A multifunctional automated spatial point tracking measurement method and device

Abstract

The present invention discloses a multifunctional automated spatial point tracking and measurement method and device, which relates to the field of tracking and measurement technology, including: obtaining N monitoring points for priority sorting, obtaining monitoring point sequence information, presetting the monitoring acquisition frequency according to the monitoring point sequence information, and generating monitoring area information; a built-in high-precision navigation module automatically navigates to the monitoring area information, fixes the target icon on the target monitoring point, and obtains the target monitoring image information of the monitoring area information; trains and constructs a point tracking module, performs target positioning and coordinate extraction, and determines the target pixel coordinate information; performs target coordinate calculation to obtain the target's actual spatial coordinates, and stores the target's actual spatial coordinates in sequence to generate N spatial point measurement results. The present invention solves the technical problems in the prior art of difficult to guarantee monitoring frequency, low measurement efficiency, and insufficient positioning accuracy, and achieves the technical effect of improving the efficiency and accuracy of spatial point tracking measurement.

Description

Technical Field

[0001] The present invention relates to the field of tracking measurement technology, and in particular to a multifunctional automated spatial point tracking measurement method and device. Background Art

[0002] During the construction process, spatial coordinate monitoring of key points is of great significance for ensuring project quality and structural safety. With the continuous improvement of industrial automation levels, the demand for high-frequency, high-precision, and continuous monitoring of spatial points is increasing, and the limitations of traditional measurement methods are becoming increasingly prominent. Existing spatial point measurement methods mostly rely on manual operation or semi-automatic equipment. Common methods include manual reading or point observation with the help of a total station. This type of method not only relies heavily on the technical level of the operator, but also has complex operating procedures and requires the collaboration of multiple people, resulting in low overall measurement efficiency. When faced with the task of continuous measurement of multiple monitoring points, traditional methods are difficult to balance the timeliness and accuracy requirements of data collection, especially in high-frequency monitoring scenarios, making it difficult to achieve continuous and automated monitoring of points. Summary of the Invention

[0003] The present application provides a multifunctional automated spatial point tracking and measurement method and device, which is used to solve the technical problems in the prior art such as difficulty in ensuring monitoring frequency, low measurement efficiency and insufficient positioning accuracy.

[0004] In view of the above problems, the present application provides a multifunctional automated spatial point tracking measurement method and device.

[0005] A first aspect of the present application provides a multifunctional automated spatial point tracking measurement method, the method comprising:

[0006] Acquire N monitoring points, prioritize the N monitoring points, obtain monitoring point sequence information, preset the monitoring acquisition frequency, input the theoretical coordinate information and preset error range of the target monitoring point in sequence according to the monitoring point sequence information, and generate monitoring area information based on the theoretical coordinate information and the preset error range; use a built-in high-precision navigation module to automatically navigate to the monitoring area information, fix a target icon on the target monitoring point, and use the target icon as a reference to acquire target monitoring image information of the monitoring area information through a high-precision telescope; train and construct a point tracking module, use the point tracking module to perform target positioning and coordinate extraction on the target monitoring image information, and determine the target pixel coordinate information; calculate the target coordinate based on the telescope camera parameters and the target pixel coordinate information to obtain the actual spatial coordinates of the target, and store the actual spatial coordinates of the target in sequence to generate N spatial point measurement results.

[0007] A second aspect of the present application provides a multifunctional automated spatial point tracking and measurement device, the device comprising:

[0008] A monitoring area information acquisition module is used to acquire N monitoring points, prioritize the N monitoring points, obtain monitoring point sequence information, preset the monitoring acquisition frequency, input the theoretical coordinate information and preset error range of the target monitoring point in sequence according to the monitoring point sequence information, and generate monitoring area information based on the theoretical coordinate information and the preset error range; a monitoring image information acquisition module is used to automatically navigate to the monitoring area information with a built-in high-precision navigation module, fix a target icon on the target monitoring point, and use the target icon as a reference to acquire the target monitoring image information of the monitoring area information through a high-precision telescope; a coordinate information confirmation module is used to train and construct a point tracking module, use the point tracking module to perform target positioning and coordinate extraction on the target monitoring image information, and determine the target pixel coordinate information; a point measurement result confirmation module is used to calculate the target coordinates based on the telescope camera parameters and the target pixel coordinate information to obtain the actual spatial coordinates of the target, and store the actual spatial coordinates of the target in sequence to generate N spatial point measurement results.

[0009] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0010] This application obtains N monitoring points, prioritizes the N monitoring points, obtains monitoring point sequence information, presets the monitoring acquisition frequency, inputs the theoretical coordinate information and preset error range of the target monitoring point in sequence according to the monitoring point sequence information, and generates monitoring area information based on the theoretical coordinate information and the preset error range; a built-in high-precision navigation module automatically navigates to the monitoring area information, fixes a target icon on the target monitoring point, and uses the target icon as a reference to acquire target monitoring image information of the monitoring area information through a high-precision telescope; trains and constructs a point tracking module, uses the point tracking module to perform target positioning and coordinate extraction on the target monitoring image information, and determines the target pixel coordinate information; calculates the target coordinates based on the telescope camera parameters and the target pixel coordinate information to obtain the actual spatial coordinates of the target, and stores the actual spatial coordinates of the target in sequence to generate N spatial point measurement results. The present invention solves the technical problems in the existing technology of difficult to guarantee monitoring frequency, low measurement efficiency and insufficient positioning accuracy. By combining automatic navigation, image recognition and spatial coordinate calculation, multiple monitoring points are identified and positioned automatically at high frequency, achieving the technical effect of improving the efficiency and accuracy of spatial point tracking measurement. BRIEF DESCRIPTION OF THE DRAWINGS

[0011] In order to more clearly illustrate the technical solutions in the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are only some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0012] Figure 1 A schematic flow chart of a multifunctional automated spatial point tracking and measurement method provided in an embodiment of the present application;

[0013] Figure 2 A schematic structural diagram of a multifunctional automated spatial point tracking and measurement device provided in an embodiment of the present application.

[0014] Description of reference numerals: monitoring area information acquisition module 11 , monitoring image information acquisition module 12 , coordinate information confirmation module 13 , point measurement result confirmation module 14 . DETAILED DESCRIPTION

[0015] This application provides a multifunctional automated spatial point tracking and measurement method and device to solve the technical problems in the existing technology such as difficulty in ensuring monitoring frequency, low measurement efficiency and insufficient positioning accuracy. By combining automatic navigation, image recognition and spatial coordinate calculation, multiple monitoring points are identified and positioned automatically at high frequency, achieving the technical effect of improving the efficiency and accuracy of spatial point tracking and measurement.

[0016] The following will be combined with the accompanying drawings in the embodiments of this application to clearly and completely describe the technical solutions in the embodiments of this application. Obviously, the embodiments described are only some of the embodiments of this application, not all of them. Based on the embodiments in this application, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of this application.

[0017] It should be noted that any variations of the terms "include" and "have" are intended to cover non-exclusive inclusions. For example, a process, method, apparatus, product or server that includes a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or modules that are not explicitly listed or are inherent to these processes, methods, products or devices.

[0018] Example 1, as Figure 1 As shown, the present application provides a multifunctional automated spatial point tracking measurement method, the method comprising:

[0019] Step S100: Obtain N monitoring points, prioritize the N monitoring points, obtain monitoring point sequence information, preset the monitoring acquisition frequency, input the theoretical coordinate information and preset error range of the target monitoring point in sequence according to the monitoring point sequence information, and generate monitoring area information based on the theoretical coordinate information and the preset error range.

[0020] In an embodiment of the present application, N pre-set monitoring points are first obtained, and based on the preset monitoring point priority factors such as criticality, urgency and monitoring difficulty, a priority factor evaluation system is constructed. Through the comprehensive evaluation of each monitoring point, N priority factor coefficient sets are obtained, and then the monitoring points are sorted to generate ordered monitoring point sequence information. The above-mentioned N monitoring points include key points in the engineering project that are sensitive to structural changes or have a greater safety impact, such as bridge supports, high pier tops, deep foundation pit slope tops, expansion joint nodes, construction cable anchor points, etc. These points carry key structural stress or displacement information and are important objects for engineering quality control and risk warning.

[0021] After obtaining the monitoring point sequence information, the monitoring acquisition frequency is further preset. This acquisition frequency is dynamically adjusted by technical experts based on the priority of the monitoring points. For example, higher sampling frequencies are assigned to more critical or frequently changing points to increase their monitoring density. Subsequently, the theoretical coordinates and preset error range of each target monitoring point are input in sequence according to the monitoring point sequence information. The theoretical coordinates are used to determine the target point's ideal position in space, while the preset error range defines the allowable spatial offset limit.

[0022] Based on the theoretical coordinates and error range, a three-dimensional cube-shaped monitoring area boundary is constructed with the theoretical coordinates as the center and the error range as the radius. The boundary is then spatially partitioned and reconstructed based on the required measurement accuracy, forming a monitoring area model. Finally, the model is verified and optimized based on the actual distribution characteristics of the monitoring points and the surrounding environmental conditions, and the final monitoring area information is output. This entire process is performed by a tracking measurement robot, which automatically loads and analyzes the monitoring point information and schedules the sampling frequency.

[0023] Furthermore, in the method provided in the embodiment of the application, obtaining the monitoring point sequence information further includes:

[0024] Preset monitoring point priority factors, wherein the monitoring point priority factors include criticality, urgency, and monitoring difficulty; establish a priority factor evaluation system based on the monitoring point priority factors, wherein the priority factor evaluation system includes specific evaluation indicators for the priority factors of each monitoring point; perform priority evaluation on the N monitoring points according to the priority factor evaluation system to obtain N priority factor coefficient sets; perform priority sorting on the N monitoring points based on the N priority factor coefficient sets to obtain monitoring point sequence information.

[0025] In this embodiment, several key factors are pre-defined to determine the importance of a monitoring point, including criticality, urgency, and monitoring difficulty. Criticality reflects the importance of a monitoring point within the overall structure or measurement, urgency indicates whether the point is experiencing rapid changes or potential risks requiring priority monitoring, and monitoring difficulty assesses the complexity of performing stable and accurate measurements of the point in a realistic environment.

[0026] To enable quantifiable evaluation of monitoring point priority factors, a priority factor evaluation system was established. This system breaks down criticality, urgency, and monitoring difficulty into specific evaluation indicators, such as structural influencing factors, historical change rates, alarm frequency, identification success rate, and degree of obstruction, ensuring a clear quantifiable basis for each factor. Based on the Analytic Hierarchy Process (AHP), weights are assigned to each priority factor to construct a priority factor weight matrix. Using the AHP, a pairwise comparison judgment matrix for priority factors is first constructed, comparing the relative importance of each pair of factors in the overall evaluation. For example, criticality is assigned a value of 3 relative to urgency, indicating greater importance, and urgency is assigned a value of 2 relative to monitoring difficulty. This complete comparison matrix is ​​then constructed. The matrix is ​​then normalized and the average value of each row is calculated to obtain the weight coefficient for each factor, forming a weight vector. A consistency check is performed on this matrix to ensure the rationality of the weight assignment and the consistency of the judgment matrix, ensuring the stability and reliability of the priority evaluation results.

[0027] After the priority factor evaluation system is constructed, a priority evaluation is performed on each of the N monitoring points. Specifically, data is collected for each monitoring point across various evaluation indicator dimensions. For example, a monitoring point with a structural impact factor of 0.85, an average daily displacement rate of 2.1 mm / d, an occlusion degree of 30%, and an identification success rate of 92% is identified. The collected indicator data is normalized, mapping the raw data to a range between 0 and 1. For example, the normalized score is calculated by taking the difference between the maximum and minimum values, eliminating dimensional and numerical scale differences between indicators. After normalization, the indicator data for all monitoring points is weighted and summed with the priority factor weight matrix to obtain a comprehensive priority score for each monitoring point. This process generates a set of priority factor coefficients encompassing all monitoring points, where each monitoring point corresponds to a set of numerical values ​​reflecting its criticality, urgency, and monitoring difficulty, as well as a final weighted total score.

[0028] After constructing the priority factor coefficient set, a ranking analysis is conducted based on these N priority factor coefficient sets. First, the normalized evaluation indicators for each monitoring point are organized by dimension to generate N priority coefficient matrices, which uniformly represent the distribution of all points across the dimensions of criticality, urgency, and monitoring difficulty. Subsequently, indicator data mining is conducted based on the existing priority factor evaluation system to construct a priority factor evaluation indicator dataset, which is used to enhance the understanding and modeling of correlations between indicators. Principal component analysis (PCA) is then applied to this indicator dataset for dimensionality reduction and feature extraction. Principal component loading coefficients for each priority factor are obtained, and PCA weights are assigned to the priority factors, forming an optimized priority factor weight matrix. A matrix weighting operation is then performed on these N priority coefficient matrices and the PCA weight matrix to calculate the monitoring priority coefficient for each monitoring point, representing its comprehensive importance based on the fusion of multiple factors. Finally, the N monitoring points are sorted in descending order based on this monitoring priority coefficient, and monitoring point sequence information is output to guide subsequent measurement scheduling and frequency configuration.

[0029] After completing the priority evaluation of all monitoring points, all monitoring points are sorted according to the total score obtained from the priority factor coefficient set, forming a monitoring point sequence information from high to low. This monitoring point sequence information not only clarifies the execution order of subsequent measurement tasks, but also directly establishes a mapping relationship with the monitoring acquisition frequency, realizing priority-based sampling strategy optimization. For example, monitoring points with a priority score higher than 0.8 are set to high-frequency sampling (such as once every 5 minutes), those with a score between 0.6 and 0.8 are set to medium-frequency sampling (such as once every 15 minutes), and those below 0.6 are set to low-frequency sampling (once every 30 minutes). This ensures that resources are focused on key points and improves the efficiency and accuracy of the overall measurement response.

[0030] Furthermore, in the method provided in the embodiment of the application, obtaining the monitoring point sequence information further includes:

[0031] Based on the N priority factor coefficient sets, N priority coefficient matrices are generated; based on the priority factor evaluation system, indicator data mining is performed to obtain a priority factor evaluation indicator data set; according to the priority factor evaluation indicator data set, PCA weight allocation is performed on the monitoring point priority factors to generate a priority factor weight matrix; weighted calculation is performed on the N priority coefficient matrices and the priority factor weight matrix to obtain N monitoring priority coefficients, and the N monitoring points are prioritized based on the N monitoring priority coefficients to obtain the monitoring point sequence information.

[0032] In the present embodiment, N priority coefficient matrices are first generated using a matrix reconstruction method based on the obtained N priority factor coefficient sets. Specifically, the normalized scores of each monitoring point in the three dimensions of criticality, urgency, and monitoring difficulty are filled into a structured two-dimensional matrix in a unified order. The rows of this matrix represent the monitoring point numbers, and the columns correspond to the priority factors, forming an N×3 numerical matrix, which facilitates unified linear operations in subsequent calculations.

[0033] After obtaining N priority coefficient matrices, based on the priority factor evaluation system, indicator data mining was conducted on existing monitoring point evaluation data. Pearson correlation analysis was then used to identify statistical correlations between different priority factors. This method calculates the Pearson correlation coefficient between any two columns of data (i.e., two factors) to determine the degree of linear correlation, thereby clarifying which factors have a high degree of shared explanatory power and which factors are relatively independent. For example, in the above matrix, if the correlation coefficient between criticality and urgency is 0.85, it indicates a high correlation between the two, while monitoring difficulty has a weak correlation with the other factors, indicating that it may represent an independent dimensional feature.

[0034] Next, principal component analysis (PCA) was applied to the priority factor evaluation indicator dataset for feature extraction and dimensionality optimization. In PCA analysis, the covariance matrix of the priority coefficient matrix was first constructed, and the principal component vectors were solved using eigenvalue decomposition. The principal component loading values ​​represent the projection weights of each original evaluation factor onto that principal component direction. Based on the loading values, the principal component that explains the greatest variance was extracted and normalized to form the priority factor weight matrix. For example, among the three factors mentioned above, if the first principal component explains 80% of the variance, its corresponding normalized loading values ​​would be (0.6, 0.5, 0.3). This vector would be used as the final weight distribution for the calculation of the comprehensive score.

[0035] Subsequently, matrix multiplication is used to linearly weight the aforementioned priority factor weight matrix and the N priority coefficient matrices. Specifically, the inner product operation is performed on the priority score vector for each monitoring point and the weight vector, resulting in a scalar value that serves as the final monitoring priority coefficient for that point. For example, point A's score vector is (0.9, 0.7, 0.4). Multiplying and summing the scores by the weight vector (0.6, 0.5, 0.3) yields a priority coefficient of 0.9 × 0.6 + 0.7 × 0.5 + 0.4 × 0.3 = 1.01. This is repeated for the other monitoring points, completing the calculation of the N priority coefficients.

[0036] Finally, a sorting operation is performed based on the obtained N monitoring priority coefficients, and a quick sorting algorithm is used to sort all monitoring points from high to low according to the priority coefficients to form monitoring point sequence information.

[0037] Furthermore, in the method provided in the embodiment of the application, generating monitoring area information further includes:

[0038] The boundary of the monitoring area is determined with the theoretical coordinate information as the center and the preset error range as the radius, and the boundary of the monitoring area is a three-dimensional cube area; the grid type and grid density are determined according to the regional measurement requirements and model complexity; the boundary of the monitoring area is spatially divided and reconstructed according to the grid type and grid density to generate a monitoring area model; the monitoring area model is verified and optimized based on the actual distribution of monitoring points and environmental conditions, and the monitoring area information is output and generated.

[0039] In an embodiment of the present application, the theoretical coordinate information of each monitoring point is first used as the center, and a three-dimensional monitoring boundary is constructed based on a preset error range. This process adopts a spatial bounding box generation method, with the theoretical coordinates as the center point of the bounding box and the error range as the boundary extension value in the three-dimensional direction. The ± error value is expanded outward in the x, y, and z directions to generate a minimum enclosing cube. This cube is defined as the boundary of the monitoring area to ensure that subsequent recognition searches for target points within a reasonable range, reducing the error rate and computational redundancy. For example, when the theoretical coordinates are (1.0, 2.0, 1.5) meters and the error range is set to 0.3 meters, the constructed cube boundary range is x∈[0.7, 1.3], y∈[1.7, 2.3], and z∈[1.2, 1.8].

[0040] After determining the boundaries, the grid type and density are determined using an empirical threshold method, taking into account the current monitoring area's requirements for measurement accuracy and model complexity. Specifically, a cubic grid is selected as the spatial discretization structure based on the target accuracy level, and the unit space division granularity is automatically set based on the grid density standard in the empirical database that matches the accuracy requirements. For example, in a centimeter-level measurement accuracy scenario, the system will set the grid density to one cell per 0.01 meter; if the accuracy is at the millimeter level, it will be set to one cell per 0.001 meter, thus ensuring that the grid modeling can meet the spatial resolution requirements of downstream coordinate solving and image recognition.

[0041] After obtaining the boundary and grid parameters, a regular gridding algorithm is used to spatially reconstruct the monitoring area boundary. This algorithm divides the cube area into equal-spaced x, y, and z axes, generating a set of three-dimensional voxel units, each representing a basic component of the monitoring area model. The spatial coordinate range of each unit is calculated sequentially, and a complete three-dimensional grid model is constructed from this. This accurately divides the entire monitoring area into regular, clearly numbered sub-areas, providing spatial support for subsequent image acquisition perspective adjustment and path planning.

[0042] After the initial grid model is constructed, the monitoring area model is verified and optimized using a model evaluation algorithm. This process utilizes an on-site image comparison method. This involves using a telescope to capture images of the current area, projecting the monitoring area model onto the image coordinate system, and comparing it with the target locations in the actual image to determine whether the model's spatial extent covers the actual target. If boundary offsets, recognition errors, or spatial occlusions are detected, the grid boundary extent is adjusted or density parameters are re-divided until the model closely matches the on-site image data. Ultimately, the output monitoring area information represents the optimized regional structure through 3D modeling and image verification.

[0043] Step S200: The built-in high-precision navigation module automatically navigates to the monitoring area information, fixes a target icon on the target monitoring point, and uses the target icon as a reference to acquire target monitoring image information of the monitoring area information through a high-precision telescope.

[0044] In this embodiment, after acquiring the monitoring area information for the target monitoring point, the tracking and measurement robot first initiates an automated navigation process using its built-in high-precision navigation module to navigate to the monitoring area where the target monitoring point is located. This navigation module combines GPS positioning with attitude sensors (such as gyroscopes and accelerometers) to achieve navigation control. It plans a route based on the theoretical coordinates of the monitoring point, controls the robot's precise movement along the set path, and makes real-time direction corrections to ensure smooth entry into the target monitoring area.

[0045] Upon reaching the monitoring area, the robot affixes a target icon at the target location. This icon is a readily recognizable marking pattern, such as a contrasting black-and-white checkered pattern or a coded identification symbol (such as an AprilTag), which allows for accurate location using image recognition algorithms. The icon is automatically attached or placed on a surface near the monitoring point by the robot's robotic arm or mounting device, ensuring its stable visibility in the image.

[0046] Once the target icons are deployed, the robot activates its high-precision telescope to capture the monitored area. The telescope includes a high-resolution camera and an adjustable lens. By controlling the lens's direction and focal length, the robot focuses and captures an image that fully captures the target icon and surrounding features for subsequent image analysis and coordinate calculation.

[0047] The image data finally obtained is the target monitoring image information of the monitoring area information, which contains clear target patterns, surrounding structures, time information when shooting, and positioning information.

[0048] Step S300: training and constructing a point tracking module, using the point tracking module to perform target positioning and coordinate extraction on the target monitoring image information, and determine target pixel coordinate information.

[0049] In an embodiment of the present application, a point tracking module is first trained and constructed. Based on a preset target icon, image samples containing the target are collected to form an initial target feature data set, and the data set is subjected to image enhancement processing, including rotation, scaling, brightness adjustment, occlusion simulation and other operations, to generate a more diverse expanded target feature data set. Subsequently, a convolutional neural network (CNN) is used to train the expanded data set to generate a target recognition subnetwork specifically for identifying target icons in images. On this basis, a target tracking subnetwork is further constructed to process the dynamic position changes of the target in continuous image frames. This subnetwork shares weights with the recognition subnetwork, that is, the same feature extraction layer is used to reduce model redundancy and improve tracking robustness. Finally, the recognition subnetwork and the tracking subnetwork are fused and connected to construct a complete point tracking module.

[0050] The collected target monitoring image information is input into the point tracking module. The module automatically processes the image, locates the specific position of the target icon in the image, and extracts and outputs the two-dimensional coordinate value of the target center, that is, the target pixel coordinate information.

[0051] Furthermore, in the method provided in the embodiment of the application, the training and construction point tracking module further includes:

[0052] Feature data is collected based on the target icon to obtain a target feature data set, and image data enhancement is performed on the target feature data set to obtain an expanded target feature data set; recognition training and verification optimization are performed on the expanded target feature data set using a convolutional neural network to generate a target recognition subnetwork; based on the target recognition subnetwork, a target tracking subnetwork is constructed, and the target tracking subnetwork and the target recognition subnetwork are shared weight networks; the target tracking subnetwork and the target recognition subnetwork are merged and connected to construct the point tracking module.

[0053] In the embodiment of the present application, firstly, based on the target icon used, the target images under different angles, lighting, distances and background conditions are collected by the camera, and the initial target feature data set is established using image acquisition technology. During the acquisition process, typical environmental scenes are selected, such as indoor, outdoor, strong light and occlusion conditions, and the target is imaged multiple times to ensure that the collected images can fully cover the various variations in actual applications and provide comprehensive sample support for subsequent model training. After obtaining the preliminary image samples, the target feature data set is expanded using image enhancement methods. The specific operations include using image processing tools such as OpenCV to perform standard enhancement operations such as image rotation (such as ±30 degree range), scaling (such as ±20%), brightness adjustment, and noise addition, and introducing processing methods such as occlusion simulation and background interference superposition to form an expanded target feature data set with rich posture changes and environmental interference features, so that the training model has the ability to maintain recognition robustness in complex scenes.

[0054] A convolutional neural network (CNN) was then used to train the expanded dataset for recognition, building a target recognition subnetwork. This process employs a typical object detection architecture, such as YOLO. By inputting an image and annotated target locations, multiple convolutional layers extract image features such as edges and corners. Combining classification and regression layers, the network outputs the target's category and location. During training, the cross-entropy loss function is used to optimize the classification results, and the IoU loss function is used to adjust the localization bounding box to ensure that the network can consistently output the accurate location of the target within the image.

[0055] After the recognition subnetwork is trained, a target tracking subnetwork is constructed using a correlation filter tracking method to continuously track targets in video streams or continuous images. This method uses a correlation filter to calculate the response of image regions. By sliding a template region within the current frame and matching it with the target feature template in the previous frame, the maximum response position is determined, thereby tracking the target's position in the time series. To improve consistency, the tracking subnetwork and the recognition subnetwork share a weight network, using the same convolutional feature extraction structure and achieving unified parameters to ensure consistent and efficient recognition and tracking results.

[0056] Finally, the trained target recognition subnetwork and target tracking subnetwork are structurally docked, and a complete point tracking module is constructed through network structure splicing and output fusion.

[0057] Furthermore, in the method provided in the embodiment of the application, the step of obtaining the expanded target feature dataset further includes:

[0058] Image enhancement factor conditions are obtained, wherein the image enhancement factor conditions include lighting conditions, background complexity, and target icon posture changes; change parameters are designed for each factor condition in the image enhancement factor conditions to generate a target image change matrix set; and image enhancement and expansion are performed on the target feature data set based on the target image change matrix set to obtain an expanded target feature data set.

[0059] In an embodiment of the present application, in order to improve the point tracking module's ability to recognize target icons in complex environments, a systematic image enhancement process is performed on the initially collected target feature data set to construct a training sample set with diversity and robustness. Specifically, first, by analyzing the possible changing factors that may occur during the image acquisition process in the actual monitoring scene, the image enhancement factor conditions are obtained, which mainly include three categories: lighting conditions, background complexity, and changes in the posture of the target icon. Among them, lighting conditions refer to changes in brightness and contrast caused by natural light or artificial light sources during image acquisition. Background complexity reflects the degree of interference in areas other than the target in the image, such as the presence or absence of debris, line textures, etc., and changes in the posture of the target icon include rotation, scaling, tilting, and other angle and size changes that may occur in icons in actual applications.

[0060] For each of the aforementioned enhancement factors, a parameter variation design is performed. This involves establishing a numerical variation range and rules for each factor, thereby forming a controllable model for target image variation. For example, in terms of lighting conditions, the brightness adjustment range is set to ±30%, and the contrast adjustment range is ±20%. In terms of background complexity, a combination of different textures or noise layers is designed. In terms of posture variation, parameters such as the icon rotation angle is set to ±45°, the scaling ratio is set to 80%-120%, and the affine distortion angle is set to no more than ±10°. These parameter variation combinations are arranged in a regular manner to form a multidimensional variation reference structure. These parameters are then encoded into a target image variation matrix set, where each row represents a complete parameter configuration for a set of image enhancement operations.

[0061] After constructing the matrix set, image augmentation is performed on the initial target feature dataset based on the transformation matrix set. Specifically, the transformation combination in the transformation matrix is ​​applied to each image in the original dataset in sequence. For example, an image may be brightness-adjusted, rotated, and then perturbed with background textures, generating multiple versions of the enhanced image. This process is executed in batches using an image processing library (such as OpenCV), ultimately generating an exponentially expanded and diverse set of images for the entire original dataset, which is then output as the augmented target feature dataset.

[0062] Step S400: Calculate the target coordinates based on the telescope camera parameters and the target pixel coordinate information to obtain the target actual space coordinates, and store the target actual space coordinates in sequence to generate N spatial point measurement results.

[0063] In the embodiments of this application, precise coordinate calculations are performed based on the telescope camera parameters and the extracted target pixel coordinate information. Specifically, the parameters of the telescope camera are first analyzed to extract key parameters for imaging and coordinate transformation, including the intrinsic parameter matrix (such as focal length and principal point position), distortion coefficients (including radial and tangential distortion), and the extrinsic parameter matrix (representing the rotation and translation relationship between the camera coordinate system and the world coordinate system). These parameters can all be obtained in one go using existing camera calibration methods such as the Zhang Zhengyou calibration method and configured during the system initialization phase.

[0064] After obtaining the camera parameters, the target pixel coordinates extracted from the image are corrected for image distortion based on the distortion coefficients to eliminate geometric distortion caused by lens non-idealities during the imaging process. The resulting coordinates are the corrected target pixel coordinates, representing the target's position under an ideal, undistorted imaging model. These corrected coordinates are then transformed and mapped based on the intrinsic parameter matrix. Using a pinhole camera model or projection transformation, these coordinates are converted into three-dimensional coordinates in the camera coordinate system. These are the target camera three-dimensional coordinates, reflecting the target's position relative to the camera.

[0065] Then, the extrinsic parameter matrix is ​​used to perform spatial transformation on the camera's three-dimensional coordinates, converting them from the camera coordinate system to a unified world coordinate system to obtain the final actual spatial coordinates of the target.

[0066] After completing the image recognition and coordinate calculation of all monitoring points, the actual spatial coordinates of each target are stored in order according to the monitoring point sequence information, and finally N spatial point measurement results are generated.

[0067] Furthermore, in the method provided in the embodiment of the application, obtaining the actual spatial coordinates of the target further includes:

[0068] Coordinate influencing parameters of the telescope camera parameters are extracted to determine an intrinsic parameter matrix, a distortion coefficient, and an extrinsic parameter matrix; radial distortion and tangential distortion correction are performed on the target pixel coordinate information based on the distortion coefficient to obtain corrected target pixel coordinates; the corrected target pixel coordinates are transformed and mapped according to the intrinsic parameter matrix to obtain the three-dimensional coordinates of the target camera; and the three-dimensional coordinates of the target camera are transformed and mapped based on the extrinsic parameter matrix to determine the actual spatial coordinates of the target.

[0069] In an embodiment of the present application, the telescope camera is first calibrated to extract the coordinate influencing parameters used for coordinate transformation, including the intrinsic parameter matrix, distortion coefficients, and extrinsic parameter matrix. The intrinsic parameter matrix is ​​mainly obtained by shooting a series of known structure images with a camera calibration plate (such as a chessboard), and automatically calculating using visual tools such as OpenCV to obtain parameters such as focal length and principal point position, which are used to describe the relationship between pixels and the geometric center of the image. The distortion coefficients are also extracted simultaneously during the calibration process, usually including three radial distortion coefficients and two tangential distortion coefficients, which are used to correct subsequent image distortion. The extrinsic parameter matrix is ​​obtained by collecting images of reference points at different positions of the camera in space, and combining with a total station or laser rangefinder for position calibration to obtain the spatial orientation and position of the camera.

[0070] After obtaining the above parameters, the identified target pixel coordinates are processed. First, based on the extracted distortion coefficients, image dedistortion methods are used to correct radial and tangential distortion of the pixels. For example, a target at the edge of an image may have an original pixel position of (1450, 980). After radial distortion correction, the position may be adjusted to (1422, 970). After tangential distortion correction, the position is finally adjusted to (1410, 965), forming the corrected target pixel coordinates. This correction process accurately restores the geometric deviations in the image caused by the lens by reconstructing the imaging geometry, and is a key step in controlling image coordinate accuracy.

[0071] After calibration is complete, the corrected pixel coordinates are normalized based on the extracted intrinsic parameter matrix by obtaining the camera's horizontal focal length, vertical focal length, and principal point coordinates. Combined with the target depth coordinates determined by the laser ranging module or known structural dimensions, the normalized pixel coordinates are normalized to obtain the target's standardized 2D coordinates within the camera's field of view. Subsequently, these normalized coordinates are multiplied by the target depth coordinates to obtain the target's 2D camera coordinates. These coordinates are then combined with the depth coordinates to obtain the target's 3D camera coordinates within the camera's coordinate system.

[0072] The camera's extrinsic matrix is ​​then used to extract the rotation matrix and translation vector, converting the aforementioned 3D camera coordinates into homogeneous coordinate format. The rotation matrix is ​​multiplied by the homogeneous coordinates and then summed with the translation vector to complete the spatial transformation from the camera coordinate system to the world coordinate system. Ultimately, the target's actual spatial coordinates within the global measurement environment are calculated.

[0073] Furthermore, in the method provided in the embodiment of the application, the step of obtaining the three-dimensional coordinates of the target camera further includes:

[0074] According to the intrinsic parameter matrix, the horizontal focal length, the vertical focal length and the horizontal principal point coordinates are obtained, and the target depth coordinates are obtained at the same time; based on the horizontal focal length, the vertical focal length and the horizontal principal point coordinates, the corrected target pixel coordinates are normalized to obtain normalized target pixel coordinates; the product of the target depth coordinates and the normalized target pixel coordinates is used as the two-dimensional coordinates of the target camera, and the two-dimensional coordinates of the target camera and the target depth coordinates are combined to obtain the three-dimensional coordinates of the target camera.

[0075] In an embodiment of the present application, the horizontal focal length, vertical focal length, and horizontal principal point coordinates are first extracted from the camera's intrinsic parameter matrix. These parameters are obtained through a standard camera calibration process, such as using a checkerboard image sequence and using the Zhang Zhengyou calibration method to complete the reverse deduction of the focal length and principal point to ensure that the parameters have sufficient geometric accuracy. Taking a set of conventional calibration results as an example, the horizontal focal length is set to 1000 pixels, the vertical focal length is set to 1000 pixels, and the horizontal principal point coordinates are set to (960, 540). Next, the target depth coordinates corresponding to the target position are obtained. The depth information is accurately measured by a laser rangefinder (such as a TOF laser sensor), or directly provided by structural data in a scene where the target height is known. Based on this example, the target depth coordinate is set to 5.0 meters.

[0076] After preparing the internal depth data, the calibrated target pixel coordinates are normalized based on the horizontal focal length, vertical focal length, and horizontal principal point coordinates. Assuming the calibrated target pixel coordinates are (u=1250, v=980), the normalization calculation process is x=(1250-960) / 1000=0.29, y=(980-540) / 1000=0.44. Through normalization, the normalized target pixel coordinates are obtained as (0.29, 0.44), which normalizes the image coordinates to a relative coordinate system aligned with the camera optical axis for the next calculation.

[0077] The product of the target depth coordinate and the normalized target pixel coordinate is then used as the target camera's 2D coordinates: X = 0.29 × 5.0 = 1.45 meters, Y = 0.44 × 5.0 = 2.20 meters, representing the target's horizontal and vertical spatial positions. The resulting target camera 2D coordinates are (1.45, 2.20) meters.

[0078] Finally, the target camera 2D coordinates and the target depth coordinates are combined to obtain the target camera 3D coordinates, which are (1.45, 2.20, 5.0) meters. These coordinates are the target camera 3D coordinates, representing the precise location of the target in the 3D space coordinate system with the camera as the origin.

[0079] Furthermore, in the method provided in the embodiment of the application, the determining of the actual spatial coordinates of the target further includes:

[0080] According to the external parameter matrix, a rotation matrix and a translation vector are obtained, and the three-dimensional coordinates of the target camera are converted into target homogeneous coordinates; based on the rotation matrix and the translation vector, the target homogeneous coordinates are multiplied and summed to determine the actual space coordinates of the target.

[0081] In an embodiment of the present application, first, the system obtains a rotation matrix and a translation vector based on the extrinsic parameter matrix. The extrinsic parameter matrix is ​​used to describe the position and posture information of the camera relative to the world coordinate system. The system collects multiple target images of known positions during the camera calibration phase and uses a single perspective point estimation method (PnP) to extract the spatial correspondence between the image points and the world coordinate points, thereby calculating the rotation matrix and the corresponding translation vector. The rotation matrix is ​​used to describe the directional transformation of the camera around each axis in space, and the translation vector represents the displacement of the camera coordinate origin in the world coordinate system.

[0082] Next, the target camera's 3D coordinates are converted to target homogeneous coordinates. To achieve unified spatial transformation operations, a coordinate extension method is used to add a fixed value to the original 3D coordinates, forming a 4D coordinate representation. This coordinate representation is widely used in 3D computer vision and can integrate rotation and translation operations into a single linear transformation, making subsequent processing more concise and unified.

[0083] The target's homogeneous coordinates are then multiplied and summed based on the rotation matrix and translation vector. In this step, the rotation matrix and translation vector are concatenated to form a complete spatial transformation structure, which is then evaluated item by item with the target's homogeneous coordinates. Specifically, the target's coordinates are substituted into each set of rotation coefficients and the corresponding translation value added to determine the target's horizontal, vertical, and height position in global 3D space.

[0084] Finally, the three coordinate values ​​in the actual space are extracted from the calculation results to determine the actual space coordinates of the target.

[0085] In the embodiments of the present application, in summary, the embodiments of the present application have at least the following technical effects:

[0086] This application obtains N monitoring points, prioritizes the N monitoring points, obtains monitoring point sequence information, presets the monitoring acquisition frequency, inputs the theoretical coordinate information and preset error range of the target monitoring point in sequence according to the monitoring point sequence information, and generates monitoring area information based on the theoretical coordinate information and the preset error range; a built-in high-precision navigation module automatically navigates to the monitoring area information, fixes a target icon on the target monitoring point, and uses the target icon as a reference to acquire target monitoring image information of the monitoring area information through a high-precision telescope; trains and constructs a point tracking module, uses the point tracking module to perform target positioning and coordinate extraction on the target monitoring image information, and determines the target pixel coordinate information; calculates the target coordinates based on the telescope camera parameters and the target pixel coordinate information to obtain the actual spatial coordinates of the target, and stores the actual spatial coordinates of the target in sequence to generate N spatial point measurement results. The present invention solves the technical problems in the existing technology of difficult to guarantee monitoring frequency, low measurement efficiency and insufficient positioning accuracy. By combining automatic navigation, image recognition and spatial coordinate calculation, multiple monitoring points are identified and positioned automatically at high frequency, achieving the technical effect of improving the efficiency and accuracy of spatial point tracking measurement.

[0087] Embodiment 2 is based on the same inventive concept as the multifunctional automated spatial point tracking and measurement method in the above embodiment. Figure 2 As shown, the present application provides a multifunctional automated spatial point tracking and measurement device. The device and method embodiments in the present application are based on the same inventive concept. The device includes:

[0088] The monitoring area information acquisition module 11 is used to acquire N monitoring points, prioritize the N monitoring points, obtain monitoring point sequence information, preset the monitoring acquisition frequency, input the theoretical coordinate information and preset error range of the target monitoring point in sequence according to the monitoring point sequence information, and generate monitoring area information based on the theoretical coordinate information and the preset error range; the monitoring image information acquisition module 12 is used to automatically navigate to the monitoring area information with a built-in high-precision navigation module, fix a target icon on the target monitoring point, and use the target icon as a reference to acquire the target monitoring image information of the monitoring area information through a high-precision telescope; the coordinate information confirmation module 13 is used to train and construct a point tracking module, use the point tracking module to perform target positioning and coordinate extraction on the target monitoring image information, and determine the target pixel coordinate information; the point measurement result confirmation module 14 is used to calculate the target coordinates based on the telescope camera parameters and the target pixel coordinate information to obtain the actual spatial coordinates of the target, and store the actual spatial coordinates of the target in sequence to generate N spatial point measurement results.

[0089] Furthermore, the device is also used to implement the following functions:

[0090] Preset monitoring point priority factors, wherein the monitoring point priority factors include criticality, urgency, and monitoring difficulty; establish a priority factor evaluation system based on the monitoring point priority factors, wherein the priority factor evaluation system includes specific evaluation indicators for the priority factors of each monitoring point; perform priority evaluation on the N monitoring points according to the priority factor evaluation system to obtain N priority factor coefficient sets; perform priority sorting on the N monitoring points based on the N priority factor coefficient sets to obtain monitoring point sequence information.

[0091] Furthermore, the device is also used to implement the following functions:

[0092] Based on the N priority factor coefficient sets, N priority coefficient matrices are generated; based on the priority factor evaluation system, indicator data mining is performed to obtain a priority factor evaluation indicator data set; according to the priority factor evaluation indicator data set, PCA weight allocation is performed on the monitoring point priority factors to generate a priority factor weight matrix; weighted calculation is performed on the N priority coefficient matrices and the priority factor weight matrix to obtain N monitoring priority coefficients, and the N monitoring points are prioritized based on the N monitoring priority coefficients to obtain the monitoring point sequence information.

[0093] Furthermore, the device is also used to implement the following functions:

[0094] The boundary of the monitoring area is determined with the theoretical coordinate information as the center and the preset error range as the radius, and the boundary of the monitoring area is a three-dimensional cube area; the grid type and grid density are determined according to the regional measurement requirements and model complexity; the boundary of the monitoring area is spatially divided and reconstructed according to the grid type and grid density to generate a monitoring area model; the monitoring area model is verified and optimized based on the actual distribution of monitoring points and environmental conditions, and the monitoring area information is output and generated.

[0095] Furthermore, the device is also used to implement the following functions:

[0096] Feature data is collected based on the target icon to obtain a target feature data set, and image data enhancement is performed on the target feature data set to obtain an expanded target feature data set; recognition training and verification optimization are performed on the expanded target feature data set using a convolutional neural network to generate a target recognition subnetwork; based on the target recognition subnetwork, a target tracking subnetwork is constructed, and the target tracking subnetwork and the target recognition subnetwork are shared weight networks; the target tracking subnetwork and the target recognition subnetwork are merged and connected to construct the point tracking module.

[0097] Furthermore, the device is also used to implement the following functions:

[0098] Image enhancement factor conditions are obtained, wherein the image enhancement factor conditions include lighting conditions, background complexity, and target icon posture changes; change parameters are designed for each factor condition in the image enhancement factor conditions to generate a target image change matrix set; and image enhancement and expansion are performed on the target feature data set based on the target image change matrix set to obtain an expanded target feature data set.

[0099] Furthermore, the device is also used to implement the following functions:

[0100] Coordinate influencing parameters of the telescope camera parameters are extracted to determine an intrinsic parameter matrix, a distortion coefficient, and an extrinsic parameter matrix; radial distortion and tangential distortion correction are performed on the target pixel coordinate information based on the distortion coefficient to obtain corrected target pixel coordinates; the corrected target pixel coordinates are transformed and mapped according to the intrinsic parameter matrix to obtain the three-dimensional coordinates of the target camera; and the three-dimensional coordinates of the target camera are transformed and mapped based on the extrinsic parameter matrix to determine the actual spatial coordinates of the target.

[0101] Furthermore, the device is also used to implement the following functions:

[0102] According to the intrinsic parameter matrix, the horizontal focal length, the vertical focal length and the horizontal principal point coordinates are obtained, and the target depth coordinates are obtained at the same time; based on the horizontal focal length, the vertical focal length and the horizontal principal point coordinates, the corrected target pixel coordinates are normalized to obtain normalized target pixel coordinates; the product of the target depth coordinates and the normalized target pixel coordinates is used as the two-dimensional coordinates of the target camera, and the two-dimensional coordinates of the target camera and the target depth coordinates are combined to obtain the three-dimensional coordinates of the target camera.

[0103] Furthermore, the device is also used to implement the following functions:

[0104] According to the external parameter matrix, a rotation matrix and a translation vector are obtained, and the three-dimensional coordinates of the target camera are converted into target homogeneous coordinates; based on the rotation matrix and the translation vector, the target homogeneous coordinates are multiplied and summed to determine the actual space coordinates of the target.

[0105] It should be noted that the order in which the embodiments of the present application are presented is for illustrative purposes only and does not necessarily represent the superiority or inferiority of the embodiments. Furthermore, the foregoing descriptions of specific embodiments of this specification are provided. The processes depicted in the accompanying drawings do not necessarily require the specific order or sequential sequence shown to achieve the desired results. In certain embodiments, multitasking and parallel processing are also possible or may be advantageous.

[0106] The above description is only a preferred embodiment of the present application and is not intended to limit the present application. Any modifications, equivalent replacements, improvements, etc. made within the spirit and principles of the present application shall be included in the scope of protection of the present application.

[0107] This specification and drawings are merely illustrative of the present application and are intended to cover any and all modifications, variations, combinations, or equivalents within the scope of this application. Obviously, those skilled in the art may make various modifications and variations to this application without departing from the scope of this application. Thus, this application is intended to include such modifications and variations as fall within the scope of this application and its equivalents.

Claims

1. A multifunctional automated spatial point tracking measurement method, characterized in that: The method comprises: Obtain N monitoring points, prioritize the N monitoring points to obtain monitoring point sequence information, preset a monitoring acquisition frequency, sequentially input theoretical coordinate information and a preset error range of the target monitoring point according to the monitoring point sequence information, and generate monitoring area information based on the theoretical coordinate information and the preset error range; The built-in high-precision navigation module automatically navigates to the monitoring area information, fixes a target icon on the target monitoring point, and uses the target icon as a reference to acquire target monitoring image information of the monitoring area information through a high-precision telescope; Training and constructing a point tracking module, using the point tracking module to perform target positioning and coordinate extraction on the target monitoring image information, and determine target pixel coordinate information; Calculating the target coordinates based on the telescope camera parameters and the target pixel coordinate information to obtain the target's actual spatial coordinates, and sequentially storing the target's actual spatial coordinates to generate N spatial point measurement results; The training and construction of the point tracking module includes: Performing feature data collection based on the target icon to obtain a target feature data set, and performing image data enhancement on the target feature data set to obtain an expanded target feature data set; Using a convolutional neural network to perform recognition training and verification optimization on the expanded target feature data set to generate a target recognition subnetwork; Constructing a target tracking subnetwork based on the target recognition subnetwork, wherein the target tracking subnetwork and the target recognition subnetwork are shared weight networks; Merging and connecting the target tracking subnetwork and the target recognition subnetwork to construct the point tracking module; The expanded target feature dataset includes: Obtaining image enhancement factor conditions, wherein the image enhancement factor conditions include lighting conditions, background complexity, and target icon posture changes; Designing the change parameters of each element condition in the image enhancement element condition to generate a target image change matrix set; performing image enhancement and expansion on the target feature dataset in sequence based on the target image change matrix set to obtain an expanded target feature dataset; The step of obtaining monitoring point sequence information includes: Preset monitoring point priority factors, including criticality, urgency, and monitoring difficulty; Establishing a priority factor evaluation system based on the priority factors of the monitoring points, wherein the priority factor evaluation system includes specific evaluation indicators of the priority factors of each monitoring point; Performing priority evaluation on the N monitoring points according to the priority factor evaluation system to obtain N priority factor coefficient sets; Prioritizing the N monitoring points based on the N priority factor coefficient sets to obtain monitoring point sequence information; Prioritizing the N monitoring points based on the N priority factor coefficient sets to obtain monitoring point sequence information includes: Generating N priority coefficient matrices according to the N priority factor coefficient sets; Performing indicator data mining based on the priority factor evaluation system to obtain a priority factor evaluation indicator data set; Performing PCA weight allocation on the priority factors of the monitoring points according to the priority factor evaluation index data set to generate a priority factor weight matrix; The N priority coefficient matrices and the priority factor weight matrix are weighted to obtain N monitoring priority coefficients, and the N monitoring points are prioritized based on the N monitoring priority coefficients to obtain the monitoring point sequence information.

2. A multifunctional automated spatial point tracking and measurement method according to claim 1, characterized in that: The generating of monitoring area information includes: Taking the theoretical coordinate information as the center and the preset error range as the radius, the monitoring area boundary is determined, and the monitoring area boundary is a three-dimensional cube area; Determine the grid type and density based on regional measurement requirements and model complexity; Performing spatial division and reconstruction on the monitoring area boundary according to the grid type and grid density to generate a monitoring area model; The monitoring area model is verified and optimized based on the actual distribution of monitoring points and environmental conditions, and the monitoring area information is output and generated.

3. A multifunctional automated spatial point tracking and measurement method according to claim 1, characterized in that: The obtaining of the actual spatial coordinates of the target includes: Extracting coordinate influencing parameters of the telescope camera parameters to determine the intrinsic parameter matrix, distortion coefficient and extrinsic parameter matrix; Performing radial distortion and tangential distortion correction on the target pixel coordinate information based on the distortion coefficient to obtain corrected target pixel coordinates; Converting and mapping the correction target pixel coordinates according to the intrinsic parameter matrix to obtain the target camera three-dimensional coordinates; The three-dimensional coordinates of the target camera are transformed and mapped based on the external parameter matrix to determine the actual spatial coordinates of the target.

4. A multifunctional automated spatial point tracking and measurement method according to claim 3, characterized in that: The obtaining of the three-dimensional coordinates of the target camera includes: According to the internal parameter matrix, the horizontal focal length, vertical focal length and horizontal principal point coordinates are obtained, and the target depth coordinates are obtained at the same time; Normalizing the correction target pixel coordinates based on the horizontal focal length, the vertical focal length, and the horizontal principal point coordinates to obtain normalized target pixel coordinates; The product of the target depth coordinate and the normalized target pixel coordinate is used as the target camera two-dimensional coordinate, and the target camera two-dimensional coordinate and the target depth coordinate are combined to obtain the target camera three-dimensional coordinate.

5. The multifunctional automated spatial point tracking and measurement method according to claim 3, wherein: Determining the actual spatial coordinates of the target includes: According to the external parameter matrix, a rotation matrix and a translation vector are obtained, and the three-dimensional coordinates of the target camera are converted into target homogeneous coordinates; The target homogeneous coordinates are multiplied and summed based on the rotation matrix and the translation vector to determine the actual space coordinates of the target.

6. A multifunctional automated spatial point tracking and measuring device, characterized in that: The device is used to implement the multifunctional automated spatial point tracking and measurement method according to any one of claims 1 to 5, and the device includes: A monitoring area information acquisition module is used to acquire N monitoring points, prioritize the N monitoring points, obtain monitoring point sequence information, preset a monitoring acquisition frequency, sequentially input theoretical coordinate information and a preset error range of the target monitoring point according to the monitoring point sequence information, and generate monitoring area information based on the theoretical coordinate information and the preset error range; A monitoring image information acquisition module is configured to automatically navigate to the monitoring area information using a built-in high-precision navigation module, fix a target icon on the target monitoring point, and use the target icon as a reference to acquire target monitoring image information of the monitoring area information through a high-precision telescope; A coordinate information confirmation module is used to train and construct a point tracking module, and use the point tracking module to perform target positioning and coordinate extraction on the target monitoring image information to determine the target pixel coordinate information; A point measurement result confirmation module is used to calculate the target coordinates based on the telescope camera parameters and the target pixel coordinate information to obtain the actual spatial coordinates of the target, and to store the actual spatial coordinates of the target in sequence to generate N spatial point measurement results; The training and construction of the point tracking module includes: Performing feature data collection based on the target icon to obtain a target feature data set, and performing image data enhancement on the target feature data set to obtain an expanded target feature data set; Using a convolutional neural network to perform recognition training and verification optimization on the expanded target feature data set to generate a target recognition subnetwork; Constructing a target tracking subnetwork based on the target recognition subnetwork, wherein the target tracking subnetwork and the target recognition subnetwork are shared weight networks; Merging and connecting the target tracking subnetwork and the target recognition subnetwork to construct the point tracking module; The expanded target feature dataset includes: Obtaining image enhancement factor conditions, wherein the image enhancement factor conditions include lighting conditions, background complexity, and target icon posture changes; Designing the change parameters of each element condition in the image enhancement element condition to generate a target image change matrix set; The target feature data set is image enhanced and expanded in sequence based on the target image change matrix set to obtain an expanded target feature data set.

Images