A smart storage method and system based on a spaceborne industrial camera

By quantifying radiation interference and generating a flip probability distribution matrix, the smoothing trend of the error fluctuation numerical sequence is determined and resources are allocated. Combined with Reed-Solomon coding, error correction and pixel-level restoration are performed, solving the problem of adaptive error correction and image restoration of spaceborne images under radiation environment, and realizing efficient resource utilization and image quality restoration.

CN122884745BActive Publication Date: 2026-11-17AXD (ANXINDA) MEMORY TECH CO LTD
View PDF 2 Cites 0 Cited by

Patent Information

Application Number
CN202611383704.0
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2026-09-08
Publication Date
2026-11-17
Estimated Expiration
2046-09-08

AI Technical Summary

Technical Problem

Existing technologies cannot perform adaptive and differentiated intelligent error correction and image restoration based on changes in the radiation environment and the actual damage state of the storage medium, resulting in excessive resource consumption or insufficient error correction capabilities, and failing to effectively guarantee the integrity of spaceborne images.

Method used

By acquiring telemetry data from the spaceborne camera and the physical topology of the storage array, radiative interference is quantified and a flip probability distribution matrix is ​​generated. The smoothing trend of the error fluctuation numerical sequence is determined and resources are allocated. Error correction is performed in combination with Reed-Solomon coding, and pixel-level repair is carried out.

Benefits of technology

It enables proactive assessment and early warning of storage arrays, dynamically allocates error correction resources, improves error correction capabilities, eliminates visual anomalies, and ensures image integrity and quality.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN122884745B_ABST
    Figure CN122884745B_ABST
Patent Text Reader

Abstract

This invention relates to the field of spacecraft data storage technology, and discloses an intelligent storage method and system based on a spaceborne industrial camera. The method includes acquiring operational telemetry data and physical topology, quantifying radiation interference, and obtaining a flip probability distribution matrix; mapping the matrix to the physical topology and performing state sampling to obtain a numerical sequence of error fluctuations; smoothing and trend determination of the sequence to generate error correction resource adjustment instructions; parsing the instructions to obtain fault frequency records and allocating resources to obtain a resource allocation ratio; performing error correction encoding based on the ratio and damaged block location information to obtain a data reconstruction matrix; and performing pixel anomaly detection and spatial interpolation filling on the matrix to obtain reconstructed image data. This method enables adaptive error correction and image integrity assurance for spaceborne cameras under space radiation environments.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] This invention relates to the field of spacecraft data storage technology, and in particular to an intelligent storage method and system based on a spaceborne industrial camera. Background Technology

[0002] Currently, during spacecraft operation in orbit, onboard industrial cameras bear the critical responsibility of recording images of space missions. In the space environment, high-energy particle radiation can penetrate the spacecraft's shielding layer, bombarding the camera's internal storage medium, causing data bit flips, and resulting in drastic fluctuations in the bit error rate. Especially when spacecraft perform critical missions such as delicate maintenance docking or planetary surface exploration, the surge in radiation can cause random tampering of image data, resulting in image tearing and severely interfering with the accurate judgment of ground personnel. How to flexibly allocate error correction resources in a dynamic radiation environment has become a core challenge in ensuring the integrity of onboard images.

[0003] In one existing technology, onboard storage systems typically pre-set fixed fault tolerance thresholds and a unified error correction coding scheme. The system first acquires radiation sensor count data through a telemetry channel to determine the current radiation level; then, based on a preset threshold, it decides whether to trigger a storage medium self-test; finally, it uses fixed-rate Reed-Solomon coding to uniformly correct errors in all storage blocks and writes the repaired data to a spare sector. However, this scheme cannot adapt to dynamic fluctuations in the bit error rate. Under normal conditions, the fixed error correction overhead leads to excessive resource consumption, and under extreme interference, the limited error correction capability is insufficient to cope with sudden large-scale bit flips. Furthermore, it cannot differentiate error correction resource allocation based on the data importance level of each logical block and the real-time bit error rate status. Moreover, it lacks pixel-level repair methods for visual anomalies such as tearing and texture breakage in the image after error correction. While current mainstream intelligent chips possess certain computing capabilities, they have not been effectively integrated into the adaptive error correction link of the storage system.

[0004] In summary, existing technologies suffer from the inability to adaptively and differentiate intelligent error correction and image restoration based on changes in the radiation environment and the actual damage state of the storage medium. Summary of the Invention

[0005] This invention provides an intelligent storage method and system based on a spaceborne industrial camera to solve the problem that existing technologies cannot perform adaptive and differentiated intelligent error correction and image restoration based on changes in the radiation environment and the actual damage state of the storage medium.

[0006] In a first aspect, to address the aforementioned technical problems, this invention provides an intelligent storage method based on a spaceborne industrial camera, comprising: The operational telemetry data of the spaceborne camera and the physical topology of the storage array are obtained. The operational telemetry data is subjected to radiative interference quantization processing to obtain the flip probability distribution matrix. The flipped probability distribution matrix is ​​mapped to the physical topology to obtain a probability mapping spectrum, and the probability mapping spectrum is sampled to obtain a numerical sequence of bit error fluctuations. The error fluctuation value sequence is smoothed to determine the trend, and an error correction resource adjustment instruction is generated based on the determination result; The error fluctuation value sequence is statistically analyzed according to the error correction resource adjustment instruction to obtain the fault frequency record, and the fault frequency record is processed for resource allocation to obtain the resource allocation ratio. Obtain the location information of the block to be corrected reported by the storage array, and perform error correction coding processing on the block to be corrected according to the resource allocation ratio and the location information to obtain a data reconstruction matrix; The image pixel data carried by the data reconstruction matrix is ​​obtained, pixel anomaly detection is performed on the image pixel data to obtain abnormal pixel markers, the neighboring pixels corresponding to the abnormal pixel markers are extracted, and spatial interpolation is performed on the neighboring pixels to obtain the reconstructed image data.

[0007] Secondly, the present invention provides an intelligent storage system based on a spaceborne industrial camera, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the method described above.

[0008] Thirdly, the present invention provides a computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the method described above.

[0009] Compared with the prior art, the present invention has the following beneficial effects: (1) This invention obtains the intensity of radiated interference and performs multi-dimensional quantitative analysis to establish a flip probability distribution matrix inside the storage medium. Then, it extracts the underlying physical bit error state within a continuous time window to form a bit error fluctuation numerical sequence. This sequence is then processed by time-series smoothing and state prediction to obtain a fluctuation monitoring sequence characterizing the future trend of error surge. This scheme overcomes the limitation of existing schemes that rely solely on current telemetry data for lagging judgment by transforming abstract radiated interference into a quantifiable regional risk distribution and introducing a time-series prediction mechanism to identify bit error deterioration trends in advance. This enables a forward-looking assessment and early warning of the health status of the storage array.

[0010] (2) This invention calculates the error distribution density of each physical storage region when the fluctuation monitoring sequence exceeds the fault tolerance threshold, matches the corresponding resource quota base according to the error distribution density, and then calculates the error distribution density of different physical storage regions to dynamically determine the resource allocation ratio for underlying data repair. This scheme introduces a differentiated resource scheduling mechanism, which enables error correction capability to be allocated on demand according to the actual degree of damage, thereby avoiding the resource waste of traditional unified threshold schemes under normal conditions and insufficient protection under extreme interference, and realizing the efficient utilization of error correction resources in the storage system.

[0011] (3) This invention obtains a data reconstruction matrix by performing polynomial verification and error correction on the error-to-correct block using Reed-Solomon coding. Then, it extracts spatial domain image tearing features from the abnormal pixel markers in the matrix and performs pixel-level filling processing using inverse distance weighted interpolation to obtain reconstructed image data with visual anomalies eliminated. This scheme solves the problem of existing schemes only completing logical error correction while ignoring residual visual distortion by superimposing image spatial domain repair on the underlying bit-level error correction, and realizes end-to-end image assurance from physical data reconstruction to visual quality restoration. Attached Figure Description

[0012] Figure 1 This is a schematic flowchart of an intelligent storage method based on a spaceborne industrial camera provided in the first embodiment of the present invention; Figure 2 This is a heatmap of the error distribution density matrix of a storage array after experiencing radiated interference, provided in the first embodiment of the present invention. Figure 3 This is a data reconstruction matrix generated after adaptive error correction coding processing, provided in the first embodiment of the present invention. Detailed Implementation

[0013] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0014] Reference Figure 1 The first embodiment of the present invention provides an intelligent storage method based on a spaceborne industrial camera, comprising the following steps: S11, acquire the operational telemetry data of the spaceborne camera and the physical topology of the storage array, perform radiative interference quantization processing on the operational telemetry data, and obtain the flip probability distribution matrix; S12, the flip probability distribution matrix is ​​mapped to the physical topology to obtain a probability mapping spectrum, and the probability mapping spectrum is sampled to obtain a numerical sequence of bit error fluctuations; S13, perform a smoothing trend determination on the bit error fluctuation numerical sequence, and generate an error correction resource adjustment instruction based on the determination result; S14, perform fault statistics on the bit error fluctuation value sequence according to the error correction resource adjustment instruction to obtain fault frequency records, and perform resource allocation processing on the fault frequency records to obtain resource allocation ratios; S15, obtain the location information of the block to be corrected reported by the storage array, and perform error correction coding processing on the block to be corrected according to the resource allocation ratio and the location information to obtain a data reconstruction matrix; S16, obtain the image pixel data carried by the data reconstruction matrix, perform pixel anomaly detection on the image pixel data to obtain abnormal pixel markers, extract the neighboring pixels corresponding to the abnormal pixel markers, perform spatial interpolation to fill the neighboring pixels, and obtain the reconstructed image data.

[0015] In step S11, the operational telemetry data of the spaceborne camera and the physical topology of the storage array are acquired. The operational telemetry data undergoes radiative interference quantization processing to obtain a flip probability distribution matrix, including: The radiated interference intensity sequence is obtained by parsing the telemetry data, and an interference feature set is extracted based on the radiated interference intensity sequence; The interference feature set is subjected to energy deposition mapping to obtain local energy deposition values. If the local energy deposition values ​​are greater than a preset interference threshold, it is determined that a medium flip has occurred, and a medium flip state record is obtained. Based on the medium flipping state record, flipping probability statistics are performed to obtain the flipping probability distribution matrix.

[0016] In this embodiment, the operational telemetry data refers to the telemetry stream data collected and transmitted in real time by the radiation dosimeter and / or single-event event monitor deployed inside the spaceborne industrial camera during its on-orbit operation. The operational telemetry data is acquired by the spaceborne camera through its built-in data transmission interface, packaging the radiation monitoring data measured by the radiation dosimeter and / or single-event event monitor into a telemetry stream according to the telemetry frame format, and then transmitting it to the ground receiving station or the spaceborne data management system via the spacecraft data transmission channel. In this embodiment, the operational telemetry data is acquired through the telemetry data receiving interface. The physical topology of the storage array refers to the address space distribution information of each physical storage unit in the storage array, including the physical row address, physical column address, and chip select addressing hierarchy relationship between each storage chip. The physical topology of the storage array is obtained by reading the address mapping table from the memory configuration register of the spaceborne camera. The address mapping table records the correspondence between logical addresses and the row and column coordinates of physical storage units. Based on the address mapping table, the row and column coordinates of each physical storage unit are determined, and a two-dimensional distribution grid of physical row addresses and physical column addresses is constructed. This two-dimensional distribution grid is then determined as the physical topology.

[0017] Subsequently, the radiated interference intensity sequence is obtained by parsing the telemetry data. Specifically, in this embodiment, the received telemetry bitstream is parsed using a frame synchronization decompression algorithm to extract the data field carrying the radiated interference intensity. The frame synchronization decompression algorithm uses a preset frame header feature code in the telemetry bitstream as the synchronization benchmark and employs a sliding window matching method to scan the bitstream bit by bit. When a data bit sequence matching the preset frame header feature code is detected, the starting boundary of the current frame is determined. The preset frame header feature code is the hexadecimal number 0xEB90, which is widely used as a frame start synchronization marker in the aerospace telemetry field. Its binary representation has good autocorrelation characteristics, and the probability of missynchronization during sliding window matching is less than 10 to the power of -6. According to the preset frame structure definition, the radiated interference intensity data field is extracted from a fixed offset position after determining the starting boundary. The extracted binary data is converted into decimal values ​​to obtain the radiated interference intensity sequence. The preset frame structure definition refers to the arrangement rules of each field in the telemetry data frame, including the length and order of the frame header field, data field, and check field. The preset frame structure definition follows the aerospace telemetry transmission frame format, setting the frame header field as a 16-bit fixed feature code, followed by a radiated interference intensity data field (16 bits), a spare field (16 bits), and a check field (16 bits), arranged sequentially from high bit to low bit. Based on this frame structure definition, the radiated interference intensity data field is extracted from bits 16 to 31 after the frame start boundary. The radiated interference intensity sequence is physically represented as a radiation dose rate value or a single-event count value. The radiation dose rate value refers to the ionizing radiation dose deposited in the space radiation environment of the spaceborne camera per unit time, measured in rads per second, and is obtained in real time by the spaceborne radiation dosimeter. The single-event count value refers to the cumulative number of single-event upset events caused by high-energy particle impacts on the spaceborne camera's storage array per unit time, and is obtained in real time by the single-event monitor.

[0018] Next, an interference feature set is extracted based on the radiated interference intensity sequence. In this embodiment, time-domain feature extraction is performed on the radiated interference intensity sequence, that is, each value in the radiated interference intensity sequence is traversed sequentially, the peak amplitude and duration of each interference pulse are extracted, and the extracted peak amplitude and duration are combined into a feature vector in chronological order; after traversing all interference pulses, all feature vectors are arranged in chronological order to form an interference feature set.

[0019] The local energy deposition value is obtained by performing energy deposition mapping on the interference feature set. Energy deposition mapping refers to mapping the time-domain characteristic parameters of the interference pulse to the radiation energy value of high-energy particles deposited in the storage medium. In this embodiment, the local energy deposition value is calculated according to the energy deposition physics formula. Specifically, for each feature vector in the interference feature set, its peak amplitude value and duration value are extracted; the peak amplitude value is multiplied by the duration value to obtain the total energy characterization value of the interference pulse; the total energy characterization value is multiplied by a preset energy deposition coefficient to obtain the local energy deposition value. In this embodiment, the preset energy deposition coefficient is set to 0.08, which is a dimensionless empirical proportionality coefficient. Its value varies with the storage medium type and irradiation conditions. In practical applications, it must be recalibrated for specific chips. The method of determination is to use a particle accelerator to generate a high-energy particle beam to irradiate the storage medium sample during the ground calibration stage of the spaceborne camera, and simultaneously record the peak amplitude and duration of the interference pulse output by the radiation monitoring circuit. The high-energy particle beam contains particles of the same type as the main radiating particles in the on-orbit operating environment of the spaceborne camera. These particles are protons, with an energy range of 10 MeV to 100 MeV, and the calibration temperature is the typical operating temperature of the on-orbit camera. For each set of known particle energy values ​​and the corresponding pulse peak amplitude and duration, the product of the pulse peak amplitude and duration is divided by the known particle energy value to obtain the energy deposition coefficient value corresponding to that set of data. The arithmetic mean of the energy deposition coefficient values ​​obtained from multiple sets of data is then taken, and this arithmetic mean is determined as the preset energy deposition coefficient.

[0020] If the local energy deposition value is greater than a preset interference threshold, a medium flip is determined to have occurred, and a medium flip state record is obtained. The preset interference threshold is determined by statistical analysis. Specifically, during the ground calibration phase of the spaceborne camera, a high-energy particle beam generated by a particle accelerator is used to irradiate the storage medium sample, while a high-precision memory read circuit monitors the logic state of the storage unit bit by bit. The minimum energy deposition value corresponding to the logic state flip of the storage unit is recorded, and the minimum value among multiple measurements is taken as the preset interference threshold. In this embodiment, the threshold is set to 10 MeV. Each local energy deposition value is compared with the preset interference threshold. If it is greater than the preset interference threshold, it is determined that the storage unit at the corresponding physical address has undergone a medium flip at that time point, and a medium flip state record is generated. The medium flip state record includes the physical address of the storage unit and the time of the flip. If the local energy deposition value is less than or equal to the preset interference threshold, it is determined that the interference pulse has not reached the energy threshold that triggers a medium flip, and no medium flip state record is generated. After traversing all local energy deposition values, all generated medium flip state records are summarized to obtain a medium flip state record set.

[0021] Based on the media flip-over state records, flip-over probability statistics are performed to obtain a flip-over probability distribution matrix. Specifically, in this embodiment, the physical topology of the storage array is obtained, and the physical address space of the storage array is divided into multiple grid regions according to the physical topology, with each grid region corresponding to a physical storage block. For each grid region, the number of media flip-overs occurring within a fixed statistical period is counted, and this number is divided by the total number of storage cells in the grid region to obtain the flip-over probability value of the grid region. After traversing all grid regions, all flip-over probability values ​​are arranged into a two-dimensional matrix according to the physical row address and column address of the grid region, and this two-dimensional matrix is ​​determined as the flip-over probability distribution matrix.

[0022] In step S12, the flip probability distribution matrix is ​​mapped to the physical topology to obtain a probability mapping spectrum, and the probability mapping spectrum is sampled to obtain a numerical sequence of bit error rate fluctuations, including: The flipped probability distribution matrix is ​​mapped to the physical topology to obtain a probability mapping map; The probability mapping spectrum is sampled within a continuous time window to obtain a set of multi-frame state slices; The underlying physical error state is obtained by parsing the multi-frame state slice set, and data bit anomaly statistics are performed on the underlying physical error state to obtain the error fluctuation numerical sequence.

[0023] In this embodiment, the flip probability distribution matrix refers to the two-dimensional matrix output in step S11 that reflects the probability of media flipping at each physical address of the storage array within a preset statistical period. The physical topology refers to the physical address space distribution information of the storage array obtained in step S11. First, the flip probability distribution matrix is ​​mapped to the physical topology to obtain a probability mapping map. Specifically, in this embodiment, the physical row address and physical column address of each physical storage unit in the physical topology are obtained; using the physical row address as the row index and the physical column address as the column index, an empty matrix with the same dimension as the two-dimensional distribution grid of the physical topology is established; each element in the flip probability distribution matrix is ​​traversed, and the row and column index of the element is mapped to the physical storage unit address corresponding to the same row and column index in the physical topology; the element value of the flip probability distribution matrix is ​​written into the same row and column index position in the empty matrix; after traversing all elements, the mapping of the flip probability distribution matrix from the statistical grid to the physical address space is completed, and the matrix after writing is determined as the probability mapping map. Each element value in the probability mapping graph represents the flip probability value of the corresponding physical storage unit address, and there is a one-to-one binding relationship between the element value and the physical storage unit address.

[0024] Next, state sampling is performed on the probability mapping spectrum within a continuous time window to obtain a multi-frame state slice set. State sampling refers to reading and recording the actual bit error state of each physical storage unit in the storage array at different time points. The continuous time window refers to the entire observation period from the start of sampling to the end of sampling. The duration of the continuous time window is set by performing state sampling and subsequent bit error fluctuation numerical sequence extraction processes with different time window durations (1 second, 2 seconds, 5 seconds, 10 seconds, and 20 seconds) during the ground verification phase of the onboard camera. The variance stability and overall computational cost of the extracted bit error fluctuation numerical sequences under each time window duration are statistically analyzed. When the time window duration is 5 seconds, the variance change rate of the bit error fluctuation numerical sequence is less than 5%, which is considered to have converged to the stable interval, and the overall computational cost does not exceed the real-time processing time limit of the onboard camera. When the time window duration is further increased to 10 seconds or 20 seconds, the variance improvement rate of the bit error fluctuation numerical sequence is less than 3%, but the overall computational cost increases by more than double. In this embodiment, the duration of the continuous time window is set to 5 seconds.

[0025] Specifically, in this embodiment, within the continuous time window, the error status of the storage units corresponding to each physical storage unit address in the probability mapping spectrum is read at fixed sampling time intervals. For each physical storage unit, a read command is sent to the physical storage unit through the read interface of the storage controller to obtain the data bits currently stored in the storage unit. The currently stored data bits are compared bit by bit with the data bits stored in the storage unit at the previous sampling time. If the comparison result indicates that there is a data bit difference, the actual error status of the physical storage unit at the current sampling time is recorded as the first status value. If the comparison result indicates that there is no data bit difference, it is recorded as the second status value. Here, the physical storage unit refers to the smallest physical storage unit addressed bit by bit in the storage array, that is, a single storage bit. The determination of data bit anomalies is performed on a bit-by-bit basis. The number of physical storage units with the error status value of the first status value is the total number of storage bits that have undergone bit flipping.

[0026] It should be noted that the first and second state values ​​are used to binarize the actual bit error state of the physical storage unit. In this embodiment, following the general convention in digital logic circuits, a value of 1 is used to represent a data bit error, and a value of 0 is used to represent no data bit error. After traversing all physical storage units, the bit error state values ​​of all physical storage units are arranged into a two-dimensional state matrix according to their physical row and column addresses. This two-dimensional state matrix is ​​defined as a state slice. The above state slice acquisition is performed once at each sampling time. Multiple state slices are continuously acquired, and all state slices are arranged in chronological order of sampling time to obtain a set of multiple state slices.

[0027] Subsequently, the multi-frame state slice set is parsed to obtain the underlying physical error state. The underlying physical error state refers to the error evolution pattern of each physical storage unit in the storage array over time. Specifically, in this embodiment, each frame state slice in the multi-frame state slice set is extracted; for each physical storage unit address, the error state value corresponding to that address is read from each frame state slice, and the error state values ​​of that physical storage unit address at each sampling time are arranged in chronological order to obtain the error state time series of that physical storage unit; after traversing all physical storage unit addresses, the error state time series of all physical storage units are arranged into a three-dimensional data structure according to the physical row address and column address, and this three-dimensional data structure is determined as the underlying physical error state; in this embodiment, each frame state slice is read sequentially for statistical analysis during the parsing process, and it is not required that all slices reside in memory simultaneously.

[0028] Finally, data bit anomaly statistics are performed on the underlying physical bit error state to obtain a bit error fluctuation value sequence. Specifically, in this embodiment, for each sampling time, the bit error state value of all physical storage units at that sampling time is extracted from the underlying physical bit error state, and the number of physical storage units with a bit error state value of the first state value is counted. This number is determined as the bit error fluctuation value at that sampling time. The bit error fluctuation values ​​of each sampling time are arranged in chronological order to obtain the bit error fluctuation value sequence. Each element value in the bit error fluctuation value sequence represents the total number of physical storage units in the storage array that experienced data bit anomalies at a sampling time. In this embodiment, the total capacity of the storage array is 4Gb, and the total number of storage units is 4×10. 9 The element value of the bit error fluctuation numerical sequence is the absolute number of physical storage units where data bit anomalies occurred.

[0029] It should be noted that the sampling time interval was determined by performing state sampling and bit error rate fluctuation numerical sequence extraction processes at different sampling time intervals (10 milliseconds, 30 milliseconds, 50 milliseconds, 80 milliseconds, and 100 milliseconds) on a spaceborne camera storage controller simulation platform. The temporal resolution of the bit error rate fluctuation numerical sequence and the overall data storage overhead were statistically analyzed for each sampling time interval. When the sampling time interval was 50 milliseconds, the bit error rate fluctuation numerical sequence could completely capture the bit error rate fluctuation characteristics caused by space radiation interference, and the number of frames collected within the continuous time window was moderate. When the sampling time interval was shortened to 10 milliseconds or 30 milliseconds, the temporal resolution improved by less than 10%, but the overall data storage overhead increased by more than three times. When the sampling time interval was extended to 80 milliseconds or 100 milliseconds, the high-frequency components in the bit error rate fluctuation characteristics were lost by more than 15%, leading to a decrease in the accuracy of subsequent trend determination. In this embodiment, the sampling time interval was set to 50 milliseconds.

[0030] It should be noted that the number of frames sampled within a continuous time window is the duration of the continuous time window divided by the sampling time interval. In this embodiment, the multi-frame state slice set contains 100 state slices. The 100 state slices are output in a frame-by-frame generation and processing manner. For each state slice, after performing data bit anomaly statistics, the complete bitmap data of the frame is not retained. Instead, the location information of the physical storage unit in the frame with the bit error state value as the first state value is compressed and recorded as a sparse list. This sparse list is then added to the cumulative frequency counter of the corresponding physical storage area. Subsequently, the storage space of the frame is released, and the next frame is generated. In this way, the storage space occupancy does not increase with the number of frames. Only the cumulative frequency counter of each physical storage area and the currently processed frame slice need to be retained. It is not required to store all 100 state slices at the same time.

[0031] In step S13, the smoothing trend of the bit error fluctuation numerical sequence is determined, and an error correction resource adjustment instruction is generated based on the determination result, including: The bit error fluctuation sequence is smoothed to obtain a time-series smoothed sequence; Extract the statistical features of the time-series smoothed sequence, and predict the bit error rate based on the statistical features to obtain a set of predicted values; If the predicted value in the predicted value set is greater than the preset surge judgment threshold, the corresponding storage unit identifier is extracted, and an error correction resource adjustment instruction is generated based on the storage unit identifier.

[0032] First, the numerical sequence of bit error rate fluctuations is smoothed to obtain a time-series smoothed sequence. The smoothing process filters out high-frequency random fluctuations caused by random thermal noise and single-event transient pulses in the space radiation environment. In this embodiment, a Kalman filter is used to smooth the numerical sequence of bit error rate fluctuations. Specifically, this embodiment constructs the state equation and observation equation of a discrete Kalman filter. The state equation is defined as the estimated value at the current time step equals the estimated value at the previous time step plus process noise, and the observation equation is defined as the observed value at the current time step equals the true value at the current time step plus observation noise. The variances of process noise and observation noise are defined by injecting a standard radiated interference signal with a known bit error rate into the storage array during the ground calibration phase of the spaceborne camera. The theoretical bit error rate fluctuation value of the standard radiated interference signal is used as the true value, and the actual bit error rate fluctuation value read from the storage array is used as the observed value. Multiple sets of data pairs of true and observed values ​​are collected. The variance of the difference between the true and observed values ​​in multiple data pairs is calculated, and this variance is determined as the variance of observation noise. The variance of the difference between the true values ​​at adjacent time points in multiple data pairs is calculated, and this variance is determined as the variance of process noise. The first value in the bit error rate fluctuation sequence is used as the initial estimate of the Kalman filter. The variances of process noise and observation noise are substituted into the prediction and update equations of the Kalman filter. The prediction and update steps are performed sequentially for each value in the bit error rate fluctuation sequence, and the filtered numerical sequence is output. This filtered numerical sequence is determined as the time-series smoothed sequence.

[0033] Next, statistical features of the time-series smoothed sequence are extracted, and bit error rate prediction is performed based on these features to obtain a set of predicted values. The statistical features are used to quantify the evolution of the time-series smoothed sequence over time. Specifically, this embodiment sets a sliding window of a preset length. The length of this sliding window is determined on a simulation platform for the onboard camera storage controller, with sliding window lengths of 10, 20, 30, 50, and 80 sampling points respectively. Multiple experiments are performed on the statistical feature extraction and bit error rate prediction for each window length, and the prediction accuracy and computational cost are statistically analyzed. When the window length is 30 sampling points, the prediction accuracy reaches over 95% with moderate computational cost. When the window length is further increased to 50 or 80 sampling points, the improvement in prediction accuracy is less than 2%, but the computational cost increases by more than double. Therefore, this embodiment sets the sliding window length to 30 sampling points. The sliding window is slid along the time axis of the time-series smoothing sequence, and 30 values ​​within the current window are extracted at each sliding sampling point as the local sequence at the current moment.

[0034] For each local sequence, this embodiment calculates the statistical characteristics of the local sequence. The statistical characteristics include the mean, variance, and linear trend slope of the local sequence. The mean is obtained by summing all the values ​​in the local sequence and dividing by the total number of values ​​in the local sequence. The variance is obtained by calculating the square of the difference between each value in the local sequence and the mean, summing all the squares, and dividing by the total number of values ​​in the local sequence. The linear trend slope is obtained by performing a univariate linear regression fit on the local sequence using the least squares method. Specifically, the sampling time sequence is used as the independent variable, and the values ​​in the local sequence are used as the dependent variable. The covariance between the independent and dependent variables is calculated and divided by the variance of the independent variable; the result is used as the linear trend slope.

[0035] The mean, variance, and linear trend slope are combined into a feature vector, which is then input into a pre-constructed prediction model to obtain the predicted value corresponding to the local sequence. The prediction model is constructed by collecting multiple sets of local sequences and their corresponding actual bit error rates at the next time step during the ground calibration phase of the spaceborne camera. The statistical feature vector of the local sequence is used as input, and the actual bit error rate is used as output. A multiple linear regression method is employed for fitting to obtain the prediction model. The prediction model includes three regression coefficients and one bias term. The three regression coefficients correspond to the mean, variance, and linear trend slope, respectively, and the bias term is a constant. The three values ​​in the feature vector are multiplied by their corresponding regression coefficients, summed, and then the bias term is added to obtain the predicted value. After sliding the window along the time axis through all sampling points, multiple predicted values ​​are obtained. All predicted values ​​are arranged in chronological order of prediction time to obtain a set of predicted values.

[0036] Finally, if the predicted value in the predicted value set is greater than the preset surge determination threshold, the corresponding storage unit identifier is extracted, and an error correction resource adjustment instruction is generated based on the storage unit identifier. The preset surge determination threshold is set by obtaining the error fluctuation value sequence of the historical data of the onboard camera under normal operating conditions, calculating the mean and standard deviation of all values ​​in the sequence; adding three times the standard deviation to the mean yields the upper control limit; this upper control limit is determined as the preset surge determination threshold. The absolute size of this threshold is related to the total capacity of the storage array. In this embodiment, the total capacity of the storage array is 4Gb, and the corresponding surge determination threshold is calibrated to 40. For storage arrays of other capacities, this threshold needs to be recalibrated using the same method. Each predicted value in the predicted value set is compared with the preset surge determination threshold. If a predicted value is greater than the preset surge determination threshold, it is determined that an error surge has occurred at that prediction time, and the storage unit identifier corresponding to the predicted value is extracted. The storage unit identifier refers to the physical address code of the physical storage unit corresponding to the predicted value, which is read from the physical topology obtained in step S11. An error correction resource adjustment instruction is generated based on all extracted storage unit identifiers. The error correction resource adjustment instruction includes a list of addresses for high-risk storage units, which instructs the storage controller to allocate additional error correction and verification resources to the high-risk storage units. If all predicted values ​​in the predicted value set are less than or equal to the preset surge determination threshold, no error correction resource adjustment instruction is generated, and the process returns to step S11 to reacquire telemetry data and continue executing the next round of monitoring.

[0037] It should be noted that the data structure of the error correction resource adjustment instruction includes an instruction header, an address list field, and a check field. The instruction header identifies the instruction type as an error correction resource adjustment instruction. The address list field stores all extracted storage unit identifiers. The check field stores the checksum obtained by performing cyclic redundancy check on the instruction data. A complete error correction resource adjustment instruction is generated by filling the storage unit identifiers into the address list field and calculating the check field.

[0038] In step S14, the error fluctuation value sequence is statistically analyzed according to the error correction resource adjustment instruction to obtain the fault frequency record, and the fault frequency record is processed for resource allocation to obtain the resource allocation ratio.

[0039] The process of allocating resources to the fault frequency records to obtain resource allocation ratios includes: An error distribution density matrix is ​​constructed based on the fault frequency records, and spatial filtering is performed on the error distribution density matrix to obtain the error distribution density value. If the error distribution density value is greater than the preset density warning threshold, then the corresponding resource quota base is matched according to the error distribution density value; The resource allocation ratio is calculated based on the resource quota base and the error distribution density.

[0040] First, the fault statistics process is triggered based on the error correction resource adjustment instruction. In this embodiment, the error correction resource adjustment instruction is parsed according to a protocol. The instruction header, address list field, and checksum field are extracted sequentially from the instruction according to a preset instruction frame structure. Cyclic redundancy check is performed on the checksum field; if the check passes, the instruction is deemed valid. After confirming the instruction's validity, the total number of data bit anomalies occurring in each storage unit within the historical observation period is read from the underlying physical error status recorded in step S12. This total number is used as the fault frequency of that storage unit. The fault frequencies of all storage units are summarized according to the physical row and column addresses of the storage units to obtain the fault frequency record for each physical storage region. A physical storage region refers to the physical address range corresponding to each grid region after the storage array is divided into multiple grid regions according to the physical topology obtained in step S11. The fault frequency record includes the region identifier of each physical storage region and its corresponding fault frequency statistics.

[0041] Next, an error distribution density matrix is ​​constructed based on the fault frequency records, and spatial filtering is performed on the error distribution density matrix to obtain the error distribution density values. Specifically, this embodiment obtains a two-dimensional distribution grid of physical row addresses and physical column addresses in the physical topology; using the physical row addresses in the physical topology as row indices and the physical column addresses as column indices, an empty matrix with the same dimension as the two-dimensional distribution grid of the physical topology is established; the fault frequency records of each physical storage region are traversed, and the fault frequency statistics of each physical storage region are filled into the corresponding row and column index positions of that region; after traversing all physical storage regions, the error distribution density matrix is ​​obtained. Each element value in the error distribution density matrix represents the cumulative frequency of data bit anomalies occurring in the corresponding physical storage region within the historical observation period.

[0042] Spatial filtering is performed on the error distribution density matrix. Spatial filtering is used to eliminate the interference of isolated noise points in the error distribution density matrix on the localization of density peaks. In this embodiment, a two-dimensional convolutional kernel is used to perform spatial smoothing filtering on the error distribution density matrix. The two-dimensional convolutional kernel has a size of 3 rows by 3 columns, with a central element value of 4, four orthogonally adjacent elements with a value of 1, four diagonally adjacent elements with a value of 0, and the sum of all element values ​​is 8. The two-dimensional convolutional kernel is slid element by element on the error distribution density matrix. For each sliding position, the value at the non-zero element position in the two-dimensional convolutional kernel is multiplied by the element value at the corresponding position in the error distribution density matrix, and the sum is calculated. The sum is divided by the sum of all non-zero element values, 8, to obtain the filtered output value at that position. After traversing all sliding positions, the filtered error distribution density matrix is ​​obtained. Then, the error distribution density value is extracted from the filtered error distribution density matrix. Specifically, in this embodiment, each element value in the filtered error distribution density matrix is ​​traversed, and the maximum element value encountered during the traversal is extracted and determined as the error distribution density value. Error distribution density numerical values ​​characterize the peak error density at the physical region in the storage array where error distribution is most concentrated.

[0043] Subsequently, if the error distribution density value exceeds a preset density warning threshold, the corresponding resource quota base is matched based on the error distribution density value. The preset density warning threshold is set by obtaining the filtered error distribution density matrix of the historical data of the spaceborne camera under normal operating conditions, extracting the maximum value in the matrix, calculating the mean and standard deviation of all historical maximum value sequences, adding three times the standard deviation to the mean to obtain the upper control limit, and determining this upper control limit as the preset density warning threshold. In this embodiment, the threshold is calibrated to 20.

[0044] The error distribution density value is compared with the preset density warning threshold. If the error distribution density value is less than or equal to the preset density warning threshold, the current error distribution density is determined to be within a safe range, resource allocation processing is not triggered, and the process returns to step S11 to reacquire telemetry data and continue to the next round of monitoring. If the error distribution density value is greater than the preset density warning threshold, the current error distribution density is determined to be outside a safe range, and the resource quota matching process is triggered. Resource quota matching refers to finding the corresponding resource quota base from a preset resource quota mapping table based on the numerical range of the error distribution density value. The resource quota mapping table is constructed on a spaceborne storage controller simulation platform, using the error distribution density value as the independent variable and iterating through the value range from 1 to 50 with a step size of 1. For each error distribution density value, the repair success rate and computation latency of the storage controller after performing error correction processing under the fault distribution conditions corresponding to that error distribution density value are tested when the error correction resource quota base is traversed from 1 to 10. Then, for each error distribution density value, the smallest error correction resource quota base that meets the constraint of a repair success rate of not less than 99% and a computation latency not exceeding the real-time processing time limit of the spaceborne storage controller is selected from the 10 candidate values ​​of the error correction resource quota base from 1 to 10 as the optimal resource quota base corresponding to that error distribution density value. After traversing all values ​​of the error distribution density value from 1 to 50, 50 sets of correspondences between error distribution density values ​​and optimal error correction resource quota bases are obtained. Finally, consecutive values ​​corresponding to the same optimal error correction resource quota base within the error distribution density value range are merged into a density level range, thereby forming a complete resource quota mapping table.

[0045] Finally, the resource allocation ratio is calculated based on the resource quota base and the error distribution density value. Specifically, in this embodiment, the error distribution density value is multiplied by the resource quota base to obtain a product; the total amount of currently available backup repair resources in the onboard storage controller is obtained, and the product is divided by the total amount of backup repair resources to obtain a quotient value, which is determined as the resource allocation ratio. The resource allocation ratio is used to instruct the storage controller to allocate a corresponding number of backup storage units from the backup repair resources for underlying data repair according to this ratio. It should be noted that the total amount of backup repair resources refers to the total number of backup physical storage units that are pre-reserved in the storage controller and are not involved in normal data storage; this value is provided by the hardware configuration register of the storage controller.

[0046] like Figure 2As shown, this figure presents a heatmap of the error distribution density matrix of the storage array after experiencing radiation interference. The horizontal and vertical axes represent the row and column addresses of the storage array, respectively. The shades of color in the figure intuitively reflect the frequency of error occurrence in each physical area. Through this visualization, the clustering effect of errors in physical space can be clearly observed, such as the dark, high-density core area in the figure. This spatial distribution characteristic provides direct visual evidence and data support for the subsequent implementation of differentiated resource allocation strategies.

[0047] In step S15, the location information of the block to be corrected reported by the storage array is obtained, and error correction encoding processing is performed on the block to be corrected according to the resource allocation ratio and the location information to obtain a data reconstruction matrix, including: Construct an error location vector based on the resource allocation ratio and the location information; A generation matrix is ​​constructed based on the error location vector, and polynomial verification is performed on the block to be corrected using the generation matrix to obtain a verification fragment. If the verification segment indicates that there is a data error, the error evaluation polynomial is extracted according to the verification segment, and error correction processing is performed according to the error evaluation polynomial to obtain the data reconstruction matrix.

[0048] First, the location information of the blocks to be corrected reported by the storage array is obtained. During each read operation, the error detection circuit of the storage array performs a Reed-Solomon check on the read data. If the check result indicates an uncorrectable data error, the storage block containing the data error is identified as a block to be corrected, and the physical index number of the block to be corrected is reported to the storage controller through the storage controller's error reporting interface. Specifically, this embodiment obtains the physical index number of the blocks to be corrected reported by the error detection circuit of the storage array; obtains the total number of all physical storage blocks in the storage array; constructs a one-dimensional vector initialized to all zeros, using the total number of all physical storage blocks as the vector dimension; for each physical index number of the block to be corrected in the location information, the element value at the corresponding position in the one-dimensional vector is set to the value 1; after traversing all the blocks to be corrected, an error location vector is obtained. Each element in the error location vector corresponds to a physical storage block; an element value of 1 indicates that a data error has occurred in the storage block, and an element value of 0 indicates that no data error has occurred in the storage block.

[0049] Next, a generator matrix is ​​constructed based on the error location vector. This generator matrix is ​​used for subsequent verification and error correction of the blocks to be corrected using Reed-Solomon coding. This embodiment uses the Reed-Solomon coding algorithm to construct the generator matrix. Specifically, this embodiment obtains a set of position indices in the error location vector where each element has a value of 1; each position index in the set is used as a eigenvalue in a finite field; and a generator matrix with a Vandermonde structure is constructed based on these eigenvalues. The number of rows in the generator matrix is ​​the number of elements in the error location vector with a value of 1, and the number of columns is the preset number of verification slices. The preset number of verification fragments was set by gradually increasing the number of verification fragments (2, 4, 6, 8, 10) during the ground calibration phase of the spaceborne camera, testing the data repair success rate and error correction computational overhead under each set of verification fragment numbers. When the number of verification fragments was 4, the data repair success rate reached over 99%, and the error correction computational overhead did not exceed the real-time processing time limit of the spaceborne storage controller. When the number of verification fragments continued to increase to 6 or more, the improvement in the repair success rate was less than 1%, but the computational overhead increased by more than double. Therefore, this embodiment sets the preset number of verification fragments to 4. These 4 verification fragments are used as the default configuration and are suitable for typical operating conditions when the resource allocation ratio is in the middle range. When the resource allocation ratio deviates from this range due to changes in the radiation environment, the number of verification fragments is re-determined according to the dynamic mapping relationship described below. The construction of the generator matrix is ​​performed under the finite field operation rules. The order of the finite field is 256, that is, each element is 8 bits of binary data.

[0050] Subsequently, polynomial verification is performed on the block to be corrected using a generator matrix to obtain verification slices. Specifically, in this embodiment, the original data symbol sequence is read from the block to be corrected. The length of the original data symbol sequence is the same as the number of columns in the generator matrix, i.e., it contains 4 original data symbols. Matrix multiplication is performed between the generator matrix and the original data symbol sequence, i.e., each row of the generator matrix is ​​multiplied by the original data symbol sequence to obtain the verification slice corresponding to that row. After performing the dot product operation on all rows, multiple verification slices are obtained, and the number of verification slices is equal to the number of rows in the generator matrix.

[0051] Subsequently, if the verification slice indicates a data error, the error evaluation polynomial is extracted based on the verification slice. Specifically, in this embodiment, the verification slice is substituted into the adjoint equation calculation to obtain the adjoint equation sequence. The adjoint equation calculation process is as follows: for each verification slice, the power of the verification slice and the corresponding eigenvalue is multiplied and added to obtain the adjoint equation value; after traversing all verification slices, the adjoint equation sequence is obtained. If all adjoint equation values ​​in the adjoint equation sequence are zero, it is determined that the block to be corrected does not have a data error, and no error correction processing is required; the original data of the block to be corrected is directly output as the data reconstruction matrix. If there are non-zero adjoint equation values ​​in the adjoint equation sequence, it is determined that the block to be corrected has a data error. Under this condition, this embodiment uses the Berlekamp-Massey algorithm to iteratively process the adjoint equation sequence to extract the error location polynomial and the error evaluation polynomial. The error evaluation polynomial is used to characterize the specific magnitude of the data error.

[0052] Finally, error correction is performed based on the error evaluation polynomial to obtain the data reconstruction matrix. Specifically, this embodiment uses the Chan's search algorithm to locate the error position based on the error position polynomial. After obtaining the error position, the error evaluation polynomial is substituted into the Forney algorithm for evaluation to calculate the error deviation value at each error position. The calculated error deviation value is then XORed with the original data at the corresponding error position in the block to be corrected to eliminate the underlying data error and complete the data repair. All the repaired data symbols are arranged in their original storage order to form a data matrix, which is then used as the data reconstruction matrix.

[0053] It should be noted that the resource allocation ratio determines the verification strength level of the error correction code. In this embodiment, the mapping relationship between the resource allocation ratio and the number of verification fragments is as follows: when the resource allocation ratio is less than or equal to 10%, the number of verification fragments is 2; when the resource allocation ratio is greater than 10% and less than or equal to 30%, the number of verification fragments is 4; when the resource allocation ratio is greater than 30% and less than or equal to 50%, the number of verification fragments is 6; and when the resource allocation ratio is greater than 50%, the number of verification fragments is 8. This mapping relationship is verified through onboard storage. On the storage controller simulation platform, with the resource allocation ratio as the independent variable, the range of values ​​from 5% to 100% is traversed in steps of 5%. For each ratio value, the repair success rate and computation latency are tested under four configurations with 2, 4, 6, and 8 verification shards. With the constraint that the repair success rate is not less than 99% and the computation latency does not exceed the real-time processing time limit of the onboard storage controller, the minimum number of verification shards that meets the constraint is selected as the optimal configuration for that ratio value. Continuous ratio ranges corresponding to the same optimal number of verification shards are merged to finally form the above mapping relationship.

[0054] like Figure 3As shown in the figure, the data reconstruction matrix generated after adaptive error correction coding is displayed. The squares of different gray levels in the figure represent the gray value distribution of image pixels, while the red circles accurately mark the data error points located and repaired by the RS error correction algorithm. This figure intuitively verifies the effectiveness of this method in pixel-level data reconstruction, showing that the system can not only restore the logical structure of the original data, but also ensure the continuity and integrity of image data in the spatial dimension, laying a high-quality data foundation for subsequent image tearing repair.

[0055] In step S16, the image pixel data carried by the data reconstruction matrix is ​​obtained, pixel anomaly detection is performed on the image pixel data to obtain anomaly pixel markers, the neighboring pixels corresponding to the anomaly pixel markers are extracted, and spatial interpolation is performed on the neighboring pixels to obtain the reconstructed image data, including: Extract pixel grayscale gradients from the image pixel data, and mark pixels whose pixel grayscale gradients are greater than a preset fluctuation threshold as abnormal pixels; A set of neighboring pixels is extracted centered on the abnormal pixel marker, and a connected component analysis is performed on the set of neighboring pixels to obtain an edge contour set; Based on the edge contour set, spatial domain image tearing features are extracted, and pixel-level filling processing is performed on the spatial domain image tearing features to obtain the reconstructed image data.

[0056] First, pixel grayscale gradients are extracted from the data reconstruction matrix, and pixels with grayscale gradients greater than a preset fluctuation threshold are marked as abnormal pixels. Pixel grayscale gradients characterize the degree of grayscale change between a pixel and its neighboring pixels in an image. In this embodiment, the Sobel operator is used to calculate the horizontal and vertical gradient values ​​at each pixel location. The absolute values ​​of the horizontal and vertical gradients are added together to obtain the pixel grayscale gradient at that location. The preset fluctuation threshold is set by acquiring the pixel grayscale values ​​of multiple frames of undamaged reference images captured by the spaceborne camera under normal operating conditions, calculating the grayscale gradient of each pixel in each frame of the reference image, and summing all pixel grayscale gradients of all reference images into a numerical set; calculating the mean and standard deviation of this numerical set; adding three times the standard deviation to the mean to obtain the upper control limit; and determining this upper control limit as the preset fluctuation threshold. In this embodiment, the threshold is calibrated to 30. The pixel grayscale gradient at each pixel location is compared with a preset fluctuation threshold. If the pixel grayscale gradient is greater than the preset fluctuation threshold, the pixel is marked as an abnormal pixel. After traversing all pixel locations, all pixel locations marked as abnormal pixels are recorded as abnormal pixel markers.

[0057] Next, a set of neighboring pixels is extracted centered on the abnormal pixel marker, and connected component analysis is performed on the neighboring pixel set to obtain the edge contour set. Specifically, for each abnormal pixel marker, this embodiment extracts all pixel grayscale values ​​within a preset-sized neighborhood window in the data reconstruction matrix, centered on the pixel position corresponding to the abnormal pixel marker, to obtain the neighborhood pixel set corresponding to the abnormal pixel marker. The size of the neighborhood window is calibrated as 5×5. Specifically, during the ground testing phase of the spaceborne camera, four window sizes, 3×3, 5×5, 7×7, and 9×9, were used to perform neighborhood pixel extraction and connected component analysis on multi-frame test images containing known tearing anomalies. The preferred conditions were that the integrity of the edge contour reached more than 95% and the processing delay of a single frame did not exceed fifty milliseconds. After testing, only the 5×5 window simultaneously met the above two requirements, so it was calibrated as 5×5.

[0058] Connectivity analysis is performed on the neighborhood pixel set. Connectivity analysis is used to aggregate adjacent pixels with similar grayscale values ​​into the same connected region. Specifically, this embodiment scans the pixels in the neighborhood pixel set based on an eight-connectivity determination rule. That is, for any pixel, if the absolute value of the difference between the grayscale value of its adjacent pixels in any of the eight directions and the grayscale value of the current pixel is less than a preset connectivity determination threshold, then the adjacent pixel and the current pixel are determined to belong to the same connected region. After traversing all pixels, the connected pixels are aggregated into a connected region. The preset connectivity determination threshold is set by obtaining the probability density distribution of the grayscale difference between adjacent pixels in a normal image from a spaceborne camera, extracting the grayscale difference corresponding to the 95th percentile of the cumulative probability distribution, and determining this grayscale difference as the connectivity determination threshold. In this embodiment, the threshold is set to 10. After aggregating all pixels in the neighborhood pixel set according to the above connectivity determination rule, multiple connected regions are obtained. The boundary pixels of each connected region are extracted, and the boundary pixels of all connected regions are summarized to obtain an edge contour set. Each contour in the edge contour set consists of a set of consecutive boundary pixels.

[0059] Subsequently, spatial domain image tearing features are extracted based on the edge contour set. Spatial domain image tearing features are used to quantify visual anomalies in an image caused by storage medium failure, such as image tearing, misalignment, or texture breakage. Specifically, this embodiment extracts pixel coordinate data and grayscale value data on both sides of each edge contour in the edge contour set; for each edge contour, the physical offset of the pixels on both sides is calculated, which is the Euclidean distance between the pixel coordinates on one side and the pixel coordinates on the other side of the contour; the grayscale difference between the pixels on both sides is calculated, which is the absolute value of the difference between the grayscale values ​​of the pixels on one side and the pixel values ​​on the other side of the contour; the physical offset and the grayscale difference are combined to form the tearing feature vector corresponding to the edge contour; after traversing all edge contours in the edge contour set, all tearing feature vectors are arranged into a feature matrix according to the spatial position of the edge contours, and this feature matrix is ​​determined as the spatial domain image tearing feature.

[0060] Subsequently, pixel-level filling processing is performed based on the spatial domain image tearing characteristics to obtain the reconstructed image data. Pixel-level filling is used to restore the grayscale values ​​of missing pixels in the torn area. This embodiment uses inverse distance weighted interpolation for pixel filling. Specifically, this embodiment determines the texture missing area based on the spatial domain image tearing characteristics; for each missing pixel in the texture missing area, an 11×11 receptive field window is extracted centered on the missing pixel; the size of the receptive field window is calibrated by performing inverse distance weighted interpolation filling on multiple test images containing known torn and missing pixels using four window sizes: 7×7, 9×9, 11×11, and 13×13, respectively, during the ground testing phase of the spaceborne camera. The preferred conditions are that the peak signal-to-noise ratio of the filled image reaches 35 dB or more and the single-frame processing delay does not exceed 100 milliseconds. After testing, only the 11×11 window simultaneously meets the above two requirements, so it is calibrated as 11×11.

[0061] Extract the grayscale values ​​of all known normal pixels within the receptive field window and their spatial distances to the missing pixels. If the spatial distance between a known normal pixel and the missing pixel is zero, use the grayscale value of that known normal pixel as the filling grayscale value for the missing pixel. If there is no known normal pixel with a distance of zero, multiply the grayscale value of each known normal pixel within the receptive field window by the reciprocal of the spatial distance between that pixel and the missing pixel, according to the principle that the closer the distance, the greater the weight. This yields the weighted grayscale value of that pixel. Sum all the weighted grayscale values ​​and divide by the sum of all the reciprocals of the spatial distances to obtain the filling grayscale value for the missing pixel. After traversing all the missing pixels in the texture missing region, fill all the filling grayscale values ​​into the corresponding positions in the data reconstruction matrix according to the original spatial positions of the missing pixels to obtain the filled complete image data. This complete image data is then identified as the reconstructed image data.

[0062] In summary, this invention achieves end-to-end intelligent storage and image integrity assurance for spaceborne cameras in complex space environments by dynamically quantifying radiation interference and adaptively allocating error correction resources, combined with underlying data error correction and image visual restoration.

[0063] The second embodiment of the present invention provides an intelligent storage system based on a spaceborne industrial camera, comprising: a memory, a processor, and a computer program stored in the memory and executable on the processor, wherein the processor executes the program to implement the method described above.

[0064] It should be noted that the intelligent storage system based on a spaceborne industrial camera provided in this embodiment of the invention is used to execute all the process steps of the intelligent storage method based on a spaceborne industrial camera in the above embodiment. The working principles and beneficial effects of the two are one-to-one, so they will not be described again.

[0065] It should be noted that the system embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. Those skilled in the art can understand and implement this without any creative effort.

[0066] The specific embodiments described above further illustrate the purpose, technical solution, and beneficial effects of the present invention. It should be understood that the above descriptions are merely specific embodiments of the present invention and are not intended to limit the scope of protection of the present invention. In particular, it should be noted that any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of the present invention should be included within the scope of protection of the present invention for those skilled in the art.

Claims

1. A smart storage method based on a spaceborne industrial camera, characterized in that, include: The operational telemetry data of the spaceborne camera and the physical topology of the storage array are obtained. The operational telemetry data is subjected to radiative interference quantization processing to obtain the flip probability distribution matrix. The flipped probability distribution matrix is ​​mapped to the physical topology to obtain a probability mapping spectrum, and the probability mapping spectrum is sampled to obtain a numerical sequence of bit error fluctuations. The error fluctuation value sequence is smoothed to determine the trend, and an error correction resource adjustment instruction is generated based on the determination result; The error fluctuation value sequence is statistically analyzed according to the error correction resource adjustment instruction to obtain the fault frequency record, and the fault frequency record is processed for resource allocation to obtain the resource allocation ratio. Obtain the location information of the block to be corrected reported by the storage array, and perform error correction coding processing on the block to be corrected according to the resource allocation ratio and the location information to obtain a data reconstruction matrix; The process involves obtaining image pixel data carried by the data reconstruction matrix, performing pixel anomaly detection on the image pixel data to obtain abnormal pixel markers, extracting the neighboring pixels corresponding to the abnormal pixel markers, performing spatial interpolation to fill in the neighboring pixels, and obtaining reconstructed image data. This includes: extracting pixel grayscale gradients from the image pixel data, and marking pixels with pixel grayscale gradients greater than a preset fluctuation threshold as abnormal pixel markers. A set of neighboring pixels is extracted centered on the abnormal pixel marker. Connectivity analysis is performed on the set of neighboring pixels to obtain an edge contour set. Spatial domain image tearing features are extracted based on the edge contour set. Pixel-level filling processing is performed on the spatial domain image tearing features to obtain the reconstructed image data.

2. The intelligent storage method based on a spaceborne industrial camera according to claim 1, characterized in that, The step of performing radiative interference quantization processing on the operational telemetry data to obtain the flip probability distribution matrix includes: The radiated interference intensity sequence is obtained by parsing the telemetry data, and an interference feature set is extracted based on the radiated interference intensity sequence; The interference feature set is subjected to energy deposition mapping to obtain local energy deposition values. If the local energy deposition values ​​are greater than a preset interference threshold, it is determined that a medium flip has occurred, and a medium flip state record is obtained. Based on the medium flipping state record, flipping probability statistics are performed to obtain the flipping probability distribution matrix.

3. The intelligent storage method based on a spaceborne industrial camera according to claim 1, characterized in that, The step of performing state sampling on the probability mapping spectrum to obtain a numerical sequence of bit error rate fluctuations includes: The probability mapping spectrum is sampled within a continuous time window to obtain a set of multi-frame state slices; The underlying physical error state is obtained by parsing the multi-frame state slice set, and data bit anomaly statistics are performed on the underlying physical error state to obtain the error fluctuation numerical sequence.

4. The intelligent storage method based on a spaceborne industrial camera according to claim 1, characterized in that, The step of smoothing the trend of the bit error fluctuation numerical sequence and generating an error correction resource adjustment instruction based on the determination result includes: The bit error fluctuation sequence is smoothed to obtain a time-series smoothed sequence; Extract the statistical features of the time-series smoothed sequence, and predict the bit error rate based on the statistical features to obtain a set of predicted values; If the predicted value in the predicted value set is greater than the preset surge judgment threshold, the corresponding storage unit identifier is extracted, and an error correction resource adjustment instruction is generated based on the storage unit identifier.

5. The intelligent storage method based on a spaceborne industrial camera according to claim 1, characterized in that, The process of allocating resources to the fault frequency records to obtain the resource allocation ratio includes: An error distribution density matrix is ​​constructed based on the fault frequency records, and spatial filtering is performed on the error distribution density matrix to obtain the error distribution density value. If the error distribution density value is greater than the preset density warning threshold, then the corresponding resource quota base is matched according to the error distribution density value; The resource allocation ratio is calculated based on the resource quota base and the error distribution density.

6. The intelligent storage method based on a spaceborne industrial camera according to claim 1, characterized in that, The step of performing error correction encoding processing on the block to be corrected based on the resource allocation ratio and the location information to obtain a data reconstruction matrix includes: Construct an error location vector based on the resource allocation ratio and the location information; A generation matrix is ​​constructed based on the error location vector, and polynomial verification is performed on the block to be corrected using the generation matrix to obtain a verification fragment. If the verification segment indicates that there is a data error, the error evaluation polynomial is extracted according to the verification segment, and error correction processing is performed according to the error evaluation polynomial to obtain the data reconstruction matrix.

7. The intelligent storage method based on a spaceborne industrial camera according to claim 2, characterized in that, The operational telemetry data includes radiation dose rate data or single-particle event count data.

8. An intelligent storage system based on a spaceborne industrial camera, comprising a memory, a processor, and a computer program stored in the memory and executable on the processor, characterized in that, When the processor executes the program, it implements the method according to any one of claims 1 to 7.

9. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the program is executed by the processor, it implements the method described in any one of claims 1 to 7.

Citation Information

Patent Citations

  • Unmanned aerial vehicle detection method and system based on visible light polarization imaging

    CN120652460A

  • Dynamic carbon sink accounting system based on multi-modal ai remote sensing monitoring and blockchain-based evidence storage

    WO2026061545A1