Search

CN-122019979-A - AI-driven low-altitude data quality intelligent verification treatment system

CN122019979ACN 122019979 ACN122019979 ACN 122019979ACN-122019979-A

Abstract

The invention relates to the technical field of data quality check and AI intelligent control, in particular to an AI-driven low-altitude data quality intelligent check control system, wherein a dynamic coupling coefficient calculation unit synchronizes GPS positioning errors and image pixel offset errors, calculates coupling coefficients of each frame through an error coupling algorithm, and ensures data continuity through error compensation and interpolation complementation; the AI driving checking and managing unit fuses the abnormal characteristics with GPS jump and image ambiguity threshold values, constructs multidimensional characteristic vectors, the time sequence convolution network identifies hidden modes of single source data which are not out of limit but are coupled abnormally, outputs an alarm after the triple conditions are met, drives an automatic calibration module to adjust exposure frequency, resets GPS filtering parameters and injects calibration parameters, and three-stage management data quality is guaranteed.

Inventors

  • DONG ZHAO
  • BAO MINGXU
  • GE XIAO

Assignees

  • 吉林财经大学

Dates

Publication Date
20260512
Application Date
20260202

Claims (10)

  1. AI drive low altitude data quality intelligence check-up treatment system, its characterized in that includes: The dynamic coupling coefficient accounting unit (2) synchronously acquires the GPS positioning error and the image pixel offset error output by the data acquisition processing unit (1) frame by frame, divides the image pixel offset error by the GPS positioning error through an error coupling algorithm, and calculates the coupling coefficient of each frame in real time; An ideal interval comparison unit (3) presets an ideal interval of the coupling coefficient in a stable flight stage, reads the coupling coefficient of the current frame from the coupling coefficient of each frame output by a dynamic coupling coefficient calculation unit (2), calculates the deviation percentage of the coupling coefficient of the current frame from the lower limit and the upper limit of the ideal interval, calculates negative deviation degree if the coupling coefficient of the current frame is lower than the lower limit of the ideal interval, calculates positive deviation degree if the coupling coefficient of the current frame is higher than the upper limit of the ideal interval, and marks the negative deviation degree and the positive deviation degree value as abnormal coupling coefficient characteristics in real time; The AI driving verification treatment unit (4) integrates the abnormal coupling coefficient characteristics into an AI model as a new verification input, the deviation percentage is fused with a GPS jump threshold and an image ambiguity threshold in the AI verification characteristic construction stage of the AI model to form a multidimensional characteristic vector, the multidimensional characteristic vector is input into a trained neural network model, the neural network model identifies a hidden abnormal mode with the deviation degree of the coupling coefficient exceeding a preset threshold through supervision and learning, and when the fact that the single source data does not exceed the threshold but the lower limit and the upper limit of the coupling coefficient deviate from an ideal interval to reach the preset threshold is identified, a hidden abnormal alarm signal is output to drive an automatic calibration module to adjust data acquisition parameters to treat data quality.
  2. 2. The AI-driven low-altitude data quality intelligent verification governance system of claim 1, wherein when the error coupling algorithm is executed in a dynamic coupling coefficient accounting unit, a time stamp synchronization mechanism is used for ensuring that GPS positioning errors and image pixel offset errors are acquired at the same moment, a floating point divider is utilized for dividing two error values to generate coupling coefficients, when the GPS positioning errors are lower than a preset minimum effective value, an error compensation module is automatically started to inject a reference compensation quantity, meanwhile, a frame level buffer queue is established for temporarily storing original error data of continuous 5 frames, and if single frame data abnormal interruption is detected, an adjacent frame interpolation algorithm is invoked for complementing a missing value.
  3. 3. The AI-driven low-altitude data quality intelligent verification and management system according to claim 1, wherein the ideal interval comparison unit (3) is characterized in that the preset coupling coefficient ideal interval is obtained through historical stable flight phase data training, and specifically comprises the following steps: When the attitude angle change rate of the unmanned aerial vehicle is continuously lower than a set angular speed threshold value and the continuous flight duration exceeds a set time window, extracting all coupling coefficient data in a time period when the continuous flight duration exceeds the set time window, taking a numerical range of a dense area in statistical distribution as an ideal interval for dynamic updating after removing a maximum minimum value, and before reading a coupling coefficient of a current frame, verifying a data frame state flag bit output by a dynamic coupling coefficient accounting unit (2), and if the flag bit is abnormal, skipping the calculation of the current frame.
  4. 4. The AI-driven low-altitude data quality intelligent verification and management system according to claim 3, wherein the deviation percentage calculation adopts a bidirectional difference ratio method, and specifically comprises the following steps: Dividing the difference between the coupling coefficient of the current frame and the lower limit of the ideal interval by the lower limit of the ideal interval to obtain a negative deviation basic value, dividing the difference between the coupling coefficient of the current frame and the upper limit of the ideal interval by the upper limit of the ideal interval to obtain a positive deviation basic value, converting the basic value into a percentage form by a normalization processor, and introducing an interval width weighting factor in the conversion process to enable the deviation percentage to reflect the relative position of the current value in the ideal interval.
  5. 5. The AI-driven low-altitude data quality intelligent verification and management system as set forth in claim 3, wherein the deviation percentage calculation adopts a bidirectional difference ratio method, specifically: Dividing the difference between the coupling coefficient of the current frame and the lower limit of the ideal interval by the lower limit of the ideal interval to obtain a negative deviation basic value, dividing the difference between the coupling coefficient of the current frame and the upper limit of the ideal interval by the upper limit of the ideal interval to obtain a positive deviation basic value, converting the basic value into a percentage form by a normalization processor, and introducing an interval width weighting factor in the conversion process to enable the deviation percentage to reflect the relative position of the current value in the ideal interval.
  6. 6. The AI-driven low-altitude data quality intelligent verification and management system of claim 3, wherein the positive bias calculation is activated when the current frame coupling coefficient is detected to be higher than the upper limit of the ideal interval, the absolute difference value of the current value minus the upper limit of the ideal interval is calculated, the basic positive ratio is obtained after dividing the absolute difference value by the upper limit of the ideal interval, the basic positive ratio is input to a gain regulator which is different from the negative bias, and the gain regulator dynamically adjusts the amplification coefficient based on the unmanned aerial vehicle flight altitude parameter.
  7. 7. The AI-driven low-altitude data quality intelligent verification and management system according to claim 3, wherein the coupling coefficient anomaly characteristic mark is realized through a characteristic encoder, the deviation percentage is converted into a binary characteristic code, a low-level alarm code is generated when the negative deviation exceeds a set warning threshold value, a high-level alarm code is generated when the positive deviation exceeds the warning threshold value, a zero value code is output if both directions are not exceeded, and the annular characteristic buffer area is written after all the characteristic codes are added with time stamps.
  8. 8. The AI-driven low-altitude data quality intelligent verification and management system of claim 3, wherein the multi-dimensional feature vector construction adopts a feature tensor splicing technology, extracts the latest coupling coefficient abnormal feature code from the annular feature buffer, combines the latest coupling coefficient abnormal feature code with a GPS jump flag bit and an image ambiguity flag bit acquired in real time to form a three-dimensional feature tuple, maps each discrete flag bit into a continuous vector through a feature embedding layer, and splices the continuous vector along a feature dimension to form a fusion feature vector with a fixed length.
  9. 9. The AI-driven low-altitude data quality intelligent verification and management system of claim 1, wherein the neural network model adopts a time sequence convolution network architecture, an input layer sequentially extracts time-dependent features, a gating activation layer filters noise features and an attention pooling layer focuses on key abnormal fragments through a causal convolution layer after receiving multidimensional feature vectors, an output layer is connected with a sigmoid classifier, and when an output value of the neural network model exceeds a decision boundary threshold, a hidden abnormal mode with an overrun coupling coefficient deviation degree is judged.
  10. 10. The AI-driven low-altitude data quality intelligent verification treatment system of claim 1, wherein the output of the implicit anomaly alert signal is required to meet three conditions simultaneously, namely that the GPS jump threshold check flag bit is not overrun, the image ambiguity threshold check flag bit is not overrun, and the neural network model outputs the anomaly determination signal, after the alert signal is generated, the automatic calibration module is driven to execute three-stage treatment, firstly, the exposure frequency compensation time asynchronous error of the image sensor is regulated, secondly, the GPS receiver signal filtering parameter is reset, and finally, the calibration parameter is injected into the data fusion algorithm to compensate the historical deviation.

Description

AI-driven low-altitude data quality intelligent verification treatment system Technical Field The invention relates to the technical field of data quality check and AI intelligent control, in particular to an AI-driven low-altitude data quality intelligent check control system. Background The intelligent data quality control and AI management method is an important technology, is particularly applied to the data quality control links of low-altitude equipment such as unmanned aerial vehicles, and has the core that the data reliability is guaranteed by accurately identifying data error coupling abnormality, the core requirements of low-altitude data on high-quality data in the scenes such as mapping, monitoring and the like are adapted, in the low-altitude data acquisition process, GPS positioning errors and image pixel offset errors are dynamically associated, the errors can form a coupling relation along with the change of flight states, and because the threshold value verification of single-source data only can judge whether each error exceeds the limit or not, hidden abnormality generated by the coupling of the error and the single-source data cannot be captured, the hidden danger of the data quality is omitted, and the accuracy of the subsequent application of the low-altitude data is further influenced. Disclosure of Invention The invention aims to provide an AI-driven low-altitude data quality intelligent verification treatment system so as to solve the problems in the background technology. In order to achieve the above object, an AI-driven low-altitude data quality intelligent verification and management system is provided, comprising: the dynamic coupling coefficient accounting unit synchronously acquires the GPS positioning error and the image pixel offset error output by the data acquisition processing unit frame by frame, divides the image pixel offset error by the GPS positioning error through an error coupling algorithm, and calculates the coupling coefficient of each frame in real time; The ideal interval comparison unit presets a coupling coefficient ideal interval in a stable flight stage, reads the coupling coefficient of the current frame from the coupling coefficient of each frame output by the dynamic coupling coefficient calculation unit, calculates the deviation percentage of the coupling coefficient of the current frame from the lower limit and the upper limit of the ideal interval, calculates the negative deviation degree if the coupling coefficient of the current frame is lower than the lower limit of the ideal interval, calculates the positive deviation degree if the coupling coefficient of the current frame is higher than the upper limit of the ideal interval, and marks the negative deviation degree and the positive deviation degree value as the abnormal coupling coefficient characteristics in real time; The AI driving verification treatment unit integrates the abnormal coupling coefficient characteristic as a new verification input into an AI model, fuses the deviation percentage with a GPS jump threshold and an image ambiguity threshold in the AI verification characteristic construction stage of the AI model to form a multidimensional characteristic vector, inputs the multidimensional characteristic vector into a trained neural network model, and monitors and learns to identify a hidden abnormal mode with the deviation of the coupling coefficient exceeding a preset threshold, and when the neural network model identifies that the single source data does not exceed the threshold but the lower limit and the upper limit of the coupling coefficient deviate from an ideal interval to reach the preset threshold, outputs a hidden abnormal alarm signal to drive an automatic calibration module to adjust data acquisition parameters so as to treat data quality. Compared with the prior art, the invention has the beneficial effects that: According to the invention, the coupling coefficient of each frame is accurately calculated through the timestamp synchronization mechanism, the error compensation and the interpolation complement function of the dynamic coupling coefficient accounting unit, so that the real continuity of the error correlation characteristic is ensured. The ideal interval comparison unit dynamically updates an ideal interval based on stable flight data, quantifies deviation degree by adopting a bidirectional difference value proportion method, combines high-low level alarming codes to clearly mark abnormal characteristics, provides accurate basis for subsequent verification, fuses coupling coefficient abnormal characteristics with GPS (global position system) and image quality indexes, constructs multidimensional characteristic vectors, accurately identifies hidden modes of single source data which are not out of limit but are abnormal in coupling by a time sequence convolution network, and three-stage condition verification avoids false alarm, and three-stage treatment by an