Marine plankton hyperspectral imaging detection system

Abstract

The invention discloses a marine plankton hyperspectral imaging detection system, relates to the technical field of spectrum detection, and is used for solving the problem of poor spectrum classification and recognition in a water body environment. According to the method, the hyperspectral image and the environmental disturbance parameters are synchronously acquired, and the disturbance mapping sequence is constructed, so that accurate alignment between the image and disturbance is realized. A disturbance contribution weight matrix is constructed based on a dominant wave band, pixel-level spectrum stripping is carried out, a plankton purification spectrum is extracted, the spectrum purity and the recognition accuracy are improved, and a spectrum abnormal drift region is precisely recognized in combination with derivative spectrum change and image texture features; a coupling relation between the drift region and background disturbance is further established, a dynamic interference factor map is generated, and identification model parameters are dynamically adjusted, so that the identification stability and classification accuracy of the system in a complex interference environment are improved, and the method is suitable for multi-scene marine ecological monitoring.

Description

Technical Field

[0001] The present invention relates to the technical field of spectrum detection, and more particularly to a marine plankton hyperspectral imaging detection system. Background Art

[0002] Currently, marine plankton, as key primary producers in marine ecosystems, are known for their abundance, population distribution, and physiological state, which significantly reflect the dynamics of the aquatic environment. In practical applications, hyperspectral imaging technology, with its ability to acquire continuous spectra and represent spatial information, has become an important means for contactless identification of plankton. Existing technologies use buoys, unmanned boats, or remote sensing platforms equipped with hyperspectral sensors to perform push-scan imaging of the water surface. This technology extracts biological pigments and structural features within the visible-to-near-infrared wavelength range, combining this image information for automatic identification and classification of plankton. This method offers advantages such as high real-time performance, good adaptability to various scenarios, and non-destructive monitoring, making it suitable for scenarios such as online marine ecological assessments, red tide warnings, and trophic level monitoring.

[0003] Deficiencies in existing technologies: In typical nearshore waters or eutrophic water bodies, hyperspectral detection of plankton still faces significant technical bottlenecks. Especially in the context of significant fluctuations in SPM (suspended particulate matter) and CDOM (dissolved organic matter) concentrations, due to the strong band selectivity of the scattering and absorption behaviors of these non-target factors, their spectral interference components highly overlap with the intrinsic reflectance characteristics of plankton, resulting in the inability of traditional spectral correction methods (such as global atmospheric correction or water body radiation normalization) to effectively remove the interference signal. This problem manifests itself in spectral data as spectral slope distortion, local peak and valley drift, and normalization error amplification, further affecting subsequent classification judgments based on spectral similarity or feature templates, so that the classification accuracy does not improve with the increase in samples, but instead creates the risk of overfitting, ultimately leading to a decrease in the reliability and lack of versatility of the plankton identification system in complex water environments. Summary of the Invention

[0004] In order to overcome the above-mentioned defects of the prior art, there is a solution as follows to solve the problem of poor spectral classification and recognition in the water environment in the above-mentioned background technology.

[0005] To achieve the above object, the present invention provides the following technical solutions:

[0006] A marine plankton hyperspectral imaging detection system includes a sea water data acquisition module, a spectrum mapping module, a purification spectrum construction module, a spectrum change analysis module, a disturbance generation module, and an adjustment recognition module. The modules are connected by signals.

[0007] The sea water data acquisition module is used to collect the hyperspectral image sequence and environmental disturbance parameters of the target sea water body, perform time sequence synchronization calibration, and generate a disturbance mapping sequence;

[0008] The spectral mapping module is used to extract the spectral perturbation morphological characteristics of the dominant bands of suspended particulate matter and dissolved organic matter in the disturbance mapping sequence, construct the perturbation contribution weight matrix and map it to the image pixel space;

[0009] The purification spectrum construction module is used to perform pixel-level spectrum stripping according to the disturbance contribution weight matrix, extract the purification spectrum curve of the plankton target area, and construct the purification spectrum feature set;

[0010] Spectral change analysis module, used to analyze the derivative spectrum change trend of the purified spectral feature set and locate the spectral abnormal drift area in combination with the local image texture intensity;

[0011] The disturbance generation module is used to establish the spatial coupling relationship between the spectral anomaly drift area and the background disturbance source of the water body, and generate a dynamic interference factor map;

[0012] The recognition module is used to adjust the weight distribution and target classification threshold of the spectral recognition model according to the steady-state fitting degree of the dynamic interference factor map and the purification spectrum feature set, so as to realize adaptive plankton recognition under different interference conditions.

[0013] In a preferred embodiment, the sea water body data acquisition module includes:

[0014] An integrated observation platform equipped with hyperspectral imaging and environmental disturbance parameter collection is deployed in the target sea area;

[0015] Hyperspectral imaging collects a sequence of water surface images at fixed time intervals and records the timestamp and geographic coordinates of each frame;

[0016] Environmental disturbance parameter collection collects suspended particulate matter concentration, dissolved organic matter concentration, incident light intensity and wave disturbance values ​​in real time to generate disturbance parameter streams;

[0017] The time synchronization controller aligns the image sequence with the perturbation parameter stream one by one using a unified time reference to build a time synchronization structure;

[0018] Each image pixel is combined with the perturbation parameter at the corresponding time point to form a perturbation mapping sequence.

[0019] In a preferred embodiment, the spectrum mapping module includes:

[0020] In the disturbance mapping sequence, spectral response analysis is performed on the disturbance parameters bound to each image pixel to extract the dominant bands of suspended particulate matter and dissolved organic matter;

[0021] Based on the dominant band, the spectrum shift value, band slope change value and reflectance fluctuation rate are calculated to construct the disturbance morphology feature vector;

[0022] Normalize the disturbance morphological feature vector to generate the disturbance contribution ratio;

[0023] The disturbance contribution ratio is mapped to the image pixel index, and the disturbance contribution weight matrix is ​​constructed to represent the distribution weight of each disturbance factor in the image space.

[0024] In a preferred embodiment, the purification spectrum building block includes:

[0025] At each image pixel, a disturbance spectrum estimation model is constructed using the disturbance contribution weight matrix to fit the disturbance component of the original reflection spectrum.

[0026] The estimated disturbance spectrum is stripped from the original reflection spectrum band by band, and the purified spectrum is output;

[0027] Perform intensity normalization and band alignment on the purified spectrum to generate a standardized spectrum curve;

[0028] The standardized spectral curves of all pixels are combined into a purified spectral feature set.

[0029] In a preferred embodiment, the spectrum change analysis module includes:

[0030] Perform first-order derivative analysis and sliding curvature extraction on each spectral curve in the purified spectral feature set to identify the spectral peak drift position and change amplitude;

[0031] Compare the band position of the curvature mutation interval with the original peak position to extract the drift band and direction;

[0032] According to the curvature mutation and the texture intensity value of the corresponding pixel extracted, the consistency evaluation index of the spectral derivative and the texture response is constructed;

[0033] The pixel area is analyzed according to the consistency evaluation index, and the spectral abnormal drift area of ​​the pixel area is marked.

[0034] In a preferred embodiment, based on the curvature mutation and the image texture intensity value of the corresponding pixel, a consistency evaluation index of the spectral derivative and the texture response is constructed, including:

[0035] At the image pixels corresponding to the spectral drift band, the texture intensity value calculated by the gray-level co-occurrence matrix is ​​extracted and a mapping relationship is established with the curvature mutation value;

[0036] The consistency evaluation index is obtained by dividing the spectral curvature mutation value of the band corresponding to each image pixel by the image texture intensity value at the same position.

[0037] In a preferred embodiment, the disturbance generation module includes:

[0038] Establishing pixel-level registration relationship between the spatial coordinate map of the spectral abnormal drift region and the spatial distribution map of each disturbance factor in the disturbance mapping sequence;

[0039] Calculate the disturbance intensity difference and relative change rate of each drift pixel and the corresponding disturbance factor to form the disturbance response gradient value;

[0040] The perturbation response gradient value and the variation trend of the derivative spectrum are analyzed for correlation, and the perturbation coupling strength is calculated;

[0041] The disturbance coupling intensity is written into the image pixel index to construct a dynamic interference factor map.

[0042] In a preferred embodiment, the disturbance coupling intensity is obtained by calculating the Pearson correlation coefficient between the disturbance response gradient value of each image pixel in the spectral anomaly drift region and the derivative spectrum change trend, the disturbance response gradient value is the change in the disturbance factor intensity per unit time, and the derivative spectrum change trend is the first-order derivative sequence of the reflectivity change in the continuous band.

[0043] In a preferred embodiment, the disturbance coupling strength is written into the image pixel index to construct a dynamic interference factor map, including:

[0044] Establishing the registration relationship between the image pixel index of the spectral abnormal drift area and the spatial distribution map of each disturbance factor in the disturbance mapping sequence;

[0045] Calculate the product of the disturbance intensity difference and the disturbance change rate of each drift image pixel under the corresponding disturbance factor to obtain the disturbance response gradient value;

[0046] The perturbation coupling strength is obtained by calculating the correlation between the perturbation response gradient value and the first-order derivative change value in the spectral derivative curve within a sliding window;

[0047] The disturbance coupling intensity is written into the image pixel index to generate a dynamic interference factor map, which is used to characterize the intensity of the influence of the disturbance source on each pixel.

[0048] In a preferred embodiment, the adjustment identification module includes:

[0049] Statistical analysis of the distribution pattern and frequency of high disturbance intensity areas in the dynamic disturbance factor map;

[0050] According to the interference intensity distribution, the weight coefficient of the corresponding image area in the plankton recognition model is adjusted;

[0051] After model prediction is performed on the purified spectral feature set, the confidence deviation between the predicted label and the historical stable label is calculated;

[0052] According to the combined results of confidence deviation and spatial interference factor, the classification threshold is dynamically updated to enhance the target category recognition under disturbance conditions.

[0053] The technical effects and advantages of the marine plankton hyperspectral imaging detection system of the present invention are as follows:

[0054] The present invention synchronously collects hyperspectral image sequences and environmental disturbance parameters such as suspended particulate matter, dissolved organic matter, incident light intensity, and wave disturbance in the target sea area, and calibrates them under a unified time reference. This accurately establishes a mapping relationship between images and environmental disturbances, effectively ensuring data temporal consistency. By extracting the dominant band disturbance morphology of suspended particulate matter and dissolved organic matter, constructing a disturbance contribution weight matrix and mapping it to the image space, the quantitative analysis of different interference factors is achieved. After performing disturbance stripping at the image pixel scale, a purified spectral characteristic curve is generated, which can significantly improve the purity of the spectral data used for subsequent plankton identification and reduce the recognition error rate.

[0055] Further combining the trend of derivative spectrum changes with image texture intensity to locate areas of spectral anomalous drift can effectively identify subtle spectral shifts and enhance the ability to perceive disturbance-induced anomalies. By establishing a coupling relationship between spectral anomalous areas and disturbance sources, a dynamic interference factor map is generated to support modeling and assessment of the interference level in the current recognition scenario. Ultimately, combining the distribution of interference factors with prediction deviations, the recognition model's weights and classification thresholds are dynamically adjusted, enabling stable identification and enhanced classification of plankton species in different interference environments. This makes it more suitable for real-time monitoring and ecological surveys of multiple types of marine plankton. BRIEF DESCRIPTION OF THE DRAWINGS

[0056] Figure 1 This is a structural schematic diagram of a marine plankton hyperspectral imaging detection system of the present invention. DETAILED DESCRIPTION

[0057] The following will clearly and completely describe the technical solutions in the embodiments of the present invention in conjunction with the accompanying drawings. Obviously, the described embodiments are only part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without making creative efforts are within the scope of protection of the present invention.

[0058] In order to achieve the above objectives, Figure 1The present invention provides a schematic structural diagram of a hyperspectral imaging detection system for marine plankton, which specifically includes a marine water data acquisition module, a spectrum mapping module, a purified spectrum construction module, a spectrum change analysis module, a disturbance generation module, and an adjustment recognition module. The modules are connected by signals.

[0059] The sea water data acquisition module is used to collect the hyperspectral image sequence and environmental disturbance parameters of the target sea water body, perform time sequence synchronization calibration, and generate a disturbance mapping sequence;

[0060] The spectral mapping module is used to extract the spectral perturbation morphological characteristics of the dominant bands of suspended particulate matter and dissolved organic matter in the disturbance mapping sequence, construct the perturbation contribution weight matrix and map it to the image pixel space;

[0061] The purification spectrum construction module is used to perform pixel-level spectrum stripping according to the disturbance contribution weight matrix, extract the purification spectrum curve of the plankton target area, and construct the purification spectrum feature set;

[0062] Spectral change analysis module, used to analyze the derivative spectrum change trend of the purified spectral feature set and locate the spectral abnormal drift area in combination with the local image texture intensity;

[0063] The disturbance generation module is used to establish the spatial coupling relationship between the spectral anomaly drift area and the background disturbance source of the water body, and generate a dynamic interference factor map;

[0064] The recognition module is used to adjust the weight distribution and target classification threshold of the spectral recognition model according to the steady-state fitting degree of the dynamic interference factor map and the purification spectrum feature set, so as to realize adaptive plankton recognition under different interference conditions.

[0065] The sea area water body data acquisition module specifically includes:

[0066] An integrated observation platform should be deployed in the target sea area to be monitored. The platform should be fixed on a floating buoy or a low-speed unmanned boat and integrated with a hyperspectral imaging module and an environmental disturbance parameter acquisition module. The platform site selection should take into account typical plankton distribution zones, such as nutrient-rich areas or areas with historical high incidence of red tides, and the observation platform should be guaranteed to have stable hovering and time synchronization capabilities.

[0067] The hyperspectral imaging module performs spectral imaging of the sea surface at fixed time intervals. The acquisition band covers the visible to near-infrared range, for example, from 400 nanometers to 1000 nanometers, every 5 nanometers, a total of 121 bands. The system performs continuous image sampling at a preset time frequency (for example, one frame is collected every 30 seconds). Each frame of the image automatically records the acquisition time and geographic coordinates, with time accuracy of seconds and spatial positioning error of less than 1 meter. The spatial resolution of the image can be set so that each pixel corresponds to a 0.5 meter × 0.5 meter water surface area, and the image size is 512 × 512 pixels.

[0068] At the same time, the environmental disturbance parameter acquisition module is started synchronously with the hyperspectral imaging, and the on-site water disturbance information is collected at equal intervals. The collected disturbance parameters include: suspended particulate matter concentration, dissolved organic matter concentration, vertical incident light intensity on the water surface, and surface wave disturbance value. Among them, the suspended particulate matter concentration is obtained by a laser turbidity meter, and the unit is milligrams per liter; the dissolved organic matter concentration is obtained by comparing the standard response curve with fluorescence detection, and the unit is one billionth of the concentration; the incident light intensity is collected by a photoelectric detection element, and the unit is watts per square meter; the wave disturbance value is measured by an acceleration sensor combined with a water level difference meter, and the unit is centimeters. The sampling period of these disturbance parameters is exactly the same as the hyperspectral image acquisition period, and has a unified timestamp label;

[0069] To achieve synchronous matching between image data and disturbance parameters, the system is equipped with a unified time controller as the time reference. The image sequence and the disturbance parameter stream are registered one-to-one on the time axis to form a synchronous structure, in which each frame of the image is bound to a set of disturbance parameters collected at the corresponding moment, and the corresponding frame number, collection time and spatial position are marked.

[0070] All pixels within each image frame are combined with the perturbation parameter data of the frame to which they belong to form an image perturbation parameter mapping relationship. For each pixel, its corresponding reflectance spectrum data is combined with the perturbation parameter values ​​(i.e., suspended particulate matter concentration, dissolved organic matter concentration, incident light intensity, and wave perturbation) at the same time point to form a perturbation mapping unit. The perturbation mapping units corresponding to all pixels in all frames are arranged in chronological order to form a perturbation mapping sequence. This perturbation mapping sequence is a triple data structure containing spatial position, spectral characteristics, and synchronous perturbation factors, which can be used in subsequent perturbation stripping, spectral line purification, and target biometric identification steps.

[0071] The spectral mapping module includes:

[0072] Based on the constructed perturbation mapping sequence, the perturbation spectral response analysis is performed on each image pixel to extract the influence characteristics of suspended particulate matter and dissolved organic matter on the hyperspectral reflectance curve, and the perturbation contribution weight matrix is ​​generated for the subsequent spectral stripping process;

[0073] In the disturbance mapping sequence, the system uses image pixels as index units, and reads the disturbance parameter values ​​bound to each pixel and its corresponding spectral reflectance curve one by one. For each disturbance factor (i.e., suspended particulate matter concentration and dissolved organic matter concentration), the system performs band-level disturbance response analysis. This process uses a fixed interference scanning range, such as 400 nanometers to 800 nanometers, to compare the changes in the pixel reflectance spectral morphology when the disturbance parameter rises or falls at different times, thereby identifying the band range that is most sensitive to the disturbance response. This band is defined as the dominant band corresponding to the disturbance factor. For example, in a set of samples, when the dissolved organic matter concentration rises from 100 ppb to 250 ppb, the pixel reflectance drops most significantly in the range of 620–700 nanometers. Based on this, the system identifies this interval as the dominant band of dissolved organic matter.

[0074] Numerical extraction of spectral morphological changes in the dominant band. Based on the reflectance curve of each pixel, three key disturbance characteristic indicators are calculated within the dominant band: spectral shift value, which indicates the degree of shift in the wavelength position of the main peak or main concave point under the action of disturbance, in nanometers; band slope change value, which indicates the speed of reflectivity change within the unit wavelength range, and is used to reflect the steepening of the curve caused by disturbance; reflectivity fluctuation rate, which is defined as the percentage of the difference between the maximum reflectivity and the minimum reflectivity in the dominant band to the average reflectivity of the entire segment, and is used to characterize the degree of spectral amplitude change caused by disturbance. For example, in a certain pixel, the main peak detected in the dominant band of dissolved organic matter was shifted by 15 nanometers, the slope increased by 1.2 times, and the reflectivity fluctuation rate was 32%. These three values ​​together constitute the disturbance morphological characteristic vector;

[0075] The disturbance morphological feature vectors are normalized to ensure that all dimensions are compared using a unified scale. Normalization uses the maximum and minimum values ​​of the entire image area during the current monitoring period as a calibration benchmark, using a linear mapping method to normalize all indicators to the range of 0–1. For example, when the reflectance fluctuation ranges from 10% to 40% across the entire image area, a pixel with a fluctuation of 25% will have a normalized result of 0.5. The normalized three-dimensional disturbance morphological features are further integrated into a disturbance contribution ratio, which represents the spectral influence of a specific disturbance factor on a pixel. A larger value indicates a stronger disturbance influence.

[0076] The disturbance contribution ratio is mapped to the image space to construct a disturbance contribution weight matrix. This matrix uses the row and column coordinates of the image pixel as the index unit, records the disturbance contribution ratio of suspended particulate matter and dissolved organic matter for each pixel, and stores it in the system storage module in matrix form. Each position in the matrix represents an image pixel, and the value at that position is the spectral impact weight of the disturbance factor at that pixel. For example, at the pixel in row 320 and column 200, the disturbance contribution ratio of suspended particulate matter is 0.72, and that of dissolved organic matter is 0.46. In the subsequent spectral stripping process, this pixel will give priority to fitting and deducting the spectral impact caused by suspended particulate matter.

[0077] The marine water data acquisition module jointly collects hyperspectral images and environmental disturbance parameters, and uses a unified time base to complete data alignment. This can achieve precise time matching between images and disturbance factors, ensure the correlation between spectral characteristics and disturbance behavior in subsequent analysis, and enhance the system's ability to analyze ocean information in a dynamic environment.

[0078] The dominant band is extracted as follows:

[0079] In the perturbation mapping sequence, each image pixel is bound to a set of perturbation parameters (such as suspended particulate matter concentration or dissolved organic matter concentration) and the corresponding hyperspectral reflectance curve. By traversing and calculating the coupling relationship between the perturbation parameter changes and the spectral response, it is determined which spectral bands a certain perturbation factor has a significant impact on. The specific method is as follows:

[0080] A sampling sequence for a disturbance factor is selected, such as the range of changes in dissolved organic matter concentration recorded during the sampling period. Multiple representative samples of disturbance strengths and weaknesses are extracted from all image pixels (e.g., three groups of low, medium, and high concentrations). For each group of samples, the average reflectance curve is calculated. In each spectral band (e.g., 400–1000 nm), the reflectance difference between samples of different disturbance intensities is calculated. The continuous band with the largest reflectance difference is identified as the dominant band for the disturbance factor. For example, if the average reflectance change in the 620–700 nm band exceeds 5% when the dissolved organic matter changes from 100 ppb to 250 ppb, while it is less than 2% in other bands, the system will identify 620–700 nm as the dominant band for dissolved organic matter.

[0081] The spectral shift value describes the amplitude of the shift in the wavelength position of the main peak (or main valley) within the dominant band before and after the disturbance conditions change. It is used to reflect the impact of the disturbance on the main characteristics of the spectral line. The calculation method is as follows: Under two states of low and high disturbance intensity, the reflectivity curve within the dominant band of the pixel is extracted respectively. In each curve, the position of the main peak or main valley value (that is, the wavelength at the maximum or minimum reflectivity value) is found, and the difference in the wavelength of the position at the two states is calculated. For example, if the main valley appears at 660nm under low disturbance conditions and at 675nm under high disturbance conditions, the spectral shift value is 15nm.

[0082] The band slope change value reflects the degree of change in the slope of the spectral curve in the dominant band, and is used to measure whether the disturbance makes the curve steeper or flatter. The calculation method is as follows: select equally spaced pairs of adjacent wavelength points in the dominant band; calculate the reflectivity slope of each pair of adjacent wavelength points (that is, the reflectivity difference between the two points divided by the wavelength interval); for the two states before and after the disturbance, calculate the mean of the slope respectively, and take the difference between the two as the band slope change value.

[0083] The reflectivity fluctuation is used to characterize the fluctuation intensity of the reflectivity curve in the dominant band, that is, the spectral line amplitude. It is calculated as follows: the maximum reflectivity value and the minimum reflectivity value are recorded in the dominant band, and the difference between the two is calculated to obtain the reflectivity amplitude. The amplitude is divided by the average reflectivity value of the band to obtain the reflectivity fluctuation, which is expressed as a percentage.

[0084] The disturbance morphological feature vector is constructed as follows: the three characteristic values ​​calculated in the dominant band for the disturbance factor corresponding to each image pixel: spectral shift value, band slope change value, and reflectance fluctuation rate are combined in a fixed order into a three-dimensional vector; the value of each dimension must maintain unit consistency and stable arrangement order; for example: the three disturbance characteristics of a certain pixel in the dominant band of dissolved organic matter are 12nm, +0.003 / nanometer, and 27.5%.

[0085] The spectral mapping module extracts the spectral response characteristics of the dominant disturbance factor and constructs a disturbance contribution weight matrix to express the disturbance weight at the pixel level, enabling the system to accurately separate the interference effect of environmental disturbance on the observed spectrum, providing a high-precision disturbance weight basis for the subsequent purification process and enhancing the system's ability to resolve complex water backgrounds.

[0086] Clean Spectrum building blocks include:

[0087] For each image pixel, the perturbation contribution weight matrix constructed in the previous stage is first called. This perturbation contribution weight matrix represents the perturbation contribution ratio of suspended particulate matter concentration (SPM) and dissolved organic matter concentration (CDOM) to the spectral reflectance change in each dominant band. Specifically, for a specific band under each pixel index, such as 620–700 nm, the perturbation contribution weight matrix stores the perturbation weight values ​​of the two perturbation factors in that band range (e.g., 0.65 for SPM and 0.25 for CDOM, with the remainder being plankton and noise interference). The values ​​are obtained by normalization, and the numerical range is limited to between 0 and 1, and the total does not exceed 1;

[0088] Based on this perturbation weight information, the system establishes a perturbation spectrum estimation model on the original hyperspectral reflectance curve of the pixel. The perturbation spectrum estimation model uses a data-driven linear approximation method: the perturbation factor and the perturbation morphological feature vector of its dominant band are weighted according to the weight, and the perturbation spectrum caused by the perturbation factor is fitted. For example, for the 620–700 nm band, if the SPM contribution reflectance morphology derived from the perturbation morphological characteristics at the pixel is [0.05, 0.06, 0.08], and the CDOM is [0.03, 0.04, 0.05], then its linear perturbation spectrum is estimated to be 0.65×SPM morphology + 0.25×CDOM morphology, and the fitted perturbation value on the corresponding band is [0.0415, 0.051, 0.067].

[0089] Within the same spectral band, the system strips the estimated perturbed reflectance values ​​from the original reflectance spectrum band by band. That is, the cleaned spectral value at each wavelength is the original reflectance value minus the corresponding perturbation contribution estimate. For example, if the original reflectance of a band is 0.32 and the perturbed spectrum estimate is 0.067, the cleaned reflectance for that band is 0.253. This operation ensures that the main spectral interference components of SPM and CDOM are stripped away at each pixel, leaving the main spectral contribution of plankton.

[0090] To ensure comparability of purified spectra across different pixels, all stripped purified spectra were subjected to intensity normalization and band alignment. Intensity normalization involves scaling each purified spectral curve by its maximum reflectance, unifying the maximum value of all curves to 1 to eliminate intensity shifts caused by variations in imaging brightness. Band alignment involves interpolating all purified spectra at standard wavelength nodes (e.g., a wavelength point every 5 nm, from 400 nm to 1000 nm) to achieve a consistent structure and facilitate subsequent statistics and analysis. The standardized purified spectral curves for all image pixels are collected to form a purified spectral feature set. This feature set is a two-dimensional matrix structure, with rows representing image pixels and columns representing purified reflectance values ​​at standard wavelength points. This feature set can serve as the input data for subsequent tasks such as plankton spectral clustering analysis, supervised classification and identification, and population lineage inference.

[0091] The purified spectrum construction module uses pixel-level disturbance stripping processing to remove disturbance components while ensuring the integrity of the spectral structure, and outputs a purified standardized spectral curve, thereby forming a basic set of spectral characteristics of plankton and improving the specificity and accuracy of the recognition algorithm for target organisms.

[0092] The spectral change analysis module includes:

[0093] After the purification spectrum feature set is constructed, the system performs a derivative spectrum analysis on each standardized purified spectrum curve. The derivative spectrum refers to the first-order derivative sequence of the reflectance of the spectrum curve as it changes with wavelength, which reflects the rate of change of the spectral morphology.

[0094] A sliding window method is used to calculate the slope of the spectral curve between adjacent wavelength points, and the average slope value of the area is recorded at the central wavelength point of each window, thereby obtaining a smooth first-order derivative curve. For example, for a spectral curve with a wavelength range of 400–1000nm and a spacing of 5nm, a window size of 3 band points is used. The slope at the central point of 500nm is the average of the slopes of the bands before and after it. The degree of fluctuation of the derivative spectrum directly indicates the jump characteristics of the reflectivity and is the basis for identifying spectral peak drift.

[0095] The sliding curvature of each band point is calculated based on the first-order derivative curve. The sliding curvature is used to describe the degree of bending of the derivative spectrum and reflect the directional mutation of the spectrum line in the local area. It is implemented by applying a sliding difference operation to the derivative spectrum curve again and counting the amplitude of the change in the slope of adjacent derivatives. If the curvature value of a band point exceeds the set threshold (for example, twice the local average value), it is considered that there is a possibility of spectrum peak position drift at this point, and the band position and direction of the mutation point (moving towards short waves or moving towards long waves) are recorded.

[0096] Compare the marked peak shift band positions with the original peak band positions of the purified spectrum one by one, extracting the specific band and direction of peak shift in each spectral curve by comparing the difference between the band index of the curvature shift point and the index of the original peak band. If a peak between 400 and 500 nm was originally located at 450 nm, but the curvature shift point is at 462 nm, it indicates that the peak has shifted toward longer wavelengths by 12 nm. This shift will be used to determine whether it is an abnormal change.

[0097] At the same time, the local image texture intensity value is extracted from the corresponding image pixel. This texture intensity value is used to characterize the stability of the image spatial structure in the area where the pixel is located, and is achieved using the contrast index or local variance of the gray-level co-occurrence matrix. Specifically, if the texture intensity of a certain area is high, it means that there are drastic changes in the edge, texture, or structure of the area; conversely, low texture intensity indicates that the image content in the area is stable and the structural changes are weak.

[0098] The system further constructs a consistency evaluation index between the spectral derivative and the image texture response. This index measures whether a sudden change in spectral peak curvature is accompanied by a change in image texture. If a pixel exhibits a dramatic curvature change in the derivative spectrum (i.e., a significant spectral peak shift), but the image texture intensity in the corresponding region is very low (i.e., the image region itself is relatively uniform and stable), the shift is more likely to be an anomalous drift caused by spectral perturbations rather than a true biological or structural boundary change. Therefore, this consistency index is evaluated based on the inverse relationship between the curvature change amplitude and the texture intensity. Pixels with large curvature changes and small texture changes are marked as anomalous spectral drift only if they meet a certain threshold (e.g., a curvature change value greater than 0.15 and a texture intensity less than 0.05). For example, in a certain image pixel, the derivative spectrum curvature change value is 0.23, far above the local average of 0.08, while its texture intensity is 0.03, below the image-wide average texture intensity of 0.07. This indicates that a strong spectral change without texture support exists at this location, thus being judged as an anomalous drift.

[0099] The specific method for constructing consistency evaluation indicators is as follows:

[0100] For each purified spectral curve in the purified spectral feature set, its first-order derivative curve is extracted, and the sliding curvature value of each band point in the derivative curve is calculated to characterize the local mutation intensity of the spectral line near that band. The image texture intensity value at the corresponding position of each image pixel is extracted. This texture intensity value is obtained by extracting the local grayscale variance based on a fixed window in the grayscale projection of the hyperspectral image or by calculating the contrast index based on the grayscale co-occurrence matrix. It is used to measure the degree of structural change in the pixel neighborhood. The sliding curvature value and the texture intensity value are jointly calculated to construct a consistency evaluation factor. The consistency evaluation factor is defined as the proportional relationship or correlation strength between the sliding curvature value and the texture intensity value. It is used to determine whether the spectral mutation is supported by the image spatial structure. When the consistency evaluation factor is higher than the set threshold and the image texture intensity is in a low range (e.g., below the 75th percentile of the sample full image mean), the spectral mutation is considered to lack texture response support and is highly likely to be an abnormal drift. This consistency evaluation factor serves as an important screening indicator for the subsequent identification of spectral abnormal drift areas and is used to eliminate non-structural spectral mutations caused by environmental disturbances.

[0101] The process of identifying abnormal spectral drift areas is as follows:

[0102] The spectral derivative values ​​between continuous bands are calculated for each standardized spectral curve in the purified spectral feature set, and the spectral curvature of each band is extracted within a sliding window of fixed length to form a spectral derivative change curve. The curvature mutation interval is identified in the spectral derivative change curve, and the position of the mutation interval is compared with the peak position in the original spectral curve to extract the spectral drift band and drift direction. At the image pixel corresponding to the spectral drift band, the texture intensity value calculated by the gray level co-occurrence matrix is ​​extracted and a one-to-one mapping relationship is established with the curvature mutation value. Based on the ratio between the curvature mutation value and the texture intensity value, a consistency evaluation index between the spectral derivative and the texture response is constructed. When the consistency evaluation index is higher than the preset threshold and the texture intensity value is lower than the set reference value, the corresponding pixel area is marked as a spectral abnormal drift area. The specific calculation process is to calculate the ratio between the curvature mutation value and the texture intensity value as the consistency evaluation index, that is, the spectral curvature mutation value of the band corresponding to each image pixel is divided by the image texture intensity value at the same position to obtain the consistency index value of the pixel. This consistency index reflects the degree of response difference when the spectrum fluctuates significantly but the texture remains unchanged, and is used to locate potential areas of spectral drift anomalies. For example, if a pixel has a sudden change in curvature of 0.18 and a texture intensity of 0.06 in a specific band, the consistency index value is 3.0. When the consistency index value is greater than 2.5 and the texture intensity value is less than 0.08, it is classified as a drift anomaly.

[0103] The spectral change analysis module combines the derivative spectrum change trend with the image texture intensity information to achieve highly sensitive positioning of weak spectral drift phenomena. It is particularly suitable for identifying areas with complex interference, stable texture but abnormal spectral response, and improving the system's early perception of hidden interference behavior.

[0104] The disturbance generation module includes:

[0105] The image space coordinates of the identified spectral anomaly drift area are registered pixel by pixel with the spatial distribution map of each disturbance factor in the disturbance mapping sequence. The registration process is based on the image coordinate index alignment method to ensure that each spectral drift pixel is associated with the suspended particulate matter concentration distribution map, dissolved organic matter concentration map, fluctuation disturbance map and other disturbance factor maps at the same location.

[0106] After registration, the perturbation intensity value of each anomalous drift pixel in the perturbation mapping sequence is extracted and subtracted from the perturbation intensity of the pixel's neighboring area to obtain the perturbation intensity difference. Simultaneously, the amplitude of the perturbation parameter change on the time axis is analyzed and divided by the duration of the change to obtain the relative rate of change of the perturbation. The perturbation intensity difference and relative rate of change are used as the perturbation response gradient value to characterize the influence of the perturbation source on the anomalous area.

[0107] Subsequently, a correlation analysis is performed between the perturbation response gradient of each pixel and the curvature mutation value in the first-order derivative spectrum of its purified spectral characteristic curve. The degree of response coupling between the two is measured using the correlation coefficient, and the perturbation coupling strength value of the pixel is output. Perturbation responses with high correlation (e.g., a correlation coefficient greater than 0.7) are judged to be strongly coupled with the spectral anomaly.

[0108] The perturbation coupling intensity value of each pixel is written to the corresponding image index to construct a complete dynamic interference factor map. This map describes the interference ability of each perturbation factor on spectral anomaly drift at the pixel level, providing input for subsequent recognition model weight adjustment and classification threshold adaptation.

[0109] A specific way to construct a dynamic interference factor map is as follows:

[0110] The construction of the dynamic interference factor map is based on the spatial coupling strength of the spectral anomaly drift region and the disturbance factor, forming a two-dimensional distribution map with pixel-level spatial resolution. The construction process is divided into the following specific steps:

[0111] In the registered image space, the disturbance parameters corresponding to each spectral abnormal drift area are analyzed pixel by pixel, including the suspended particulate matter concentration (such as SPM), dissolved organic matter concentration (such as CDOM) and incident light intensity disturbance value. For each image pixel, its disturbance parameter value and the disturbance change rate at the corresponding time point are extracted.

[0112] Next, a Pearson correlation analysis is performed based on the first-order derivative change value (i.e., the trend of spectral shape change) extracted from the spectral derivative curve and the change amplitude of the aforementioned perturbation parameter. This analysis is performed within a sliding window, and a linear fit is performed between each perturbation factor and the change value of the derivative curvature, and the correlation coefficient is recorded. If the correlation coefficient exceeds a preset threshold (e.g., 0.6), it is determined that the perturbation factor has a significant coupling with the spectral drift of the current pixel.

[0113] For each perturbation factor, a coupling strength distribution map is generated in the image space. The coupling strength is defined as the correlation score between the perturbation response gradient (i.e., the product of the perturbation value difference and the perturbation change rate) and the spectral curvature change value. Its value range can be normalized to between 0 and 1, with larger values ​​representing higher perturbation strength.

[0114] The coupling intensity distribution maps of all disturbance factors are weighted and superimposed to form a comprehensive dynamic disturbance factor map. The weight assignment can be set based on the frequency and intensity of each disturbance factor's impact on spectral anomaly drift in historical observation data. For example, if CDOM disturbance dominates spectral anomaly drift in most scenarios, the superposition weight of its corresponding coupling map is set to 0.5, SPM to 0.3, and light intensity to 0.2.

[0115] The constructed atlas is a two-dimensional matrix of the same size as the original image. Each pixel contains an interference factor value between 0 and 1, which is used to characterize the degree to which it is affected by background disturbance sources in the current environment. The atlas can be used for subsequent tasks such as model weight adjustment, adaptive classification threshold control, and interference robustness enhancement.

[0116] The perturbation coupling strength is calculated for each image pixel within the spectral anomaly drift region. First, for each pixel, the perturbation response gradient value corresponding to the perturbation mapping sequence is obtained. The perturbation response gradient value represents the product of the intensity change of the perturbation factor and the time gradient. This value is obtained by calculating the ratio of the perturbation factor's difference per unit time to the intensity difference at the pixel's spatial position. For example, for wave perturbations, the response gradient value is the rate of change of wave height per unit time. The derivative spectrum trend corresponding to this pixel is extracted from the purified spectral feature set. The derivative spectrum trend represents the sequence of first-order derivatives of reflectivity over continuous bands, expressed in units of reflectivity change per nanometer. The sliding window method is commonly used to extract continuous derivative changes. The window length can be set to 5 bands. The average derivative change in each window is calculated. When establishing correlation analysis, the disturbance response gradient sequence and the derivative spectrum change trend sequence are time-aligned, and the Pearson correlation coefficient is calculated on each pixel. The correlation coefficient is defined as the ratio of the covariance of the two sequences to the product of their respective standard deviations. The value range is [-1, 1], which represents the strength and direction of the linear relationship between the two.

[0117] The disturbance generation module builds a dynamic interference factor map by establishing a pixel-level spatial coupling relationship between the abnormal drift area and the disturbance factor, enabling the system to dynamically perceive and quantify the interference intensity of the disturbance source, effectively enhancing the system's ability to trace and analyze the cause of the drift.

[0118] The adjustment identification module includes:

[0119] Extract high-disturbance intensity areas from the dynamic interference factor map. By setting a disturbance intensity threshold (for example, areas above the 80th percentile in the map), identify image pixels with significant interference, and count the spatial distribution pattern and frequency of these high-disturbance areas in the entire image. The distribution pattern can be quantified by cluster center density and dispersion to characterize the concentration and localization trend of the disturbance source.

[0120] Based on the statistical distribution of high-disturbance regions, the weight coefficient distribution of the spectral recognition model in the image space is adjusted. Specifically, pixels in high-disturbance regions are assigned a lower recognition model attention weight to reduce the interference of disturbances on the recognition results; while the weight of low-disturbance regions is maintained or increased to enhance the discriminative contribution of stable regions in recognition. For example, the weight coefficient of high-disturbance regions is set to the original value of 0.6, and that of low-disturbance regions is set to 1.2 to achieve a differentiated response.

[0121] The cleaned spectral feature set is input into the current recognition model to predict the plankton target, obtaining the model's predicted label. Based on this, the stable labels corresponding to the same area in the historical samples are extracted and compared. The difference between the two is then calculated to determine the confidence bias. The confidence bias is the absolute value of the difference between the predicted probability and the stable probability of the historical label, reflecting the confidence level of the current recognition result. For example, if the current model predicts a pixel's probability of belonging to the "Phytoplankton A" category as 0.72, while the historical stable sample predicts it as 0.91, the confidence bias is 0.19.

[0122] Finally, the model classification threshold is dynamically adjusted based on a combined score of confidence deviation and interference factor strength. The combined score is calculated by multiplying the normalized confidence deviation and interference strength value for each pixel. A higher combined score indicates a pixel that is more susceptible to interference and unstable model recognition. The system then adjusts the adaptive classification threshold based on the combined score. For example, if the combined score is greater than 0.15, the pixel classification threshold is raised from 0.5 to 0.6 to improve recognition conservatism in uncertain areas. If the score is less than 0.05, the threshold can be appropriately relaxed to improve recognition sensitivity.

[0123] The adjustment recognition module adaptively adjusts the recognition model by fusing the dynamic interference spectrum and the steady-state characteristics of the purified spectrum. It can automatically optimize the recognition strategy according to the interference conditions, achieve stable recognition of plankton in multiple disturbance scenarios, and improve the environmental adaptability and classification accuracy of the system.

[0124] The above formulas are all dimensionless and numerical calculations. The formula is a formula that is closest to the actual situation obtained by collecting a large amount of data and performing software simulation. The preset parameters in the formula are set by technicians in this field according to actual conditions.

[0125] The above embodiments may be implemented in whole or in part through software, hardware, firmware, or any other combination thereof. When implemented using software, the above embodiments may be implemented in whole or in part in the form of a computer program product.

[0126] Those skilled in the art will appreciate that the modules and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered beyond the scope of this application.

[0127] In addition, each functional module in each embodiment of the present application may be integrated into one processing module, or each module may exist physically separately, or two or more modules may be integrated into one module.

[0128] The above description is merely a specific embodiment of the present application, but the scope of protection of the present application is not limited thereto. Any changes or substitutions that can be easily conceived by a person skilled in the art within the technical scope disclosed in this application should be included in the scope of protection of this application. Therefore, the scope of protection of this application should be based on the scope of protection of the claims.

[0129] Finally: The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present invention should be included in the scope of protection of the present invention.

Claims

1. A hyperspectral imaging detection system for marine plankton, characterized by: It includes a sea water data acquisition module, a spectrum mapping module, a purification spectrum construction module, a spectrum change analysis module, a disturbance generation module, and an adjustment recognition module. Each module is connected through signals. The sea water data acquisition module is used to collect the hyperspectral image sequence and environmental disturbance parameters of the target sea water body, perform time sequence synchronization calibration, and generate a disturbance mapping sequence; The spectral mapping module is used to extract the spectral perturbation morphological characteristics of the dominant bands of suspended particulate matter and dissolved organic matter in the disturbance mapping sequence, construct the perturbation contribution weight matrix and map it to the image pixel space; The purification spectrum construction module is used to perform pixel-level spectrum stripping according to the disturbance contribution weight matrix, extract the purification spectrum curve of the plankton target area, and construct the purification spectrum feature set; Spectral change analysis module, used to analyze the derivative spectrum change trend of the purified spectral feature set and locate the spectral abnormal drift area in combination with the local image texture intensity; The disturbance generation module is used to establish the spatial coupling relationship between the spectral anomaly drift area and the background disturbance source of the water body, and generate a dynamic interference factor map; The recognition module is used to adjust the weight distribution and target classification threshold of the spectral recognition model according to the steady-state fitting degree of the dynamic interference factor map and the purification spectrum feature set, so as to realize adaptive plankton recognition under different interference conditions.

2. The marine plankton hyperspectral imaging detection system according to claim 1, characterized in that: The sea area water body data acquisition module includes: An integrated observation platform equipped with hyperspectral imaging and environmental disturbance parameter collection is deployed in the target sea area; Hyperspectral imaging collects a sequence of water surface images at fixed time intervals and records the timestamp and geographic coordinates of each frame; Environmental disturbance parameter collection collects suspended particulate matter concentration, dissolved organic matter concentration, incident light intensity and wave disturbance values ​​in real time to generate disturbance parameter streams; The time synchronization controller aligns the image sequence with the perturbation parameter stream one by one using a unified time reference to build a time synchronization structure; Each image pixel is combined with the perturbation parameter at the corresponding time point to form a perturbation mapping sequence.

3. The marine plankton hyperspectral imaging detection system according to claim 2, characterized in that: The spectral mapping module includes: In the disturbance mapping sequence, spectral response analysis is performed on the disturbance parameters bound to each image pixel to extract the dominant bands of suspended particulate matter and dissolved organic matter; Based on the dominant band, the spectrum shift value, band slope change value and reflectance fluctuation rate are calculated to construct the disturbance morphology feature vector; Normalize the disturbance morphological feature vector to generate the disturbance contribution ratio; The disturbance contribution ratio is mapped to the image pixel index, and the disturbance contribution weight matrix is ​​constructed to represent the distribution weight of each disturbance factor in the image space.

4. The marine plankton hyperspectral imaging detection system according to claim 3, characterized in that: Clean Spectrum building blocks include: At each image pixel, a disturbance spectrum estimation model is constructed using the disturbance contribution weight matrix to fit the disturbance component of the original reflection spectrum. The estimated disturbance spectrum is stripped from the original reflection spectrum band by band, and the purified spectrum is output; Perform intensity normalization and band alignment on the purified spectrum to generate a standardized spectrum curve; The standardized spectral curves of all pixels are combined into a purified spectral feature set.

5. The marine plankton hyperspectral imaging detection system according to claim 4, characterized in that: The spectral change analysis module includes: Perform first-order derivative analysis and sliding curvature extraction on each spectral curve in the purified spectral feature set to identify the spectral peak drift position and change amplitude; Compare the band position of the curvature mutation interval with the original peak position to extract the drift band and direction; According to the curvature mutation and the texture intensity value of the corresponding pixel extracted, the consistency evaluation index of the spectral derivative and the texture response is constructed; The pixel area is analyzed according to the consistency evaluation index, and the spectral abnormal drift area of ​​the pixel area is marked.

6. The marine plankton hyperspectral imaging detection system according to claim 5, characterized in that: According to the curvature mutation and the image texture intensity value of the corresponding pixel, the consistency evaluation index of the spectral derivative and texture response is constructed, including: At the image pixels corresponding to the spectral drift band, the texture intensity value calculated by the gray-level co-occurrence matrix is ​​extracted and a mapping relationship is established with the curvature mutation value; The consistency evaluation index is obtained by dividing the spectral curvature mutation value of the band corresponding to each image pixel by the image texture intensity value at the same position.

7. The marine plankton hyperspectral imaging detection system according to claim 6, characterized in that: The disturbance generation module includes: Establishing pixel-level registration relationship between the spatial coordinate map of the spectral abnormal drift region and the spatial distribution map of each disturbance factor in the disturbance mapping sequence; Calculate the disturbance intensity difference and relative change rate of each drift pixel and the corresponding disturbance factor to form the disturbance response gradient value; The perturbation response gradient value and the variation trend of the derivative spectrum are analyzed for correlation, and the perturbation coupling strength is calculated; The disturbance coupling intensity is written into the image pixel index to construct a dynamic interference factor map.

8. The marine plankton hyperspectral imaging detection system according to claim 7, characterized in that: The perturbation coupling intensity is obtained by calculating the Pearson correlation coefficient between the perturbation response gradient value of each image pixel in the spectral anomaly drift region and the derivative spectrum change trend. The perturbation response gradient value is the change in the perturbation factor intensity per unit time, and the derivative spectrum change trend is the first-order derivative sequence of the reflectivity change in continuous bands.

9. The marine plankton hyperspectral imaging detection system according to claim 8, characterized in that: Write the disturbance coupling intensity into the image pixel index to construct a dynamic interference factor map, including: Establishing the registration relationship between the image pixel index of the spectral abnormal drift area and the spatial distribution map of each disturbance factor in the disturbance mapping sequence; Calculate the product of the disturbance intensity difference and the disturbance change rate of each drift image pixel under the corresponding disturbance factor to obtain the disturbance response gradient value; The perturbation coupling strength is obtained by calculating the correlation between the perturbation response gradient value and the first-order derivative change value in the spectral derivative curve within a sliding window; The disturbance coupling intensity is written into the image pixel index to generate a dynamic interference factor map, which is used to characterize the intensity of the influence of the disturbance source on each pixel.

10. The marine plankton hyperspectral imaging detection system according to claim 9, characterized in that: The adjustment identification module includes: Statistical analysis of the distribution pattern and frequency of high disturbance intensity areas in the dynamic disturbance factor map; According to the interference intensity distribution, the weight coefficient of the corresponding image area in the plankton recognition model is adjusted; After model prediction is performed on the purified spectral feature set, the confidence deviation between the predicted label and the historical stable label is calculated; According to the combined results of confidence deviation and spatial interference factor, the classification threshold is dynamically updated to enhance the target category recognition under disturbance conditions.

Images