the artificial marker point, and uses the sub-pixel corner point positioning method to obtain the precise image plane coordinates of the center of the artificial marker point.

[0068] The basic principle of Harris corner detection is to use a local window to move on the image to determine the grayscale changes to detect corners. When a window moves on the image and the window does not change in all directions, the window area is considered to be a smooth area; if the window does not change in one direction but changes significantly in another direction, then this area may have an edge; if the window has a large change in grayscale in all directions, then this area may have a corner. Harris corner detection uses this intuitive physical phenomenon to determine whether it is a corner by the degree of change of the window in all directions.

[0069] The basic principle of sub-pixel corner location is that the vector from the sub-pixel corner point to the surrounding pixels should be perpendicular to the grayscale gradient of the image, and the coordinate value with sub-pixel accuracy is obtained by an iterative method that minimizes the error function.

[0070] This embodiment performs stereo matching on the landmarks of the photos taken by the two cameras of the binocular camera. When the binocular vision system shoots a common object, for a landmark on the left photo, there must be a landmark position corresponding to it on the right photo, and the landmark of the right photo must be on the epipolar line. Therefore, for the landmark of the left photo, as long as a number of pending landmarks are extracted on the right photo, and then these landmarks are screened, the right landmark whose distance to the epipolar line is less than a threshold and the closest distance and the smallest parallax is selected, can be used as the landmark of the right photo matched by the landmark of the left photo.

[0071] In this embodiment, the binocular vision system is subjected to Bouguet epipolar correction, so that the phase planes on the left and right sides are coplanar and aligned, and the baseline formed by the optical center of the camera is parallel to the phase plane. If the left mark point is known, the corresponding point in the right photo must be on a specific straight line. It is only necessary to extract the mark points in the right photo, and select the mark points whose distance to the specific straight line is less than a certain threshold and the closest distance, and the smallest parallax, which are the matching mark points.

[0072] The three-dimensional measurement of the marker point in this embodiment. After the above technical steps, the coordinates of the corresponding points of the marker point in the left and right photos can be obtained. According to the internal and external orientation elements obtained by calibration, the three-dimensional coordinates of the marker point can be calculated by formula. At this time, the three-dimensional coordinates are the coordinates of the camera coordinate system with the optical center on the left side of the camera as the origin.

[0073] This embodiment combines the GNSS satellite positioning receiving system with the binocular vision monitoring system, and calibrates them to obtain the conversion parameters of the left camera coordinate system relative to the CGCS2000 national geodetic coordinate system. According to the coordinate conversion principle, since there is no scale transformation between the two coordinate systems, the landmark points in the left camera coordinate system can be converted into landmark points in the Earth-centered Earth-fixed coordinate system, and then converted into Gaussian projection coordinates and their geodetic heights.

[0074] Finally, the deformation and displacement of the slope are analyzed through the time-series coordinate changes of the landmark points, and early warning rules are formulated.

[0075] The above contents are further detailed descriptions of the present invention in combination with specific preferred embodiments, and it cannot be determined that the specific implementation of the present invention is limited to these descriptions. For ordinary technicians in the technical field to which the present invention belongs, several simple deductions and substitutions can be made without departing from the concept of the present invention, which should be regarded as belonging to the protection scope of the present invention.

Claims

1. A slope displacement monitoring system combining GNSS and binocular vision, characterized in that: The invention comprises an artificial marker (4), a binocular camera (2), a microcomputer (1), a GNSS positioning system (3), a solar cell (5) and a data center; the artificial marker (4) is arranged on a target slope, the artificial marker (4) is used to provide a marker point for the binocular camera, the binocular camera (2) is used to collect images of the target slope, the GNSS positioning system (3) is used to obtain the absolute coordinates of the slope displacement monitoring system in real time, the microcomputer (1) is used to control the slope displacement monitoring system and store and transmit images, the solar cell (5) is used to provide electric energy to the slope displacement monitoring system, and the data center is used to process images transmitted by the microcomputer.

2. The slope displacement monitoring system combining GNSS and binocular vision according to claim 1 is characterized in that: The slope displacement monitoring system combining GNSS and binocular vision also includes a bracket, a binocular camera, a microcomputer and a GNSS positioning system are all arranged on the bracket, and the binocular camera, the GNSS positioning system and the solar cell are all connected to the microcomputer; the artificial markers are arranged on the points of interest of the target slope, and a horizontal section line is arranged on the target slope at a horizontal distance of 30 to 50 meters, and at least 3 points of interest are arranged on each section line.

3. A slope displacement monitoring method combining GNSS and binocular vision, characterized in that: The slope displacement monitoring method combining GNSS and binocular vision adopts the slope displacement monitoring system combining GNSS and binocular vision as claimed in any one of claims 1 to 2, comprising the following steps: Step 1: Set up artificial markers on the target slope, and set up a binocular camera, a microcomputer, a GNSS positioning system and a solar cell in a non-deformed area at a certain distance from the target slope; Step 2: The binocular camera takes a picture of the artificial marker and performs camera calibration and correction; Step 3: The images of artificial marker points taken by the binocular camera are transmitted to the data center through a microcomputer; Step 4: The data center identifies and locates the landmarks in the transmitted image and calculates the three-dimensional coordinates; Step 5: Analyze the deformation and displacement of the slope through the time-series coordinate changes of the landmark points and formulate early warning rules.

4. The slope displacement monitoring method combining GNSS and binocular vision according to claim 3 is characterized in that: In the step 1, a binocular camera, a microcomputer, a GNSS positioning system and a solar cell are arranged in a non-deformed area 20 to 300 m away from the target slope.

5. The slope displacement monitoring method combining GNSS and binocular vision according to claim 3 is characterized in that: In the step 2, the Zhang Zhengyou calibration method is used to simultaneously perform monocular camera calibration, camera distortion correction, and stereo calibration on the binocular camera; and epipolar correction is performed on the binocular camera; In step 3, the images captured by the binocular camera are transmitted to the data center via a 4G / 5G mobile communication network; In the step 4, the landmark points are extracted by deep learning landmark detection, image preprocessing and corner point detection; the landmark points of the photos taken by the two cameras of the binocular camera are stereo matched; the three-dimensional coordinates of the landmark points are calculated according to the corresponding point coordinates of the landmark points in the photos taken by the two cameras; In step 5, the GNSS positioning system is combined with the calculated three-dimensional coordinates of the marker point, and coordinate transformation is performed to obtain the Gaussian projection coordinates of the marker point and its geodetic height.

6. The slope displacement monitoring method combining GNSS and binocular vision according to claim 5 is characterized in that: The Bouguet epipolar correction method is used to perform epipolar correction on the binocular camera.

7. The slope displacement monitoring method combining GNSS and binocular vision according to claim 5 is characterized in that: Deep learning landmark detection includes: using a deep learning framework to build a neural network model, building a training set and a validation set, and training the neural network model to obtain a landmark detection algorithm model.

8. The slope displacement monitoring method combining GNSS and binocular vision according to claim 7 is characterized in that: The deep learning framework is the YOLOv5 neural network framework.

9. The slope displacement monitoring method combining GNSS and binocular vision according to claim 5 is characterized in that: The image preprocessing includes: image grayscale and image binarization.

10. The slope displacement monitoring method combining GNSS and binocular vision according to claim 5, characterized in that: The corner point detection adopts Harris corner point detection algorithm to extract the center of the artificial marker point, and uses sub-pixel corner point positioning method to obtain the precise image plane coordinates of the center of the artificial marker point.

Images