Remote island reference-free control network autonomous construction and transmission method

Abstract

The invention discloses a remote island reference-free control network autonomous construction and transmission method, which belongs to the technical field of geodetic survey and surveying and mapping engineering, and comprises the following steps: receiving sea surface laser ranging, land gravity, mobile platform images and accelerometer data; performing statistics on peak values of the laser ranging histograms to extract local average sea surface elevation; a relative geometric point cloud is generated based on image texture tracking in combination with an accelerometer calculation scale; calculating a vertical line inclination angle by using a difference value between the gravity data and an earth normal gravity field; and correcting a point cloud vertical component by using a difference value between a water line and a sea surface elevation, correcting a plane component by using an inclination angle, and outputting a control point. According to the invention, a mode of fusing sea surface laser ranging, mobile platform images, inertial data and land gravity data is adopted, and a local three-dimensional control network with unified scale, elevation reference and vertical line direction can be autonomously constructed under the condition of no external reference dependence.

Description

Technical Field

[0001] This invention relates to the field of geodesy and surveying engineering technology, and in particular to a method for the autonomous construction and transmission of a baseline control network for remote islands. Background Technology

[0002] Island and reef surveying is a fundamental task for marine resource development, engineering construction, and the protection of maritime rights. In the process of acquiring geographic information for islands and reefs, establishing a high-precision spatial control network is a prerequisite for accurate digital reconstruction of the island and reef topography. For remote sea areas far from the mainland, due to the lack of known measurement control points, obtaining reliable three-dimensional spatial benchmark data has always been a key focus in the surveying and mapping field.

[0003] In related technologies, Chinese invention patent application CN115326011A discloses a method for transferring near-island and reef elevations based on neural networks. The method includes: firstly, conducting GNSS joint measurements of control points on the land and islands and reefs, and conducting elevation joint measurements of the land control points to obtain the GNSS / leveling elevation anomaly on the land; then, using the EGM2008 gravity field model to obtain the residual elevation anomaly; constructing a BP neural network; using the residual elevation anomaly to train the neural network; and using the trained neural network to obtain the residual elevation anomaly of the islands and reefs. Finally, using the GNSS geodetic height results from the joint measurements of the islands and reefs, the EGM2008 gravity field model, and the residual elevation anomaly of the islands and reefs, the elevation of the island and reef control points is obtained.

[0004] Regarding the aforementioned technologies, the inventors believe that this method is primarily applicable to "nearshore" areas relatively close to the mainland, and its core logic relies on high-precision land control point data as training samples for the neural network. For remote islands far from the mainland, due to the large spatial span, the correlation of gravitational field characteristics between the island / reef and the land weakens, making it difficult to guarantee the prediction accuracy of the neural network model. Furthermore, this method is essentially a benchmark "transfer" technology, which must rely on existing external land benchmarks, and cannot meet the requirements of autonomously constructing the overall three-dimensional geometry of the island / reef and jointly correcting it in the planar and vertical directions under conditions without any external benchmark data support. Summary of the Invention

[0005] To address the aforementioned issues, this invention provides a method for the autonomous construction and transfer of a local three-dimensional control network without a reference datum on remote islands. By integrating sea surface laser ranging, mobile platform imagery and inertial data, and land gravity data, it is possible to autonomously construct a local three-dimensional control network with a unified scale, elevation datum, and vertical direction without relying on external references.

[0006] The above objectives can be achieved through the following approach:

[0007] A method for autonomously constructing and transmitting a baseline-free control network for remote islands includes receiving a sea surface laser ranging signal stream, a land gravimeter reading sequence, serialized image data from a mobile platform, and accelerometer readings; statistically analyzing a distance distribution histogram through a continuous observation window based on the sea surface laser ranging signal stream, and extracting the peak distance data from the histogram as the local average sea level elevation; tracking corresponding texture pixels based on the serialized image data from the mobile platform, and calculating the spatial displacement scale of the pixels in conjunction with the accelerometer readings to generate a relative geometric point cloud of the islands and reefs; comparing the difference between the land gravimeter reading sequence and the Earth's normal gravity field data to calculate the tilt angle values ​​of the gravity vertical in the north-south and east-west directions; calculating the average height difference between the waterline points in the relative geometric point cloud of the islands and reefs and the local average sea level elevation, using the average height difference to correct the vertical coordinate components of the relative geometric point cloud of the islands and reefs, and using the tilt angle values ​​to correct the planar coordinate components of the relative geometric point cloud of the islands and reefs, and outputting a set of control point coordinates.

[0008] Optionally, receiving the sea surface laser ranging signal stream, the land gravimeter reading sequence, the mobile platform serialized image data, and the accelerometer readings includes: establishing a first communication channel with the mobile platform by calling the wireless gateway interface, and establishing a second communication channel with the shore-based equipment by calling the multi-channel serial communication interface; reading binary bit streams through the first communication channel and the second communication channel respectively, and mapping the binary bit streams to the mobile platform serialized image data, accelerometer readings, the sea surface laser ranging signal stream, and the land gravimeter reading sequence according to the device identifier.

[0009] Optionally, the step of using the local average sea surface elevation includes: performing discretization binning on the sea surface laser ranging signal stream through a continuous observation window, counting the number of sampling points falling into each bin, and constructing a discrete histogram vector; constructing a Gaussian probability density function, performing nonlinear least squares fitting on the Gaussian probability density function using the discrete histogram vector, and solving for the optimal position parameters; calculating the expected peak value of the distance distribution histogram using the optimal position parameters, and outputting the local average sea surface elevation.

[0010] Optionally, the calculation of the optimal position parameters includes: establishing a residual vector using the discrete histogram vector and the Gaussian probability density function, and constructing a sum-of-squares objective function based on the residual vector; performing partial derivative calculations on the sum-of-squares objective function with respect to the position parameters to generate a gradient descent direction vector, and performing parameter iterative update operations along the gradient descent direction vector; monitoring the L2 value using the residual vector, stopping the iteration when the rate of change of the L2 value converges and stabilizes, and outputting the current position parameters as the optimal position parameters.

[0011] Optionally, generating the relative geometric point cloud of islands and reefs includes: extracting common texture points from two adjacent frames of images based on the serialized image data of the mobile platform, and calculating the pixel coordinate position difference using the common texture points; performing a double integration operation in the time dimension based on the accelerometer readings to calculate the physical movement distance of the mobile platform within the shooting interval of the two adjacent frames; using the physical movement distance as a reference length, and combining it with the pixel coordinate position difference to perform geometric calculations to determine the three-dimensional spatial coordinates of the common texture points, and aggregating to generate the relative geometric point cloud of islands and reefs.

[0012] Optionally, the method further includes: calculating a numerical ratio using the difference between the pixel coordinate positions and the physical movement distance, and mapping the numerical ratio to a distribution histogram; performing mode extraction on the distribution histogram to perform cross-validation of multi-source data, and calculating the scale reliability coefficient.

[0013] Optionally, the calculation of the tilt angle values ​​of the gravity vertical line in the north-south and east-west directions includes: extracting the geographic coordinates of each observation point based on the land gravimeter reading sequence; calculating the normal gravity value of the Earth ellipsoid using the geographic coordinates; calculating the numerical difference between the land gravimeter reading sequence and the normal gravity value of the Earth ellipsoid to generate discrete gravity anomaly data; performing spatial gridding interpolation on the discrete gravity anomaly data to cover the island and reef areas to generate a gravity anomaly grid matrix; traversing the gravity anomaly grid matrix, establishing concentric annular zones with grid nodes as the calculation center, calculating the spatial distance from the remaining grid nodes in the gravity anomaly grid matrix to the calculation center, and generating a distance inverse weighting factor using the spatial distance; performing convolution summation based on the gravity anomaly grid matrix and the distance inverse weighting factor to output the tilt angle values ​​of the gravity vertical line in the north-south and east-west directions.

[0014] Optionally, the output control point coordinate set includes: performing multiplicative scaling on all coordinate values ​​of the island-reef relative geometric point cloud based on the scale reliability coefficient; calculating the numerical difference between the average height of the waterline points in the island-reef relative geometric point cloud and the local mean sea level elevation, superimposing the numerical difference onto the vertical coordinate component of the island-reef relative geometric point cloud to generate a corrected vertical coordinate component; multiplying the corrected vertical coordinate component by the tilt angle value to calculate the horizontal offset value of the island-reef relative geometric point cloud in the planar direction, and compensating the horizontal offset value to the planar coordinate component of the island-reef relative geometric point cloud, and outputting the control point coordinate set.

[0015] Optionally, calculating the numerical difference between the average height of the waterline points in the relative geometric point cloud of the island / reef and the local mean sea level includes: sorting the vertical coordinate components in ascending order and extracting the set of coordinate values ​​less than or equal to the lower quartile; calculating the arithmetic mean of the set of coordinate values ​​as the average height of the waterline points, and calculating the scalar difference between the average height of the waterline points and the local mean sea level as the numerical difference.

[0016] Based on the same inventive concept, this invention also provides an autonomous construction and transmission system for a remote island control network without a reference point. The system includes: a multi-source data receiving module for receiving sea surface laser ranging signal streams, land gravimeter reading sequences, serialized image data from a mobile platform, and accelerometer readings; a sea surface elevation extraction module for statistically analyzing a distance distribution histogram through a continuous observation window based on the sea surface laser ranging signal stream, and extracting the peak distance data from the distance distribution histogram as the local average sea surface elevation; and a geometric point cloud generation module for tracing corresponding texture pixels based on the serialized image data from the mobile platform, combined with the accelerometer readings. The gravimeter readings are used to calculate the spatial displacement scale of pixels and generate a relative geometric point cloud of islands and reefs. The vertical tilt angle calculation module is used to compare the difference between the land gravimeter reading sequence and the Earth's normal gravity field data to calculate the tilt angle values ​​of the gravity vertical line in the north-south and east-west directions. The control point output module is used to calculate the average height difference between the waterline points in the relative geometric point cloud of islands and reefs and the local mean sea level elevation, use the average height difference to correct the vertical coordinate components of the relative geometric point cloud of islands and reefs, and use the tilt angle values ​​to correct the planar coordinate components of the relative geometric point cloud of islands and reefs, and output the control point coordinate set.

[0017] Compared with the prior art, the present invention has the following advantages:

[0018] 1. By integrating sensor data from multiple sensors based on different physical principles, it achieves fully autonomous positioning and mapping in an environment without external reference, freeing it from dependence on global navigation satellite system signals or preset ground control points, and greatly expanding the applicable scope of high-precision surveying and mapping operations, especially in remote areas where signals are limited or inaccessible.

[0019] 2. Multi-dimensional self-calibration of the coordinate frame was achieved through joint calculation. A robust local elevation benchmark was established using sea surface laser ranging data. The true scale of the geometric model was determined by fusing visual and inertial data. The vertical direction deviation was corrected using gravity measurement data. This solved the three major uncertainties of scale, elevation and direction in independent mapping, and ensured the physical authenticity and internal consistency of the final control point coordinates.

[0020] 3. A complete automated processing flow from raw data acquisition to final coordinate output is proposed, integrating the complex geophysical and surveying calculation processes to improve the efficiency of basic surveying data acquisition in remote island areas. This method features a clear operation process, a high degree of automation, and reduces the requirements for complex on-site operations by professional surveyors, demonstrating strong engineering practicality and scalability.

[0021] Other features and advantages of the invention will be set forth in the description which follows, and will be apparent in part from the description, or may be learned by practicing the invention. The objects and other advantages of the invention may be realized and obtained by means of the structures pointed out in the description, claims and drawings. Attached Figure Description

[0022] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0023] Figure 1 This is a flowchart illustrating the method for autonomous construction and transmission of a baseline-free control network for remote islands according to an embodiment of the present invention.

[0024] Figure 2 This is a graph showing the probability density distribution and fitting results of the sea surface laser ranging signal flow according to an embodiment of the present invention.

[0025] Figure 3 This is a time-series scatter plot and probability density distribution of the visual inertial scale ratio according to an embodiment of the present invention.

[0026] Figure 4 This is a schematic diagram of the structure of the autonomous construction and transmission system for a remote island control network without a reference point, according to an embodiment of the present invention. Detailed Implementation

[0027] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0028] Reference Figure 1One embodiment of the present invention proposes a method for autonomous construction and transfer of a reference-free control network for remote islands. By fusing sea surface laser ranging, mobile platform imagery and inertial data, and land gravity data, a local three-dimensional control network with a unified scale, elevation reference, and vertical direction can be autonomously constructed without external reference dependence.

[0029] The method described in this embodiment specifically includes:

[0030] Receives sea surface laser ranging signal streams, land gravimeter reading sequences, serialized image data from mobile platforms, and accelerometer readings;

[0031] Based on the distance distribution histogram of the sea surface laser ranging signal stream through a continuous observation window, the peak distance data of the distance distribution histogram is extracted as the local average sea surface elevation.

[0032] Based on the serialized image data of the mobile platform, trace the texture pixels with the same name, and combine the accelerometer readings to calculate the spatial displacement scale of the pixels to generate a relative geometric point cloud of islands and reefs.

[0033] The land gravimeter reading sequence is compared with the Earth's normal gravity field data to calculate the tilt angle of the gravity vertical in the north-south and east-west directions.

[0034] The average height difference between the waterline points in the relative geometric point cloud of the islands and reefs and the local average sea level is calculated. The average height difference is used to correct the vertical coordinate components of the relative geometric point cloud of the islands and reefs, and the tilt angle value is used to correct the planar coordinate components of the relative geometric point cloud of the islands and reefs, and the control point coordinate set is output.

[0035] Optionally, the receiving of the sea surface laser ranging signal stream, the land gravimeter reading sequence, the mobile platform serialized image data, and the accelerometer readings includes:

[0036] The system calls the wireless gateway interface to establish a first communication channel with the mobile platform, and calls the multi-channel serial communication interface to establish a second communication channel with the shore-based device.

[0037] The industrial control computer first performs hardware interface initialization. For mobile platforms, such as drones or unmanned surface vessels (USVs), due to their high-speed mobility and location on the sea or in the air, it calls the underlying wireless gateway interface to handshake with the mobile platform's onboard communication module, establishing a first communication channel. This first communication channel is typically configured as a high-bandwidth, low-latency wireless link, such as Wi-Fi 5.8GHz or 5G, to meet the high throughput transmission requirements of image data. For shore-based devices, due to their fixed location and relatively small data volume but extremely high stability requirements, a multi-channel serial communication interface is used to establish physical connections with each shore-based device, forming a second communication channel. When establishing a connection, communication parameters such as baud rate, data bits, stop bits, and parity bits need to be configured to ensure the synchronization of physical layer signal transmission.

[0038] For example, in a construction task targeting an uninhabited island reef in the South China Sea, the mobile platform is a quadcopter drone equipped with a monocular camera and a MEMS inertial measurement unit. It connects to a high-power omnidirectional antenna gateway via a gigabit Ethernet port, and establishes a first communication channel with the drone by calling the Socket API function connect("192.168.1.100", 8080), using the UDP protocol to reduce video stream transmission latency. Simultaneously, shore-based equipment includes a high-frequency laser rangefinder and a relative gravimeter, which are connected to the industrial serial port card of the data acquisition and control unit via shielded twisted-pair cables. The serial port driver is used to open COM1 (for the laser rangefinder) with a baud rate set to 115200bps, and COM2 (for the gravimeter) with a baud rate set to 9600bps, thereby establishing a stable second communication channel to ensure continuous data transmission even in rough seas.

[0039] The binary bit stream is read through the first communication channel and the second communication channel respectively, and the binary bit stream is split and mapped into mobile platform serialized image data, accelerometer readings, sea surface laser ranging signal stream and land gravimeter reading sequence according to the device identifier.

[0040] After the channel is established, an asynchronous listening thread is started to continuously and in parallel read the raw binary bitstream through the first and second communication channels. These bitstreams are unprocessed byte sequences containing time-slice data from various sensors. To recover independent physical quantities from the mixed bitstream, an application-layer communication protocol is defined, specifying the frame structure of data packets, including frame headers, device identifiers, data lengths, payloads, and checksums. A state machine logic is used to scan the read binary bitstream byte by byte. First, a specific frame header byte is searched for to achieve frame synchronization; then, the following device identifier byte is read; based on the device identifier, the subsequent payloads are distributed to different memory circular buffers. When the identifier of the mobile platform camera is identified, the payload is mapped to the image decoding queue; when the identifier of the accelerometer is identified, the payload is parsed into a floating-point number and stored in the IMU data vector; similarly, the data corresponding to the identifiers of the laser rangefinder and gravimeter are mapped to the ranging signal stream and gravity reading sequence, respectively. This process completes the unpacking and distribution from the bitstream to structured data.

[0041] For example, suppose the binary bit stream transmitted through the first communication channel is: AA 55 01 04 3F 8000 00 CS ... AA 55 02 .... First, lock the frame header AA 55, read the next byte 01, and according to the protocol configuration table, identify that the packet belongs to the accelerometer reading. Read the length byte 04, indicating that there are 4 bytes of payload following. Extract the payload 3F80 00 00, convert it to a floating-point number 1.0 according to the IEEE 754 standard, representing an acceleration value of 1g, and add a receiving timestamp before storing it in the accelerometer reading sequence. Subsequently, continue scanning until the next AA 55 identifier 02 is found, which is identified as an image data packet. The payload data is then extracted and sent to the video decoder to restore a frame of mobile platform serialized image data. In this way, the image and inertial navigation data, which were originally mixed in the wireless link, are accurately separated and do not interfere with each other.

[0042] Optionally, the local average sea level elevation includes:

[0043] The sea surface laser ranging signal stream is discretized and binned through a continuous observation window. The number of sampling points falling into each bin is counted, and a discrete histogram vector is constructed.

[0044] First, time slicing is applied to the received sea surface laser ranging signal stream. Since the sea surface is constantly fluctuating due to wave influence, a single measurement cannot represent the average sea surface. Therefore, a continuous observation window is used, with a duration sufficient to cover at least 3 to 5 complete wave cycles to ensure the representativeness of the statistical sample. Discretization binning is then performed on all ranging values ​​within this window. This process maps continuously changing floating-point distance data to a finite number of discrete numerical intervals. A distance resolution is set as the bin width, dividing the range between the maximum and minimum observed distances into several consecutive equal-width intervals. Next, each ranging data point within the window is traversed, its interval index is calculated, and the counter for the corresponding interval is incremented. Finally, the ordered array of counts from all intervals constitutes a discrete histogram vector, which mathematically approximates the probability distribution of the sea surface distance data.

[0045] For example, in a deployment targeting an island reef in the South China Sea, a laser rangefinder operated at a frequency of 10 Hz. The continuous observation window was set to 60 seconds, and a total of 600 distance sampling points were collected. Assuming these sampling points were distributed between 12.5 meters and 13.5 meters, and the bin width was set to 1 centimeter, the 12.5-meter to 13.5-meter range was divided into 100 bins. For a sampling point with a value of 12.854 meters, it was assigned to the bin "12.85 meters to 12.86 meters," and the count value of that bin was incremented by 1. After statistical analysis of the 600 points, a discrete histogram vector containing 100 elements was generated, where the 35th element had a value of 15, indicating that 15 sea surface echoes were detected within the 12.85-meter interval.

[0046] Construct a Gaussian probability density function, and use the discrete histogram vector to perform nonlinear least squares fitting on the Gaussian probability density function to solve for the optimal position parameters;

[0047] By utilizing statistical principles to reconstruct the physical baseline from the noise, and considering that the surface elevation distribution of deep-water waves theoretically follows a Gaussian distribution under gravity, a parameterized Gaussian probability density function model is constructed to fit the aforementioned discrete histogram vector. The mathematical expression of this model is shown below:

[0048] ,

[0049] in, This represents the frequency or probability density of sampling points at a specific distance predicted by the model. The distance value representing the laser ranging over the sea surface is calculated using the center value of each sub-bin interval; The peak amplitude of the Gaussian distribution curve is positively correlated with the total number of sampling points and the degree of concentration of the distribution within the continuous observation window; The distance represents the center of symmetry of the Gaussian distribution, and is the static mean sea surface distance after removing the disturbance of wave fluctuations. The standard deviation represents the roughness of the sea surface or the significant wave height, i.e., the dispersion of wave undulation. A nonlinear least squares fitting algorithm is used, with the actual statistical frequencies in the discrete histogram vector as the observed values, and the formula is used to calculate... The value is used as the predicted value. Parameters are adjusted iteratively. , and The numerical value of is intended to minimize the sum of squared residuals between observed and predicted values. When the iteration converges, the set of parameters output by the algorithm is the optimal solution.

[0050] For example, if the discrete histogram vector is read to show a bell-shaped distribution with a high center and low sides, initialize the fitting parameters and set the initial position parameters. The distance corresponding to the maximum value of the histogram is 13.0 meters. Then, a nonlinear least squares fitting iteration is initiated. After 15 iterations, the algorithm converges, calculating the optimal location parameters. It is 13.025 meters, with a standard deviation of 13.025 meters. The standard deviation was 0.15 meters. This means that although the measured data fluctuated wildly between 12.5 meters and 13.5 meters, the "still" sea surface distance determined by statistical fitting was precisely at 13.025 meters, and the 0.15-meter standard deviation reflects the presence of light waves with a wave height of approximately 30 to 50 centimeters at that time. Figure 2 As shown, the center of symmetry of the probability density curve obtained by nonlinear least squares fitting is the local mean sea surface elevation after removing wave noise.

[0051] The expected peak value is calculated using the optimal location parameters for the distance distribution histogram, and the local mean sea level elevation is output.

[0052] Directly extract the optimal position parameters calculated during the fitting process. Since the mathematical expectation of the Gaussian distribution is its center of symmetry, this parameter physically represents the average sea level position after eliminating random wave noise. The value of this optimal position parameter is marked as the local average sea level elevation and used as the zero elevation reference benchmark in the vertical direction, that is, the sensor distance corresponding to the 0-meter elevation surface.

[0053] Optionally, the calculation of the optimal position parameters includes:

[0054] A residual vector is established using the discrete histogram vector and the Gaussian probability density function, and a sum-of-squares objective function is constructed based on the residual vector.

[0055] First, the actual count value of each bin in the discrete histogram vector is read. Then, using the position, amplitude, and scale parameters from the current iteration, the Gaussian probability density function is substituted to calculate the theoretical prediction value at the center distance of the corresponding bin. Each element of the residual vector is the difference between the actual count value and the theoretical prediction value of that bin. Subsequently, the sum-of-squares objective function SSE is constructed. This function is defined as the sum of the squares of all elements in the residual vector, and its mathematical expression is as follows:

[0056] ,

[0057] in, The result of the calculation of the sum of squares objective function represents the overall fitting error energy under the current model parameters; This represents the total number of bins in the discrete histogram vector, which is the set number of bins. Representing the The actual number of sampling points in each sub-box; Representing the The center-to-center distance of each compartment; This represents the inclusion of positional parameters. Amplitude parameters and scale parameters The vector of parameters to be optimized; Represents using the current parameter vector The theoretical frequency calculated using the Gaussian probability density function.

[0058] For example, suppose that in one iteration of the calculation, for a bin with a distance of 13.0 meters, The discrete histogram vector records 20 actual sampling points. The theoretical prediction value calculated from the current model parameters is 18.5. Calculate the residual of this bin. If there are a total of 100 bins, the squared residuals of all 100 bins will be calculated and summed. Assuming the sum is 450.5, this value is the current sum of squares objective function value. .

[0059] The partial derivatives of the sum of squares objective function with respect to the position parameters are calculated to generate a gradient descent direction vector, and parameter iterative update operations are performed along the gradient descent direction vector;

[0060] First, using the principles of calculus, calculate the sum of squares and the objective function. For position parameters Amplitude parameters and scale parameters The partial derivatives of the gradient vector are used to determine the direction of the fastest error growth. To reduce the error, a gradient descent direction vector (the negative direction of the gradient vector) is generated, and the parameters are updated according to the step size. The mathematical formula for the parameter iterative update operation is as follows:

[0061] ,

[0062] in, This represents the set of new parameters calculated in this iteration; This represents the set of old parameters at the start of this iteration; This represents the gradient of the objective function at the current parameter point; The step size coefficient representing the parameter update is typically set to 1. to The positive real numbers between these values ​​can be used, or the adaptive algorithm Adam can be used to dynamically adjust the values ​​to ensure that the iteration process can converge quickly without oscillating near the extreme points.

[0063] The residual vector is used to monitor the L2 norm value. When the rate of change of the L2 norm value converges and stabilizes, the iteration stops, and the current position parameter is output as the optimal position parameter.

[0064] Calculate the L2 norm of the residual vector, which is the square root of the sum of squared residuals (L2-Norm). This value reflects the average error level of the fit. Monitor the relative rate of change of this L2 norm value between two consecutive iterations. When the rate of change is less than the convergence threshold or when the maximum number of iterations is reached, the fitting process is considered to have converged and stabilized. At this point, stop the iteration, and update the position parameters in the parameter vector obtained from the last update. Extract it and output it as the optimal position parameter.

[0065] For example, setting the learning rate The value is 0.01, and the convergence threshold is... In the 50th iteration, the calculated L2 norm of the residual vector was 5.2000; in the 51st iteration, after updating the parameters, the calculated L2 norm was 5.1998. The relative rate of change was calculated. Because the rate of change is less than the convergence threshold. The determination result has stabilized, and the iteration is terminated. At this point, the position parameters in the parameter vector... The value is 13.025 meters, which is output as the optimal position parameter, representing the average elevation of the current sea surface.

[0066] Optionally, the generation of the relative geometric point cloud of islands and reefs includes:

[0067] Based on the serialized image data of the mobile platform, common texture points are extracted from two adjacent frames of images, and the pixel coordinate position difference is calculated using the common texture points.

[0068] First, the received video stream is frame-by-frame extracted, selecting the frames that are consecutive in time. Frame image and the first For each frame of the image, a feature point detection algorithm, such as ORB, SIFT, or SURF, is used to identify corner regions with significant grayscale changes or texture gradients between the two frames, marking them as common texture points. Then, feature descriptor matching or optical flow tracing methods are used to establish the first... a point in the frame With the Corresponding points in the frame The same-name correspondence, among which Representing the horizontal and vertical pixel coordinates on the image plane, calculate the pixel displacement vector of this common texture point between two frames, i.e., the difference in pixel coordinate positions. The calculation formula is as follows:

[0069] ,

[0070] in, The disparity modulus of common texture points on the image plane is represented in pixels. Representing the The x and y coordinates of feature points in a frame image; Representing the The x and y coordinates of the same feature points in the frame image.

[0071] For example, suppose the camera on the mobile platform has a resolution of In the 100th frame image, a feature point at the tip of a reef was identified, with coordinates as follows: In frame 101, the imaging position of the reef tip shifted due to the drone's forward flight. Calculate the difference in pixel coordinates. The number of pixels indicates that the feature point generates a vertical parallax of 20 pixels between the two frames.

[0072] Based on the accelerometer readings, a double integration operation is performed in the time dimension to calculate the physical distance traveled by the mobile platform within the interval between two adjacent image frames.

[0073] Since monocular vision cannot perceive absolute distance, the physical displacement of the camera is calculated using synchronously acquired accelerometer readings. First, the [data / image / section] is captured... Frame and the The time interval between frame captures All accelerometer readings within the range are subjected to a discrete double integration operation. The first integration converts the accumulated acceleration readings into a velocity change, and the second integration converts the accumulated velocity into a displacement change. The physical operation process can be approximated by the following discrete formula:

[0074] ,

[0075] in, This represents the absolute displacement length of the mobile platform within the interval between two adjacent shooting frames, in meters. Representing the The instantaneous velocity estimate of the mobile platform at each frame time is usually derived from the Kalman filter state estimate of the previous time step; Represents the time difference between two adjacent frames; This represents the total number of accelerometer sampling points within that time interval; Representing the The acceleration measurements at each sampling time need to have the gravity component deducted beforehand and converted to the world coordinate system; This represents the time step between two accelerometer readings.

[0076] For example, suppose the drone camera has a frame rate of 20fps, that is, the interval between two adjacent frames. The accelerometer sampling rate is 200Hz, meaning 10 acceleration readings were collected between two frames. Assuming the drone is in uniformly accelerated linear flight, the initial velocity... m / s, the average effective acceleration of these 10 acceleration readings after coordinate transformation is meters per second², calculating the physical distance traveled. Meters, this value is the true baseline length of the camera between these two frames.

[0077] Using the physical movement distance as the reference length, and combining it with the pixel coordinate position difference, geometric calculation is performed to determine the three-dimensional spatial coordinates of the common texture points, and aggregated to generate a relative geometric point cloud of islands and reefs.

[0078] The calculated physical movement distance is used as the baseline length in the stereo vision model. Combined with the camera's focal length parameters and pixel parallax, the spatial depth of the feature points is calculated. The calculation formula for the parallel optical axis model is as follows:

[0079] ,

[0080] in, Represents the vertical distance from the common texture point to the camera's optical center, in meters; Represents the camera's equivalent focal length, expressed in pixels, and is obtained from the camera's intrinsic parameter calibration matrix; That is, the calculated baseline length; That is, the calculated disparity. Using this depth value... The image coordinates of the feature points are used to calculate the three-dimensional spatial coordinates of the points based on the back projection of the pinhole camera model. Finally, all the 3D coordinate points calculated between frames are converged into the same local coordinate system, forming a discrete set of points with a true physical scale, namely the relative geometric point cloud of the islands and reefs.

[0081] For example, the camera focal length is known. Pixels, physical distance traveled Meters, the difference in pixel coordinates of a feature point Pixel, calculate the depth of the feature point. This means the tip of the reef is approximately 12.51 meters from the drone. The three-dimensional coordinates of this point are then generated and added to the point cloud dataset.

[0082] Optionally, the method further includes:

[0083] The numerical ratio is calculated by using the difference in pixel coordinates and the physical movement distance, and the numerical ratio is mapped to a distribution histogram.

[0084] For each pair of image frames acquired by the mobile platform, i.e. the first... Frame and the From each frame, two key physical quantities are extracted: the physical distance traveled, obtained through double integration from the accelerometer, and the difference in pixel coordinates, obtained through visual feature matching. The ratio of these two values ​​is used to generate the single-frame scale assumption, calculated using the following formula:

[0085] ,

[0086] in, Representing the The scale factor for each frame pair, under ideal conditions where the flight altitude is constant and the camera focal length is fixed, should be proportional to the scene depth. The displacement is calculated based on accelerometer integration, and the unit is meters. The disparity is calculated based on common texture points, in pixels. Due to accelerometer drift noise and potential mismatches in visual matching, directly averaging is susceptible to extreme values; therefore, a histogram method is used. First, the numerical distribution range of the ratio sequence is analyzed, and a fixed bin width is set, for example, 1% to 5% of the expected ratio values. Then, the entire sequence is traversed, and the number of ratios falling into each numerical interval is counted, constructing a distribution histogram. This histogram reflects the degree of "consensus" between visual and inertial data over a long time series. If the two are highly consistent, the histogram will show a sharp, single-peaked shape; if there is severe interference, it may show a flat or multi-peaked shape.

[0087] For example, in a mission where a drone flies along the edge of an island reef, for the 50th frame pair, the accelerometer calculates a physical distance traveled of 0.2 meters, and the difference in pixel coordinates of common texture points is 16 pixels. Calculate the numerical ratio of this frame pair. Meters per pixel: For the 51st frame pair, the physical movement distance is 0.21 meters, and the pixel coordinate difference is 16.5 pixels. The calculated ratio is... Meters per pixel. This calculation is repeated for 1000 frame pairs throughout the entire flight, generating a ratio sequence containing 1000 elements.

[0088] The mode is extracted from the distribution histogram to perform cross-validation of multi-source data and calculate the scale reliability coefficient.

[0089] First, identify the interval with the highest frequency in the distribution histogram, i.e., the mode interval, and determine the center value of this interval as the optimal scale ratio. The mode represents the most statistically reliable physical visual ratio relationship, effectively eliminating outliers caused by momentary shocks or matching errors. Based on this mode, calculate the scale reliability coefficient to quantify the reliability of the current measurement environment. The calculation formula is as follows:

[0090] ,

[0091] in, The quality evaluation index representing the final output is a dimensionless real number between 0 and 1. The total percentage of values ​​in the distribution histogram that fall within the confidence interval centered on the mode value; This represents the total number of frame pairs included in the statistics. The closer this coefficient is to 1, the more consistent the ratio between inertial and visual measurements is with each other, indicating high data quality. Conversely, a smaller coefficient indicates the presence of significant noise or scale drift.

[0092] For example, a histogram was constructed on the numerical ratios of 1000 frame pairs, and the peak was found to be located at... The interval is set as the confidence interval. The mode Statistical analysis revealed that out of these 1000 data points, 850 numerical ratios fell within this confidence interval, i.e. The remaining 150 data points are discretely distributed outside the interval, and the scaling reliability coefficient is calculated. Since this coefficient is greater than the acceptable threshold, the physical scale information collected in this study is deemed reliable and can be used for point cloud coordinate correction. For example... Figure 3As shown, the scatter plot on the left displays the single-frame scale estimate and effective confidence interval as the image frame sequence changes, while the density plot on the right reflects the statistical central tendency of the overall data and the optimal scale mode.

[0093] Optionally, the calculation of the tilt angle values ​​of the gravity vertical line in the north-south and east-west directions includes:

[0094] Geographic coordinates of each observation point are extracted based on the land gravimeter reading sequence. The normal gravity value of the Earth ellipsoid is calculated using the geographic coordinates. The numerical difference between the land gravimeter reading sequence and the normal gravity value of the Earth ellipsoid is calculated to generate discrete gravity anomaly data.

[0095] First, the land gravimeter reading sequence obtained from the second communication channel is analyzed. Each data packet in this sequence contains the absolute gravity observation value and the geographic coordinates of the observation point. The normal gravity value of the Earth's ellipsoid at that latitude is then calculated. The calculation formula is as follows:

[0096] ,

[0097] in, Representing an ideal ellipsoid at latitude The theoretical gravity value at that location, usually measured in milligal (mGal); It is a physical constant, with a value of mGal; Values ; Values ; This is the geographical latitude of the observation point. Perform numerical difference calculation:

[0098] ,

[0099] in, These are the measured gravity values ​​from the land gravimeter reading sequence. (Calculated...) This refers to the gravity anomaly at that point, and all observation points... The set constitutes the discrete data of gravity anomalies.

[0100] For example, suppose at observation point A on an island reef, at latitude 10 degrees North, the reading of the land gravimeter is... mGal, substitute into the formula to calculate the normal gravity value of the Earth's ellipsoid. Calculate the numerical difference mGal, this value mGal represents the discrete gravity anomaly data at that point, indicating that the density of matter below that point is less than that in the theoretical model.

[0101] Spatial gridding interpolation is performed on the discrete gravity anomaly data to cover the island and reef region to generate a gravity anomaly grid matrix;

[0102] Due to terrain limitations, measurement points are often irregularly distributed during field observations, making direct convolution operations impossible. First, a rectangular region covering the entire island / reef and extending outwards to a certain buffer zone is defined, and a grid resolution is set. Using limited discrete gravity anomaly data, the gravity anomaly value of each grid node within the rectangular region is estimated, generating a two-dimensional array, which is the gravity anomaly grid matrix. This matrix comprehensively describes the mass distribution characteristics of the island / reef region.

[0103] Traverse the gravity anomaly grid matrix, establish concentric annular zones with grid nodes as the calculation center, calculate the spatial distance from the remaining grid nodes in the gravity anomaly grid matrix to the calculation center, and use the spatial distance to generate a distance inverse weighting factor;

[0104] Each node in the gravity anomaly grid matrix is ​​selected sequentially as the computation center. For except Any remaining grid node other than the point Calculate the spatial distance and azimuth between the two. A distance-inverse weighting factor is generated. Under the near-distance approximation, this factor is inversely proportional to the distance. Its mathematical expression and physical meaning are as follows:

[0105] ,

[0106] in, Represents distance The contribution weight of unit gravity anomaly at a given location to vertical deviation; It is a classic function in physical geodesy that describes the relationship between gravitational potential and geoid height; Computing Center With remaining grid nodes The spherical angular distance between them. In small-scale island and reef surveys, this factor can be simplified to the following calculation: This form reflects the physical law that the closer the distance, the more significant the torsional effect of gravitational anomalies on the vertical direction.

[0107] Based on the gravity anomaly grid matrix and the distance inverse weighting factor, convolution accumulation is performed to output the tilt angle values ​​of the gravity vertical line in the north-south and east-west directions.

[0108] The discretized convolution integral operation aims to convert the scalar field into a vector field, and calculate the north-south components separately. and east-west component The calculation formula is as follows:

[0109] ,

[0110] in, These represent the angles of the gravity plumb line relative to the ellipsoid normal in the north-south and east-west directions, respectively, in radians; It is the average normal gravity value at the calculation point; It is the total number of nodes in the gravity anomaly grid matrix; It is the first Gravity anomaly values ​​of each grid node; It is the first The distance-inverse weighting factor corresponding to each node; It is the first The azimuth angle of each node relative to the computing center; This represents the area element of a single grid cell. This accumulation process is repeated for each grid node, outputting two matrices of the same dimension as the gravity anomaly grid matrix, storing the tilt angle values ​​of each point on the island / reef in the north-south and east-west directions, respectively.

[0111] For example, suppose in a computing center There is a gravity anomaly 100 meters to the due north. mGal. Due to The high point has an east-west component No contribution, but for the north-south component Due to the presence of a large positive anomaly, it generates a significant gravitational pull, causing the vertical to tilt northward. Through convolutional summation calculations, if the solution... arc second, Arcseconds means that at this point, the actual physical perpendicular is 10 arcseconds north and 2 arcseconds east compared to the geometric normal. Point cloud corrections must offset this deviation; otherwise, the elevation projection onto the plane will produce an error on the order of several meters.

[0112] Optionally, the set of output control point coordinates includes:

[0113] Based on the scale reliability coefficient, perform multiplicative scaling and scale alignment on all coordinate values ​​of the relative geometric point cloud of the islands and reefs.

[0114] First, the scale reliability coefficient is read and compared with the engineering-available threshold. If the coefficient is higher than the threshold, the mode extracted from the distribution histogram, i.e., the optimal numerical ratio, is defined as the global scale scaling factor. Then, each point in the relative geometric point cloud of the islands and reefs is traversed, and its original three-dimensional coordinate values ​​are obtained. By multiplying by the global scale scaling factor, this multiplication operation maps the point cloud from the dimensionless "pixel / relative space" to the "metric space", thus completing the absolutization of scale.

[0115] For example, suppose that in the relative geometric point cloud of islands and reefs, the original relative coordinates of a certain reef vertex are... The mode of the distribution histogram is And the scale reliability coefficient is 0.92, performing multiplicative scaling: rice, rice, Meters. At this point, the coordinates of the point change to It possesses a real physical length meaning.

[0116] Calculate the numerical difference between the average height of the waterline points in the relative geometric point cloud of the island and reef and the local average sea level elevation, and superimpose the numerical difference onto the vertical coordinate component of the relative geometric point cloud of the island and reef to generate the corrected vertical coordinate component.

[0117] Since the origin of the point cloud coordinate system for visual reconstruction is usually located at the optical center of the first frame camera, Since axis values ​​do not correspond to altitude, we first identify the waterline points in the point cloud and calculate the average altitude of these points in the scaled-up point cloud. Simultaneously, we retrieve the local mean sea level elevation and calculate the difference between the two. This difference represents the vertical error of the point cloud coordinate system relative to the physical sea level. It is added to the vertical coordinate component of each point in the point cloud to achieve elevation normalization.

[0118] For example, the average height of the waterline points in the scaled point cloud. for This likely means the camera was launched 2 meters above the water surface, measuring the local average sea level elevation. for Meters, that is, defining the sea level as the 0-meter reference, to calculate the numerical difference. For the top of the reef, the numerical differences are summed. The meter indicates that the actual elevation of the reef's summit is 4.5 meters.

[0119] Multiply the corrected vertical coordinate component by the tilt angle value to calculate the horizontal offset value of the island / reef relative geometric point cloud in the planar direction, and compensate the horizontal offset value to the planar coordinate component of the island / reef relative geometric point cloud to output the control point coordinate set.

[0120] In physical geodesy, traditional instruments such as total stations and theodolites are leveled based on gravity plumb lines, while GPS / GNSS coordinates are based on the Earth's ellipsoid normal. There is a small angle between the two, known as plumb line deviation. For a target point with a certain elevation, this angle will cause a shift in its projected position on the horizontal plane. This shift is compensated for using the tilt angle value, calculated using the following formula:

[0121] ,

[0122] in, These represent the distances that need to be compensated for in the east-west and north-south directions, respectively, in meters; The point cloud height value that has been aligned with elevation is based on the physical principle that the higher the elevation, the longer the horizontal projection arm caused by the tilt of the vertical line. These represent the angles of inclination of the gravity plumb line in the east-west and north-south directions, respectively. : Tangent trigonometric function. Before substituting into the calculation, the angle value in arcseconds must be converted to radians. and The coordinates of each point are accumulated and added to the planar coordinate components to complete the final three-dimensional correction. The resulting set is then output as the control point coordinate set.

[0123] For example, suppose the corrected vertical coordinate components of a control point on a mountain peak are... Meters, at that position the vertical line tilts eastward. Arcseconds, tilted northwards. Arcseconds are used to calculate the horizontal offset in the east-west direction. This means that due to the non-uniform gravitational field, the point would have a 1-centimeter planar position error if not corrected. This 0.01-meter error is compensated for by adding it to the eastern coordinates of the point, thus obtaining the coordinates of the control point that conform to the geodetic surveying standard.

[0124] Optionally, calculating the numerical difference between the average elevation of the waterline points in the relative geometric point cloud of the island / reef and the local mean sea level elevation includes:

[0125] The vertical coordinate components are sorted in ascending order, and the set of coordinate values ​​less than or equal to the lower quartile is extracted.

[0126] Since the island and reef point cloud encompasses all terrain features from beaches to mountain peaks, directly averaging the values ​​would introduce a significant positive bias. First, the scale-aligned relative geometric point cloud of the islands and reefs is traversed, extracting the vertical coordinate components of each point to construct a one-dimensional elevation array. Then, this array is sorted in ascending order using quicksort or mergesort. In the sorted array, the lower quartile index is calculated, defined as the position at 25% of the total array length. The physical significance of the lower quartile lies in its representation of the upper bound of the smaller 25% of values ​​in the dataset, effectively eliminating elevation anomalies such as highlands, vegetation, and buildings on the islands and reefs, while retaining data from beaches or reef bases close to the water surface. All values ​​from the starting position to this lower quartile index are extracted from the array to form a set of coordinate values.

[0127] For example, suppose the relative geometric point cloud of an island or reef contains 1000 points. Extracting the Z-coordinates of these 1000 points reveals that their values ​​range from -5 meters to +50 meters. The origin of this coordinate system is the camera's initial position, not yet aligned with the sea surface. Arranging these 1000 values ​​in ascending order, the 250th value is -1.5 meters. This means that 250 points have Z-coordinates less than or equal to -1.5 meters. Extracting these 250 values ​​as a set of coordinate values, these points physically likely correspond to low-lying waterfront areas at the edge of the island or reef.

[0128] The arithmetic mean of the set of coordinate values ​​is calculated as the average elevation of the waterline points, and the scalar difference between the average elevation of the waterline points and the local mean sea level is calculated as the numerical difference.

[0129] First, calculate the arithmetic mean of the above coordinate values, defining it as the average elevation of the waterline points. Compared to directly taking the minimum value, averaging effectively smooths out errors caused by individual underwater reflection noise or potholes. Then, read the calculated local mean sea level elevation and calculate the scalar difference between the two, using the following formula:

[0130] ,

[0131] in, This represents the elevation system deviation to be determined, in meters. This value will be used to correct the vertical coordinates of the entire point cloud. The statistical average elevation of low-altitude areas in the point cloud data; The value represents the actual physical sea level height, derived from optimal position parameters obtained through statistical analysis of sea surface laser ranging signal flow. For example, the average Z-coordinate of 250 low-elevation points is calculated to determine the average height of the waterline points. Meters. Meanwhile, laser ranging results show that the local average sea level elevation... Corresponding to a sensor reading of 13.0 meters, assuming the sensor is located at a high position and the sea surface is 13 meters below. In practical engineering, it is necessary to unify the definition of the coordinate system. The laser rangefinder measures "distance," while the Z-axis of the point cloud is the "coordinate." If downward is defined as negative, assuming the laser rangefinder is located in the point cloud coordinate system... The actual sea level is located at [location]. Meters. At this point, calculate the numerical difference. This means that the "water surface" in the point cloud is 11.2 meters higher than the "real sea surface." To align the point cloud to the real sea surface, the inverse of this difference, or the corresponding adjustment amount defined according to the coordinate system, is superimposed on the point cloud. and coincide.

[0132] Based on the same inventive concept, this invention also provides an autonomous construction and transmission system for remote islands without a reference control network, such as... Figure 4 As shown, the system includes:

[0133] The multi-source data receiving module is used to receive sea surface laser ranging signal streams, land gravimeter reading sequences, serialized image data from mobile platforms, and accelerometer readings.

[0134] The sea surface elevation extraction module is used to extract the peak distance data of the distance distribution histogram as the local average sea surface elevation based on the distance distribution histogram of the sea surface laser ranging signal stream through a continuous observation window.

[0135] The geometric point cloud generation module is used to track textured pixels with the same name based on the serialized image data of the mobile platform, and calculate the spatial displacement scale of the pixels by combining the accelerometer readings to generate a relative geometric point cloud of islands and reefs.

[0136] The vertical tilt angle calculation module is used to compare the difference between the land gravimeter reading sequence and the Earth's normal gravity field data, and calculate the tilt angle values ​​of the gravity vertical in the north-south and east-west directions.

[0137] The control point output module is used to calculate the average height difference between the waterline points in the relative geometric point cloud of the islands and reefs and the local average sea level, use the average height difference to correct the vertical coordinate components of the relative geometric point cloud of the islands and reefs, and use the tilt angle value to correct the planar coordinate components of the relative geometric point cloud of the islands and reefs, and output the control point coordinate set.

[0138] It should be noted that the functional division and information interaction between the various modules described above are logical, but in terms of physical implementation, they can be integrated on the same software platform or deployed in a distributed manner. The connections between them represent data flow and control flow, aiming to collaboratively achieve the objectives of this invention. The above descriptions are merely exemplary embodiments of this invention and should not be construed as limiting the scope of protection of this invention.

Claims

1. A method for autonomous construction and transmission of a control network without a reference point on remote islands, characterized in that: The method includes: Receives sea surface laser ranging signal streams, land gravimeter reading sequences, serialized image data from mobile platforms, and accelerometer readings; Based on the distance distribution histogram of the sea surface laser ranging signal stream through a continuous observation window, the peak distance data of the distance distribution histogram is extracted as the local average sea surface elevation. Based on the serialized image data of the mobile platform, trace the texture pixels with the same name, and combine the accelerometer readings to calculate the spatial displacement scale of the pixels to generate a relative geometric point cloud of islands and reefs. The land gravimeter reading sequence is compared with the Earth's normal gravity field data to calculate the tilt angle of the gravity vertical in the north-south and east-west directions. The average height difference between the waterline points in the relative geometric point cloud of the islands and reefs and the local average sea level is calculated. The average height difference is used to correct the vertical coordinate components of the relative geometric point cloud of the islands and reefs, and the tilt angle value is used to correct the planar coordinate components of the relative geometric point cloud of the islands and reefs, and the control point coordinate set is output.

2. The method for autonomous construction and transmission of a reference-free control network for remote islands according to claim 1, characterized in that, The received sea surface laser ranging signal stream, land gravimeter reading sequence, mobile platform serialized image data, and accelerometer readings include: The system calls the wireless gateway interface to establish a first communication channel with the mobile platform, and calls the multi-channel serial communication interface to establish a second communication channel with the shore-based device. The binary bit stream is read through the first communication channel and the second communication channel respectively, and the binary bit stream is split and mapped into mobile platform serialized image data, accelerometer readings, sea surface laser ranging signal stream and land gravimeter reading sequence according to the device identifier.

3. The method for autonomous construction and transmission of a reference-free control network for remote islands according to claim 1, characterized in that, The local average sea level elevation includes: The sea surface laser ranging signal stream is discretized and binned through a continuous observation window. The number of sampling points falling into each bin is counted, and a discrete histogram vector is constructed. Construct a Gaussian probability density function, and use the discrete histogram vector to perform nonlinear least squares fitting on the Gaussian probability density function to solve for the optimal position parameters; The expected peak value is calculated using the optimal location parameters for the distance distribution histogram, and the local mean sea level elevation is output.

4. The method for autonomous construction and transmission of a reference-free control network for remote islands according to claim 3, characterized in that, The solution for the optimal position parameters includes: A residual vector is established using the discrete histogram vector and the Gaussian probability density function, and a sum-of-squares objective function is constructed based on the residual vector. The partial derivatives of the sum of squares objective function with respect to the position parameters are calculated to generate a gradient descent direction vector, and parameter iterative update operations are performed along the gradient descent direction vector; The residual vector is used to monitor the L2 norm value. When the rate of change of the L2 norm value converges and stabilizes, the iteration stops, and the current position parameter is output as the optimal position parameter.

5. The method for autonomous construction and transmission of a reference-free control network for remote islands according to claim 1, characterized in that, The generated island / reef relative geometric point cloud includes: Based on the serialized image data of the mobile platform, common texture points are extracted from two adjacent frames of images, and the pixel coordinate position difference is calculated using the common texture points. Based on the accelerometer readings, a double integration operation is performed in the time dimension to calculate the physical distance traveled by the mobile platform within the interval between two adjacent image frames. Using the physical movement distance as the reference length, and combining it with the pixel coordinate position difference, geometric calculation is performed to determine the three-dimensional spatial coordinates of the common texture points, and aggregated to generate a relative geometric point cloud of islands and reefs.

6. The method for autonomous construction and transmission of a reference-free control network for remote islands according to claim 5, characterized in that, The method further includes: The numerical ratio is calculated using the difference in pixel coordinates and the physical movement distance, and the numerical ratio is mapped to a distribution histogram. The mode is extracted from the distribution histogram to perform cross-validation of multi-source data and calculate the scale reliability coefficient.

7. The method for autonomous construction and transmission of a reference-free control network for remote islands according to claim 1, characterized in that, The calculated tilt angles of the gravity plumb line in the north-south and east-west directions include: Geographic coordinates of each observation point are extracted based on the land gravimeter reading sequence. The normal gravity value of the Earth ellipsoid is calculated using the geographic coordinates. The numerical difference between the land gravimeter reading sequence and the normal gravity value of the Earth ellipsoid is calculated to generate discrete gravity anomaly data. Spatial gridding interpolation is performed on the discrete gravity anomaly data to cover the island and reef region to generate a gravity anomaly grid matrix; Traverse the gravity anomaly grid matrix, establish concentric annular zones with grid nodes as the calculation center, calculate the spatial distance from the remaining grid nodes in the gravity anomaly grid matrix to the calculation center, and use the spatial distance to generate a distance inverse weighting factor; Based on the gravity anomaly grid matrix and the distance inverse weighting factor, convolution accumulation is performed to output the tilt angle values ​​of the gravity vertical line in the north-south and east-west directions.

8. The method for autonomous construction and transmission of a reference-free control network for remote islands according to claim 6, characterized in that, The set of output control point coordinates includes: Based on the scale reliability coefficient, perform multiplicative scaling and scale alignment on all coordinate values ​​of the relative geometric point cloud of the islands and reefs; Calculate the numerical difference between the average height of the waterline points in the relative geometric point cloud of the island and reef and the local average sea level elevation, and superimpose the numerical difference onto the vertical coordinate component of the relative geometric point cloud of the island and reef to generate the corrected vertical coordinate component. Multiply the corrected vertical coordinate component by the tilt angle value to calculate the horizontal offset value of the island / reef relative geometric point cloud in the planar direction, and compensate the horizontal offset value to the planar coordinate component of the island / reef relative geometric point cloud to output the control point coordinate set.

9. The method for autonomous construction and transmission of a reference-free control network for remote islands according to claim 8, characterized in that, The calculation of the numerical difference between the average elevation of the waterline points in the relative geometric point cloud of the islands and reefs and the local average sea level elevation includes: The vertical coordinate components are sorted in ascending order, and the set of coordinate values ​​less than or equal to the lower quartile is extracted. The arithmetic mean of the set of coordinate values ​​is calculated as the average elevation of the waterline points, and the scalar difference between the average elevation of the waterline points and the local mean sea level is calculated as the numerical difference.

10. A system for autonomous construction and transmission of a control network without a reference network for remote islands, applied to the method for autonomous construction and transmission of a control network without a reference network for remote islands as described in any one of claims 1-9, characterized in that, The system includes: The multi-source data receiving module is used to receive sea surface laser ranging signal streams, land gravimeter reading sequences, serialized image data from mobile platforms, and accelerometer readings. The sea surface elevation extraction module is used to extract the peak distance data of the distance distribution histogram as the local average sea surface elevation based on the distance distribution histogram of the sea surface laser ranging signal stream through a continuous observation window. The geometric point cloud generation module is used to track textured pixels with the same name based on the serialized image data of the mobile platform, and calculate the spatial displacement scale of the pixels by combining the accelerometer readings to generate a relative geometric point cloud of islands and reefs. The vertical tilt angle calculation module is used to compare the difference between the land gravimeter reading sequence and the Earth's normal gravity field data, and calculate the tilt angle values ​​of the gravity vertical in the north-south and east-west directions. The control point output module is used to calculate the average height difference between the waterline points in the relative geometric point cloud of the islands and reefs and the local average sea level, use the average height difference to correct the vertical coordinate components of the relative geometric point cloud of the islands and reefs, and use the tilt angle value to correct the planar coordinate components of the relative geometric point cloud of the islands and reefs, and output the control point coordinate set.

Images