Aerial photography surveying and mapping error compensation method and system

Abstract

The invention discloses an aerial photography surveying and mapping error compensation method and system, relates to the technical field of photogrammetry, and aims at dividing an inherent error and a random error according to an aerial surveying and mapping environment and an airborne surveying and mapping system, establishing a chaos attraction area based on error characteristic elements through chaos error distribution, and compensating a chaos attraction area along with an aerial surveying and mapping process. Obtaining surveying and mapping data and transmitting the data back to an aerial surveying and mapping system, performing partitioning based on the chaotic attraction region through element verification, introducing a region error condition and importing the region error condition into the error discriminator, performing first-order judgment on an inherent error and a random error, performing positive and negative error second-order judgment based on the random error, and performing negative and positive error judgment based on the random error. The method and the device are used for solving the technical problem of limited surveying and mapping precision caused by insufficient error compensation flexibility and adaptation degree in a complex surveying and mapping scene in the prior art, can effectively improve the scene integrating degree of error compensation in surveying and mapping measurement service, and further improve the surveying and mapping precision.

Description

Technical Field

[0001] The present invention relates to the technical field of photogrammetry, and in particular to an aerial photogrammetry error compensation method and system. Background Art

[0002] In the field of aerial photogrammetry, geospatial data is acquired through flying platforms equipped with high-precision imaging equipment and is widely used in scenarios such as land surveys, urban planning, and disaster monitoring. However, in the actual mapping process, errors are prone to occur due to a variety of internal and external factors. For example, errors stemming from inherent sensor defects, optical distortion, and platform attitude deviations can exhibit strong regularity; while errors caused by uncontrollable factors such as weather changes, flight disturbances, and cloud cover can exhibit unstable and nonlinear characteristics.

[0003] Existing error compensation methods primarily include parameter correction based on physical modeling, error fitting based on statistical methods, and hardware error suppression methods that rely on high-performance equipment. However, these methods generally suffer from poor adaptability, limited compensation accuracy, and difficulty distinguishing complex error sources. In particular, in dynamic flight environments and with multi-source heterogeneous data, error modeling and compensation strategies lack systematic coordination and intelligent optimization mechanisms.

[0004] Although some studies have attempted to introduce machine learning for error prediction, they mostly focus on a single type of error and lack a unified dynamic classification and directional compensation framework.

[0005] Therefore, it is urgent to propose a complete aerial photogrammetry error compensation system to solve the problems of insufficient flexibility and adaptability of error compensation in existing technologies, especially in complex mapping scenarios, so as to improve mapping accuracy and operation reliability. Summary of the Invention

[0006] The present application provides an aerial photogrammetry error compensation method and system, which is used to solve the technical problem in the prior art of insufficient flexibility and adaptability of error compensation in complex mapping scenarios, resulting in limited mapping accuracy.

[0007] In view of the above problems, the present application provides an aerial photogrammetry error compensation method and system.

[0008] In a first aspect, the present application provides an aerial photogrammetry error compensation method, the method comprising: dividing inherent errors and random errors based on the aerial photogrammetry environment and the airborne photogrammetry system, wherein the inherent errors include environmental inherent errors and system inherent errors; establishing a chaotic attraction zone based on error characteristic elements through chaotic error distribution, wherein the chaotic attraction zone maps the error to different attraction zones, and the exit end of each attraction zone is connected to an error discriminator; acquiring surveying data as the aerial photogrammetry process progresses and transmitting it back to the aerial photogrammetry system, performing partitioning based on the chaotic attraction zone through element verification, introducing zone error conditions and importing the error discriminator, performing first-order judgment on inherent errors and random errors, performing second-order judgment on positive and negative errors based on random errors, and determining compensated surveying data through error directional compensation and fitting, wherein discrimination constraints are performed by introducing dynamic synaptic weights.

[0009] In the second aspect, the present application provides an aerial photography surveying and mapping error compensation system, the system comprising: an error division unit, used to divide inherent errors and random errors based on the aerial surveying and mapping environment and the airborne surveying and mapping system, wherein the inherent errors include environmental inherent errors and system inherent errors; an attraction zone establishment unit, used to establish a chaotic attraction zone based on error characteristic elements through chaotic error distribution, wherein the chaotic attraction zone maps the error to different attraction zones, and the outlet end of each attraction zone is connected to an error discriminator; a surveying and mapping compensation unit, used to obtain surveying and mapping data as the aerial surveying and mapping process progresses and transmit it back to the aerial surveying and mapping system, perform partitioning based on the chaotic attraction zone through element verification, introduce zone error conditions and import the error discriminator, perform first-order judgment on inherent errors and random errors, perform second-order judgment on positive and negative errors based on random errors, and determine the compensated surveying and mapping data by error directional compensation and fitting, wherein the discrimination constraints are performed by introducing dynamic synaptic weights.

[0010] One or more technical solutions provided in this application have at least the following technical effects or advantages:

[0011] An embodiment of the present application provides an aerial photogrammetry error compensation method, which divides inherent errors and random errors based on the aerial photogrammetry environment and the airborne photogrammetry system, wherein the inherent errors include environmental inherent errors and system inherent errors; establishes a chaotic attraction zone based on error characteristic elements through chaotic error distribution, wherein the chaotic attraction zone maps errors to different attraction zones, and the outlet of each attraction zone is connected to an error discriminator; acquires surveying data as the aerial photogrammetry process progresses and transmits it back to the aerial photogrammetry system, performs partitioning based on the chaotic attraction zone through element verification, introduces zone error conditions and imports the error discriminator, performs first-order judgment on inherent errors and random errors, and performs second-order judgment on positive and negative errors based on random errors, and determines compensated surveying data through error directional compensation and fitting, wherein dynamic synaptic weights are introduced to perform discrimination constraints, which is used to solve the technical problem of insufficient flexibility and adaptability of error compensation in complex surveying scenarios in the prior art, resulting in limited surveying accuracy, and can effectively improve the scenario fit of error compensation in surveying and mapping metrology services and improve surveying accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0012] Figure 1 A schematic flow chart of an aerial photogrammetry error compensation method is provided for this application;

[0013] Figure 2 A schematic diagram of the process of dividing inherent error and random error in an aerial photogrammetry error compensation method is provided for this application;

[0014] Figure 3 A schematic diagram of the structure of an aerial photogrammetry error compensation system is provided for this application.

[0015] Description of the reference numerals: error division unit 11 , attraction region establishment unit 12 , surveying and mapping compensation unit 13 . DETAILED DESCRIPTION

[0016] This application provides an aerial photography surveying and mapping error compensation method and system to solve the technical problem in the prior art of insufficient error compensation flexibility and adaptability in complex surveying and mapping scenarios, resulting in limited surveying and mapping accuracy.

[0017] Example 1: Figure 1 As shown, the present application provides an aerial photogrammetry error compensation method, the method comprising:

[0018] S1: Based on the aerial mapping environment and the airborne mapping system, the inherent error and random error are divided, wherein the inherent error includes the environment inherent error and the system inherent error.

[0019] In an embodiment of the present invention, firstly, based on the combined structure of the operating environment involved in aerial surveying and mapping and the airborne surveying and mapping system, the error sources are systematically divided, thereby constructing an error basic classification system that can be used for compensation modeling.

[0020] Specifically, all errors that may affect surveying and mapping accuracy are divided into two first-level error categories according to their source attributes: inherent error and random error. Intrinsic error refers to the error component that exists stably and has a certain degree of predictability under a specific environment or equipment configuration, while random error refers to the error component caused by uncontrollable disturbances or instantaneous events, which is non-repeatable and difficult to predict in time and space.

[0021] Furthermore, the inherent errors are divided into environmental inherent errors and system inherent errors, depending on the conditions under which they occur. Environmental inherent errors refer to errors caused by the stable parameters of the external environment in which aerial surveying occurs. For example, during flight at constant high altitude, very low altitude, or in highly turbulent environments, the constant shift in atmospheric refractive index and the steady change in temperature gradient can lead to systematic distortion of aerial images. These errors are related to the stability of environmental conditions.

[0022] For example, when aerial mapping missions are carried out in low-pressure areas in winter, the ambient atmospheric density is lower than that of the conventional standard atmospheric model, which can easily cause changes in the optical path difference and thus cause image projection offset. This phenomenon is attributed to environmental inherent errors.

[0023] Intrinsic system errors refer to errors arising from the construction of the surveying and mapping platform or the characteristics of the sensor components, such as radial distortion of optical camera lenses, long-term drift of inertial navigation systems, and antenna position errors of GNSS equipment. These errors are highly repeatable and stable and can usually be initially compensated through factory calibration and sensor internal parameter adjustments.

[0024] For example, if the IMU component generates a gyro bias stability offset after long-term operation, the drift trend can be fitted through historical data for use in compensation processing.

[0025] In summary, this not only clarifies the root cause of the error, but also provides a classification basis for the subsequent implementation of dynamic error identification and targeted compensation path design, ensuring that the entire error compensation process has hierarchy and logical consistency.

[0026] Further, such as Figure 2 As shown, based on the aerial surveying and mapping environment and the airborne surveying and mapping system, inherent error and random error are divided. Step S1 of this application includes:

[0027] Taking the surveying altitude and meteorological environment as the aerial surveying environment, and based on the changes in error characteristic elements, the environment is divided to determine a first error classification, wherein the first error classification includes the environmental inherent error and environmental random error of each divided environment; for the airborne surveying system, a second error classification is determined, wherein the second error classification includes the system inherent error and system random error based on the scale of the surveying installed capacity; the first error classification and the second error classification are integrated to calibrate the inherent error and random error.

[0028] In an embodiment of the present invention, key external condition parameters obtainable during the surveying and mapping trip are first used as a basis for environmental division, specifically including the surveying and mapping flight altitude and the meteorological environment of the corresponding area.

[0029] Among them, the surveying altitude typically affects the sensor's imaging angle of view and atmospheric path length, changes in which are directly related to differences in image distortion and radiation inversion accuracy. The meteorological environment, which primarily includes factors such as temperature, humidity, air pressure, and wind speed, is a significant factor in atmospheric refraction, flight attitude disturbances, and image stability. Based on the dynamic changes in these environmental parameters, a classification structure for environmental conditions is constructed to identify the environmental state within different flight sections or operating periods, and to classify errors accordingly, thus forming the first error classification.

[0030] Specifically, the first error classification classifies the error components within each environmental category into two categories: inherent environmental error and random environmental error. Intrinsic environmental error refers to error factors that are highly stable and regular within a given environmental category. For example, when operating at altitudes above 3,000 meters, the optical path refraction error caused by the significant decrease in atmospheric density exhibits a relatively consistent shift trend. Random environmental error, on the other hand, refers to short-term nonlinear error fluctuations within the same environmental category caused by local disturbances, such as airflow fluctuations and sudden humidity changes.

[0031] Subsequently, an independent analysis was conducted on the airborne surveying and mapping systems that performed surveying and mapping tasks, and a second error classification was constructed based on the composition of the airborne equipment and the installed capacity.

[0032] In a specific embodiment, the installed capacity in the present invention refers to the number, type and layout of the surveying and mapping sensors carried, such as a single optical camera + inertial navigation system, or a multispectral + lidar + IMU combination system. Different configurations introduce different device response modes and signal coupling characteristics, which in turn affect the system error performance.

[0033] Under this category, system inherent error refers to the stability error caused by the hardware structure and its parameter settings, such as the fixed pixel offset in high-precision camera sensors or the long-term temperature drift deviation of IMU; system random error includes measurement error fluctuations caused by incidental factors such as short-term electromagnetic interference, mechanical vibration, signal loss of synchronization, etc. during the operation of surveying and mapping equipment.

[0034] In a preferred embodiment, according to the above classification requirements, historical aerial survey data is used to mine error characteristics and perform variability probability analysis to perform classification assessment.

[0035] Finally, the first error classification under the above-mentioned environmental dimension and the second error classification under the airborne system dimension are organically integrated to establish a full-domain error classification system, and the inherent errors and random errors that fully cover the operation process are calibrated.

[0036] In summary, a unified data structure support is provided for subsequent error judgment, matching and directional compensation strategies, which significantly improves the adaptability, accuracy and execution efficiency of the error compensation method.

[0037] S2: establishing a chaotic attraction zone based on error characteristic elements through chaotic error distribution, wherein the chaotic attraction zone maps the error to different attraction zones, and the exit end of each attraction zone is connected to an error discriminator.

[0038] In the embodiment of the present invention, in order to achieve precise identification and structured classification of complex error features, a chaotic error allocation strategy is adopted to construct a chaotic attraction zone system based on error feature elements.

[0039] Specifically, chaotic error allocation addresses the high-dimensional, multivariate, and nonlinear error data found in aerial surveying and mapping. By introducing chaotic mapping, the error data is transformed into an attractive distribution with fractal structure and local stability in the feature space, facilitating differentiated classification and dynamic response scheduling. The proposed method effectively reveals the nonlinear evolution of errors as they change with environmental perturbations and system coupling, avoiding the recognition failures of traditional linear classification methods in complex multivariate data.

[0040] In the specific implementation, a multidimensional feature vector space is first constructed based on error characteristic elements extracted from historical aerial survey mission data. These error characteristic elements include but are not limited to flight attitude parameters (such as pitch angle and roll angle), sensor bias parameters (such as CCD displacement and focal length drift), meteorological parameters (such as humidity gradient and wind shear), and geometric deformation parameters of survey and mapping images. Using this feature vector as the initial input, an evolutionary architecture based on chaotic mapping, such as the Logistic mapping, Tent mapping, or Lorenz system, is constructed. Multiple stable attraction regions are divided in the error evolution space, forming an error attraction mechanism with a dynamic aggregation effect.

[0041] Among them, each chaotic attraction region represents a type of region with a common error behavior pattern.

[0042] Examples include: systematic distortion-dominated, disturbance-type offset-dominated, or composite interference-dominated, etc. In order to further identify the specific error type and trigger the compensation path, the exit of each attraction zone is connected to an error discriminator for further judgment and analysis.

[0043] In a specific implementation, the error discriminator receives error aerial survey data that falls into the attraction area, identifies the type of input error based on the embedded recognition rules, and outputs the belonging category label and directional indicator (such as positive deviation or negative disturbance).

[0044] In summary, not only is the heterogeneity identification and clustering of multi-dimensional errors realized, but a precise mapping relationship between error types and subsequent compensation channels is also established, providing a pre-judgment basis for implementing advanced compensation control strategies such as branch selective compensation and synaptic weight adjustment, effectively improving the accuracy and response efficiency of error identification and correction.

[0045] Furthermore, a chaotic attraction region based on error characteristic elements is established. Step S2 of this application includes:

[0046] Historical aerial survey data is acquired, error characteristic elements are defined, and element error conditions are mined, wherein the error characteristic elements at least include flight elements, sensor elements, and atmospheric elements based on the aerial surveying and mapping environment and the airborne surveying and mapping system; the chaotic attraction zone is deployed by using the error characteristic elements as a threshold, by embedding the element error conditions and using conditional initialization as a processing standard.

[0047] In an embodiment of the present invention, to accurately construct the chaotic attraction region, it is first necessary to obtain historical aerial survey data covering multiple mission scenarios. The historical aerial survey data includes, but is not limited to, flight control information (such as flight attitude, speed, and altitude) recorded in multiple batches of flight missions, sensor status data (such as camera exposure parameters, IMU drift records, and GNSS reception status), and meteorological monitoring data (such as temperature and humidity, wind speed, and air pressure) within the corresponding time period. This data covers the entire chain of information from the external environment to the internal system during the execution of the surveying and mapping mission, providing a data foundation for subsequent error feature extraction and classification modeling.

[0048] Furthermore, based on the above historical aerial survey data, error characteristic elements that are representative of errors are defined and extracted.

[0049] In the present invention, the error characteristic factors include but are not limited to the following: first, flight factors, such as flight trajectory stability, pitch angle change rate, route deviation angle, etc. These parameters reflect the impact of the platform motion state on the image geometric accuracy; second, sensor factors, such as CCD pixel response non-uniformity, focal length thermal drift, IMU attitude solution error, etc., which reflect the inherent contribution of internal components of the surveying and mapping system to the measurement results; third, atmospheric factors, including atmospheric refractive index fluctuations, humidity mutations, wind shear index, etc., which directly affect the light propagation path and image clarity.

[0050] Subsequently, after completing the definition of the error characteristic elements, the error conditions of the elements are further explored based on the historical performance patterns and statistical laws corresponding to the above elements.

[0051] In a specific embodiment, the element error conditions include threshold-type rules, such as a wind speed change exceeding 3m / s within a certain flight altitude range will cause image distortion; or multi-parameter joint conditions, such as when the temperature is ≥30°C and the IMU usage time is ≥2 hours, the probability of attitude drift increases significantly, etc., constituting a set of prior knowledge rules for error behavior judgment.

[0052] Subsequently, the error characteristic elements are used as threshold variables and as boundary conditions for constructing chaotic attraction zones. The chaotic evolution space is partitioned and deployed using the mapping relationship between the error characteristic elements and their error conditions as initialization input. During the deployment process, each combination of elements and conditions is used as a processing criterion to trigger the activation of the corresponding attraction zone and form identifiable error evolution blocks. The boundaries of different attraction zones are defined by the error threshold triggering condition and have a certain fuzzy tolerance to accommodate actual error fluctuations.

[0053] In summary, the dynamic construction of the chaotic attraction region is completed under the constraints of multi-source characteristics and nonlinear boundary conditions, so that subsequent error samples can be automatically preliminarily classified before entering the discriminator, which significantly improves the error recognition efficiency and classification accuracy, and provides a prior distribution basis for accurate compensation path selection.

[0054] Furthermore, the construction of the error discriminator, step S2 of this application includes:

[0055] For the inherent error, a first normal compensation branch is established; for the random error, a second optional compensation branch is established with directional triggering and compensation; the first normal compensation branch and the second optional compensation branch are connected in parallel as an error compensation module and are subsequently connected to the error discriminator.

[0056] In an embodiment of the present invention, in order to implement a refined compensation strategy for different types of errors, an error compensation module with a dual-channel parallel structure is constructed after the error discriminator to adapt to the response characteristics and correction requirements of inherent errors and random errors respectively.

[0057] Specifically, a first normal compensation branch is established for stability compensation of the inherent error identified by the error discriminator. This first normal compensation branch presets a fixed compensation method based on the predictability and stability trend of the error. For example, real-time correction based on polynomial fitting, sensor calibration matrix, or empirical distortion model is used as the compensation mode. This compensation method uses historical sample training, with pre-compensation mapping samples as input and post-compensation mapping samples as output, and is constructed through supervised training until convergence.

[0058] Preferably, the first normal compensation branch of the present invention is suitable for regular error scenarios such as image system distortion repair and flight attitude deviation correction. This branch structure has the advantages of constant compensation parameters, fixed model structure, and fast computational convergence, making it suitable for deployment in the high-frequency main compensation path.

[0059] At the same time, a second optional compensation branch is constructed for the random error part identified by the error discriminator.

[0060] Among them, the random error has non-deterministic and dynamic drift characteristics, so this branch adopts a directional trigger and compensation mechanism: that is, through the error direction (such as offset direction) and scale information (such as error amplitude) output by the error discriminator, the optimal processing path is dynamically selected from multiple candidate compensation strategies. This branch can integrate a variety of filtering compensation algorithms (such as adaptive Kalman filtering, wavelet transform denoising, dynamic Bayesian correction, etc.), construct multiple compensation channels, and adopt a two-step architecture of directional trigger judgment and compensation channel triggering to quickly switch in combination with the current error characteristics. The training method is similar, and the training is supervised by sample data under the above logic until convergence.

[0061] For example, when the random error in directional trigger judgment manifests as image jitter caused by posture disturbance, the motion compensation channel will be automatically called for spatial alignment processing; if the error manifests as high-frequency noise output by the sensor, the filtering compensation channel will be activated for signal smoothing.

[0062] Furthermore, the first normal compensation branch and the second optional compensation branch form a parallel structure and are provided after the error discriminator as a unified error compensation module.

[0063] During implementation, the error compensation module receives error classification results and selects the corresponding branch to perform compensation processing based on the error type. The parallel deployment of the branch structure ensures that different types of errors can be processed simultaneously without blocking each other, improving the overall compensation response speed and processing throughput. It also provides a clear path separation basis for subsequent compensation data fusion and error residual analysis.

[0064] The method proposed in the present invention ensures that the error compensation module has high adaptability and robustness when facing complex surveying and mapping environments and various error scenarios.

[0065] Furthermore, during the aerial surveying process, before obtaining the surveying data, step S2 of this application includes:

[0066] An aerial surveying and mapping task is acquired, and a task-inherent error class is defined according to full-cycle characteristics of the task; and the first normal compensation branch is initialized according to the task-inherent error class.

[0067] In an embodiment of the present invention, in order to ensure that the error compensation process has adaptability and forward adjustability to a specific surveying and mapping task, firstly, the full cycle information of the current aerial surveying and mapping task is obtained before the task is executed.

[0068] Specifically, the aerial surveying and mapping mission usually includes multiple phased parameters, such as the operation time period, terrain characteristics of the flight area, operation altitude range, preset route planning, equipment configuration and sensor operation strategy. This type of information can be extracted and summarized through the task scheduling system, flight plan or surveying and mapping scheduling platform to form a task description data set.

[0069] After completing the task data acquisition, a pre-analysis of possible systematic errors is conducted based on the full-cycle characteristics of the task, and the task-inherent error class is defined accordingly. The task-inherent error class proposed in this application refers to a set of errors that are closely coupled with the task's preset environment and equipment configuration and are highly predictable.

[0070] For example, during long-duration aerial photography missions on plateaus, the attitude drift of the pre-set IMU will increase exponentially with rising temperatures. Alternatively, during midday operations in hot and humid regions, sensor focal length thermal drift and atmospheric refraction errors will significantly increase. These inherent errors typically have characteristics that can be defined a priori through historical statistics, experience, or physical derivation, forming a mission-level error prediction template.

[0071] Furthermore, based on the structured definition of the above-mentioned task-inherent error class, an initialization operation will be performed on the first normal compensation branch before the formal surveying and mapping begins to ensure the task scenario compatibility of the first normal compensation branch.

[0072] In one feasible embodiment, the initialization process includes: calling the error model parameters that match the inherent error class of the current task, loading the corresponding calibration matrix, fitting coefficients, or static correction curve; configuring the error matching threshold to limit the compensation range and accuracy; and setting the time synchronization reference and attitude reference required by the branch. This initialization ensures that the first normal compensation branch has a responsive mechanism in place at the beginning of the surveying and mapping task, and can quickly respond to subsequent error identification results.

[0073] In summary, by utilizing information from the entire mission cycle and predefining inherent error classes, a closed-loop structure of pre-measurement setting and execution during measurement is realized, enabling the compensation module to have scene perception capabilities, significantly improving the adaptability and compensation accuracy in cross-regional, cross-seasonal, and cross-platform surveying and mapping tasks.

[0074] S3: As the aerial surveying and mapping process progresses, the surveying and mapping data is acquired and transmitted back to the aerial surveying and mapping system. Partitioning is performed based on the chaotic attraction zone through element verification. Zone error conditions are introduced and imported into the error discriminator. First-order judgment of inherent error and random error is performed. Second-order judgment of positive and negative errors based on random error is performed. Compensated surveying and mapping data is determined by error-oriented compensation and fitting. Discrimination constraints are performed by introducing dynamic synaptic weights.

[0075] In an embodiment of the present invention, multi-source mapping data generated in the current aerial survey phase is acquired in real time during the execution of the aerial survey task. The mapping data includes but is not limited to aerial image data; flight attitude and position data, such as IMU and GNSS outputs; sensor operating status information and environmental monitoring data, such as atmospheric pressure, humidity, wind speed, etc.

[0076] Preferably, the above surveying and mapping data are transmitted back to the aerial surveying and mapping system center in a time-stamp synchronized manner as the input basis for subsequent error identification and compensation processes.

[0077] Furthermore, after receiving the surveying and mapping data, the key characteristic variables are first subjected to element verification. The element verification proposed in this application involves extracting preset error characteristic elements from the surveying and mapping data and, combined with the established chaotic attraction zone discrimination criteria, classifying the position of the current error sample in the feature space. This step automatically maps the surveying and mapping data to the corresponding chaotic attraction zones, forming multiple error partitions divided by characteristic clusters.

[0078] The data in each partition will be further introduced into the preset zone error conditions, which are based on historical experience and the error behavior threshold under a specific environment-equipment combination to enhance the robustness and accuracy of classification judgment.

[0079] Subsequently, after completing the partition mapping and conditional guidance, the set of surveying and mapping data is imported into the error discriminator to perform the first level of error type identification.

[0080] Specifically, the first-order decision operation is used to determine whether the current error sample is an inherent error or a random error. Specifically, by comparing the stability indicators of the characteristic dimensions (such as the trend, periodicity, and amplitude of the error change) and whether the regularity labels in the error zone conditions are met, long-term stable distortions and drifts can be classified as inherent errors, while disturbances that are non-repeatable or change dramatically within a short period can be classified as random errors.

[0081] Furthermore, for the data classified as random error in the first-order judgment, a second-order judgment is performed to identify the characteristics of the error in terms of direction and mode of action. The second-order judgment includes two dimensions: positive and negative error judgment and error scale judgment. Positive error judgment is used to distinguish whether the error causes an overestimation or underestimation of the observation value, often in combination with the direction of the residual mean offset. Error scale judgment is used to estimate the scope of the error in the target coordinate space or at the image pixel level.

[0082] In a preferred embodiment, the second-order recognition can be achieved by using the activation layer to output a direction label, or by extracting the feature gradient direction.

[0083] Subsequently, the first-order and second-order judgment results are structurally fused to obtain complete error classification information, which is passed as input to the compensation module to perform error directional compensation and fitting processing.

[0084] The first normal compensation branch performs static correction for inherent errors, while the second optional compensation branch triggers filtering or correction strategies based on the error direction and magnitude. This process ultimately generates compensated mapping data with optimized accuracy, which serves as the foundation for subsequent map generation and geographic information product output.

[0085] This application preferably proposes that, to improve the adaptability of the discriminator in long-term tasks and control the dynamic adjustment of error classification boundaries, this embodiment introduces a dynamic synaptic weight mechanism. Specifically, based on a neural-like structure, this mechanism assigns adjustable synaptic weights to each error discrimination path, automatically adjusting the weight distribution based on the expected matching degree of the current compensation data and the residual fluctuation.

[0086] For example, when a certain type of random error is continuously overestimated, the synaptic weight of the corresponding pathway will be down-regulated to suppress the discrimination bias of the classification path; while for sudden errors that are not fully recognized, the weight of the corresponding path will be up-regulated to enhance the response sensitivity.

[0087] In summary, through this dynamic adjustment mechanism, the error discriminator has the ability of adaptive learning, which enables it to learn from tasks and optimize in feedback, thereby improving the intelligence level and processing accuracy of the entire compensation system.

[0088] Furthermore, partitioning based on the chaotic attraction zone is performed, and zone error conditions are introduced. Step S3 of this application includes:

[0089] Acquire surveying and mapping data as the aerial surveying and mapping task progresses; transmit the surveying and mapping data back to the aerial surveying and mapping system, perform element verification, import data partitions based on chaotic attraction zones, perform dynamic condition calibration based on target attraction zones, and determine calibrated surveying and mapping data, wherein the calibrated surveying and mapping data includes partitioned surveying and mapping data-zone error conditions.

[0090] In an embodiment of the present invention, as the aerial surveying and mapping mission continues to advance, surveying and mapping data are collected and updated in real time. The surveying and mapping data include aerial image data corresponding to the current moment, attitude and position information (such as pitch angle, GPS coordinates, heading angle, etc.), sensor operating status (such as focal length change, voltage and current fluctuations) and external environmental parameters (such as wind speed, temperature, humidity).

[0091] During the optional execution process, the data is synchronously collected through a standardized interface and transmitted back to the aerial mapping system in real time according to the time series structure for error identification and dynamic analysis.

[0092] After receiving the surveying data, key parameters in the data are first subjected to element verification. This verification process, based on a pre-set table of error characteristic elements, selects characteristic indicators with significant error. Specifically, typical elements include flight attitude stability indicators (such as attitude angle change rate), sensor output noise coefficient, and environmental disturbance factors (such as humidity gradient and wind shear). Each identified element sample is then fed into the chaotic attraction region as an input feature vector.

[0093] Then, based on the extracted eigenvectors, data partitioning and importing based on chaotic attraction zones is performed. This partitioning and importing process involves matching the surveying and mapping data with the constructed attraction zone distribution according to its error characteristics, so that each data set is classified into the target attraction zone that best matches its evolution trajectory.

[0094] In the implementation of the present application, the division method of the chaotic attraction zone is based on the high-dimensional nonlinear mapping space formed by historical error samples. In different attraction zones, the error samples show similar dynamic aggregation and disturbance evolution trends, which facilitates subsequent classification, judgment and compensation scheduling.

[0095] Then, after completing the data partitioning, dynamic condition calibration is performed on the data entering each target attraction area.

[0096] The dynamic condition calibration proposed in this application refers to the quantitative correction or conditional marking of surveying and mapping data based on the error evolution characteristics of the current attraction zone and its corresponding zone error conditions (such as error offset direction, disturbance frequency threshold, influencing factor coupling rules, etc.).

[0097] In a preferred embodiment, the calibration process not only considers the static properties of the current sample, but also combines the aggregation rules of historical samples in the attraction area and the feedback results of known compensation strategies to add error condition labels to each surveying and mapping data, such as high-amplitude positive disturbances, environmentally triggered distortions, etc.

[0098] Ultimately, the mapping data after dynamic calibration is defined as calibrated mapping data. This data consists of two key components: the partitioned mapping data, which represents the original data successfully mapped to a specific chaotic attraction zone; and the zone error condition, which represents the set of dynamic error discrimination conditions that match this attraction zone.

[0099] The structured data format proposed in this application enables the active classification and regularized expression of aerial survey data in the nonlinear error space, providing a sufficient basis for the error discriminator's decision path scheduling and establishing a precise precondition framework for the execution logic of the compensation branch. This approach significantly enhances the compensation system's ability to perceive complex error dynamics and the accuracy of its discriminative response.

[0100] Furthermore, a first-order determination of the inherent error and the random error is performed, and a second-order determination of the positive and negative errors based on the random error is performed. Step S3 of the present application includes:

[0101] The calibration surveying and mapping data is imported into the error discriminator, and a first-order judgment based on inherent error and random error is performed to determine the first classification data; for the random error class in the first classification data, a second-order judgment based on positive error and negative error is performed with the zone error condition as a constraint to determine the second classification data, which includes direction judgment and scale judgment; the first classification data and the second classification data are integrated to determine the classified surveying and mapping data.

[0102] In the embodiment of the present invention, after the dynamic condition calibration of the surveying and mapping data is completed, the obtained calibrated surveying and mapping data is input into the constructed error discriminator to start executing the hierarchical discrimination process.

[0103] The following is a feasible construction method for the error discriminator provided in this application: deploy a first classification node based on inherent error and random error, and a second classification node based on positive error and negative error, and cascade the nodes to form a discriminant architecture. Using a sample training method, with raw sample data as input and classified identified sample data as output, the discriminant architecture is supervised and trained until convergence is achieved. In an optional training method, adversarial network training can be used.

[0104] Specifically, the error discriminator has a pre-set determination path for identifying error types. Combining the aforementioned chaotic attraction zone partitioning and zone error conditions, it first performs a first-order determination on the input data based on inherent and random errors. This first-order determination process is based on the evolutionary stability, periodicity, and offset trends of the error characteristic elements, combined with the historical error evolution trajectory recorded in the attraction zone and empirical rules, to identify whether the current sample error belongs to a predictable inherent error category (such as systematic lens distortion and constant sensor drift) or a random error category with large uncontrollable fluctuations (such as flight attitude jumps caused by wind field disturbances and image blur caused by sudden humidity changes).

[0105] Subsequently, after the determination is completed, the first classification data is output, which includes a set of surveying and mapping data marked as inherent errors and a set of surveying and mapping data marked as random errors. For the random error class in the first classification data, the internal secondary discrimination is further called to perform a second-order determination based on positive errors and negative errors. In the preferred embodiment provided by the present application, the second-order determination step uses the previously bound zone error conditions as constraints, for example, to determine whether there is a positive drift characteristic value in the attraction zone that exceeds a preset upper limit, or whether there is a negative disturbance amplitude that is higher than the historical interval mean, etc.

[0106] In the specific implementation of the second-order judgment, the first step is to perform a directional judgment to identify the direction of the error in the spatial or parameter dimension, distinguishing whether it causes the measurement value to float (such as the image center shifting to the far end) or sink (such as attitude error causing elevation to be underestimated). Then, a scale judgment is performed to assess the impact of the error on the surveying and mapping data. This is usually expressed in quantitative indicators such as pixel offset value, attitude angle deviation, and GNSS error radius. For example, if an image shows a yaw angle change of 1.2° due to wind shear, it corresponds to a positive error judgment and is determined to be a mesoscale disturbance interval.

[0107] Subsequently, after the directionality and scale identification is completed, the second classification data is output, wherein each classification data has a direction label, such as positive or negative; and an error intensity level, such as small, medium, large, etc.

[0108] Finally, the first classification data and the second classification data are structurally integrated to generate complete classified surveying and mapping data. Preferably, in this data structure, each surveying and mapping data sample not only clarifies the error source type (inherent or random), but also refines the directionality and scale of the random error, providing a decision-making basis for the subsequent selection of compensation paths and adjustment of weight coefficients, ensuring that the compensation mechanism can achieve optimal accuracy and execution efficiency.

[0109] Furthermore, the compensation surveying and mapping data is determined by error orientation compensation and fitting. Step S3 of this application includes:

[0110] For the classified surveying and mapping data, the inherent error class in the first classified data is input into the initialized first normal compensation branch, and normal compensation is performed to determine the first compensation data; for the second classified data, by executing class-directional compensation guidance, the second optional compensation branch is selectively triggered and compensated to determine the second compensation data; based on the first compensation data and the second compensation data, the compensated surveying and mapping data is determined.

[0111] In an embodiment of the present invention, after completing the error classification and identification of the aerial survey data, structured classified surveying and mapping data is obtained, which includes the first classification data (i.e., the inherent error class) and the second classification data (i.e., the random error class, with direction and scale labels attached) output after processing by the error discriminator. To achieve differentiated compensation processing for corresponding error types, the classification results are directed to different compensation branches, and targeted error correction is performed according to type. The specific implementation method is as follows:

[0112] First, the inherent error data within the first classification is fed into the first normal compensation branch, which has already completed the task initialization configuration. This compensation branch stores the error correction architecture and calibration parameters corresponding to the current surveying and mapping task scenario, such as optical system distortion correction, GNSS fixed offset correction, and IMU attitude linear drift compensation curves. Based on the error label of each input data item, the matching correction method is invoked, performing parameter substitution or incremental fitting operations to output the corrected high-precision aerial survey data as the first compensation data.

[0113] For example, if the constant lateral distortion offset caused by CCD non-uniformity corresponding to a certain image is +3 pixels, linear inverse interpolation will be called to reset the image points to generate the compensated coordinates.

[0114] Then, the random error data contained in the second classification data is guided into the second optional compensation branch, which has multiple optional compensation channels, each of which is suitable for different types of nonlinear and non-stationary disturbances.

[0115] For example, the Kalman filter channel is suitable for continuous-time disturbance compensation, the channel based on the time series convolutional network is suitable for short-term mutation feature recognition, and the Bayesian estimation channel is suitable for probability domain error reconstruction. Based on the directionality and scale level information contained in the second classification data, a class-oriented compensation guidance mechanism is executed to select the most suitable channel from multiple candidate channels and trigger compensation processing.

[0116] For example, for error samples with large positive disturbances and short-period fluctuations, the time domain filtering path is preferentially selected and a nonlinear threshold constraint strategy is added for error suppression and fitting, and finally the corrected second compensation data is output.

[0117] Subsequently, after the two compensation branches output their corresponding compensation results, the first and second compensation data are synchronously fused, performing timestamp-level data alignment and spatial position vector splicing to eliminate data splitting and coordinate mismatches that may arise during error processing. The unified data structure obtained after the fusion is completed is the final compensated mapping data. This data combines the consistency of error correction with the real-time performance of random perturbation cancellation, with significantly improved geometric accuracy and data stability. It can be directly used as the basic input for subsequent scenarios, such as aerial survey product generation, post-processing mapping, or 3D reconstruction, comprehensively improving the accuracy and application reliability of aerial surveying and mapping results.

[0118] Furthermore, step S3 of the present application includes: each error class corresponds to a synaptic weight; by evaluating the compensation mapping data, the unexpected compensation part is located; the synaptic weight of the unexpected compensation part is increased, and the synaptic weight of the expected compensation part is decreased, and the synaptic weight of the error discriminator is dynamically adjusted.

[0119] In an embodiment of the present invention, in order to enhance the adaptability of the error discriminator to complex dynamic mapping scenarios, a synaptic weight mechanism is introduced to simulate the self-learning ability of a neural network.

[0120] Specifically, within the error discriminator, each error discrimination path is assigned a separate synaptic weight. This weight characterizes the activation priority and response strength of the current path in the discrimination decision. Each synaptic weight establishes a one-to-one correspondence with a specific error category (such as system-specific error, environmental random error, and posture perturbation error), serving as a dynamic control parameter in the error discrimination logic.

[0121] After completing the output of the compensated surveying and mapping data, a retrospective analysis of the compensation effect is performed. For example, error residual analysis, accuracy index comparison, coordinate offset statistics, etc. are performed on the surveying and mapping data before and after compensation to extract areas where there is insufficient error compensation or misjudged paths, forming an identification result of the unexpected compensation part.

[0122] For example, if a certain type of error that is determined to be system distortion still has significant residual offset after normal compensation, the compensation failure part is marked as unexpected compensation.

[0123] Then, based on the recognition result, a directional dynamic adjustment is performed on the synaptic weight of the corresponding error path.

[0124] Specifically, for the error compensation guided by the judgment path that appears as an unexpected compensation part, its associated synaptic weight will be increased to increase the sensitivity and responsiveness of the path in subsequent similar inputs; on the contrary, for the expected compensation part with good compensation results and minimal residuals, the synaptic weight of the corresponding path will be decreased to prevent over-activation or misidentification of the judgment path.

[0125] In an optional embodiment, during execution, weight adjustment is implemented using an exponential smoothing or gradient descent mechanism to ensure the continuity and stability of synaptic weight updates and avoid structural oscillations caused by short-term fluctuations in the discriminator.

[0126] In summary, the error discriminator has the closed-loop capability of perception-compensation-feedback-re-optimization. After executing multiple tasks, it can gradually accumulate optimization experience, adjust the response mechanism of the discrimination path, and significantly improve the accuracy of error classification and the adaptability of compensation path selection, thereby realizing the adaptive error identification evolution mechanism in long-term tasks and meeting the intelligent needs of high-precision surveying and mapping data processing in complex flight environments.

[0127] The present application provides an aerial photography surveying and mapping error compensation method with the following technical effects:

[0128] 1. Error Classification and Chaotic Mapping: Aerial surveying and mapping errors are divided into inherent errors and random errors. Chaotic error allocation is used to construct attraction zones, mapping errors to different regions and achieving orderly classification of error characteristics. This systematically sorts out complex error sources, laying the foundation for precise compensation, avoiding blind error handling, and improving compensation efficiency. A dual-branch error discrimination architecture: A normal compensation branch for inherent errors and an optional compensation branch for random errors are constructed, and the error discriminator is connected in parallel to implement differentiated compensation strategies. Compensation logic is automatically matched according to error type to reduce ineffective compensation, ensuring stable correction of inherent errors and flexible handling of random errors. A task-driven initialization mechanism: Inherent error classes are defined based on the characteristics of aerial surveying and mapping tasks, and the normal compensation branch is initialized in advance to adapt to different task scenarios. Task-level customization of compensation strategies is achieved, reducing repeated calculations and improving surveying and mapping task execution efficiency and compensation adaptability.

[0129] 2. Dynamic Partitioning and Second-Order Determination: Surveying and mapping data is partitioned based on chaotic attraction zones, and first-order error type determination and second-order random error direction / scale determination are performed based on zone error conditions. This refines error analysis dimensions, accurately pinpoints error properties and characteristics, provides a quantitative basis for compensation strategies, and enhances targeted compensation. Dynamic Optimization of Synaptic Weights: Synaptic weights are assigned to each error class, and the weights of unexpected and expected compensation components are dynamically adjusted based on compensation results to optimize the error discriminator. This continuously improves error discrimination accuracy, iteratively optimizes the compensation model, and ensures the accuracy of surveying and mapping data.

[0130] Embodiment 2: Based on the same inventive concept as the aerial photography surveying and mapping error compensation method in the above embodiment, Figure 3 As shown, the present application provides an aerial photogrammetry error compensation system, the system comprising:

[0131] The error division unit 11 is used to divide the inherent error and the random error based on the aerial mapping environment and the airborne mapping system, wherein the inherent error includes the environment inherent error and the system inherent error;

[0132] An attraction zone establishment unit 12 is used to establish a chaotic attraction zone based on error characteristic elements by chaotic error distribution, wherein the chaotic attraction zone maps the error to different attraction zones, and an error discriminator is connected to the exit end of each attraction zone;

[0133] The surveying and mapping compensation unit 13 is used to obtain surveying and mapping data and transmit it back to the aerial surveying and mapping system as the aerial surveying and mapping process progresses. The unit performs partitioning based on the chaotic attraction zone through element verification, introduces zone error conditions and imports the error discriminator, performs first-order judgment on inherent error and random error, performs second-order judgment on positive and negative errors based on random error, and determines compensated surveying and mapping data through error-oriented compensation and fitting. The dynamic synaptic weight is introduced to perform discrimination constraints.

[0134] Furthermore, the error division unit 11 is used to perform the following steps: using the surveying altitude and meteorological environment as the aerial surveying environment, performing environmental division based on the changes in error characteristic elements, and determining a first error classification, wherein the first error classification includes the environmental inherent error and environmental random error of each divided environment; determining a second error classification for the airborne surveying system, wherein the second error classification includes the system inherent error and system random error based on the scale of surveying installed capacity; integrating the first error classification and the second error classification, and calibrating the inherent error and random error.

[0135] Furthermore, the attraction zone establishment unit 12 is used to perform the following steps: building a first normal compensation branch for the inherent error; building a second optional compensation branch for the random error with directional triggering and compensation; and running the first normal compensation branch and the second optional compensation branch in parallel as an error compensation module and post-connecting them to the error discriminator.

[0136] Furthermore, the attraction zone establishing unit 12 is configured to perform the following steps: obtaining an aerial surveying and mapping task, defining a task inherent error class according to the full cycle characteristics of the task; and initializing the first normal compensation branch according to the task inherent error class.

[0137] Furthermore, the surveying and mapping compensation unit 13 is used to perform the following steps: obtaining surveying and mapping data as the aerial surveying and mapping task progresses; transmitting the surveying and mapping data back to the aerial surveying and mapping system, performing element verification, importing data partitions based on the chaotic attraction zone, performing dynamic condition calibration based on the target attraction zone, and determining calibrated surveying and mapping data, wherein the calibrated surveying and mapping data includes partitioned surveying and mapping data-zone error conditions.

[0138] Furthermore, the surveying and mapping compensation unit 13 is used to perform the following steps: importing the calibrated surveying and mapping data into the error discriminator, performing a first-order judgment based on inherent error and random error, and determining first classified data; for the random error class in the first classified data, with the zone error condition as a constraint, performing a second-order judgment based on positive error and negative error, and determining second classified data, which includes direction judgment and scale judgment; integrating the first classified data and the second classified data to determine classified surveying and mapping data.

[0139] Furthermore, the surveying and mapping compensation unit 13 is used to perform the following steps: for the classified surveying and mapping data, the inherent error class in the first classified data is input into the initialized first normal compensation branch, and normal compensation is performed to determine the first compensation data; for the second classified data, by executing class-directional compensation guidance, the second optional compensation branch is selectively triggered and compensated to determine the second compensation data; and the compensated surveying and mapping data is determined based on the first compensation data and the second compensation data.

[0140] Furthermore, the attraction zone establishment unit 12 is used to perform the following steps: obtaining historical aerial survey data, defining error characteristic elements and mining element error conditions, wherein the error characteristic elements at least include flight elements, sensor elements, and atmospheric elements based on the aerial surveying and mapping environment and the airborne surveying and mapping system; using the error characteristic elements as the threshold, by building in the element error conditions and using conditional initialization as the processing standard, the chaotic attraction zone is deployed.

[0141] Furthermore, the mapping compensation unit 13 is configured to perform the following steps: each error class corresponds to a synaptic weight; an unexpected compensation part is located by evaluating the compensation mapping data; the synaptic weight of the unexpected compensation part is increased, and the synaptic weight of the expected compensation part is decreased, to dynamically adjust the synaptic weight of the error discriminator.

[0142] Through the above detailed description of an aerial photogrammetry error compensation method in this specification, those skilled in the art can clearly understand an aerial photogrammetry error compensation method and system in this embodiment. For the device disclosed in the embodiment, since it corresponds to the method disclosed in the embodiment, the description is relatively simple, and the relevant parts can be referred to the method part description.

[0143] The above description of the disclosed embodiments is intended to enable one skilled in the art to implement or use the present application. Various modifications to these embodiments will be readily apparent to one skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the present application. Therefore, the present application is not limited to the embodiments shown herein, but is intended to conform to the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for compensating aerial photogrammetry errors, characterized in that: The method comprises: Based on the aerial surveying and mapping environment and the airborne surveying and mapping system, the inherent error and random error are divided, wherein the inherent error includes the environmental inherent error and the system inherent error; By distributing chaotic errors, a chaotic attraction zone based on error characteristic elements is established, wherein the chaotic attraction zone maps the error to different attraction zones, and the exit of each attraction zone is connected to an error discriminator; As the aerial surveying and mapping process progresses, the surveying and mapping data is acquired and transmitted back to the aerial surveying and mapping system. Partitioning is performed based on the chaotic attraction zone through element verification. Zone error conditions are introduced and imported into the error discriminator. First-order judgment of inherent error and random error is performed. Second-order judgment of positive and negative errors based on random error is performed. Error-oriented compensation and fitting are used to determine compensated surveying and mapping data. Discrimination constraints are performed by introducing dynamic synaptic weights.

2. The aerial photogrammetry error compensation method according to claim 1, wherein: Based on the aerial surveying and mapping environment and the airborne surveying and mapping system, the inherent errors and random errors are divided into: Taking the surveying altitude and the meteorological environment as the aerial surveying environment, dividing the environment according to the change of the error characteristic elements, and determining a first error classification, wherein the first error classification includes the environmental inherent error and the environmental random error of each divided environment; Determining a second error classification for the airborne surveying and mapping system, wherein the second error classification includes a system inherent error and a system random error based on the scale of surveying and mapping installed capacity; The first error classification and the second error classification are integrated to calibrate the inherent error and the random error.

3. The aerial photogrammetry error compensation method according to claim 1, wherein: The construction of the error discriminator includes: Aiming at the inherent error, a first normal compensation branch is established; Aiming at the random error, a second optional compensation branch is built with directional triggering and compensation; The first normal compensation branch and the second optional compensation branch are connected in parallel as an error compensation module and are subsequently connected to the error discriminator.

4. The aerial photogrammetry error compensation method according to claim 3, wherein: As aerial surveying progresses, before obtaining surveying data, the following steps are required: Obtain aerial surveying and mapping tasks and define the inherent error classes of the tasks based on the full cycle characteristics of the tasks; The first normality compensation branch is initialized according to the task inherent error class.

5. The aerial photogrammetry error compensation method according to claim 4, wherein: Partitioning based on the chaotic attraction region and introducing a region error condition include: Acquiring surveying and mapping data as the aerial surveying and mapping mission progresses; The surveying and mapping data is transmitted back to the aerial surveying and mapping system, and through element verification, data partitioning import based on the chaotic attraction zone is performed, and dynamic condition calibration based on the target attraction zone is performed to determine the calibrated surveying and mapping data, wherein the calibrated surveying and mapping data includes the partitioned surveying and mapping data-zone error condition.

6. The aerial photogrammetry error compensation method according to claim 5, wherein: Perform first-order judgment on inherent error and random error, and perform second-order judgment on positive and negative errors based on random error, including: Importing the calibration surveying and mapping data into the error discriminator, performing a first-order judgment based on inherent error and random error, and determining first classification data; For the random error class in the first classification data, a second-order determination based on positive errors and negative errors is performed with the area error condition as a constraint to determine the second classification data, which includes direction determination and scale determination; The first classification data and the second classification data are integrated to determine classification surveying and mapping data.

7. The aerial photogrammetry error compensation method according to claim 6, wherein: Determine the compensated surveying and mapping data by error orientation compensation and fitting, including: For the classified surveying and mapping data, inputting the inherent error class in the first classified data into the initialized first normal compensation branch, performing normal compensation to determine first compensation data; For the second classified data, performing class-oriented compensation guidance, selectively triggering and compensating the compensation channel of the second optional compensation branch, and determining second compensation data; The compensated surveying and mapping data is determined according to the first compensation data and the second compensation data.

8. The aerial photogrammetry error compensation method according to claim 1, wherein: Establish a chaotic attraction zone based on error characteristic elements, including: Acquire historical aerial survey data, define error characteristic elements, and mine element error conditions, wherein the error characteristic elements at least include flight elements, sensor elements, and atmospheric elements based on the aerial surveying and mapping environment and the airborne surveying and mapping system; The chaotic attraction zone is deployed by taking the error characteristic element as a threshold, embedding the element error condition and taking conditional initialization as a processing standard.

9. The aerial photogrammetry error compensation method according to claim 1, wherein: Each error class corresponds to a synaptic weight; locating unintended compensation portions by evaluating the compensation mapping data; The synaptic weight of the error discriminator is dynamically adjusted by increasing the synaptic weight of the unexpected compensation part and decreasing the synaptic weight of the expected compensation part.

10. An aerial photogrammetry error compensation system, characterized in that: A system for executing an aerial photogrammetry error compensation method according to any one of claims 1 to 9, comprising: An error division unit is used to divide the inherent error and the random error based on the aerial surveying and mapping environment and the airborne surveying and mapping system, wherein the inherent error includes the environment inherent error and the system inherent error; An attraction zone establishment unit is used to establish a chaotic attraction zone based on error characteristic elements through chaotic error distribution, wherein the chaotic attraction zone maps the error to different attraction zones, and an error discriminator is connected to the exit end of each attraction zone; The surveying and mapping compensation unit is used to obtain surveying and mapping data and transmit it back to the aerial surveying and mapping system as the aerial surveying and mapping process progresses, perform partitioning based on the chaotic attraction zone through element verification, introduce zone error conditions and import the error discriminator, perform first-order judgment on inherent error and random error, perform second-order judgment on positive and negative errors based on random error, and determine compensated surveying and mapping data through error-oriented compensation and fitting, wherein the discrimination constraint is performed by introducing dynamic synaptic weights.

Images