Radiation calibration-based multispectral image variable-depth adaptive mapping method and system

Abstract

The application belongs to the technical field of deep space exploration, and particularly relates to a multispectral image variable-depth self-adaptive mapping method and system based on radiation calibration, which comprises the following steps: acquiring original black field data with an input depth of N and original flat field data with an input depth of N; calculating row offset and row response after shielding bad pixels according to the original black field data and the original flat field data; sequentially determining a global scale according to the row offset and the row response, calculating an integrated slope, and generating a lookup table according to the integrated slope; acquiring real-time monitoring input, reading an M-bit output value corresponding to the real-time monitoring input according to the lookup table, and correcting and updating the lookup table according to the M-bit output value. Based on the parameterized calculation process (derivation of the global scale and the integrated slope from the row offset and the row response), different input depth and output depth requirements can be adapted by modifying input parameters, avoiding the fixed limitations caused by hard coding, and improving the configurability of cross-platform tasks.

Description

Technical Field

[0001] The present invention belongs to the technical field of deep space exploration, and particularly relates to a multi-spectral image variable depth adaptive mapping method and system based on radiometric calibration. Background Technique

[0002] With the in-depth development of deep space exploration missions (such as lunar, Mars, and asteroid exploration), multi-spectral imaging technology has become a key means to obtain the surface material composition, mineral distribution, and evolutionary history of celestial bodies. To ensure the quantitative accuracy of spectral analysis, deep space multi-spectral payloads usually require extremely high radiation resolution, and the original data bit depth output by their detectors is usually N bits (such as 14 bits or 16 bits). However, limited by the extremely limited downlink communication bandwidth and storage resources between the detector and the ground station, these high-bit-depth original data must be subjected to variable depth mapping processing at the payload end before being downloaded, and compressed to the target bit depth M bits (M < N, such as 8 bits or 10 bits).

[0003] In the existing deep space multi-spectral data processing system, the generally known process of "ground test calibration + step-by-step execution at the payload end" is usually followed:

[0004] Ground test calibration stage: In a controlled laboratory environment, multiple frames of black field and flat field calibration sequences are collected using a standard light source such as an integrating sphere. By calculating the arithmetic mean of the sequences, the offset and gain parameters of each pixel in each band of the detector are calculated and stored in the payload memory as preset calibration parameters.

[0005] Payload end step-by-step execution stage: During the transmission of real-shot images, the processing unit first calls the calibration parameters to perform floating-point or high-bit-width multiplication and addition operations on the original N-bit pixels to complete non-uniformity correction; subsequently, the corrected result is then subjected to secondary quantization or global linear scaling, and mapped into M-bit data for packet download.

[0006] Although this process can complete basic data compression, there are significant technical bottlenecks when facing the extreme deep space environment and high-precision scientific inversion requirements.

[0007] The above existing step-by-step processing architecture has the following defects and deficiencies in engineering practice, which directly limit the scientific detection efficiency of the detection payload:

[0008] 1. Computational overhead and resource redundancy in real-time processing at the payload end

[0009] In existing step-by-step processing flows, radiometric correction and bit-depth mapping are two logically separate steps. During transmission, the hardware must first perform complex scaling arithmetic operations on each pixel before quantization and compression. This "calculate first, compress later" serial architecture generates significant computational latency and consumes a large amount of Field-Programmable Gate Array (FPGA) multiplier resources when processing multispectral, high-frame-rate images. Given the limited resources of deep space probes, this redundant computation not only increases power consumption but also reduces the system's real-time processing capabilities.

[0010] 2. Insufficient robustness of calibration parameter extraction.

[0011] During ground-based testing or in-orbit calibration, calibration sequences are susceptible to environmental noise in the laboratory, cosmic rays, or transient anomalies in the detector. Existing techniques often employ simple arithmetic averaging to extract parameters, failing to identify and remove outliers (such as hot pixels or spike noise) from the sequence. This leads to contamination of the extracted bias and gain parameters, resulting in artificial residual bands or pseudo-textures appearing in the image when these parameters are subsequently used for mapping processing, thus reducing the radiometric accuracy of the data.

[0012] 3. The information fidelity of weak signal bands in bit depth compression is low.

[0013] The transmittance of different channels in a multispectral detector varies significantly. Current techniques for deep mapping (N-to-M) often neglect the impact of these channel response differences on quantization noise. If a simple global scaling method is used, weaker bands will occupy only a small number of quantization levels after mapping, resulting in a severe loss of signal-to-noise ratio (SNR). Conversely, if independent channel enhancement is used (each band filling the entire range), the physical radiation ratio between different bands will be completely disrupted, leading to spectral distortion and making quantitative mineral spectral analysis impossible.

[0014] 4. Static calibration architecture cannot adapt to dynamic deviations caused by environmental temperature drift.

[0015] Dramatic temperature variations in deep space environments (such as those encountered during lunar exploration) can cause real-time drift in the dark current and gain of probes. Current technologies primarily rely on static calibration parameters acquired from the ground, which is an open-loop processing mode. When the physical state of the probe experiences temperature drift, the preset parameters become mismatched with the actual state. Since compression mapping has already been completed, the ground-based system cannot effectively compensate for the stripe noise in this "pressed-in" mapped data, severely impacting the effective image quality.

[0016] 5. The mathematical description of deep displacement mapping relationships lacks universality.

[0017] Existing technologies typically employ hard-coded logic designs for fixed input bit depth N and output bit depth M. When communication bandwidth requirements are adjusted or detection accuracy requirements change, necessitating dynamic changes in bit depth M, existing mathematical models struggle to provide a universal mapping mechanism that simultaneously balances radiation calibration accuracy and quantization noise optimization, resulting in poor cross-platform compatibility. Summary of the Invention

[0018] The purpose of this invention is to provide a method and system for adaptive mapping of multispectral image displacement depth based on radiometric calibration, so as to solve the problems of resource redundancy, low accuracy and poor adaptability of the existing step-by-step serial architecture.

[0019] To achieve the above objectives, the present invention adopts the following technical solution:

[0020] In a first aspect, the present invention provides a method for adaptive mapping of multispectral image displacement depth based on radiometric calibration, comprising:

[0021] Obtain the original black field data and the original flat field data with an input bit depth of N;

[0022] After masking bad pixels based on the original black field data and the original flat field data, calculate the row offset and row response;

[0023] The global scale is determined sequentially based on row offset and row response, the ensemble slope is calculated, and a lookup table is generated based on the ensemble slope.

[0024] Obtain the real-time monitoring input, read the corresponding M-bit output value according to the lookup table, and update the lookup table according to the M-bit output value.

[0025] Preferably, the original black field data with an input bit depth of N is continuously acquired in a completely dark environment for no less than 50 frames; the original flat field data with an input bit depth of N is acquired under uniform light source illumination, with the brightness adjusted to the maximum value within the linear range of the photodetector, and continuously acquired for no less than 50 frames.

[0026] Preferably, the step of calculating the line offset and line response after masking bad pixels based on the original black field data and the original flat field data includes:

[0027] Determine bad pixels based on the original black field data and the original flat field data, and generate a bad pixel mask based on the bad pixels;

[0028] On the bad pixel mask, the row offset B is calculated for each row of black data using Sigma-Clipping;

[0029] For each row of flat data, the row response W is calculated using Sigma-Clipping.

[0030] Preferably, the criterion for judging bad pixels based on the original black field data and the original flat field data is as follows:

[0031] Within the same row, if a pixel is in a flat field, the average DN value of the pixel is < 0.05 × 2. N Or, at black levels, the average DN value of the pixels is >0.2×2. N If the pixel is not a bad pixel, then the pixel is a bad pixel.

[0032] Preferably, the step of calculating the row bias B for each row of black data using Sigma-Clipping includes:

[0033] Calculate the initial average value µ of the black level data for the valid pixels in the same row. B and the standard deviation σ of the black field data B ;

[0034] Keep all those falling within [µ] B -3σ B µ B +3σ B DN values ​​within the interval, remove outliers;

[0035] Calculate [µ] B -3σ B µ B +3σ B The row offset B is obtained by averaging the DN values ​​within the interval.

[0036] The calculation of the row response W using Sigma-Clipping for each row of flat data includes:

[0037] Calculate the initial average value µ of the flat field data for the effective pixels in the same row. W Standard deviation σ of flat field data W ;

[0038] Keep all those falling within [µ] W -3σ W µ W +3σ W ] DN values ​​within the interval, remove outliers;

[0039] Calculate [µ] W -3σ W µ W +3σ W The average value of DN within the interval is used to obtain the row response W.

[0040] Preferably, the step of determining the global scale, calculating the integration slope, and generating a lookup table based on the integration slope, in sequence according to the row offset and row response, includes:

[0041] Iterate through all rows and find (2N The maximum value of -1 - B) / R is denoted as the global scale Max; where R = WB;

[0042] Calculate the slope K of each row = (2 M -1) / (Max × R;

[0043] For each row, pre-calculate 0 to (2) using the following formula. N -1) The corresponding output yields the lookup table:

[0044] DN out = round(K ×(DN in - B))

[0045] In the formula, N is the bit depth of the input image; M is the output bit depth; W is the line response; B is the line offset; K is the slope; Max is the global scale; R is the line sensitivity including transmittance and response; DN out For output value; DN in The input value is 'round'; 'round' indicates rounding to the nearest integer.

[0046] Preferably, the step of acquiring real-time monitoring input, reading the M-bit output value corresponding to the real-time monitoring input according to the lookup table, and correcting and updating the lookup table according to the M-bit output value includes:

[0047] Obtain real-time monitoring input and read the corresponding M-bit output value from the lookup table based on the row coordinates of the real-time monitoring input;

[0048] Calculate the average brightness of each line of the current frame's M-bit output value. ;

[0049] Based on average brightness Calculate the expected target value :

[0050]

[0051] Based on average brightness and expected target value The criterion for determining whether temperature drift bands exist in each row is: if If so, then the line is determined to have a temperature drift band;

[0052] For rows with temperature drift bands, calculate the correction factor according to the following formula, update the row parameters based on the correction factor, and refresh the entries in the lookup table:

[0053]

[0054]

[0055] In the formula, This represents the expected target value for the i-th row; This represents the average brightness of the i-th row; This represents the average brightness of the (i-1)th row; This represents the average brightness of the (i-2)th row; This represents the average brightness of the (i+1)th row; This represents the average brightness of the (i+2)th row; This indicates that the median of the values ​​in parentheses is taken. Indicates the correction factor; This represents the slope of the i-th row after the new correction; This represents the slope of the i-th row before the correction.

[0056] In a second aspect, the present invention provides a multispectral image displacement depth adaptive mapping system based on radiometric calibration, comprising:

[0057] The ground calibration unit is used to acquire the original black field data with an input depth of N and the original flat field data with an input depth of N.

[0058] The robust parameter calculation unit is used to calculate the line offset and line response after masking bad pixels based on the original black field data and the original flat field data.

[0059] The lookup table generation unit is used to determine the global scale and calculate the integration slope in sequence based on the row offset and row response, and generate a lookup table based on the integration slope.

[0060] The real-time execution and dynamic closed-loop feedback unit is used to acquire real-time monitoring input, read the corresponding M-bit output value of the real-time monitoring input according to the lookup table, and correct and update the lookup table according to the M-bit output value.

[0061] In a third aspect, the present invention provides an electronic device including a processor and a memory, the processor being configured to execute a computer program stored in the memory to implement the radiometrically calibrated multispectral image displacement depth adaptive mapping method described in any one of the preceding claims.

[0062] In a fourth aspect, the present invention provides a computer-readable storage medium storing at least one instruction that, when executed by a processor, implements the radiometrically calibrated multispectral image displacement depth adaptive mapping method described in any one of the preceding claims.

[0063] Compared with the prior art, the beneficial effects of the present invention are as follows:

[0064] This application discloses a radiometrically calibrated multispectral image depth-adaptive mapping method that integrates radiometric correction and depth mapping by generating an integrated slope based on row offset and row response and constructing a lookup table. During execution, the payload only needs to read the output value through the lookup table, eliminating the need for real-time multiplication and addition operations, thus significantly reducing FPGA logic resource consumption and dynamic power consumption. In the parameter extraction stage, by masking bad pixels and calculating row offset and row response separately, the contamination of calibration parameters by abnormal pixels is avoided, ensuring the accuracy of the reference data used for subsequent mapping and eliminating image pseudo-textures at the source. By determining a global scale based on row offset and row response and calculating an asymmetric integrated slope, differentiated mapping curves are generated for channels with different response characteristics, ensuring that weak signal bands still occupy sufficient quantization steps after compression, thereby preserving the scientific information of weak signals. A closed-loop feedback mechanism is constructed by acquiring real-time monitoring input and updating the lookup table based on the output value. This enables the system to dynamically adjust mapping parameters according to the statistical characteristics of the current image output, compensating online for radiometric deviations caused by environmental changes such as temperature drift. This method is based on a parameterized computation process (derived from row bias, row response to global scale and integrated slope). By modifying the input parameters, it can adapt to different input bit depth and output bit depth requirements, avoiding the inherent limitations of hard coding and improving the configurability of cross-platform tasks. Attached Figure Description

[0065] The accompanying drawings, which form part of this application, are used to provide a further understanding of the invention. The illustrative embodiments of the invention and their descriptions are used to explain the invention and do not constitute an undue limitation of the invention. In the drawings:

[0066] Figure 1 This is a schematic diagram of the method flow according to an embodiment of the present invention;

[0067] Figure 2 This is a system structure block diagram according to an embodiment of the present invention;

[0068] Figure 3 This is a structural block diagram of an electronic device according to an embodiment of the present invention. Detailed Implementation

[0069] The present invention will now be described in detail with reference to the accompanying drawings and embodiments. It should be noted that, unless otherwise specified, the embodiments and features described in this application can be combined with each other.

[0070] The following detailed description is exemplary and intended to provide further detailed explanation of the invention. Unless otherwise specified, all technical terms used in this invention have the same meaning as commonly understood by one of ordinary skill in the art to which this application pertains. The terminology used in this invention is for the purpose of describing particular embodiments only and is not intended to limit the scope of exemplary embodiments according to the invention.

[0071] The DN value is the raw value directly recorded by the remote sensing sensor; its full name is Digital Number. It represents the radiance of a ground target detected by the sensor in that wavelength band, and is usually expressed as a grayscale value.

[0072] Sigma-Clipping is an iterative outlier filtering algorithm based on data distribution (standard deviation).

[0073] See Figure 1 This application discloses a multispectral image displacement depth adaptive mapping method based on radiometric calibration, including:

[0074] S1: Obtain the original black field data and the original flat field data with an input bit depth of N;

[0075] S2: After masking bad pixels based on the original black field data and the original flat field data, calculate the line offset and line response;

[0076] S3: Determine the global scale and calculate the integration slope in sequence based on the row offset and row response, and generate a lookup table based on the integration slope;

[0077] S4: Obtain the real-time monitoring input, read the corresponding M-bit output value of the real-time monitoring input according to the lookup table, and update the lookup table according to the M-bit output value.

[0078] First, it significantly reduces the computational burden on the payload, transforming the original serial operation of calibration followed by compression into a lookup table, saving substantial multiplier resources and power consumption in the FPGA, allowing the photodetector to operate more smoothly under limited hardware conditions. Second, it improves the reliability of calibration parameters because bad pixels and abnormal noise are removed during preprocessing, resulting in cleaner biases and responses, preventing noise from being introduced into the image during subsequent mapping. Third, it preserves the scientific value of weak signal bands. By calculating different mapping slopes for channels with different response characteristics, even bands with weak signals can still occupy sufficient quantization levels after compression, preventing them from being submerged. Fourth, it achieves self-adaptation to changes in the space environment. By monitoring the output in real time and dynamically correcting the lookup table, the system can automatically compensate for drift-induced deviations during drastic temperature fluctuations, avoiding the failure of traditional static calibration parameters. Fifth, it has high adaptability; the input and output bit depths can be configured, allowing a single architecture to adapt to the needs of different detection missions, from high-precision mineral composition analysis to high-frame-rate terrain surveys.

[0079] In some embodiments, the original black field data with an input bit depth of N is continuously acquired in a completely dark environment for no less than 50 frames; the original flat field data with an input bit depth of N is acquired under uniform light source illumination, with the brightness adjusted to the maximum value within the linear range of the photodetector, and continuously acquired for no less than 50 frames, ensuring that the signal is strong enough and that the signal-to-noise ratio is in the optimal range.

[0080] In some embodiments, the step of calculating the line offset and line response after masking bad pixels based on the original black field data and the original flat field data includes:

[0081] Bad pixels are identified based on the original black field data and the original flat field data, and a bad pixel mask is generated based on the bad pixels. This avoids the problem of bad pixels dragging down the parameters of the entire row in traditional methods, and eliminates fixed stripes in the image from the root.

[0082] On the bad pixel mask, Sigma-Clipping is used to calculate the row bias B for each row of black field data. Sigma-Clipping eliminates these outliers one by one through multiple rounds of iteration, and what remains are physically consistent real signals, resulting in cleaner and more reliable parameters.

[0083] The row response W is calculated using Sigma-Clipping for each row of flat-field data. Calculating by row precisely captures the actual physical characteristics of the photodetector readout circuit, controlling the computational load while preserving the necessary correction accuracy.

[0084] In some embodiments, the criterion for determining bad pixels based on the original black field data and the original flat field data is as follows:

[0085] Within the same row, if a pixel is in a flat field, the average DN value of the pixel is < 0.05 × 2. N Or, at black levels, the average DN value of the pixels is >0.2×2. N If the pixel is not found in the image, then it is considered a bad pixel. This transforms bad pixel identification into a quantifiable and automatically executable rule, eliminating the need for manual inspection of each frame to determine which pixel is problematic. This significantly improves the automation and repeatability of the calibration process. Furthermore, the averaging of 100 frames eliminates some random noise interference, greatly reducing the false positive rate compared to using single-frame data.

[0086] In some embodiments, calculating the row bias B for each row of black data using Sigma-Clipping includes:

[0087] Calculate the initial average value µ of the black level data for the valid pixels in the same row. B and the standard deviation σ of the black field data B ;

[0088] Keep all those falling within [µ] B -3σ B µ B +3σ B ] DN values ​​within the interval, remove outliers;

[0089] Calculate [µ] B -3σ B µ B +3σ B The row offset B is obtained by averaging the DN values ​​within the interval.

[0090] The calculation of the row response W using Sigma-Clipping for each row of flat data includes:

[0091] Calculate the initial average value µ of the flat field data for the effective pixels in the same row. W Standard deviation σ of flat field data W ;

[0092] Keep all those falling within [µ] W -3σ W µ W +3σ W ] DN values ​​within the interval, remove outliers;

[0093] Calculate [µ] W -3σ W µ W +3σ WThe average value of DN within the interval is used to obtain the row response W. The Sigma-Clipping algorithm makes the final output adaptive and highly robust. The two core parameters B and W, which are input to the subsequent lookup table generation stage, are of very high quality, ensuring the radiation accuracy of the entire displacement deep mapping process from the source.

[0094] In some embodiments, the step of determining the global scale, calculating the integration slope, and generating a lookup table based on the integration slope in sequence according to the row offset and row response includes:

[0095] Iterate through all rows and find (2 N The maximum value of -1 - B) / R is denoted as the global scale Max; where R = WB;

[0096] Calculate the slope K of each row = (2 M -1) / (Max × R;

[0097] For each row, pre-calculate 0 to (2) using the following formula. N -1) The corresponding output yields the lookup table:

[0098] DN out = round(K ×(DN in - B))

[0099] In the formula, N is the bit depth of the input image; M is the output bit depth; W is the line response; B is the line offset; K is the slope; Max is the global scale; R is the line sensitivity including transmittance and response; DN out For output value; DN in The input value is 'round'; 'round' represents the rounding operation. This scheme, through the unified constraint of the global scale Max, ensures that all row mappings are based on the same physical reference. The compressed data still maintains the original radiation ratio, meeting the spectral fidelity requirements of scientific inversion. The slope K calculation formula in the scheme naturally allocates a larger amplification factor to weak response bands, effectively suppressing quantization noise and significantly improving the signal-to-noise ratio of weak signal bands. By pre-generating a lookup table, complex multiplication and addition operations are transformed into simple table lookup operations. The FPGA at the payload end no longer needs to call the multiplier for floating-point calculations in real time, greatly reducing hardware resource consumption and dynamic power consumption. It also eliminates the processing delay caused by the serial architecture of correction followed by compression, improving the real-time response capability of deep space exploration missions. The parameters are flexibly configurable; if the communication bandwidth changes and the output bit depth needs to be changed from 8 bits to 10 bits, only the value of M needs to be modified, and the entire calculation process automatically adapts without redesigning the hardware logic.

[0100] In some embodiments, the step of acquiring real-time monitoring input, reading the M-bit output value corresponding to the real-time monitoring input according to a lookup table, and correcting and updating the lookup table according to the M-bit output value includes:

[0101] Obtain real-time monitoring input and read the corresponding M-bit output value from the lookup table based on the row coordinates of the real-time monitoring input;

[0102] Calculate the average brightness of each line of the current frame's M-bit output value. ;

[0103] Based on average brightness Calculate the expected target value :

[0104]

[0105] Based on average brightness and expected target value The criterion for determining whether temperature drift bands exist in each row is: if If so, then the line is determined to have a temperature drift band;

[0106] For rows with temperature drift bands, calculate the correction factor according to the following formula, update the row parameters based on the correction factor, and refresh the entries in the lookup table:

[0107]

[0108]

[0109] In the formula, This represents the expected target value for the i-th row; This represents the average brightness of the i-th row; This represents the average brightness of the (i-1)th row; This represents the average brightness of the (i-2)th row; This represents the average brightness of the (i+1)th row; This represents the average brightness of the (i+2)th row; This indicates that the median of the values ​​in parentheses is taken. Indicates the correction factor; This represents the slope of the i-th row after the new correction; This represents the slope of the i-th row before the correction.

[0110] Example 1

[0111] A radiometrically calibrated multispectral image displacement depth adaptive mapping method includes:

[0112] Phase 1: Ground Calibration

[0113] Black frame capture: With the lens cap on, continuously capture 100 frames in a completely dark environment. Obtain raw black frame data with an input bit depth of N;

[0114] Flat field acquisition: Uniform light source illumination, brightness adjusted to 80% of full scale (approximately 0.8 × 2). N The system continuously acquires 100 frames to obtain raw flat-field data with an input bit depth of N. The 80% setting is to avoid photodetector saturation and nonlinear distortion. When a photodetector approaches full scale, its photoelectric conversion response typically loses linearity (i.e., doubling the input photons does not necessarily double the output DN value). Here, the brightness is adjusted to the maximum value within the photodetector's linear range, typically 80%.

[0115] Phase Two: Robust Parameter Calculation

[0116] 1. Bad Pixel Masking: Compare black and flat areas. Within the same row, if a pixel has an average DN value < 0.05 × 2 when flat. N Or, at black levels, the average DN value of the pixels is >0.2×2. N If it does not, then mark it as a bad pixel.

[0117] Result: A bad pixel mask is generated, and subsequent calculations skip these points.

[0118] Calculate row bias: Apply Sigma-Clipping to each row of black data (remove noise that deviates from the mean by more than 3 standard deviations).

[0119] (1) Calculate the initial average value µ of the black field data for all valid pixels in the row. B and the standard deviation σ of the black field data B ;

[0120] (2) Perform σ-truncation. Retain all values ​​falling within [µ]. B -3σ B µ B +3σ B DN values ​​within the interval, removing outliers.

[0121] (3) Calculate the final average value of the remaining pixels to obtain the row offset B of the row.

[0122] B = The average value of the valid pixels in this row.

[0123] Calculate row response: Sigma-Clipping is also used for each row of flat data (to remove noise that deviates from the mean by more than three standard deviations).

[0124] (1) Calculate the initial average value µ of the flat field data for all valid pixels in the row. W Standard deviation σ of flat field dataW ;

[0125] (2) Perform σ-truncation. Retain all values ​​falling within [µ]. W -3σ W µ W +3σ W DN values ​​within the interval, outliers removed.

[0126] (3) Calculate the final average value of the remaining pixels to obtain the row response W of the row.

[0127] R = WB.

[0128] R represents the line photosensitivity, including transmittance and response.

[0129] Phase 3: Generation of Integrated Lookup Tables

[0130] 1. Determine the global scale: Traverse all rows and find (2 N The maximum value of (-1 - B) / R is denoted as the global scale Max. N is the bit depth of the input image.

[0131] 2. Calculate the integration slope: The slope K of each row = (2 M -1) / (Max × R), where M is the output bit depth;

[0132] In rows with low transmittance, the slope R is small, and the slope K automatically increases (high stretching ratio), thereby compensating for energy loss.

[0133] 3. Generate a lookup table: For each row, pre-calculate 0 to (2... N -1) corresponding output:

[0134] DN out = round(K ×(DN in - B))

[0135] A photodetector row number × 2 N A matrix of columns;

[0136] In the formula, N is the bit depth of the input image; M is the output bit depth; W is the line response; B is the line offset; K is the slope; Max is the global scale; R is the line sensitivity including transmittance and response; DN out For output value; DN in The input value is 'round'; 'round' indicates rounding to the nearest integer.

[0137] Phase 4: Real-time execution and dynamic closed-loop feedback (temperature drift adaptive)

[0138] During instrument operation, the strip noise caused by temperature drift is dynamically corrected by monitoring the output results in real time.

[0139] 1. Real-time mapping

[0140] Input N-bit raw pixel values, and directly read the corresponding M-bit output values ​​from the lookup table matrix based on their row coordinates i.

[0141] 2. Closed-loop monitoring and prediction

[0142] Calculate the row mean: Calculate the average brightness of each row in the current frame's M-bit image. .

[0143] Spatial smoothing prediction: A 5-point median filter is used to smooth the mean sequence of each row to obtain the desired target value. .

[0144]

[0145] 3. Deviation Correction and Lookup Table Update

[0146] Calculation and judgment: If If so, it is determined that there is a temperature drift band in that row.

[0147] Correction factor: Calculation .

[0148] Instant update: Updates the parameters in this row. And only refresh the lookup table entries for that row.

[0149] Key advantage: No need for full calculation, only dynamic gain compensation is performed on the bands that generate the stripes.

[0150] In the formula, This represents the expected target value for the i-th row; This represents the average brightness of the i-th row; This represents the average brightness of the (i-1)th row; This represents the average brightness of the (i-2)th row; This represents the average brightness of the (i+1)th row; This represents the average brightness of the (i+2)th row; This indicates that the median of the values ​​in parentheses is taken. Indicates the correction factor; This represents the slope of the i-th row after the new correction; This represents the slope of the i-th row before the correction.

[0151] Phase 5: Noise Analysis

[0152] The noise in an M-bit image consists of two parts:

[0153] 1. Original noise σ NNoise inherent in the original N-bit image (readout noise, shot noise).

[0154] 2. Quantization noise σ q Because we forcibly categorize continuous values ​​into 0—(2 M The error generated at integer steps of -1). For uniform quantization, the standard deviation of the quantization noise is constant:

[0155] σq = (M-bit DN unit)

[0156] Signal-to-noise ratio loss calculation model

[0157] Original N-bit SNR: SNR N = (DN) in - B)\ σ N

[0158] Compressed M-bit SNR: SNR M =

[0159] (Note: DN out For M-bit output values, ×σ N (This is the original noise that was transmitted)

[0160] Signal-to-noise ratio loss ratio:

[0161] Loss = =

[0162] Where, σ N The original noise; σ q For quantization noise; SNR N The original N-bit signal-to-noise ratio loss; SNR M This represents the M-bit signal-to-noise ratio loss after compression; DN out Output value in M-bit format; ×σ N It is the original noise that has been transmitted; The slope is denoted as .

[0163] In summary, this application has the following improvements.

[0164] 1) Significantly reduced real-time computing overhead and power consumption at the payload end.

[0165] This invention pre-couples ground calibration parameters to generate an integrated lookup table that combines radiation calibration and displacement deep mapping.

[0166] Reasoning Process: Existing technologies require a serial logic process of "floating-point scaling followed by quantization and compression" during data transmission, involving a large number of multiplier resources. This invention, through mathematical pre-coupling, streamlines the complex... The calculation process is fixed in the lookup table.

[0167] Technical advantages: The payload hardware (such as FPGA) only needs to perform a single address index to complete both correction and compression simultaneously. This "memory-for-computation" architecture greatly simplifies the real-time processing chain, reduces logic resource consumption and dynamic power consumption, and improves the real-time processing efficiency of deep space exploration payloads.

[0168] 2) Effectively shields calibration source noise, ensuring the purity of calibration parameters.

[0169] This invention introduces a method in the calibration parameter extraction stage. -Truncation Robust Statistical Iterative Algorithm.

[0170] Reasoning Process: Calibration sequences in deep space environments often contain isolated noise points generated by cosmic rays. Traditional mean-based methods cannot remove these outliers, leading to deviations in the extracted bias B and response rate R. This invention iteratively removes outliers exceeding the mean. The outliers ensured the physical consistency of the statistical sample.

[0171] Technical advantages: This improvement eliminates false textures in images caused by "dirty parameters" at the source, significantly improving the reference accuracy of radiometric calibration and laying a solid foundation for subsequent high-precision spectral inversion.

[0172] 3) It preserves scientific information and signal-to-noise ratio in weak signal bands to the greatest extent possible.

[0173] This invention constructs a global physical scale. The slope K of the asymmetric gain enables the optimal allocation of quantified resources.

[0174] Reasoning process: Traditional global linear compression does not consider the response differences between channels. This invention utilizes the formula K = (2 M -1) / (Max × R), automatically assign a larger mapping slope K to channels with weaker responses.

[0175] Technical advantages: This allows the weak signal band to occupy more quantization steps when compressed to the M-bit space, ensuring that the quantization noise is always lower than the original noise of the photodetector. Compared with existing technologies, this improvement significantly enhances the signal-to-noise ratio fidelity of the weak signal band without compromising spectral proportion consistency.

[0176] 4) Real-time online compensation for dynamic temperature drift strips was achieved.

[0177] This invention innovatively introduces a closed-loop feedback mechanism based on a 5-point median filter at the mapping output.

[0178] Reasoning: Since the on-orbit temperature drift of ground test parameters is unpredictable, static calibration will inevitably fail. This invention uses real-time statistical analysis of the row average of the M-bit output image, employs 5-point median filtering to extract the desired target value reflecting the true spectral trend, and uses this to dynamically correct and obtain the new, corrected slope of the i-th row. .

[0179] Technical advantages: This improvement enables the system to have online self-sensing and error correction capabilities, which can instantly eliminate strip noise caused by temperature gradient changes, ensure the radiation uniformity of the transmitted image under extreme temperature difference conditions, and overcome the lag of traditional open-loop calibration schemes.

[0180] 5) Possesses extremely strong adaptability to deep tasks and architectural versatility.

[0181] The mapping model of this invention adopts a parameterized variable architecture ( ).

[0182] Inference process: By setting the input bit depth N and output bit depth M as configurable parameters, the global scale and slope calculation in the mapping model can be automatically adapted.

[0183] Technical advantages: Compared with the hard-coded logic of existing technologies, this invention can flexibly adjust the output bit depth according to the changes in communication bandwidth or the requirements of detection accuracy, realizing a single architecture that can adapt to various detection mission objectives (such as high-precision component analysis or high-frame-rate geomorphological survey), and has extremely high engineering application value.

[0184] Example 2

[0185] like Figure 2 As shown, based on the same inventive concept as the above embodiments, the present invention also provides a multispectral image displacement depth adaptive mapping system based on radiometric calibration, characterized in that it includes:

[0186] The ground calibration unit is used to acquire the original black field data with an input depth of N and the original flat field data with an input depth of N.

[0187] The robust parameter calculation unit is used to calculate the line offset and line response after masking bad pixels based on the original black field data and the original flat field data.

[0188] The lookup table generation unit is used to determine the global scale and calculate the integration slope in sequence based on the row offset and row response, and generate a lookup table based on the integration slope.

[0189] The real-time execution and dynamic closed-loop feedback unit is used to acquire real-time monitoring input, read the corresponding M-bit output value of the real-time monitoring input according to the lookup table, and correct and update the lookup table according to the M-bit output value.

[0190] Example 3

[0191] like Figure 3 As shown, the present invention also provides an electronic device 100 for implementing a radiometrically calibrated multispectral image displacement depth adaptive mapping method;

[0192] The electronic device 100 includes a memory 101, at least one processor 102, a computer program 103 stored in the memory 101 and executable on at least one processor 102, and at least one communication bus 104.

[0193] The memory 101 can be used to store the computer program 103. The processor 102 implements the steps of the radiometrically calibrated multispectral image displacement depth adaptive mapping method by running or executing the computer program stored in the memory 101 and calling the data stored in the memory 101.

[0194] The memory 101 may primarily include a program storage area and a data storage area. The program storage area may store the operating system, application programs required for at least one function (such as sound playback function, image playback function, etc.), etc.; the data storage area may store data created based on the use of the electronic device 100 (such as audio data), etc. In addition, the memory 101 may include non-volatile memory, such as hard disk, RAM, plug-in hard disk, smart media card (SMC), secure digital (SD) card, flash card, at least one disk storage device, flash memory device, or other non-volatile solid-state storage device.

[0195] At least one processor 102 may be a Central Processing Unit (CPU), or other general-purpose processors, digital signal processors (DSPs), application-specific integrated circuits (ASICs), field-programmable gate arrays (FPGAs), or other programmable logic devices, discrete gate or transistor logic devices, discrete hardware components, etc. Processor 102 may be a microprocessor or any conventional processor. Processor 102 is the control center of electronic device 100, connecting various parts of electronic device 100 via various interfaces and lines.

[0196] The memory 101 in the electronic device 100 stores multiple instructions to implement a radiometrically calibrated multispectral image displacement depth adaptive mapping method, and the processor 102 can execute multiple instructions to achieve the following:

[0197] S1: Obtain the original black field data and the original flat field data with an input bit depth of N;

[0198] S2: After masking bad pixels based on the original black field data and the original flat field data, calculate the line offset and line response;

[0199] S3: Determine the global scale and calculate the integration slope in sequence based on the row offset and row response, and generate a lookup table based on the integration slope;

[0200] S4: Obtain the real-time monitoring input, read the corresponding M-bit output value of the real-time monitoring input according to the lookup table, and update the lookup table according to the M-bit output value.

[0201] Example 4

[0202] If the modules / units integrated in the electronic device 100 are implemented as software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, all or part of the processes in the methods of the above embodiments of the present invention can also be implemented by a computer program instructing related hardware. The computer program can be stored in a computer-readable storage medium, and when executed by a processor, it can implement the steps of the various method embodiments described above. The computer program includes computer program code, which can be in the form of source code, object code, executable files, or certain intermediate forms. The computer-readable medium can include: any entity or device capable of carrying computer program code, recording media, USB flash drives, portable hard drives, magnetic disks, optical disks, computer memory, and read-only memory (ROM).

[0203] Those skilled in the art will understand that embodiments of the present invention can be provided as methods, systems, or computer program products. Therefore, the present invention can take the form of a completely hardware embodiment, a completely software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention can take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0204] This invention is described with reference to flowchart illustrations and / or block diagrams of methods, apparatus (systems), and computer program products according to embodiments of the invention. It will be understood that each block of the flowchart illustrations and / or block diagrams, and combinations of blocks in the flowchart illustrations and / or block diagrams, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, special-purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, generate instructions for implementing the flowchart illustrations and / or block diagrams. Figure 1 One or more processes and / or boxes Figure 1 A device that provides the functions specified in one or more boxes.

[0205] These computer program instructions may also be stored in a computer-readable storage medium that can direct a computer or other programmable data processing device to function in a particular manner, such that the instructions stored in the computer-readable storage medium produce an article of manufacture including instruction means, which are implemented in a process Figure 1 One or more processes and / or boxes Figure 1 The function specified in one or more boxes.

[0206] These computer program instructions may also be loaded onto a computer or other programmable data processing equipment to cause a series of operational steps to be performed on the computer or other programmable equipment to produce a computer-implemented process, thereby providing instructions that execute on the computer or other programmable equipment for implementing the process. Figure 1 One or more processes and / or boxes Figure 1 The steps of the function specified in one or more boxes.

[0207] In the description of this specification, references to terms such as "an embodiment," "example," "specific example," etc., indicate that a specific feature, structure, material, or characteristic described in connection with that embodiment or example is included in at least one embodiment or example of the invention. In this specification, illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples.

[0208] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and not to limit it. Although the present invention has been described in detail with reference to the above embodiments, those skilled in the art should understand that modifications or equivalent substitutions can still be made to the specific implementation of the present invention. Any modifications or equivalent substitutions that do not depart from the spirit and scope of the present invention should be covered within the scope of protection of the claims of the present invention.

Claims

1. A method for adaptive mapping of multispectral image displacement depth based on radiometric calibration, characterized in that, include: Obtain the original black field data and the original flat field data with an input bit depth of N; After masking bad pixels based on the original black field data and the original flat field data, the line offset and line response are calculated; specifically including: Determine bad pixels based on the original black field data and the original flat field data, and generate a bad pixel mask based on the bad pixels; On the bad pixel mask, the row offset B is calculated for each row of black data using Sigma-Clipping; For each row of flat data, the row response W is calculated using Sigma-Clipping; The global scale is determined sequentially based on row offset and row response, the ensemble slope is calculated, and a lookup table is generated based on the ensemble slope. Obtain the real-time monitoring input, read the corresponding M-bit output value according to the lookup table, and update the lookup table according to the M-bit output value.

2. The method for adaptive mapping of multispectral image displacement depth based on radiometric calibration according to claim 1, characterized in that, The original black field data with an input bit depth of N is continuously acquired for no less than 50 frames in a completely dark environment; the original flat field data with an input bit depth of N is acquired continuously for no less than 50 frames under uniform light illumination with the brightness adjusted to the maximum value within the linear range of the photodetector.

3. The method for adaptive mapping of multispectral image displacement depth based on radiometric calibration according to claim 1, characterized in that, The criteria for determining bad pixels based on the original black field data and the original flat field data are as follows: Within the same row, if a pixel is in a flat field, the average DN value of the pixel is < 0.05 × 2. N Or, in black, the average DN value of the pixel is >0.2×2. N If the pixel is not a bad pixel, then the pixel is a bad pixel.

4. The method for adaptive mapping of multispectral image displacement depth based on radiometric calibration according to claim 1, characterized in that, The calculation of the row offset B using Sigma-Clipping for each row of black data includes: Calculate the initial average value µ of the black level data for the valid pixels in the same row. B and the standard deviation σ of the black field data B ; Keep all those falling within [µ] B -3σ B µ B +3σ B ] DN values ​​within the interval, remove outliers; Calculate [µ] B -3σ B µ B +3σ B The row offset B is obtained by averaging the DN values ​​within the interval. The calculation of the row response W using Sigma-Clipping for each row of flat data includes: Calculate the initial average value µ of the flat field data for the effective pixels in the same row. W Standard deviation σ of flat field data W ; Keep all those falling within [µ] W -3σ W µ W +3σ W ] DN values ​​within the interval, remove outliers; Calculate [µ] W -3σ W µ W +3σ W The average value of DN within the interval is used to obtain the row response W.

5. The method for adaptive mapping of multispectral image displacement depth based on radiometric calibration according to claim 1, characterized in that, The process of determining the global scale and calculating the integration slope based on row offset and row response, and generating a lookup table based on the integration slope, includes: Iterate through all rows and find (2 N The maximum value of -1 - B) / R is denoted as the global scale Max; where R = WB; Calculate the slope K of each row = (2 M -1) / (Max × R; For each row, pre-calculate 0 to (2) using the following formula. N -1) The corresponding output yields the lookup table: DN out = round(K ×(DN in - B)) In the formula, N is the bit depth of the input image; M is the output bit depth; W is the line response; B is the line offset; K is the slope; Max is the global scale; R is the line sensitivity including transmittance and response; DN out For output value; DN in The input value is 'round'; 'round' indicates rounding to the nearest integer.

6. The method for adaptive mapping of multispectral image displacement depth based on radiometric calibration according to claim 1, characterized in that, The process of acquiring real-time monitoring input, reading the corresponding M-bit output value according to a lookup table, and updating the lookup table based on the M-bit output value includes: Obtain real-time monitoring input and read the corresponding M-bit output value from the lookup table based on the row coordinates of the real-time monitoring input; Calculate the average brightness of each line of the current frame's M-bit output value. ; Based on average brightness Calculate the expected target value : Based on average brightness and expected target value The criterion for determining whether temperature drift bands exist in each row is: if If so, it is determined that the row contains temperature drift bands; For rows with temperature drift bands, calculate the correction factor according to the following formula, update the row parameters based on the correction factor, and refresh the entries in the lookup table: In the formula, This represents the expected target value for the i-th row; This represents the average brightness of the i-th row; This represents the average brightness of the (i-1)th row; This represents the average brightness of the (i-2)th row; This represents the average brightness of the (i+1)th row; This represents the average brightness of the (i+2)th row; This indicates that the median of the values ​​in parentheses is taken. Indicates the correction factor; This represents the slope of the i-th row after the new correction; This represents the slope of the i-th row before the correction.

7. A multispectral image displacement depth adaptive mapping system based on radiometric calibration, characterized in that, include: The ground calibration unit is used to acquire the original black field data with an input depth of N and the original flat field data with an input depth of N. The robust parameter calculation unit is used to calculate the line offset and line response after masking bad pixels based on the original black field data and the original flat field data. The lookup table generation unit is used to determine the global scale and calculate the integration slope in sequence based on the row offset and row response, and generate a lookup table based on the integration slope. The real-time execution and dynamic closed-loop feedback unit is used to acquire real-time monitoring input, read the corresponding M-bit output value of the real-time monitoring input according to the lookup table, and correct and update the lookup table according to the M-bit output value.

8. An electronic device, characterized in that, It includes a processor and a memory, the processor being used to execute a computer program stored in the memory to implement the radiometrically calibrated multispectral image displacement depth adaptive mapping method as described in any one of claims 1 to 6.

9. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores at least one instruction that, when executed by a processor, implements the radiometrically calibrated multispectral image displacement depth adaptive mapping method as described in any one of claims 1 to 6.

Images