Fish diversity drawing method, device and equipment based on hyperspectral data and eDNA

Abstract

The invention relates to the technical field of remote sensing ecology, in particular to a fish diversity charting method and device based on hyperspectral data and eDNA, and the method comprises the steps: obtaining an eDNA sample and a hyperspectral remote sensing image of a target water area, calculating a fish diversity index of the target water area according to the eDNA sample, and calculating the fish diversity index of the target water area according to the eDNA sample; extracting spectral band information corresponding to the eDNA sample from the hyperspectral remote sensing image; a spectral feature library of fish diversity is constructed according to spectral band information corresponding to the eDNA samples, a fish diversity prediction model is constructed based on the spectral feature library and a machine learning regression model, and the fish diversity prediction model comprises a mapping relation between spectral features and fish diversity established based on the machine learning regression model; utilizing the fish diversity prediction model to predict the fish diversity of any water area, and generating a fish diversity spatial distribution diagram of any water area according to the fish diversity. Therefore, the problems of limited sampling points, poor characterization effect of the fish habitat and the like in related technologies are solved.

Description

Technical Field

[0001] This application relates to the field of remote sensing ecology technology, and in particular to a method, apparatus, device and medium for fish diversity mapping based on hyperspectral data and eDNA. Background Technology

[0002] With the increasing demand for fish diversity monitoring, traditional methods such as electrofishing and netting are unsuitable for large-scale, non-destructive monitoring due to their high cost and significant interference. eDNA (environmental deoxyribonucleic acid) technology, with its non-invasiveness and high sensitivity, can achieve simultaneous detection of multiple species through DNA extraction and high-throughput sequencing in water, significantly improving monitoring efficiency. Hyperspectral remote sensing technology can indirectly reflect fish habitat conditions by acquiring environmental parameters such as water quality and biomass. Combining eDNA with hyperspectral remote sensing can overcome spatial and temporal limitations, enabling precise, large-scale remote sensing mapping of fish diversity.

[0003] However, the relevant technologies are limited by the small number of sampling points and insufficient spatial representativeness, and are mostly used for terrestrial biological monitoring; at the same time, multispectral remote sensing has limited resolution, making it difficult to accurately reflect subtle changes in water body attributes, which affects the accuracy of fish habitat characterization. Summary of the Invention

[0004] This application provides a method, apparatus, device, and medium for fish diversity mapping based on hyperspectral data and eDNA, in order to solve the problems of limited sampling points and poor fish habitat characterization in related technologies.

[0005] The first aspect of this application provides a method for mapping fish diversity based on hyperspectral data and eDNA, comprising the following steps: acquiring eDNA samples and hyperspectral remote sensing images of a target water area; calculating a fish diversity index of the target water area based on the eDNA samples; extracting spectral band information corresponding to the eDNA samples from the hyperspectral remote sensing images; calculating and optimizing a vegetation index based on the spectral band information corresponding to the eDNA samples; constructing a spectral feature library of fish diversity; constructing a fish diversity prediction model based on the spectral feature library and a machine learning regression model, wherein the fish diversity prediction model includes a mapping relationship between spectral features and fish diversity established based on the machine learning regression model; predicting fish diversity in any water area using the fish diversity prediction model; and generating a spatial distribution map of fish diversity in any water area based on the fish diversity.

[0006] Optionally, in one embodiment of this application, calculating the fish diversity index of the target water area based on the eDNA sample includes: performing high-throughput sequencing on the eDNA sample to obtain raw sequence data; performing at least one processing on the raw sequence data, including adapter removal, chimera removal, noise reduction, and sequence clustering, to obtain a species information expression matrix; and calculating the fish diversity index of the target water area based on the species information expression matrix.

[0007] Optionally, in one embodiment of this application, extracting the spectral band information corresponding to the eDNA sample from the hyperspectral remote sensing image includes: performing radiometric calibration, atmospheric correction, geometric correction, and other processing on the hyperspectral remote sensing image to obtain the spectral reflectance of the hyperspectral remote sensing image; after image registration is completed, projecting the eDNA sample sampling point onto the hyperspectral remote sensing image according to the geographical coordinates of the eDNA sample sampling point, and extracting the spectral band information of the corresponding pixel.

[0008] Optionally, in one embodiment of this application, a spectral feature library of fish diversity is constructed by calculating and selecting vegetation indices based on the spectral band information corresponding to the eDNA sample, including: calculating vegetation indices based on the spectral band information corresponding to the sample points; constructing an original feature vector space based on the calculated vegetation indices; selecting features by calculating feature importance; and constructing a spectral feature library of fish diversity.

[0009] Optionally, in one embodiment of this application, constructing a fish diversity prediction model based on a spectral feature library and a machine learning regression model includes: using a spectral feature library as input and a fish diversity index as output, constructing a fish diversity prediction model using a machine learning regression model, training the fish diversity prediction model, and using a cross-validation strategy to evaluate the prediction accuracy of the fish diversity prediction model during the training process.

[0010] Optionally, in one embodiment of this application, predicting the fish diversity value of any water body using a fish diversity prediction model includes: acquiring a hyperspectral remote sensing image of any water body; extracting corresponding spectral band information from each pixel of the hyperspectral remote sensing image; calculating and selecting spectral features based on the spectral band information to construct a fish diversity spectral feature library; inputting the spectral features into the fish diversity prediction model, and the fish diversity prediction model outputting the fish diversity value of any water body.

[0011] Optionally, in one embodiment of this application, after generating a spatial distribution map of fish diversity in any water area based on fish diversity, the method further includes: image smoothing and spatial smoothing of the fish diversity spatial distribution map.

[0012] A second aspect of this application provides a fish diversity mapping device based on hyperspectral data and eDNA, comprising: a preprocessing module for acquiring eDNA samples and hyperspectral remote sensing images of a target water area, calculating a fish diversity index of the target water area based on the eDNA samples, and extracting spectral band information corresponding to the eDNA samples from the hyperspectral remote sensing images; a modeling module for calculating and optimizing vegetation indices based on the spectral band information corresponding to the eDNA samples, constructing a spectral feature library of fish diversity, and constructing a fish diversity prediction model based on the spectral feature library and a machine learning regression model, wherein the fish diversity prediction model includes a mapping relationship between spectral features and fish diversity established based on the machine learning regression model; and a prediction module for predicting fish diversity in any water area using the fish diversity prediction model, and generating a spatial distribution map of fish diversity in any water area based on the fish diversity.

[0013] Optionally, in one embodiment of this application, the preprocessing module is further used to perform high-throughput sequencing on the eDNA sample to obtain raw sequence data; to perform at least one of the following processing on the raw sequence data: adapter removal, chimera removal, noise reduction, and sequence clustering, to obtain a species information expression matrix; and to calculate the fish diversity index of the target water area based on the species information expression matrix.

[0014] Optionally, in one embodiment of this application, the preprocessing module is further used to perform radiometric calibration, atmospheric correction, geometric correction and other processing on the hyperspectral remote sensing image to obtain the spectral reflectance of the hyperspectral remote sensing image; after the image registration is completed, the eDNA sample sampling points are projected onto the hyperspectral remote sensing image according to the geographic coordinates of the eDNA sample sampling points, and the spectral band information of the corresponding pixels is extracted.

[0015] Optionally, in one embodiment of this application, the modeling module is further used to calculate the vegetation index based on the spectral band information corresponding to the sample points; construct the original feature vector space based on the calculated vegetation index; and construct a spectral feature library of fish diversity by optimizing the features through calculating feature importance.

[0016] Optionally, in one embodiment of this application, the modeling module is further configured to construct a fish diversity prediction model using a machine learning regression model with a spectral feature library as input and a fish diversity index as output, train the fish diversity prediction model, and use a cross-validation strategy to evaluate the prediction accuracy of the fish diversity prediction model during the training process.

[0017] Optionally, in one embodiment of this application, the prediction module is further configured to acquire a hyperspectral remote sensing image of any water body; extract corresponding spectral band information from each pixel of the hyperspectral remote sensing image; calculate and select spectral features based on the spectral band information to construct a fish diversity spectral feature library; input the spectral features into a fish diversity prediction model, and the fish diversity prediction model outputs fish diversity values ​​for any water body.

[0018] Optionally, in one embodiment of this application, the prediction module is further used for image smoothing and spatial smoothing of the fish diversity spatial distribution map.

[0019] A third aspect of this application provides an electronic device, including: a memory, a processor, and a computer program stored in the memory and executable on the processor. The processor executes the program to implement the fish diversity mapping method based on hyperspectral data and eDNA as described in the above embodiments.

[0020] A fourth aspect of this application provides a computer-readable storage medium having a computer program stored thereon, which is executed by a processor to implement the fish diversity mapping method based on hyperspectral data and eDNA as described in the above embodiments.

[0021] Therefore, this application has the following beneficial effects: First, eDNA samples and corresponding hyperspectral remote sensing images of the target water area are acquired. The fish diversity index of the target water area is calculated by analyzing the eDNA samples. Then, spectral band information corresponding to the eDNA sampling locations is extracted from the hyperspectral images to characterize the environmental features of the area. Next, spectral indices are calculated based on the extracted diagnostic spectral bands, constructing a spectral feature library of fish diversity. A machine learning regression algorithm is then used to establish a mapping relationship between spectral features and fish diversity, forming a fish diversity prediction model. Finally, this prediction model is used to predict fish diversity in any target water area, and a spatial distribution map of fish diversity is drawn based on the prediction results, achieving large-scale, high-resolution remote sensing mapping of diversity. This solves the problems of limited sampling points and poor fish habitat characterization in related technologies.

[0022] Additional aspects and advantages of this application will be set forth in part in the description which follows, and in part will be obvious from the description, or may be learned by practice of this application. Attached Figure Description

[0023] The above and / or additional aspects and advantages of this application will become apparent and readily understood from the following description of the embodiments taken in conjunction with the accompanying drawings, wherein: Figure 1 This is a flowchart of a fish diversity mapping method based on hyperspectral data and eDNA according to an embodiment of this application; Figure 2 This is a schematic diagram illustrating the construction process of a fish diversity mapping method based on hyperspectral data and eDNA according to an embodiment of this application. Figure 3 This is an example diagram of a fish diversity mapping device based on hyperspectral data and eDNA according to an embodiment of this application; Figure 4 This is a schematic diagram of the structure of an electronic device according to an embodiment of this application. Detailed Implementation

[0024] The embodiments of this application are described in detail below. Examples of the embodiments are shown in the accompanying drawings, wherein the same or similar reference numerals denote the same or similar elements or elements having the same or similar functions throughout. The embodiments described below with reference to the accompanying drawings are exemplary and intended to explain this application, and should not be construed as limiting this application.

[0025] The following describes a method, apparatus, device, and medium for fish diversity mapping based on hyperspectral data and eDNA, according to embodiments of this application, with reference to the accompanying drawings. Addressing the problems mentioned in the background art, this application provides a method for fish diversity mapping based on hyperspectral data and eDNA. In this method, firstly, eDNA samples and corresponding hyperspectral remote sensing images of a target water area are acquired. By analyzing the eDNA samples, a fish diversity index for the target water area is calculated. Subsequently, spectral band information corresponding to the eDNA sampling location is extracted from the hyperspectral image to characterize the environmental characteristics of the area. Next, spectral indices are calculated based on the extracted diagnostic spectral bands, constructing a spectral feature library for fish diversity. Combined with a machine learning regression algorithm, a mapping relationship between spectral features and fish diversity is established, forming a fish diversity prediction model. Finally, using this prediction model, fish diversity is predicted for any target water area, and a spatial distribution map of fish diversity is drawn based on the prediction results, achieving large-scale, refined diversity remote sensing mapping. This solves the problems of limited sampling points and poor fish habitat characterization in related technologies.

[0026] Specifically, Figure 1 This is a flowchart illustrating a fish diversity mapping method based on hyperspectral data and eDNA, provided in an embodiment of this application.

[0027] like Figure 1 As shown, this fish diversity mapping method based on hyperspectral data and eDNA includes the following steps: In step S101, eDNA samples and hyperspectral remote sensing images of the target water area are acquired, the fish diversity index of the target water area is calculated based on the eDNA samples, and the spectral band information corresponding to the eDNA samples is extracted from the hyperspectral remote sensing images.

[0028] eDNA samples are DNA fragments extracted from environmental media. These fragments originate from cells, mucus, and excrement released by organisms such as fish in the environment and are used to detect the presence and diversity of fish in specific water bodies. Hyperspectral remote sensing images are a type of remote sensing image with continuous, narrow-band spectral channels, capable of precisely capturing the spectral reflectance characteristics of target objects at different wavelengths. Compared to multispectral remote sensing, hyperspectral images provide higher spectral resolution, helping to identify subtle environmental differences in water bodies. Fish diversity indices are indicators that measure the diversity of fish in a given water body. Commonly used indicators include the Shannon diversity index and the Simpson diversity index, calculated by analyzing the fish species and abundance identified in eDNA, reflecting the ecological complexity of the water body. Spectral band information is the reflectance data corresponding to the eDNA sampling point location extracted from the hyperspectral image, used to characterize the physical and chemical properties of the water body at that location, such as chlorophyll concentration, suspended solids content, and water color.

[0029] Understandably, by extracting eDNA samples from target waters and corresponding hyperspectral remote sensing images, a precise correspondence between fish diversity information and environmental spectral characteristics is achieved. On the one hand, eDNA samples can efficiently and non-invasively acquire fish diversity indices, overcoming the shortcomings of traditional methods in terms of cost, interference, and spatial coverage. On the other hand, calculating the spectral characteristics of eDNA sampling points based on remote sensing images provides crucial data support for constructing a mapping relationship between "spectral characteristics and fish diversity."

[0030] In one embodiment of this application, calculating the fish diversity index of a target water area based on an eDNA sample includes: performing high-throughput sequencing on the eDNA sample to obtain raw sequence data; performing at least one of the following processing on the raw sequence data: adapter removal, chimera removal, noise reduction, and sequence clustering to obtain a species information expression matrix; and calculating the fish diversity index of the target water area based on the species information expression matrix.

[0031] High-throughput sequencing is a technique for rapidly and parallelly determining large amounts of DNA sequences. It is widely used in microbial community analysis and species identification in environmental samples. Adapter removal removes adapter sequences introduced during library construction from sequencing data. Chimera removal identifies and eliminates "chimera" sequences generated during sequencing due to PCR amplification errors; these sequences do not represent actual species. Noise reduction filters out low-quality or erroneous sequences resulting from sequencing errors. Sequence clustering groups DNA sequences with a similarity threshold into a single class to represent a potential species or taxa in the environment. A species information representation matrix (SEM) is a data structure used in biodiversity research to record the number of each species detected in different samples.

[0032] Understandably, by performing high-throughput sequencing and standardized sequence processing on eDNA samples from target water areas, constructing a species information expression matrix, and calculating a diversity index, it is possible not only to comprehensively reflect the composition and diversity characteristics of fish communities in the water, but also to serve as a high-quality biological response variable for hyperspectral remote sensing modeling.

[0033] In this embodiment, after collecting eDNA samples from outdoor water bodies, amplicon sequencing was performed using a high-throughput sequencing platform to obtain raw sequence data. Subsequently, using general quality control tools, steps such as adapter removal, chimera removal, noise reduction, and sequence clustering were performed to generate an effective species information representation matrix, such as an OTU (Operational Taxonomic Unit) table. Finally, based on this matrix, biodiversity indicators such as the Shannon index and Simpson index were calculated and used as response variables in subsequent modeling.

[0034] Amplicon sequencing is a high-throughput sequencing technology that specifically amplifies target DNA fragments using polymerase chain reaction (PCR) and is commonly used for the analysis of microbial and macrobial community structures. OTU is a unit representing microorganisms or biological groups. The Shannon index is an indicator used to assess species diversity in a community; a higher value indicates greater diversity. The Simpson index is an indicator used to assess community diversity; a lower value indicates higher diversity. The response variable, used as the predicted or explained variable in modeling, is, in this embodiment, a diversity index of a fish community, used to establish a relationship model with spectral characteristics.

[0035] In one embodiment of this application, extracting the spectral band information corresponding to the eDNA sample from a hyperspectral remote sensing image includes: performing radiometric calibration, atmospheric correction, geometric correction, and other processing on the hyperspectral remote sensing image to obtain the spectral reflectance of the hyperspectral remote sensing image; after image registration is completed, projecting the eDNA sample sampling points onto the hyperspectral remote sensing image according to the geographic coordinates of the eDNA sample sampling points, and extracting the spectral band information of the corresponding pixels.

[0036] Radiometric calibration is the process of converting the raw digital values ​​of remote sensing images into physically meaningful radiance or reflectance, used to eliminate inconsistencies in sensor response. Atmospheric correction processes remote sensing images to remove interference from atmospheric scattering and absorption, restoring the true spectral reflectance of ground surfaces. Geometric correction registers pixel locations in remote sensing images with geographic coordinates, ensuring that the spatial locations of ground features in the image match their actual geographic locations, facilitating spatial analysis and overlay. Spectral reflectance represents the ability of a ground surface to reflect incident light. Pixel spectral band information is a set of spectral reflectance values ​​for a given image pixel across all bands, forming the pixel's "spectral fingerprint," used to identify or characterize the attributes of ground features in that area. Image registration spatially aligns remote sensing images acquired from multiple sources or at different times, ensuring their geographic locations are consistent.

[0037] Understandably, by performing radiometric calibration, atmospheric correction, and geometric correction on hyperspectral remote sensing images, the spectral accuracy and spatial positioning accuracy of the images can be effectively improved, ensuring that the acquired spectral reflectance truly reflects the aquatic environment. Furthermore, by precisely matching eDNA sampling points to corresponding pixels in the image and extracting their spectral band information, a precise spatial correspondence between biological sampling data and remote sensing spectral data is achieved, providing data support for subsequent construction of a fish diversity spectral feature library and modeling.

[0038] In this embodiment, the hyperspectral remote sensing image used for modeling is first preprocessed with standard procedures such as radiometric calibration, atmospheric correction, and geometric correction to obtain accurate spectral reflectance data. Specifically, a hyperspectral remote sensing image (such as GF-5A) covering the target water area is acquired, radiance data is calculated based on the radiometric calibration coefficient, atmospheric correction is performed using the FLAASH module, and geometric correction is performed by selecting spatial correction points using Sentinel-2 remote sensing imagery to ensure accurate and consistent spatial positioning of the remote sensing image.

[0039] Subsequently, based on the GPS coordinates of the eDNA sampling points, the spectral information of the corresponding pixels in the remote sensing image was extracted. Specifically, the nearest neighbor method was used to determine the nearest pixel of the sampling point, and the spectral reflectance curve of that pixel was extracted to provide accurate spatial matching data for subsequent analysis.

[0040] In step S102, vegetation indices are calculated and optimized based on the spectral band information corresponding to the eDNA samples, and a spectral feature library of fish diversity is constructed. A fish diversity prediction model is constructed based on the spectral feature library and a machine learning regression model. The fish diversity prediction model includes the mapping relationship between spectral features and fish diversity established based on the machine learning regression model.

[0041] Understandably, by constructing a spectral feature library of fish diversity and establishing a mapping relationship between remote sensing spectra and fish diversity based on this feature library and a machine learning regression model, a fish diversity prediction model that can be widely applied can be formed, enabling the automatic estimation of fish diversity indices in waters from remote sensing data. This improves monitoring efficiency, reduces reliance on large-scale eDNA field sampling, and enables dynamic monitoring over a wider range and with higher timeliness.

[0042] In one embodiment of this application, a spectral feature library of fish diversity is constructed by calculating and selecting vegetation indices based on the spectral band information corresponding to the eDNA sample. This includes: calculating vegetation indices based on the spectral band information corresponding to the sample points; constructing an original feature vector space based on the calculated vegetation indices; selecting features by calculating feature importance; and constructing a spectral feature library of fish diversity.

[0043] Among them, the vegetation index is a remote sensing feature parameter calculated based on the combination of multiple bands in hyperspectral or multispectral remote sensing images. In this application, it is used to characterize the spectral features related to fish distribution in water bodies.

[0044] It is understood that the embodiments of this application can effectively compress redundant hyperspectral data dimensions, enhance the stability and generalization ability of the model, and improve modeling efficiency, providing a reliable data foundation for large-scale fish diversity remote sensing inversion.

[0045] In this embodiment of the application, after sample pairing is completed, the fish diversity index of the eDNA locus is used as the response variable, and the corresponding hyperspectral vegetation index (such as NDWI (Normalized Difference Water Index) and MNDWI (Modified Normalized Difference Water Index)) is used as the explanatory variable to form a fish diversity-spectral feature associated dataset.

[0046] To improve the model's generalization ability and prediction accuracy, feature selection methods such as random forest feature importance assessment and Pearson correlation analysis were employed. Vegetation indices, calculated using bands, were selected from the original spectral features. Furthermore, a representative and concise spectral feature library was constructed, ultimately forming the feature input set used for modeling.

[0047] in, X It is a set of multiple feature inputs of a sample. It is the first input n One characteristic.

[0048] In this context, the response variable is the variable that is predicted or explained in statistical modeling or machine learning; in this embodiment, it is the fish diversity index. The explanatory variable is the independent variable used to explain or predict the response variable; in this embodiment, it is the hyperspectral vegetation index. NDWI is a remote sensing index used to extract water body information, effectively highlighting water body characteristics and reducing interference from vegetation and soil. MNDWI is an improved form of NDWI, replacing the near-infrared band with the mid-infrared band, enhancing the ability to identify water bodies in complex backgrounds such as urban areas and shallow water areas. Random forest is an ensemble learning method used for classification and regression tasks. Pearson correlation analysis is a statistical method used to measure the degree of linear correlation between two variables, with a correlation coefficient ranging from -1 to 1. Physically interpretable spectral bands refer to bands whose spectral signals have a clear scientific connection with the specific physical or biological characteristics of water bodies or ecological environments, reasonably reflecting environmental changes and facilitating the understanding and interpretation of model results.

[0049] In one embodiment of this application, a fish diversity prediction model is constructed based on a spectral feature library and a machine learning regression model, including: using the spectral feature library as input and the fish diversity index as output, constructing a fish diversity prediction model using a machine learning regression model, training the fish diversity prediction model, and using a cross-validation strategy to evaluate the prediction accuracy of the fish diversity prediction model during the training process.

[0050] Cross-validation is a model validation technique that divides the data into multiple subsets, uses a subset of the data to train the model in turn, and uses the other subset to validate the model's performance in order to evaluate the model's generalization ability and prediction accuracy.

[0051] Understandably, by constructing a fish diversity prediction model based on a spectral feature library and machine learning, the embodiments of this application achieve a precise mapping between remote sensing data and fish diversity, improving the accuracy and efficiency of monitoring. Cross-validation ensures the stability and generalization ability of the model, making it suitable for dynamic monitoring of large-scale water areas.

[0052] Based on the calculation of water-related bands and the optimization of spectral feature libraries such as vegetation indices, the constructed spectral feature library is used as input. X The fish diversity index calculated using eDNA is used as the output target. Y Establish a supervised regression model: This model is used to predict fish diversity. Model training employs machine learning regression algorithms such as LightGBM (Light Gradient Boosting Machine), XGBoost (Extreme Gradient Boosting), and Random Forest, and optimizes model parameters through grid search to obtain the optimal model structure.

[0053] To ensure the stability and generalization ability of the model, embodiments of this application employ 5-fold cross-validation or leave-one-out cross-validation to evaluate the model. The coefficient of determination is calculated. Indicators such as root mean square error and mean absolute error are used, and errors are analyzed by combining scatter plots of predicted and measured values ​​to ensure that the model has good predictive accuracy and physical interpretability. Ultimately, the optimal output model can be used for spatial prediction and dynamic monitoring of fish diversity in large-scale waters.

[0054] Supervised regression models are a type of model training method in supervised learning, using known inputs and outputs to learn the mapping relationship and predict values ​​in unknown data. LightGBM is a high-performance gradient boosting framework for regression and classification problems on large-scale datasets and high-dimensional feature data. XGBoost is an ensemble learning algorithm based on gradient boosting trees. Grid search is a hyperparameter optimization method that exhaustively searches all combinations within a set parameter range to select the optimal parameter configuration, thereby improving model performance. 5-fold cross-validation divides the data into 5 parts, using one part for testing in turn, and the rest for training, repeating 5 times and averaging. Leave-one-out cross-validation means that only one sample is left for testing in each round, and the rest are used as the training set; this is suitable for situations with small sample sizes. Coefficient of determination. This is a statistic that measures the consistency between model predictions and actual observations, ranging from 0 to 1. A higher value indicates a better fit. Root mean square error (RMSE) represents the average difference between predicted and actual values, with units consistent with the original variable; a smaller value indicates higher prediction accuracy. Mean absolute error (MAO) is the average of the absolute values ​​of all prediction errors, directly measuring prediction bias; a smaller value indicates higher accuracy.

[0055] In step S103, the fish diversity prediction model is used to predict the fish diversity of any water body, and a spatial distribution map of fish diversity of any water body is generated based on the fish diversity.

[0056] It is understood that, through the fish diversity prediction model, the embodiments of this application can achieve rapid and accurate prediction of fish diversity in different waters, breaking the limitations of traditional monitoring that relies on on-site sampling. The spatial distribution map generated based on the prediction results can intuitively reflect the spatial variation pattern of fish diversity.

[0057] In one embodiment of this application, the fish diversity prediction model is used to predict the fish diversity value of any water body, including: acquiring a hyperspectral remote sensing image of any water body; extracting the corresponding spectral band information from each pixel of the hyperspectral remote sensing image; calculating and selecting spectral features based on the spectral band information to construct a fish diversity spectral feature library; inputting the spectral features into the fish diversity prediction model, and the fish diversity prediction model outputting the fish diversity value of any water body.

[0058] A pixel is the smallest unit in a remote sensing image, representing the spectral information of a specific spatial location on the ground.

[0059] This application embodiment utilizes hyperspectral remote sensing technology to map fish diversity. First, a trained fish diversity prediction model is applied to hyperspectral remote sensing image data of the target area. For each pixel in the image, the spectral features of that pixel are calculated and optimized, and these spectral features are input into the prediction model to calculate the corresponding fish diversity index point by point. Finally, the prediction results of all pixels are integrated to generate a continuous spatial distribution map of fish diversity within the target area.

[0060] Specifically, for each pixel i Calculate spectral characteristics X i Input the model to obtain the predicted diversity value Y i = f (X i ) This allows for the construction of a spatial distribution map of fish diversity at the regional scale. R(Y) .

[0061] In one embodiment of this application, after generating a spatial distribution map of fish diversity in any water area based on fish diversity, the method further includes: image smoothing and spatial smoothing of the fish diversity spatial distribution map.

[0062] Image smoothing utilizes filtering algorithms to locally average or adjust the values ​​of image pixels, reducing local noise or outliers and improving image continuity and visual quality. Spatial smoothing is a spatial analysis method in image processing that enhances the geographic consistency of prediction results through smoothing.

[0063] Understandably, in order to improve the spatial continuity and ecological interpretability of fish diversity prediction maps, embodiments of this application employ a filtering and smoothing method to process the preliminary prediction map. Specifically, by applying spatial filtering techniques, the pixel values ​​in the prediction map are smoothed, effectively reducing the impact of isolated outliers and noise, thereby improving image continuity and ecological interpretability.

[0064] In summary, the hyperspectral remote sensing mapping process for fish diversity based on the fusion of hyperspectral remote sensing and eDNA data in this application embodiment is as follows: Figure 2 As shown: The first stage involves data acquisition, which includes collecting field samples (including fish eDNA and latitude / longitude information) and acquiring hyperspectral remote sensing images of the corresponding regions. The next stage is data preprocessing, which involves subsequence amplification, adapter removal, and sequence splicing of the eDNA data. Simultaneously, the remote sensing images undergo radiometric calibration, atmospheric correction, and geometric correction to ensure the accuracy and usability of their spectral information.

[0065] Subsequently, a spectral feature library was constructed and optimized. By matching sample points, spectral reflectance data corresponding to eDNA sample points were extracted from hyperspectral images. Relevant vegetation indices (such as NDWI, MNDWI, etc.) were calculated. Combined with feature screening methods (such as random forest feature importance, correlation analysis, etc.), a spectral feature set that is highly correlated with fish diversity and has physical interpretability was constructed, and finally a spectral feature library for modeling was formed.

[0066] Next, we move on to the regression prediction model construction stage. Based on the constructed spectral feature library (as input) and the fish diversity index calculated by eDNA (as output), we use machine learning regression algorithms such as LightGBM, Random Forest and XGBoost to build a supervised regression model, and perform parameter tuning and training to complete the construction of the prediction model.

[0067] Finally, in the hyperspectral remote sensing mapping stage of fish diversity, the trained model is applied to pixel-level prediction of the entire remote sensing image, outputting the fish diversity index corresponding to each pixel, resulting in a continuous spatial distribution map. To improve image coherence and ecological interpretability, a spatial smoothing method is further used to filter the prediction results, ultimately outputting a high-quality fish diversity remote sensing map. This process enables efficient extrapolation of fish diversity spatial information from point samples to regional scales, providing technical support for ecological monitoring and conservation.

[0068] According to the fish diversity mapping method based on hyperspectral data and eDNA proposed in this application, firstly, eDNA samples and corresponding hyperspectral remote sensing images of the target water area are acquired. By analyzing the eDNA samples, a fish diversity index of the target water area is calculated. Subsequently, spectral band information corresponding to the eDNA sampling location is extracted from the hyperspectral image to characterize the environmental features of the area. Next, based on the extracted diagnostic spectral bands, a spectral index is calculated to construct a spectral feature library of fish diversity. Combined with a machine learning regression algorithm, a mapping relationship between spectral features and fish diversity is established, forming a fish diversity prediction model. Finally, using this prediction model, fish diversity is predicted for any target water area, and a spatial distribution map of fish diversity is drawn based on the prediction results, achieving large-scale, refined diversity remote sensing mapping. This solves the problems of limited sampling points and poor fish habitat characterization in related technologies. Next, with reference to the accompanying drawings, a fish diversity mapping device based on hyperspectral data and eDNA proposed in this application is described.

[0069] Figure 3 This is a block diagram of a fish diversity mapping device based on hyperspectral data and eDNA, according to an embodiment of this application.

[0070] like Figure 3 As shown, the fish diversity mapping device 10 based on hyperspectral data and eDNA includes: a preprocessing module 101, a modeling module 102, and a prediction module 103.

[0071] The preprocessing module 101 is used to acquire eDNA samples and hyperspectral remote sensing images of the target water area, calculate the fish diversity index of the target water area based on the eDNA samples, and extract the spectral band information corresponding to the eDNA samples from the hyperspectral remote sensing images. The modeling module 102 is used to construct a spectral feature library of fish diversity based on the spectral feature information corresponding to the eDNA samples, and construct a fish diversity prediction model based on the spectral feature library and a machine learning regression model. The fish diversity prediction model includes the mapping relationship between spectral features and fish diversity established based on the machine learning regression model. The prediction module 103 is used to predict the fish diversity of any water area using the fish diversity prediction model, and generate a spatial distribution map of fish diversity in any water area based on the fish diversity.

[0072] In one embodiment of this application, the preprocessing module 101 is further used to perform high-throughput sequencing on the eDNA sample to obtain raw sequence data; to perform at least one of the following processing on the raw sequence data: adapter removal, chimera removal, noise reduction, and sequence clustering, to obtain a species information expression matrix; and to calculate the fish diversity index of the target water area based on the species information expression matrix.

[0073] In one embodiment of this application, the preprocessing module 101 is further used to perform radiometric calibration, atmospheric correction, geometric correction and other processing on the hyperspectral remote sensing image to obtain the spectral reflectance of the hyperspectral remote sensing image; after the image registration is completed, the eDNA sample sampling points are projected onto the hyperspectral remote sensing image according to the geographical coordinates of the eDNA sample sampling points, and the spectral band information of the corresponding pixels is extracted.

[0074] In one embodiment of this application, the modeling module 102 is further used to calculate the vegetation index based on the spectral band information corresponding to the sample points; construct the original feature vector space based on the calculated vegetation index, and construct a spectral feature library of fish diversity by optimizing the features by calculating the feature importance.

[0075] In one embodiment of this application, the modeling module 102 is further configured to construct a fish diversity prediction model using a spectral feature library as input and a fish diversity index as output, and to train the fish diversity prediction model. During the training process, a cross-validation strategy is used to evaluate the prediction accuracy of the fish diversity prediction model.

[0076] In one embodiment of this application, the prediction module 103 is further configured to acquire a hyperspectral remote sensing image of any water body; extract corresponding spectral band information from each pixel of the hyperspectral remote sensing image; calculate and select spectral features based on the spectral band information to construct a fish diversity spectral feature library; input the spectral features into a fish diversity prediction model, and the fish diversity prediction model outputs fish diversity values ​​for any water body.

[0077] In one embodiment of this application, the prediction module 103 is further used for processing the image smoothing and spatial smoothing of the fish diversity spatial distribution map.

[0078] According to the fish diversity mapping device based on hyperspectral data and eDNA proposed in this application, firstly, eDNA samples and corresponding hyperspectral remote sensing images of the target water area are acquired. By analyzing the eDNA samples, a fish diversity index of the target water area is calculated. Subsequently, spectral band information corresponding to the eDNA sampling location is extracted from the hyperspectral image to characterize the environmental features of the area. Next, based on the extracted diagnostic spectral bands, a spectral index is calculated to construct a spectral feature library of fish diversity. Combined with a machine learning regression algorithm, a mapping relationship between spectral features and fish diversity is established, forming a fish diversity prediction model. Finally, using this prediction model, fish diversity is predicted for any target water area, and a spatial distribution map of fish diversity is drawn based on the prediction results, achieving large-scale, high-resolution remote sensing mapping of diversity. This solves the problems of limited sampling points and poor fish habitat characterization in related technologies.

[0079] Figure 4A schematic diagram of the structure of an electronic device provided in an embodiment of this application. The electronic device may include: The memory 401, the processor 402, and the computer program stored on the memory 401 and capable of running on the processor 402.

[0080] When the processor 402 executes the program, it implements the fish diversity mapping method based on hyperspectral data and eDNA provided in the above embodiments.

[0081] Furthermore, electronic devices also include: Communication interface 403 is used for communication between memory 401 and processor 402.

[0082] The memory 401 is used to store computer programs that can run on the processor 402.

[0083] The memory 401 may include high-speed RAM (Random Access Memory) memory, and may also include non-volatile memory, such as at least one disk storage.

[0084] If the memory 401, processor 402, and communication interface 403 are implemented independently, then the communication interface 403, memory 401, and processor 402 can be interconnected via a bus to complete communication between them. The bus can be an ISA (Industry Standard Architecture) bus, a PCI (Peripheral Component Interconnect) bus, or an EISA (Extended Industry Standard Architecture) bus, etc. The bus can be divided into address bus, data bus, control bus, etc. For ease of representation, Figure 4 The bus is represented by a single thick line, but this does not mean that there is only one bus or one type of bus.

[0085] Optionally, in a specific implementation, if the memory 401, processor 402, and communication interface 403 are integrated on a single chip, then the memory 401, processor 402, and communication interface 403 can communicate with each other through an internal interface.

[0086] Processor 402 may be a CPU (Central Processing Unit), an ASIC (Application Specific Integrated Circuit), or one or more integrated circuits configured to implement the embodiments of this application.

[0087] This application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described method for mapping fish diversity based on hyperspectral data and eDNA.

[0088] In the description of this specification, the references to terms such as "one embodiment," "some embodiments," "example," "specific example," or "some examples," etc., refer to specific features, structures, materials, or characteristics described in connection with that embodiment or example, which are included in at least one embodiment or example of this application. In this specification, the illustrative expressions of the above terms do not necessarily refer to the same embodiment or example. Furthermore, the specific features, structures, materials, or characteristics described may be combined in any suitable manner in one or more embodiments or examples. Moreover, without contradiction, those skilled in the art can combine and integrate the different embodiments or examples described in this specification, as well as the features of different embodiments or examples.

[0089] Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of technical features indicated. Thus, a feature defined as "first" or "second" may explicitly or implicitly include at least one of that feature. In the description of this application, "N" means at least two, such as two, three, etc., unless otherwise explicitly specified.

[0090] Any process or method described in the flowchart or otherwise herein can be understood as representing a module, segment, or portion of code comprising one or N executable instructions for implementing custom logic functions or processes, and the scope of the preferred embodiments of this application includes additional implementations in which functions may be performed not in the order shown or discussed, including substantially simultaneously or in reverse order depending on the functions involved, as should be understood by those skilled in the art to which embodiments of this application pertain.

[0091] It should be understood that various parts of this application can be implemented using hardware, software, firmware, or a combination thereof. In the above embodiments, steps or methods can be implemented using software or firmware stored in memory and executed by a suitable instruction execution system. For example, if implemented in hardware, as in another embodiment, it can be implemented using any of the following techniques known in the art, or a combination thereof: discrete logic circuits having logic gates for implementing logical functions on data signals, application-specific integrated circuits (ASICs) having suitable combinational logic gates, programmable gate arrays (FPGAs), field-programmable gate arrays (FPGAs), etc.

[0092] Those skilled in the art will understand that all or part of the steps of the methods implementing the above embodiments can be implemented by a program instructing related hardware. The program can be stored in a computer-readable storage medium, and when executed, the program includes one or a combination of the steps of the method embodiments.

[0093] Although embodiments of this application have been shown and described above, it is understood that the above embodiments are exemplary and should not be construed as limiting this application. Those skilled in the art can make changes, modifications, substitutions and variations to the above embodiments within the scope of this application.

Claims

1. A method for mapping fish diversity based on hyperspectral data and eDNA, characterized in that, Includes the following steps: Acquire eDNA samples and hyperspectral remote sensing images of the target water area, calculate the fish diversity index of the target water area based on the eDNA samples, and extract the spectral band information corresponding to the eDNA samples from the hyperspectral remote sensing images; Based on the spectral band information corresponding to the eDNA sample, vegetation indices are calculated and optimized to construct a spectral feature library of fish diversity. A fish diversity prediction model is constructed based on the spectral feature library and a machine learning regression model. The fish diversity prediction model includes a mapping relationship between spectral features and fish diversity established based on the machine learning regression model. The fish diversity prediction model is used to predict the fish diversity of any water body, and a spatial distribution map of fish diversity in the arbitrary water body is generated based on the fish diversity.

2. The fish diversity mapping method based on hyperspectral data and eDNA according to claim 1, characterized in that, The calculation of the fish diversity index of the target water area based on the eDNA sample includes: The eDNA sample was subjected to high-throughput sequencing to obtain raw sequence data; The original sequence data is subjected to at least one process, including connector removal, chimera removal, noise reduction, and sequence clustering, to obtain a species information representation matrix; The fish diversity index of the target water area is calculated based on the species information representation matrix.

3. The method for fish diversity mapping based on hyperspectral data and eDNA according to claim 1, characterized in that, The step of extracting the spectral band information corresponding to the eDNA sample from the hyperspectral remote sensing image includes: The hyperspectral remote sensing image is processed by radiometric calibration, atmospheric correction, geometric correction, etc., to obtain the spectral reflectance of the hyperspectral remote sensing image; After image registration is completed, the eDNA sample sampling points are projected onto the hyperspectral remote sensing image based on their geographic coordinates, and the spectral band information of the corresponding pixels is extracted.

4. The method for fish diversity mapping based on hyperspectral data and eDNA according to claim 1, characterized in that, The step of calculating and optimizing vegetation indices based on the spectral band information corresponding to the eDNA sample to construct a spectral feature library of fish diversity includes: The vegetation index is calculated based on the spectral band information corresponding to the sample points. The original feature vector space is constructed based on the calculated vegetation index. After selecting the best features by calculating the importance of the features, a spectral feature library of fish diversity is constructed.

5. The method for fish diversity mapping based on hyperspectral data and eDNA according to claim 1, characterized in that, The construction of a fish diversity prediction model based on the spectral feature library and machine learning regression model includes: Using the spectral feature library as input and the fish diversity index as output, the fish diversity prediction model is constructed using the machine learning regression model. The fish diversity prediction model is trained, and during the training process, a cross-validation strategy is used to evaluate the prediction accuracy of the fish diversity prediction model.

6. The method for fish diversity mapping based on hyperspectral data and eDNA according to claim 1, characterized in that, The method of predicting fish diversity values ​​for any body of water using the fish diversity prediction model includes: Acquire hyperspectral remote sensing images of any body of water; Extract the corresponding spectral band information from each pixel of the hyperspectral remote sensing image; A database of spectral features for fish diversity is constructed by calculating and optimizing spectral features based on spectral band information. The spectral features are input into the fish diversity prediction model, and the fish diversity prediction model outputs the fish diversity value of the arbitrary water area.

7. The method for fish diversity mapping based on hyperspectral data and eDNA according to claim 1, characterized in that, After generating a spatial distribution map of fish diversity for the arbitrary water area based on the fish diversity, the method further includes: Image smoothing and spatial smoothing processing of the spatial distribution map of fish diversity.

8. A fish diversity mapping device based on hyperspectral data and eDNA, characterized in that, include: The preprocessing module is used to acquire eDNA samples and hyperspectral remote sensing images of the target water area, calculate the fish diversity index of the target water area based on the eDNA samples, and extract the spectral band information corresponding to the eDNA samples from the hyperspectral remote sensing images. The modeling module is used to calculate and select vegetation indices based on the spectral band information corresponding to the eDNA sample, construct a spectral feature library of fish diversity, and construct a fish diversity prediction model based on the spectral feature library and a machine learning regression model. The fish diversity prediction model includes a mapping relationship between spectral features and fish diversity established based on the machine learning regression model. The prediction module is used to predict the fish diversity of any water body using the fish diversity prediction model, and generate a spatial distribution map of fish diversity in the arbitrary water body based on the fish diversity.

9. An electronic device, characterized in that, include: A memory, a processor, and a computer program stored in the memory and executable on the processor, the processor executing the program to implement the fish diversity mapping method based on hyperspectral data and eDNA as described in any one of claims 1-7.

10. A computer-readable storage medium having a computer program or instructions stored thereon, characterized in that, When the computer program or instructions are executed, they implement the fish diversity mapping method based on hyperspectral data and eDNA as described in any one of claims 1-7.

Images