An intelligent identification method and device for distinguishing dry sand from wet sand by using multi-spectral data

Abstract

The application discloses an intelligent identification method and device for distinguishing dry sand from wet sand by using multispectral data, and belongs to the technical field of remote sensing monitoring, and comprises the following steps: acquiring a geometric distortion coefficient based on an instantaneous field of view angle of imaging, resolution and ground surface wind speed; acquiring a background heterogeneity coefficient based on the bulk density of sand surface, porosity and salinity; acquiring a surface pollution coefficient based on the coverage rate of algal and microbial film and the coverage degree of coexisting impurities; further combining atmospheric aerosol optical thickness, sand particle size median value and volume percentage of main mineral components to construct an atmosphere-particle-component matching model and acquire matching parameters; and finally, acquiring the target effective spectral channel number by a channel optimization model based on the matching parameters, the current effective spectral channel number and the background heterogeneity coefficient. The application realizes the fusion evaluation of multi-source influencing factors through multi-level system modeling, and significantly improves the precision and environmental adaptability of dry and wet sand identification.

Description

Technical Field

[0001] This invention belongs to the field of remote sensing monitoring technology, and in particular relates to an intelligent identification method and device for distinguishing between dry sand and wet sand using multispectral data. Background Technology

[0002] With the development of remote sensing technology, surface condition identification using multispectral data has become an important tool in fields such as environmental monitoring and geological exploration. Distinguishing between dry and wet sand is particularly significant for desertification control, water resource assessment, and remote sensing inversion of soil moisture. Current research has attempted to identify dry and wet sand based on differences in spectral reflectance or moisture-sensitive bands; however, most methods rely on only single or limited spectral features, failing to systematically consider the combined effects of imaging conditions, sand body physical properties, and environmental interference, resulting in insufficient accuracy and robustness in practical applications.

[0003] In existing technologies, common methods for identifying dry and wet sand are mostly based on reflectance thresholds or spectral indices (such as the normalized moisture index) in specific bands. These methods are effective in ideal environments, but in complex real-world scenarios, they are easily affected by factors such as imaging geometric distortion, heterogeneous sand background, surface contaminants, and atmospheric conditions, often resulting in significant errors in the identification results. Furthermore, existing methods typically do not construct quantitative evaluation models that couple multiple factors, making it difficult to dynamically optimize spectral channel configurations, thus limiting their adaptability and reliability in different environments. Summary of the Invention

[0004] To address the shortcomings of existing technologies, this invention provides an intelligent identification method and device that uses multispectral data to distinguish between dry and wet sand, thus solving the aforementioned problems.

[0005] To achieve the above objectives, the present invention provides the following technical solution: an intelligent identification method for distinguishing between dry and wet sand using multispectral data, comprising: Geometric distortion coefficients are obtained by constructing a geometric distortion model based on the instantaneous field of view, imaging resolution, and surface wind speed. A background heterogeneity model was constructed based on the sand body surface bulk density, sand body porosity, and sand body salinity to obtain the background heterogeneity coefficient. A surface contamination status model was constructed based on the coverage of algae and microbial films and the coverage of coexisting impurities to obtain the surface contamination coefficient. Based on the geometric distortion coefficient and surface pollution coefficient, atmospheric aerosol optical thickness, median sand particle size distribution, and volume percentage of major mineral components, an atmosphere-particle size-composition matching model is constructed to obtain the atmosphere-particle size-composition matching performance. Based on atmospheric-particle-composition matching, the current number of effective spectral channels, and the background heterogeneity coefficient, a channel optimization model is constructed to obtain the target number of effective spectral channels.

[0006] Based on the above technical solutions, the present invention also provides the following optional technical solutions: A further technical solution: The channel optimization model adjusts the difference between the preset target atmospheric-particle-composition matching and the current atmospheric-particle-composition matching based on the background heterogeneity coefficient, and performs nonlinear scaling and rounding up on the current effective spectral channel number based on the adjusted result to output the target effective spectral channel number.

[0007] A further technical solution: The method for obtaining the atmosphere-particle size-composition matching is as follows: Obtain atmospheric aerosol optical thickness, median sand particle size distribution, and volume percentage of major mineral components; The thickness index, particle size distribution index, and component volume index are obtained by comparing the absolute differences between atmospheric aerosol optical thickness, median sand particle size distribution, and volume percentage of major mineral components and their corresponding ideal values ​​with the allowable deviation values. Import the geometric distortion coefficient and surface contamination coefficient into the preset background noise model to obtain the background noise coefficient; The background noise model performs a weighted combination of the geometric distortion coefficient and the surface contamination coefficient, and calculates the combination result with a reference value, so that the background noise coefficient is negatively correlated with the ideality of imaging conditions and surface condition; The background noise figure, thickness index, particle size distribution index, and component volume index are imported into the atmosphere-particle size-component matching to obtain the atmosphere-particle size-component matching. The atmosphere-particle size-composition matching model integrates the thickness index, particle size distribution index, and component volume index to characterize the degree of matching between the atmosphere, sand particle size, and mineral composition and the ideal spectral moisture response conditions. It is also combined with the background noise coefficient for calculation to finally obtain the atmosphere-particle size-composition matching degree. The larger the atmosphere-particle size-composition matching degree, the better the spectral response to moisture.

[0008] Further technical solution: The surface contamination coefficient is obtained as follows: Obtain the coverage rate of algae and microorganisms as well as the coverage rate of coexisting impurities; The biofilm coverage rate and coexisting impurity coverage rate were compared with the corresponding reference values ​​to obtain the biofilm coverage rate index and the coexisting impurity coverage index. The biofilm coverage index and the coexisting impurity coverage index are imported into the surface contamination state model to obtain the surface contamination state coefficient.

[0009] Further technical solution: The method for obtaining the background heterogeneity coefficient is as follows: Obtain the surface bulk density, porosity, and salinity of the sand body; The absolute differences between the sand surface bulk density, sand porosity, and sand salinity and the corresponding reference values ​​are compared with the corresponding reference values ​​to obtain the bulk density index, porosity index, and salinity index. The density index, porosity index, and salinity index were imported into the background heterogeneity model to obtain the background heterogeneity coefficient.

[0010] A further technical solution: The geometric distortion coefficient is obtained as follows: Acquire the instantaneous field of view, imaging resolution, and surface wind speed; The wind speed index is obtained by comparing the ground wind speed with the reference wind speed. The absolute difference between the instantaneous field of view and the imaging resolution and the corresponding optimal value is compared with the corresponding allowable deviation from the optimal value to obtain the field of view deviation index and the resolution deviation index. The wind speed index, field of view deviation index, and resolution deviation index are imported into the geometric distortion model to obtain the geometric distortion coefficients.

[0011] A further technical solution: The surface contamination state model performs a nonlinear function transformation on the square root of the product of the biofilm contamination index and the impurity contamination index, and the complement of the result is the surface contamination coefficient. The larger the coefficient value, the lower the degree of surface contamination.

[0012] A further technical solution: The background heterogeneity model uses the weighted sum of the bulk density index, porosity index, and salinity index as the inverse function of the denominator to calculate the background heterogeneity coefficient. The larger the coefficient value, the better the background homogeneity of the sand body.

[0013] A further technical solution: The geometric distortion model performs a comprehensive calculation on the ratio of the surface wind speed to the reference wind speed, the normalized deviation of the instantaneous field of view and the imaging resolution from their respective optimal values, so that the geometric distortion coefficient decreases as the wind speed increases and the field of view and resolution deviate from their optimal values. The larger the coefficient value, the more ideal the imaging geometric conditions.

[0014] A smart identification device for distinguishing between dry and wet sand using multispectral data includes a processor and a memory. The memory stores a computer program, and when the computer program is executed by the processor, it implements the aforementioned smart identification method for distinguishing between dry and wet sand using multispectral data.

[0015] This invention provides an intelligent identification method and device for distinguishing between dry and wet sand using multispectral data, which has the following advantages compared with the prior art: 1. By constructing multi-level coupling models such as geometric distortion, background heterogeneity, surface contamination, and atmospheric-particle-composition matching, key factors such as imaging conditions, sand body properties, and environmental interference are systematically quantified and integrated, overcoming the limitations of traditional methods that rely on single spectral features, thereby achieving more accurate and reliable differentiation between dry and wet sand in complex real-world scenarios. 2. The coefficients output by each model (such as geometric distortion coefficient, surface contamination coefficient, etc.) have clear physical meanings and adjustable weight parameters, which facilitates calibration and adaptation according to different regions or sensors. The final output of "target effective spectral channel number" provides a direct quantitative basis for the optimized configuration of multispectral sensors or data screening, enhancing the engineering application value of the method. 3. This invention integrates geometric and pollution noise through a background noise model and subtracts it in the atmosphere-particle size-composition matching model. The core model focuses on enhancing the spectral signal that is sensitive to moisture, thereby increasing the proportion of effective information at the algorithm level, which is beneficial for the extraction of weak moisture features. Attached Figure Description

[0016] Figure 1 This is a schematic diagram of the process of the present invention. Detailed Implementation

[0017] To make the objectives, technical solutions, and advantages of this invention clearer, the invention will be further described in detail below with reference to the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the invention.

[0018] The specific implementation of the present invention will be described in detail below with reference to specific embodiments.

[0019] Please see Figure 1 The present invention provides an intelligent identification method for distinguishing between dry and wet sand using multispectral data, comprising: Geometric distortion coefficients are obtained by constructing a geometric distortion model based on the instantaneous field of view, imaging resolution, and surface wind speed. A background heterogeneity model was constructed based on the sand body surface bulk density, sand body porosity, and sand body salinity (expressed as mass fraction or volume fraction) to obtain the background heterogeneity coefficient; A surface pollution state model is constructed based on the coverage of algae and microbial films (biological disturbances with spectra similar to vegetation or dark substances) and the coverage of coexisting impurities (physical disturbances such as garbage and fallen leaves) to obtain the surface pollution coefficient. Based on the geometric distortion coefficient and surface pollution coefficient, atmospheric aerosol optical thickness, median sand particle size distribution, and volume percentage of major mineral components (quartz), an atmosphere-particle size-composition matching model is constructed to obtain the atmosphere-particle size-composition matching performance. Based on atmospheric-particle-composition matching, the current number of effective spectral channels, and the background heterogeneity coefficient, a channel optimization model is constructed to obtain the target number of effective spectral channels.

[0020] This invention systematically quantifies the impact of imaging conditions, sand body characteristics, and environmental disturbances on multispectral data by introducing geometric distortion models, background heterogeneity models, and surface contamination state models, in stark contrast to existing techniques that neglect the coupling effects of multiple factors. By constructing an atmosphere-particle size-composition matching model, this invention combines the aforementioned influencing factors with atmospheric conditions, sand grain characteristics, and mineral composition to comprehensively evaluate the matching of multispectral data to the moisture response of the sand body. This quantitative evaluation model involving multi-factor coupling is lacking in existing technologies. By calculating the atmosphere-particle size-composition matching, this method can accurately determine the potential of current multispectral data in distinguishing between dry and wet sand, avoiding blind identification under unsuitable conditions. Finally, based on the atmosphere-particle size-composition matching, the current number of effective spectral channels, and the background heterogeneity coefficient, a channel optimization model is constructed to dynamically obtain the target number of effective spectral channels. This enables the method to intelligently select the most effective spectral information for processing according to actual environmental conditions and sand body characteristics. Compared to existing methods that fix or empirically select spectral channels, the dynamic channel optimization mechanism of this invention significantly improves the adaptability and reliability of the method, effectively reducing the impact of noise and interference, thereby achieving more accurate and robust identification of dry and wet sand in complex scenarios.

[0021] Preferably, the geometric distortion coefficient is obtained in the following way: Acquire the instantaneous field of view, imaging resolution, and surface wind speed; The wind speed index is obtained by comparing the ground wind speed with the reference wind speed. The absolute difference between the instantaneous field of view and the imaging resolution and the corresponding optimal value is compared with the corresponding allowable deviation from the optimal value to obtain the field of view deviation index and the resolution deviation index. The wind speed index, field of view deviation index, and resolution deviation index are imported into the geometric distortion model to obtain the geometric distortion coefficients. The surface contamination state model performs a nonlinear function transformation on the square root of the product of the biofilm contamination index and the impurity contamination index. The complement of the result is the surface contamination coefficient. The larger the coefficient value, the lower the degree of surface contamination. The geometric distortion model is represented as follows:

[0022] in, Represents the geometric distortion coefficient. Indicates wind speed index, Indicates the field of view deviation index. Indicates the resolution deviation index, the Furthermore, the larger the value, the more ideal the imaging conditions.

[0023] Imaging instantaneous field of view (IFOV) is the angular range of the ground area that a remote sensing sensor can observe at a given moment, determining the image's detail capture capability. Imaging resolution refers to the smallest distinguishable ground feature size in the image, directly reflecting the image's fineness. These two parameters can be obtained by consulting the sensor's technical specifications or through ground calibration experiments. Surface wind speed is an important environmental factor affecting surface sand grain movement and atmospheric transport characteristics. It can be estimated through measurements at on-site weather stations, real-time monitoring with anemometers, or by combining numerical weather prediction models. The wind speed index is obtained by comparing the surface wind speed with a reference wind speed. Its purpose is to quantify the deviation of current wind conditions from a certain benchmark wind speed, thus reflecting the intensity of the potential impact of wind speed on imaging geometric distortion. For example, the currently measured surface wind speed can be compared with a preset reference wind speed that is generally considered to have the least impact on imaging, or normalized, so that the wind speed index can intuitively represent the degree of interference of wind speed on imaging stability. The absolute differences between the instantaneous field of view and imaging resolution and their corresponding optimal values ​​are compared with the corresponding allowable deviations from the optimal values ​​to obtain the field of view deviation index and resolution deviation index. The purpose is to quantify the degree to which the current imaging parameters deviate from ideal imaging conditions. The optimal values ​​typically refer to the instantaneous field of view and resolution that best guarantee image quality under sensor design or application scenarios, while the allowable deviation values ​​set the acceptable range of variation for these parameters without significantly affecting image quality. By calculating the absolute difference between the current value and the optimal value and comparing it with the allowable deviation value, a dimensionless index can be obtained. The larger the index, the more severe the deviation from the ideal imaging conditions, and the greater the possibility of geometric distortion. The wind speed index, field of view deviation index, and resolution deviation index are imported into the geometric distortion model to obtain the geometric distortion coefficient. This step comprehensively considers the influence of multiple factors on imaging geometric distortion and quantifies them into a unified coefficient. This step uses a predefined mathematical model, taking these independent indices as input, and outputs a value that reflects the overall degree of geometric distortion. The geometric distortion model is represented as follows: ,in Represents the geometric distortion coefficient. Indicates wind speed index, Indicates the field of view deviation index. The resolution deviation index is represented by an exponential function. This model uses an exponential function to nonlinearly combine the wind speed index, the field-of-view deviation index, and the resolution deviation index to reflect their combined impact on geometric distortion. The term represents the effect of wind speed on imaging stability, when the wind speed index... When the value is close to 1 (ideal wind speed condition), this term approaches 0, and the geometric distortion coefficient... A value close to 1 indicates ideal imaging conditions; when the wind speed index... When the wind speed is relatively low (high wind speed, unstable), this term increases, making... Decrease. The terms comprehensively reflect the degree to which the instantaneous field of view and imaging resolution deviate from their optimal values. The overall model design makes... The value of is in the range of (0,1], and the larger the value, the more ideal the imaging conditions and the smaller the geometric distortion.

[0024] The following is a specific example to illustrate this. The following steps can be used to obtain the geometric distortion coefficient. First, the instantaneous field of view and imaging resolution are obtained using a multispectral sensor mounted on a remote sensing platform. For example, a sensor in a specific operating mode might have an instantaneous field of view of 0.5 milliradians and an imaging resolution of 5 meters. Simultaneously, the ground-based automatic weather station or an anemometer mounted on a drone measures the surface wind speed in real time, for example, a current surface wind speed of 8 meters per second. Next, a reference wind speed is set, for example, 2 meters per second. The measured 8 meters per second is compared to 2 meters per second to obtain the wind speed index. The optimal value for the instantaneous field of view is set to 0.3 milliradians, with an allowable deviation of 0.1 milliradians; the optimal imaging resolution is set to 3 meters, with an allowable deviation of 1 meter. The absolute difference (0.2 milliradians) between the current instantaneous field of view of 0.5 milliradians and the optimal value of 0.3 milliradians is compared to the allowable deviation of 0.1 milliradians to obtain the field of view deviation index. Similarly, the absolute difference (2 meters) between the current imaging resolution of 5 meters and the optimal value of 3 meters is compared with the allowable deviation of 1 meter to obtain the resolution deviation index. Finally, the calculated wind speed index, field of view deviation index, and resolution deviation index are substituted into the geometric distortion model. The current geometric distortion coefficient can be calculated from this. For example, if the calculated wind speed index is 0.25, the field of view deviation index is 2, and the resolution deviation index is 2, then the geometric distortion coefficient... = 0.0498. This coefficient reflects the degree of geometric distortion under the current imaging conditions, providing a quantitative basis for subsequent background noise assessment.

[0025] Through the above technical solution, this application can accurately quantify the impact of instantaneous field of view, imaging resolution, and surface wind speed on the geometric distortion of remote sensing images. By introducing wind speed index, field of view deviation index, and resolution deviation index, and integrating them into the geometric distortion model, an objective and accurate geometric distortion coefficient can be obtained. The accurate acquisition of this coefficient effectively avoids the problem of inaccurate geometric distortion assessment caused by changes in imaging conditions and environmental factors, thus providing a more reliable input for the construction of the background noise model. This makes the calculation of the background noise coefficient more accurate, thereby improving the assessment accuracy of atmosphere-particle size-composition matching, and ultimately significantly improving the overall accuracy and robustness of the intelligent identification method for distinguishing between dry and wet sand using multispectral data, especially under complex and variable imaging and environmental conditions, enabling more stable differentiation between dry and wet sand.

[0026] Preferably, the background heterogeneity coefficient is obtained as follows: Obtain the surface bulk density, porosity, and salinity of the sand body (expressed as mass fraction or volume fraction). The absolute differences between the sand surface bulk density, sand porosity, and sand salinity and the corresponding reference values ​​are compared with the corresponding reference values ​​to obtain the bulk density index, porosity index, and salinity index. The bulk density index, porosity index, and salinity index are imported into the background heterogeneity model to obtain the background heterogeneity coefficient. The background heterogeneity model uses the weighted sum of bulk density index, porosity index and salinity index as the reciprocal function of the denominator to calculate the background heterogeneity coefficient. The larger the coefficient value, the better the background homogeneity of the sand body. The background heterogeneity model is represented as follows:

[0027] in, Indicates the background heterogeneity coefficient. This indicates the bulk density index. Indicates the porosity index. Indicates the salinity index. Represents the weight coefficient and The Furthermore, the larger the value, the better the background homogeneity.

[0028] Among them, the surface bulk density of sand refers to the mass of sand per unit volume, reflecting the compactness of the sand and directly affecting its spectral reflectance characteristics. Sand porosity refers to the proportion of pore volume to total volume in sand, determining its permeability and water retention capacity, thus affecting its wet and dry states and spectral response. Sand salinity refers to the content of soluble salts in sand, expressed as a mass fraction or volume fraction. The presence of salt alters the hygroscopicity of sand and may produce specific absorption or reflection characteristics in the spectrum. These parameters can be obtained in various ways. For example, they can be precisely measured in the laboratory after on-site sampling, such as using the ring sampler method to measure bulk density, the saturation method or drying method to measure porosity, and the conductivity meter or ion chromatography method to measure salinity. Alternatively, on-site sensors, such as soil densitometers, TDR (time domain reflectometer) sensors, or conductivity sensors, can be used for real-time or near-real-time data acquisition. The background heterogeneous model can be constructed in various forms, such as, as described in this invention, using a weighted sum model in reciprocal form, where the weight coefficients... The model can be trained and optimized using expert experience, regression analysis, or machine learning methods (such as genetic algorithms and particle swarm optimization) to improve the baseline heterogeneity coefficient of the output model. It can accurately reflect the actual homogeneity of sand bodies. In addition, a neural network-based model can be constructed, using bulk density index, porosity index, and salinity index as inputs, and learning the background heterogeneity coefficient through training.

[0029] The following is a specific example to illustrate this. When obtaining the surface bulk density, porosity, and salinity of a sand body, sand samples can be collected from the target area and sent to a laboratory for analysis. For example, the surface bulk density of the sand body is measured to be 1.65 g / cm³ using the ring sampler method, the porosity is measured to be 0.35 using the saturation method, and the salinity is measured to be 0.15% (mass fraction) using the conductivity method. Assuming a preset reference bulk density of 1.5 g / cm³, a reference porosity of 0.4, and a reference salinity of 0.1% (mass fraction), then the bulk density index... It can be calculated as |1.65 - 1.5| / 1.5 = 0.1; porosity index This can be calculated as |0.35 - 0.4| / 0.4 = 0.125; Salinity Index This can be calculated as |0.15 - 0.1| / 0.1 = 0.5. These indices are then imported into the background heterogeneous model, assuming weighting coefficients... =0.4, =0.3, =0.3, then the background heterogeneity coefficient = 0.8147. This calculated background heterogeneity coefficient This will be used as input for subsequent channel optimization models to guide the selection of effective spectral channels.

[0030] Through the above technical solution, this application can accurately quantify the background heterogeneity of sand bodies, thereby providing more accurate and representative input for subsequent channel optimization models. This refined consideration of the inherent physicochemical properties of sand bodies enables the effective reduction of spectral response uncertainty caused by sand body heterogeneity in complex and variable sand body environments, significantly improving the accuracy and reliability of multispectral data in distinguishing between dry and wet sand.

[0031] Preferably, the surface contamination coefficient is obtained in the following way: Obtain the coverage of algae and microbial film (biological disturbances with spectra similar to vegetation or dark substances) and the coverage of coexisting impurities (physical disturbances such as garbage and fallen leaves). The biofilm coverage rate and coexisting impurity coverage rate were compared with the corresponding reference values ​​to obtain the biofilm coverage rate index and the coexisting impurity coverage index. The biofilm coverage index and the coexisting impurity coverage index are imported into the surface contamination state model to obtain the surface contamination state coefficient. The geometric distortion model performs a comprehensive calculation on the ratio of the surface wind speed to the reference wind speed, the normalized deviation of the instantaneous field of view and the imaging resolution from their respective optimal values, so that the geometric distortion coefficient decreases as the wind speed increases and the field of view and resolution deviate from their optimal values. The larger the coefficient value, the more ideal the imaging geometry conditions. The surface contamination state model is represented as follows:

[0032] in, Represents the surface contamination state coefficient. Indicates the biofilm coverage index. The index represents the coverage of coexisting impurities. The higher the value, the better the surface condition.

[0033] Among these, algae and microbial biofilms, due to their spectral characteristics similar to vegetation or dark substances, significantly alter the spectral reflectance of sand bodies; coexisting impurities, due to their spectral irregularities, introduce random noise. Accurately obtaining these coverage rates is fundamental to assessing the degree of surface contamination. This step can be achieved through visual interpretation of high-resolution remote sensing imagery or automatic identification and quantification based on image segmentation algorithms. For example, vegetation indices can be used to identify algae and microbial biofilms, and texture features or spectral classification methods can be combined to identify coexisting impurities, then calculating their coverage ratio on the sand surface. Alternatively, a combination of on-site sampling and laboratory analysis can be used. For instance, several sampling areas can be delineated on the sand surface, samples can be collected, and the coverage area of ​​algae and microbial biofilms can be observed and counted under a microscope, while the coverage of physical impurities can be measured manually or through image processing techniques. Reference values ​​can be determined based on historical data, experimental data, or empirical values ​​under specific application scenarios. The core step of this scheme is to import the biofilm coverage index and the coexisting impurity coverage index into the surface contamination state model to obtain the surface contamination state coefficient. This model integrates the two different types of contamination indices and calculates the final surface contamination state coefficient through a nonlinear model. The model aims to quantify the combined impact of surface contamination on the spectral response of sand bodies, where A larger value indicates a better surface condition. The hyperbolic tangent function in the model can map the input value to a specific range and provide a nonlinear response, so that small changes in the degree of pollution have a greater impact when the pollution level is low, while the impact tends to saturate when the pollution level is high, which is more in line with the actual situation.

[0034] The following example illustrates this concept. Suppose we are conducting multispectral data collection in a desert area and need to assess the pollution status of the sand surface. First, we can use a drone equipped with a hyperspectral imager to acquire image data of the area. By processing these hyperspectral images—for example, using an improved normalized vegetation index or a specially designed algae index—we can identify and calculate the coverage rate of algae and microbial biofilms. Simultaneously, we can use image classification algorithms combined with texture and spectral features to identify coexisting impurities such as litter and fallen leaves in the images and calculate their coverage. For instance, if image analysis shows that the algae and microbial biofilm coverage rate is 5% and the coexisting impurity coverage rate is 2%, the preset reference values ​​can be set as follows: a reference coverage rate of 10% for algae and microbial biofilms and a reference coverage rate of 5% for coexisting impurities. Then, the biofilm coverage index... This can be calculated as 5% / 10% = 0.5, the coexisting impurity coverage index. This can be calculated as 2% / 5% = 0.4. Then, these two indices are imported into the surface contamination state model for calculation. = 0.5801. This calculated surface contamination state coefficient The closer the value is to 1, the better the surface condition; the closer it is to 0, the more severe the pollution. In this example, 0.5801 represents a moderately favorable surface condition. This coefficient can then be passed to the background noise model for subsequent atmospheric-particle-composition matching calculations.

[0035] Through the above technical solution, this application provides a more accurate and quantitative method for obtaining surface contamination coefficients. This method meticulously distinguishes between biological and physical contaminants, standardizes their respective coverage rates into exponents, and then uses a nonlinear model for comprehensive evaluation, effectively avoiding the coarseness of traditional methods in surface contamination assessment. In particular, this model can capture the nonlinear effects and potential synergistic effects of different types of pollutants on the spectral response of sand bodies, making the calculated surface contamination state coefficient more accurate. This allows for a more accurate reflection of the actual pollution level on the sand surface. This directly improves the accuracy of the surface pollution coefficient, thereby optimizing the input to the background noise model. Ultimately, this enables the intelligent identification method based on multispectral data to more effectively eliminate interference from surface pollution, significantly improving the accuracy and reliability of identifying the moisture state of the sand.

[0036] Preferably, the method for obtaining the atmosphere-particle size-composition matching is as follows: Obtain atmospheric aerosol optical thickness, median sand particle size distribution, and volume percentage of major mineral components (quartz). The thickness index, particle size distribution index, and component volume index are obtained by comparing the absolute differences between the atmospheric aerosol optical thickness, the median sand particle size distribution, and the volume percentage of the main mineral components (quartz) and the corresponding ideal values ​​with the allowable deviation values. Import the geometric distortion coefficient and surface contamination coefficient into the preset background noise model to obtain the background noise coefficient; The background noise model performs a weighted combination of the geometric distortion coefficient and the surface contamination coefficient, and calculates the combination result with a reference value, so that the background noise coefficient is negatively correlated with the ideality of imaging conditions and surface condition; The background noise model is represented as follows:

[0037] in, Represents the background noise coefficient. Represents the geometric distortion coefficient. Indicates the surface contamination coefficient. Represents the weight coefficient and ; The background noise figure, thickness index, particle size distribution index, and component volume index are imported into the atmosphere-particle size-component matching to obtain the atmosphere-particle size-component matching. The atmosphere-particle size-composition matching model integrates the thickness index, particle size distribution index, and composition volume index to characterize the degree of matching between the atmosphere, sand particle size, and mineral composition and the ideal spectral moisture response conditions. It is also combined with the background noise coefficient for calculation to finally obtain the atmosphere-particle size-composition matching degree. The larger the atmosphere-particle size-composition matching degree, the better the spectral response to moisture. The atmosphere-particle size-composition matching model is represented as follows:

[0038] in, Indicates atmospheric-particle-composition matching. Represents the background noise coefficient. Indicates the thickness index. Indicates the particle size distribution index. The volume index represents the component. Furthermore, the larger the value, the better the spectrum responds to moisture.

[0039] Atmospheric aerosol optical thickness reflects the degree of atmospheric attenuation of electromagnetic waves, directly affecting the surface reflection signals received by sensors. The median sand particle size distribution is an important indicator of the physical properties of sand bodies, as sand particles of different sizes exhibit different light scattering and absorption characteristics. The volume percentage of major mineral components (quartz) characterizes the chemical composition of the sand body; quartz, as a major component, significantly influences spectral characteristics. These parameters can be obtained in various ways. For example, atmospheric aerosol optical thickness can be obtained through inversion from ground-based or spaceborne remote sensing data, or estimated using atmospheric radiative transfer models combined with meteorological data. The median sand particle size distribution and the volume percentage of major mineral components can be obtained through on-site sampling analysis, laboratory spectral measurements combined with inversion algorithms, or mineral mapping using hyperspectral remote sensing data. The purpose of obtaining the thickness index, particle size distribution index, and component volume index by comparing the absolute differences between atmospheric aerosol optical thickness, median sand particle size distribution, and the volume percentage of major mineral components (quartz) and their corresponding ideal values ​​with allowable deviations is to quantify the deviation between actual environmental parameters and ideal conditions and standardize them into dimensionless indices. The ideal value represents the parameter value that is most favorable to the spectral response or closest to the standard conditions, while the allowable deviation value sets the acceptable error range. By calculating the ratio of the absolute difference between the actual value and the ideal value to the allowable deviation value, the thickness index, particle size distribution index, and composition volume index can be obtained. The larger these indices are, the greater the deviation between the actual parameters and the ideal conditions. The geometric distortion coefficient and the surface contamination coefficient are imported into a preset background noise model to obtain the background noise coefficient. The background noise model is represented as... The aim is to quantify the impact of background noise caused by imaging conditions and surface pollution on the spectral response. Geometric distortion coefficient. It reflects the ideal level of image quality, while the surface contamination coefficient These coefficients characterize the degree to which the surface is covered by biofilm and impurities. These two coefficients directly affect the signal-to-noise ratio of the effective signal in multispectral data. The preset background noise model calculates the background noise coefficient by weighted combination of these two coefficients. ,in, and These are weighting coefficients used to adjust the relative contributions of geometric distortion and surface contamination to background noise, and their sum is 1. These values ​​can be assigned through expert experience or obtained through the analytic hierarchy process (AHP). This model indicates that the more ideal the imaging conditions (…), the better. The larger the surface area, the less contamination it contains. The larger the background noise figure, the higher the background noise figure. The smaller the value, the larger the value. The atmosphere-particle size-composition matching model is the core matching calculation process, which integrates the degree of deviation of environmental parameters and the influence of background noise. Atmospheric-particle size-composition matching... Through an exponential decay function and a thickness exponent Particle size distribution index and component volume index Correlation means that the greater the deviation of environmental parameters from the ideal value, the worse the matching. Meanwhile, the denominator... The term introduces the influence of background noise, and the background noise coefficient is... Larger size means better atmospheric-particle-composition matching. The smaller the size, the better the design of the model makes The value of is between (0, 1], and the larger the value, the better the spectral response to moisture, that is, the greater the potential of multispectral data to distinguish between dry and wet sand under the current environmental and land cover conditions. This model can be constructed based on physical radiative transfer theory and empirical models, and can be verified and optimized through a large amount of measured data.

[0040] The following is a specific example to illustrate this. As a concrete implementation method, obtaining atmospheric-particle-composition matching can be performed as follows: First, the atmospheric aerosol optical thickness of the current area is obtained through inversion from satellite remote sensing data. Then, the median sand particle size distribution and the volume percentage of quartz are determined through on-site sampling and laboratory analysis. Assume the ideal atmospheric aerosol optical thickness is 0.1, with an allowable deviation of 0.05; the ideal median sand particle size distribution is 200 micrometers, with an allowable deviation of 50 micrometers; and the ideal quartz volume percentage is 90%, with an allowable deviation of 5%. If the actual measured atmospheric aerosol optical thickness is 0.12, the median sand particle size distribution is 260 micrometers, and the quartz volume percentage is 80%, then the thickness index can be calculated as |0.12-0.1| / 0.05 = 0.4, the particle size distribution index as |260-200| / 50 = 1.2, and the composition volume index as |80-90| / 5 = 2.0. Meanwhile, it is assumed that the geometric distortion coefficients have been obtained through the geometric distortion model and the surface contamination state model. The surface contamination coefficient is 0.85. The weighting factor is 0.7 in the preset background noise model. and The background noise figures are 0.6 and 0.4 respectively. = 0.21, and finally, the calculated background noise coefficient is... = 0.21, Thickness Index =0.4, Particle size distribution index =1.2 and component volume index Substituting 2.0 into the atmosphere-particle size-composition matching model, then =0.0226, this calculation result This refers to the atmospheric-particle-composition matching under current conditions, and its value reflects the potential of current multispectral data to distinguish between dry and wet sand.

[0041] Through the above technical solution, this method can accurately quantify the atmospheric-particle-composition matching. This scheme comprehensively considers multiple factors, including the atmospheric environment, the physicochemical properties of the sand body, and background noise from imaging and surface pollution, quantifying them into a thickness index, particle size distribution index, compositional volume index, and background noise coefficient. By organically integrating these indices and coefficients into the atmospheric-particle-composition matching model, a matching index that comprehensively reflects the current multispectral data's response to sand moisture can be obtained. This avoids the limitations of single-factor evaluation in traditional methods, significantly improving the accuracy and reliability of matching assessment. This precise matching quantification provides high-quality input for subsequent channel optimization models, making the selection of the target effective spectral channels more scientific and reasonable, thereby effectively improving the overall performance and recognition accuracy of intelligent identification methods that use multispectral data to distinguish between dry and wet sand.

[0042] Preferably, the channel optimization model adjusts the difference between the preset target atmosphere-particle-composition matching and the current atmosphere-particle-composition matching based on the background heterogeneity coefficient, and performs nonlinear scaling and rounding up on the current effective spectral channel number based on the adjusted result to output the target effective spectral channel number. The channel optimization model is expressed as follows:

[0043] in, Indicates the number of effective spectral channels of the target. Indicates the current number of valid spectral channels. Indicates the background heterogeneity coefficient. Indicates the target atmosphere-particle size-composition matching. This indicates the matching of atmosphere, particle size, and composition.

[0044] Among them, the number of effective spectral channels of the target This refers to the optimal number of spectral channels used to distinguish between dry and wet sand after optimization. It is not a fixed value but is dynamically adjusted based on environmental parameters to improve the accuracy and efficiency of identification. Current effective spectral channel count. This refers to the actual number of usable spectral channels in the multispectral data before channel optimization. This is typically determined by the hardware design of the multispectral sensor. For example, a sensor might have 100 spectral channels, but in a specific application, only a portion may be considered currently valid due to noise or data quality issues. This number can be obtained directly from the sensor specifications or statistically derived after data preprocessing steps (such as noise filtering and bad band removal). Background heterogeneity coefficient. This reflects the homogeneity of the sand's background environment; a higher value indicates a more homogeneous background and less interference with the spectral response. In the channel optimization model, it acts as a regulating factor, affecting the sensitivity of atmospheric-particle-composition matching to adjustments in the target number of channels. Target atmospheric-particle-composition matching. Atmosphere-particle-composition matching refers to the degree of matching between the atmosphere, sand particle size distribution, and major mineral components (such as quartz) and their spectral response under ideal or desired conditions. It represents a benchmark used to measure whether the matching under current conditions meets expectations. This matching can be calibrated experimentally under standard conditions or by setting a threshold based on expert experience. This reflects the combined influence of atmospheric aerosol optical thickness, median sand grain size distribution, and the volume percentage of major mineral components (quartz) on the multispectral data response under the current actual environment. A higher value indicates a better match between the current environment and the spectral response, which is more conducive to distinguishing between dry and wet sand. In the channel optimization model, its matching with the target is... The difference determines the direction and magnitude of the channel number adjustment. The hyperbolic tangent function (tanh) plays a non-linear adjustment role in this model, with its output value ranging from -1 to 1, capable of... The input values ​​are mapped to a finite range, thus smoothly controlling the adjustment range of the number of channels. When near When the value of the tanh function is close to 0, the adjustment range of the number of channels is small; when and When the difference is large, the value of the tanh function approaches 1 or -1, resulting in a larger adjustment range for the number of channels, but it will not increase indefinitely, ensuring the stability of the adjustment. (Rounding up sign) This means rounding the calculation result up to the nearest integer. Since the number of spectral channels must be an integer, this operation ensures the target effective number of spectral channels. It is a usable integer value.

[0045] The solution proposed in this application quantifies and dynamically adjusts the target effective spectral channel number by introducing a specific mathematical model. The core of this model is that it uses the current effective spectral channel number as a basis for its calculation. Based on this, the final target effective spectral channel number is dynamically determined through an adjustment factor. The adjustment factor is determined by the background heterogeneity coefficient. Target Atmosphere-Particle Size-Composition Matching and actual atmospheric-particle-composition matching A joint decision. Specifically, when Below When this occurs, it indicates that the current environment does not match the spectral response well, and more spectral channels may be needed to capture subtle differences. Therefore, the adjustment factor will be increased, thereby increasing the number of effective target spectral channels. Conversely, when Higher than or close to When environmental conditions are relatively ideal, accurate identification may not require a large number of spectral channels. The adjustment factor will decrease or remain stable, thereby reducing or maintaining the number of effective spectral channels for the target. Background heterogeneity coefficient It acts as a sensitivity modulator in this process, when the background heterogeneity is relatively good (i.e., When the value is large, the model is more sensitive to differences in matching, and the adjustment range of the number of channels is also larger. The introduction of the hyperbolic tangent function (tanh) ensures that this adjustment is a smooth and bounded nonlinear process, avoiding drastic fluctuations in the number of channels and thus improving the robustness of the model. Finally, by rounding up, the target effective spectral channel number is ensured. It is always a positive integer that can be practically applied. This model quantifies the impact of environmental factors on the spectral response and incorporates it into the dynamic optimization of the number of channels, enabling the system to adaptively select the most suitable combination of spectral channels according to real-time environmental changes. This solves the problem of how to efficiently and accurately determine the optimal number of spectral channels in complex and ever-changing environments.

[0046] The following is a concrete example to illustrate this. Assume that the current multispectral sensor has 100 effective spectral channels, i.e. = 100. Our target atmospheric-particle-composition matching = 0.9. As a specific implementation method, when the background heterogeneity coefficient of the sand body... = 0.7, and the actual atmospheric-particle-composition matching value calculated by the atmospheric-particle-composition matching model is 0.7. When the value is 0.8, it indicates that the current environment's compatibility is slightly lower than the ideal value. = 107. This means that under the current environment, in order to better distinguish between dry and wet sand, the system adjusts the number of effective spectral channels of the target to 107. For example, if the background heterogeneity coefficient of the sand body... = 0.9 (better background homogeneity), while the actual atmospheric-particle-composition matching = 0.5 (poor match), then = 135. This indicates that when the background homogeneity is good but the environmental matching is significantly reduced, the model will increase the number of channels more significantly to compensate for the information loss caused by insufficient matching. The above examples demonstrate that the model can dynamically and reasonably adjust the number of effective spectral channels according to different environmental parameters to adapt to complex and ever-changing application scenarios.

[0047] Through the above technical solution, this application provides a method for accurately quantifying and dynamically adjusting the number of effective spectral channels for a target. This method introduces a specific mathematical model to organically combine key environmental parameters such as background heterogeneity coefficient, target atmospheric-particle-composition matching, and actual atmospheric-particle-composition matching, achieving adaptive optimization of the number of spectral channels. This enables the system to intelligently select the most effective combination of spectral channels to distinguish between dry and wet sand based on changes in real-time environmental conditions, avoiding the problems of decreased recognition accuracy or resource waste that may result from a fixed number of channels. Especially in situations with complex and variable environments, high background heterogeneity of sand bodies, or unstable atmospheric-particle-composition matching, this model can provide more robust and accurate recognition performance, significantly improving the accuracy and efficiency of distinguishing between dry and wet sand.

[0048] A smart identification device for distinguishing between dry and wet sand using multispectral data is characterized by comprising a processor and a memory, wherein the memory stores a computer program, and when the computer program is executed by the processor, it implements the aforementioned smart identification method for distinguishing between dry and wet sand using multispectral data.

[0049] It should be noted that, in this document, relational terms such as "first" and "second" are used only to distinguish one entity or operation from another, and do not necessarily require or imply any such actual relationship or order between these entities or operations. Furthermore, the terms "comprising," "including," or any other variations thereof are intended to cover non-exclusive inclusion, such that a process, method, article, or apparatus that comprises a list of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such a process, method, article, or apparatus.

[0050] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.

Claims

1. A smart identification method for distinguishing between dry and wet sand using multispectral data, characterized in that, include: Geometric distortion coefficients are obtained by constructing a geometric distortion model based on the instantaneous field of view, imaging resolution, and surface wind speed. A background heterogeneity model was constructed based on the sand body surface bulk density, sand body porosity, and sand body salinity to obtain the background heterogeneity coefficient. A surface contamination status model was constructed based on the coverage of algae and microbial films and the coverage of coexisting impurities to obtain the surface contamination coefficient. Based on the geometric distortion coefficient and surface pollution coefficient, atmospheric aerosol optical thickness, median sand particle size distribution, and volume percentage of major mineral components, an atmosphere-particle size-composition matching model is constructed to obtain the atmosphere-particle size-composition matching performance. Based on atmospheric-particle-composition matching, the current number of effective spectral channels, and the background heterogeneity coefficient, a channel optimization model is constructed to obtain the target number of effective spectral channels.

2. The intelligent identification method for distinguishing between dry and wet sand using multispectral data according to claim 1, characterized in that, The channel optimization model adjusts the difference between the preset target atmospheric-particle-composition matching and the current atmospheric-particle-composition matching based on the background heterogeneity coefficient, and performs nonlinear scaling and rounding up on the current effective spectral channel number based on the adjusted result to output the target effective spectral channel number.

3. The intelligent identification method for distinguishing between dry and wet sand using multispectral data according to claim 2, characterized in that, The method for obtaining the atmospheric-particle-composition matching is as follows: Obtain atmospheric aerosol optical thickness, median sand particle size distribution, and volume percentage of major mineral components; The thickness index, particle size distribution index, and component volume index are obtained by comparing the absolute differences between atmospheric aerosol optical thickness, median sand particle size distribution, and volume percentage of major mineral components and their corresponding ideal values ​​with the allowable deviation values. Import the geometric distortion coefficient and surface contamination coefficient into the preset background noise model to obtain the background noise coefficient; The background noise model performs a weighted combination of the geometric distortion coefficient and the surface contamination coefficient, and calculates the combination result with a reference value, so that the background noise coefficient is negatively correlated with the ideality of imaging conditions and surface condition; The background noise figure, thickness index, particle size distribution index, and component volume index are imported into the atmosphere-particle size-component matching to obtain the atmosphere-particle size-component matching. The atmosphere-particle size-composition matching model integrates the thickness index, particle size distribution index, and component volume index to characterize the degree of matching between the atmosphere, sand particle size, and mineral composition and the ideal spectral moisture response conditions. It is also combined with the background noise coefficient for calculation to finally obtain the atmosphere-particle size-composition matching degree. The larger the atmosphere-particle size-composition matching degree, the better the spectral response to moisture.

4. The intelligent identification method for distinguishing between dry and wet sand using multispectral data according to claim 3, characterized in that, The surface contamination coefficient is obtained as follows: Obtain the coverage rate of algae and microorganisms as well as the coverage rate of coexisting impurities; The biofilm coverage rate and coexisting impurity coverage rate were compared with the corresponding reference values ​​to obtain the biofilm coverage rate index and the coexisting impurity coverage index. The biofilm coverage index and the coexisting impurity coverage index are imported into the surface contamination state model to obtain the surface contamination state coefficient.

5. The intelligent identification method for distinguishing between dry and wet sand using multispectral data according to claim 3, characterized in that, The method for obtaining the background heterogeneity coefficient is as follows: Obtain the surface bulk density, porosity, and salinity of the sand body; The absolute differences between the sand surface bulk density, sand porosity, and sand salinity and the corresponding reference values ​​are compared with the corresponding reference values ​​to obtain the bulk density index, porosity index, and salinity index. The density index, porosity index, and salinity index were imported into the background heterogeneity model to obtain the background heterogeneity coefficient.

6. The intelligent identification method for distinguishing between dry and wet sand using multispectral data according to claim 3, characterized in that, The geometric distortion coefficients are obtained as follows: Acquire the instantaneous field of view, imaging resolution, and surface wind speed; The wind speed index is obtained by comparing the ground wind speed with the reference wind speed. The absolute difference between the instantaneous field of view and the imaging resolution and the corresponding optimal value is compared with the corresponding allowable deviation from the optimal value to obtain the field of view deviation index and the resolution deviation index. The wind speed index, field of view deviation index, and resolution deviation index are imported into the geometric distortion model to obtain the geometric distortion coefficients.

7. The intelligent identification method for distinguishing between dry and wet sand using multispectral data according to claim 4, characterized in that, The surface contamination state model performs a nonlinear function transformation on the square root of the product of the biofilm contamination index and the impurity contamination index. The complement of the result is the surface contamination coefficient. The larger the coefficient value, the lower the degree of surface contamination.

8. The intelligent identification method for distinguishing between dry and wet sand using multispectral data according to claim 5, characterized in that, The background heterogeneity model calculates the background heterogeneity coefficient by using the weighted sum of the bulk density index, porosity index, and salinity index as the inverse function of the denominator. The larger the coefficient value, the better the background homogeneity of the sand body.

9. The intelligent identification method for distinguishing between dry and wet sand using multispectral data according to claim 6, characterized in that, The geometric distortion model performs a comprehensive calculation on the ratio of the surface wind speed to the reference wind speed, the normalized deviation of the instantaneous field of view and the imaging resolution from their respective optimal values, so that the geometric distortion coefficient decreases as the wind speed increases and the field of view and resolution deviate from their optimal values. The larger the coefficient value, the more ideal the imaging geometry conditions.

10. An intelligent identification device that uses multispectral data to distinguish between dry and wet sand, characterized in that, It includes a processor and a memory, the memory storing a computer program, which, when executed by the processor, implements the intelligent identification method for distinguishing between dry and wet sand using multispectral data as described in any one of claims 1 to 9.

Images