Polymetallic nodule distribution prediction method and device, electronic equipment and storage medium

Abstract

The invention discloses a multi-metal nodule distribution prediction method and device, electronic equipment and a storage medium. The method comprises the following steps: preprocessing initial multi-source seabed observation data to obtain target multi-source seabed observation data; performing sample balance on the to-be-enhanced sample data set by adopting a synthetic minority class oversampling technology to obtain a synthetic sample data set; constructing a training sample data set according to the target multi-source seabed observation data and the synthetic sample data set; inputting the training sample data set into a random forest classifier for training to obtain a multi-metal nodule distribution prediction model; and inputting the to-be-predicted data set and the to-be-predicted tuberculosis abundance classification threshold into the multi-metal tuberculosis distribution prediction model to obtain a multi-metal tuberculosis distribution prediction result in the research area. The method can reduce the modeling deviation caused by uneven distribution of tuberculosis abundance extreme value samples, improves the recognition capability and prediction precision of the model for various samples, improves the spatial resolution of a prediction result, and can be widely applied to the technical field of deep sea resource prediction.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of deep-sea resource prediction, and in particular to a polymetallic nodule distribution prediction method and device, an electronic device, and a storage medium. BACKGROUND

[0002] Polymetallic nodules are an important mineral resource, and nodule abundance is a key parameter for evaluating polymetallic nodule resources. Currently, when estimating the amount of deep-sea polymetallic nodules, traditional methods mainly rely on sparse point sampling data combined with Kriging interpolation for resource prediction, which has limitations such as low resolution and the need for a large number of samples. Methods based on machine learning for estimating the amount of deep-sea polymetallic nodules have limited ability to represent the nonlinear relationship between input features and target variables.

[0003] To sum up, the technical problems in the related art need to be improved. SUMMARY

[0004] Embodiments of the present application aim to at least partially solve one of the technical problems in the related art. To this end, the main purpose of the embodiments of the present application is to propose a polymetallic nodule distribution prediction method and device, an electronic device, and a storage medium, which can reduce the modeling bias caused by uneven distribution of nodule abundance extreme value samples, improve the recognition ability and prediction accuracy of the model for various samples, and improve the spatial resolution of the prediction results.

[0005] To achieve the above purpose, one aspect of the embodiments of the present application proposes a polymetallic nodule distribution prediction method, which comprises the following steps:

[0006] Obtain initial multi-source seabed observation data of a deep-sea study area, and pre-process the initial multi-source seabed observation data to obtain target multi-source seabed observation data;

[0007] Perform data analysis on the target multi-source seabed observation data to determine a to-be-enhanced sample data set;

[0008] Perform sample balancing processing on the to-be-enhanced sample data set using a synthetic minority over-sampling technique to obtain a synthetic sample data set;

[0009] Construct a training sample data set according to the target multi-source seabed observation data and the synthetic sample data set;

[0010] Input the training sample data set into a random forest classifier for model training to obtain a polymetallic nodule distribution prediction model;

[0011] Input a to-be-predicted data set and a to-be-predicted nodule abundance classification threshold into the polymetallic nodule distribution prediction model to obtain a polymetallic nodule distribution prediction result of the study area.

[0012] In some embodiments, the initial multi-source seabed observation data includes initial seabed topography data, initial echo intensity data, and sample position information, the preprocessing of the initial multi-source seabed observation data to obtain target multi-source seabed observation data includes:

[0013] The water depth data in the initial seabed topography data is subjected to Mercator projection processing to obtain a first plane rectangular coordinate system;

[0014] The initial echo intensity data is subjected to Mercator projection processing to obtain a second plane rectangular coordinate system;

[0015] The first plane rectangular coordinate system is subjected to gridding processing to obtain first grid data;

[0016] The second plane rectangular coordinate system is subjected to gridding processing to obtain second grid data;

[0017] The terrain slope data and the terrain roughness data are calculated according to the water depth data in the first grid data;

[0018] The candidate multi-source seabed observation data is constructed according to the water depth data in the first grid data, the second grid data, the terrain slope data, the terrain roughness data, and the sample position information;

[0019] The candidate multi-source seabed observation data is subjected to data cleaning processing to obtain the target multi-source seabed observation data.

[0020] In some embodiments, the data analysis of the target multi-source seabed observation data to determine a to-be-enhanced sample data set includes:

[0021] The target multi-source seabed observation data is subjected to sample classification processing according to a preset test nodule abundance classification threshold to obtain a positive sample data set and a negative sample data set;

[0022] The to-be-enhanced sample data set is determined according to the number of positive samples corresponding to the positive sample data set and the number of negative samples corresponding to the negative sample data set.

[0023] In some embodiments, the sample balancing processing of the to-be-enhanced sample data set by using a synthetic minority over-sampling technique to obtain a synthetic sample data set includes:

[0024] Based on the synthetic minority over-sampling technique, the feature space distance of each to-be-enhanced sample data in the to-be-enhanced sample data set is calculated to generate the nearest neighbor sample data corresponding to each to-be-enhanced sample data;

[0025] Determine a candidate adjacent sample data set from the nearest adjacent sample data corresponding to each of the sample data to be enhanced by using a random selection mechanism;

[0026] Generate an initial synthetic data set based on the candidate neighboring sample data set;

[0027] A quality check is performed on the initial synthetic data set to determine the synthetic sample data set.

[0028] In some embodiments, inputting the training sample dataset into a random forest classifier for model training to obtain a polymetallic nodule distribution prediction model includes:

[0029] Inputting the training sample data set into the random forest classifier;

[0030] According to the training sample data set, the random forest classifier is trained using a preset cross-validation method to obtain the polymetallic nodule distribution prediction model.

[0031] In some embodiments, inputting the dataset to be predicted and the classification threshold of nodule abundance to be predicted into the polymetallic nodule distribution prediction model to obtain the prediction result of polymetallic nodule distribution in the study area includes:

[0032] Inputting the to-be-predicted data set and the to-be-predicted nodule abundance classification threshold into the polymetallic nodule distribution prediction model;

[0033] Based on the nodule abundance classification threshold to be predicted, the polymetallic nodule distribution prediction model is guided to perform nodule abundance classification prediction according to the data set to be predicted, and the polymetallic nodule distribution prediction result of the study area corresponding to the nodule abundance classification threshold to be predicted is obtained.

[0034] In some embodiments, after inputting the dataset to be predicted and the nodule abundance classification threshold to be predicted into the polymetallic nodule distribution prediction model to obtain the prediction result of the polymetallic nodule distribution in the study area, the method further includes:

[0035] According to the prediction results of the polymetallic nodule distribution in the study area, a polymetallic nodule distribution prediction table is generated, and the polymetallic nodule distribution prediction table is displayed through a visual interface.

[0036] To achieve the above objectives, another aspect of the present invention provides a device for predicting the distribution of polymetallic nodules, the device comprising the following modules:

[0037] An initial data preprocessing module is used to obtain initial multi-source seabed observation data of the deep-sea research area and preprocess the initial multi-source seabed observation data to obtain target multi-source seabed observation data;

[0038] a target data analysis module configured to perform data analysis on the target multi-source seabed observation data to determine a to-be-enhanced sample data set;

[0039] a sample balance processing module configured to perform sample balance processing on the to-be-enhanced sample data set by using a synthetic minority over-sampling technique to obtain a synthetic sample data set;

[0040] a training data set construction module configured to construct a training sample data set according to the target multi-source seabed observation data and the synthetic sample data set;

[0041] a prediction model training module configured to input the training sample data set into a random forest classifier to perform model training, and obtain a polymetallic nodule distribution prediction model;

[0042] a polymetallic nodule distribution prediction module configured to input a to-be-predicted data set and a to-be-predicted nodule abundance classification threshold into the polymetallic nodule distribution prediction model to obtain a polymetallic nodule distribution prediction result of a study area.

[0043] To achieve the above object, another aspect of the embodiments of the present application provides an electronic device, which comprises a memory and a processor, the memory stores a computer program, and the processor implements the method described above when executing the computer program.

[0044] To achieve the above object, another aspect of the embodiments of the present application provides a computer readable storage medium, which stores a computer program, and the computer program is executed by a processor to implement the method described above.

[0045] The embodiments of the present application include at least the following beneficial effects: the present application provides a method, device, electronic device and storage medium for predicting the distribution of polymetallic nodules. The scheme obtains initial multi-source seabed observation data of a deep-sea study area and pre-processes the initial multi-source seabed observation data to obtain target multi-source seabed observation data; performs data analysis on the target multi-source seabed observation data to determine a sample data set to be enhanced; uses synthetic minority class oversampling technology to perform sample balancing processing on the sample data set to be enhanced to obtain a synthetic sample data set; constructs a training sample data set based on the target multi-source seabed observation data and the synthetic sample data set; inputs the training sample data set into a random forest classifier for model training to obtain a polymetallic nodule distribution prediction model; inputs the data set to be predicted and the classification threshold of the nodule abundance to be predicted into the polymetallic nodule distribution prediction model to obtain a prediction result of the polymetallic nodule distribution in the study area. The embodiment of the present application obtains multi-source seabed observation data to construct a training sample data set, which can overcome the limitations of a single data source and significantly improve the spatial coverage and information dimension of the initial data. In addition, by adopting synthetic minority class oversampling technology to perform sample balancing processing on the minority class sample data set, it solves the class imbalance problem that may exist in the polymetallic nodule distribution data, makes the number of samples of different categories more balanced, avoids the model from neglecting minority class samples during the training process, and effectively alleviates the modeling bias caused by the uneven distribution of samples with extreme nodule abundance values, thereby improving the model's recognition ability and prediction accuracy for various types of samples. In addition, predictions and decisions are made based on the patterns and rules in the learning data of the random forest classifier, which reduces the influence of subjective factors, realizes the multi-scale fusion of geophysical surface data and point sampling data, and improves the spatial resolution of the prediction results. BRIEF DESCRIPTION OF THE DRAWINGS

[0046] Figure 1 This is a flowchart of the steps of the polymetallic nodule distribution prediction method provided in an embodiment of the present application;

[0047] Figure 2 This is a schematic diagram of a sample balancing process provided by an embodiment of the present application;

[0048] Figure 3 This is a schematic diagram of the technical framework of the polymetallic nodule distribution prediction method provided in an embodiment of the present application;

[0049] Figure 4 1 is a flow chart of a method for predicting the distribution of polymetallic nodules provided in an embodiment of the present application;

[0050] Figure 5 Schematic diagram of the confusion matrix of the polymetallic nodule abundance prediction results provided in the embodiment of the present application;

[0051] Figure 6is a schematic diagram of an interface of a multi-metal nodule abundance prediction model provided by an embodiment of the present application;

[0052] Figure 7 is a schematic diagram of a multi-metal nodule distribution prediction result provided by an embodiment of the present application;

[0053] Figure 8 is a comparison schematic diagram of a multi-metal nodule distribution prediction method provided by an embodiment of the present application;

[0054] Figure 9 is a structural schematic diagram of a multi-metal nodule distribution prediction device provided by an embodiment of the present application;

[0055] Figure 10 is a hardware structural schematic diagram of an electronic device provided by an embodiment of the present application. DETAILED DESCRIPTION

[0056] In order to make the objectives, technical solutions and advantages of the present application clearer, the present application 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 only used to explain the present application and are not intended to limit the present application. When the following description refers to the accompanying drawings, the same numbers in different drawings represent the same or similar elements unless otherwise indicated. The implementation described in the following exemplary embodiments does not represent all the implementations consistent with the embodiments of the present application. They are only examples of devices and methods consistent with some aspects of the embodiments of the present application as detailed in the appended claims.

[0057] It can be understood that the terms "first", "second", and the like used in the present application can be used herein to describe various concepts, but unless specifically stated, these concepts are not limited by these terms. These terms are only used to distinguish one concept from another. For example, without departing from the scope of the embodiments of the present application, the first information can also be referred to as the second information, and similarly, the second information can also be referred to as the first information. Depending on the context, the word "if" as used herein can be interpreted as "when" or "when" or "in response to determining".

[0058] The terms "at least one", "multiple", "each", "any" and the like used in the present application include one, two or more than two, multiple includes two or more than two, each refers to each of the corresponding multiple, and any refers to any one of the multiple.

[0059] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as understood by those skilled in the art to which the present application belongs. The terms used herein are only for the purpose of describing the embodiments of the present application and are not intended to limit the present application.

[0060] Polymetallic nodule is an important mineral resource, which generally occurs in deep-sea basins with a water depth of 4000-6500 meters. The abundance of polymetallic nodules is a key parameter for evaluating the resource of polymetallic nodules, which refers to the mass of polymetallic nodules per unit area of seabed. Therefore, according to the exploration target and actual demand of polymetallic nodules, it is of great significance to determine the distribution of nodules within a certain abundance range and the continuity of the region for resource exploration. At present, the acquisition of nodule abundance information mainly relies on box sampling and seabed photography. Due to the limitation of equipment and the huge consumption of manpower, material and financial resources, only point or line sampling and imaging data can be obtained, which cannot cover the study area. The study area refers to a specific geographical area for polymetallic nodule exploration and research. The study area may be divided into smaller sub-areas or grids for systematic investigation and sampling. These sub-areas or grids can help researchers better understand the distribution, abundance and continuity of polymetallic nodules, thereby providing data support for resource assessment and mining planning.

[0061] As an example, when estimating the amount of deep-sea polymetallic nodules, the traditional method mainly relies on sparse point sampling data combined with Kriging interpolation for resource prediction, which has limitations such as low resolution, inability to effectively utilize geophysical data, and the need for a large number of samples. In recent years, due to the problems of feature selection bias and insufficient sample representativeness, the machine learning method has limited representation ability for the nonlinear relationship between input features and target variables, and does not consider the data imbalance problem proposed in the embodiments of the present application.

[0062] In view of this, the embodiment of the present application provides a multi-metal nodule distribution prediction method and device, an electronic device and a storage medium. The method comprises the following steps: obtaining initial multi-source seabed observation data of a deep sea research area, and preprocessing the initial multi-source seabed observation data to obtain target multi-source seabed observation data; performing data analysis on the target multi-source seabed observation data to determine a to-be-enhanced sample data set; performing sample balancing processing on the to-be-enhanced sample data set by using a synthetic minority over-sampling technique to obtain a synthetic sample data set; constructing a training sample data set according to the target multi-source seabed observation data and the synthetic sample data set; inputting the training sample data set into a random forest classifier for model training to obtain a multi-metal nodule distribution prediction model; and inputting a to-be-predicted data set and a to-be-predicted nodule abundance classification threshold into the multi-metal nodule distribution prediction model to obtain a multi-metal nodule distribution prediction result of the research area. The embodiment of the present application can overcome the limitation of a single data source, significantly improve the spatial coverage and information dimension of the initial data, and solve the class imbalance problem that may exist in the multi-metal nodule distribution data by performing sample balancing processing on the minority class sample data set by using the synthetic minority over-sampling technique, so that the number of samples of different classes is more balanced, the model ignores the minority class samples in the training process is avoided, the modeling deviation caused by the uneven distribution of nodule abundance extreme value samples is effectively alleviated, and the recognition ability and prediction accuracy of the model for various samples are improved. In addition, the random forest classifier is used to learn the patterns and rules in the data for prediction and decision-making, the influence of subjective factors is reduced, the multi-scale fusion of geophysical surface data and point sampling data is realized, and the spatial resolution of the prediction result is improved.

[0063] The multi-metal nodule distribution prediction method provided by the embodiment of the present application relates to the technical field of deep sea resource prediction. The multi-metal nodule distribution prediction method provided by the embodiment of the present application can be applied to a terminal, a server, or software running in a terminal or a server. In some embodiments, the terminal can be a smart phone, a tablet computer, a notebook computer, a desktop computer, a smart speaker, a smart watch, a vehicle-mounted terminal, or the like, but is not limited thereto. The server can be configured as a standalone physical server, a server cluster composed of multiple physical servers, or a distributed system, and can also be configured as a cloud server providing basic cloud computing services such as cloud service, cloud database, cloud computing, cloud function, cloud storage, network service, cloud communication, middleware service, domain name service, security service, CDN (Content Delivery Network), and big data and artificial intelligence platform. The server can also be a node server in a blockchain network. The software can be an application that implements the multi-metal nodule distribution prediction method, but is not limited to the above forms.

[0064] The application is operable in a multitude of generic or specific computer system environments or configurations. For example: personal computer, server computer, handheld or portable device, tablet device, multiprocessor systems, microprocessor-based systems, set top boxes, programmable consumer electronics, network PC (Personal Computer), minicomputers, mainframe computers, distributed computing environments that include any of the above systems or devices, and the like. The application can be described in the general context of computer-executable instructions, such as program modules, being executed by a computer. Generally, program modules include routines, programs, objects, components, data structures, and the like, that perform particular tasks or implement particular abstract data types. The application can also be practiced in distributed computing environments where tasks are performed by remote processing devices that are linked through a communications network. In a distributed computing environment, program modules can be located in local and remote computer storage media including memory storage devices.

[0065] Referring to Figure 1 , Figure 1 is an optional step flowchart of the method for predicting the distribution of multi-metal nodules provided by the embodiments of the application, Figure 1 The method in the step S101 to the step S106 can include but is not limited to the steps.

[0066] In step S101, initial multi-source seabed observation data of a deep-sea research area is acquired, and the initial multi-source seabed observation data is preprocessed to obtain target multi-source seabed observation data.

[0067] In some embodiments, the step of preprocessing the initial multi-source seabed observation data to obtain the target multi-source seabed observation data can include: performing Mercator projection processing on water depth data in the initial seabed topography data to obtain a first plane rectangular coordinate system; performing Mercator projection processing on the initial echo intensity data to obtain a second plane rectangular coordinate system; performing gridding processing on the first plane rectangular coordinate system to obtain first grid data; performing gridding processing on the second plane rectangular coordinate system to obtain second grid data; calculating topography slope data and topography roughness data according to the water depth data in the first grid data; constructing candidate multi-source seabed observation data according to the water depth data in the first grid data, the second grid data, the topography slope data, the topography roughness data, and sample position information; and performing data cleaning processing on the candidate multi-source seabed observation data to obtain the target multi-source seabed observation data.

[0068] The initial multi-source seabed observation data refers to observation data that has been subjected to equipment calibration and basic format standardization processing, and the initial multi-source seabed observation data is the basic training data for training the model. The basic training data is composed of measured station polymetallic nodules abundance data at sample positions of nodule box sampling stations, seabed topographic data, echo intensity data and the like. The seabed topographic data and echo intensity data at sample positions are extracted from the multi-source seabed observation data of the study area according to the coordinates. In subsequent model training and prediction, the range of polymetallic nodules abundance to be identified can be divided into two categories (above a certain abundance value / below a certain abundance value) according to the obtained measured station polymetallic nodules abundance data.

[0069] It can be understood that the initial multi-source seabed observation data can include but is not limited to the following data: initial seabed topographic data, initial echo intensity data, measured station polymetallic nodules abundance data and sample position information. However, the initial multi-source seabed observation data must include the three data of initial seabed topographic data, initial echo intensity data and measured station polymetallic nodules abundance data. It should be noted that, since the measured station polymetallic nodules abundance data is used as label data of the training sample to guide the model to learn the mapping relationship between the feature data and the output, so that the model can learn the real data distribution, the measured station polymetallic nodules abundance data does not participate in the subsequent preprocessing and sample enhancement process, but the measured station polymetallic nodules abundance data needs to exist in the training sample data set as label data of the training sample for training the model.

[0070] In a specific implementation, first, the seabed topographic data and echo intensity data in the study area (such as a certain area of the ocean) are subjected to Mercator projection in Arcgis to convert the seabed topographic data and echo intensity data to a plane rectangular coordinate system; then, the natural neighborhood method is used to uniformly interpolate the data on the plane rectangular coordinate system to form regular grid data (such as 500m x 500m); next, the slope and terrain roughness are calculated according to the water depth data in the seabed topographic data; finally, all the preprocessed data are subjected to data cleaning processing (including but not limited to operations such as removing outliers and completing missing values), to obtain the target multi-source seabed observation data. Arcgis refers to Geographic Information System (GIS) software.

[0071] Exemplarily, according to the longitude and latitude of the existing sampling station (sample position information), the topographic features (water depth, slope, roughness) of the grid unit where each station is located, the echo intensity value and the position information are extracted, and combined with the measured nodule abundance data of the station to form a training sample set. Exemplarily, assuming that there are 40 stations, the water depth, slope, roughness and echo intensity data at each station can be extracted according to the positions of the existing 40 stations, and the longitude and latitude information corresponding to the water depth, slope, roughness, echo intensity and station position are used as model input variables, and the measured multi-metal nodule abundance data of the station is used as the prediction output variable of the model to form a required data set (i.e. training label data).

[0072] In step S102, data analysis is performed on the target multi-source seabed observation data to determine a to-be-enhanced sample data set.

[0073] In some embodiments, step S102 can include: performing sample classification processing on the target multi-source seabed observation data according to a preset test nodule abundance classification threshold to obtain a positive sample data set and a negative sample data set; and determining the to-be-enhanced sample data set according to the number of positive samples corresponding to the positive sample data set and the number of negative samples corresponding to the negative sample data set.

[0074] In a specific implementation, assuming that the multi-metal nodule abundance boundary value is 15 kg / m 2 , the sample data greater than the boundary value is the nodule abundance that needs to be identified, i.e. the abnormal value, and the sample data less than the boundary value is the background value, which needs to identify the abnormal value from the background value, and the two types of samples greater than and less than the multi-metal nodule boundary value are distinguished.

[0075] Exemplarily, assuming that the multi-metal nodule abundance classification threshold is 8 kg / m 2 , samples higher than the multi-metal nodule abundance classification threshold (8 kg / m2) are divided into positive samples, and a positive sample data set is constructed based on the positive samples, and similarly, samples lower than the multi-metal nodule abundance classification threshold (8 kg / m2) are divided into negative samples, and a negative sample data set is constructed based on the negative samples. The to-be-enhanced sample data set is the data set with fewer samples in the positive sample data set and the negative sample data set. It can be understood that if the number of positive samples corresponding to the positive sample data set is less than the number of negative samples corresponding to the negative sample data set, the data set with fewer samples is the positive sample data set (i.e. the to-be-enhanced sample data set), and similarly, if the number of positive samples corresponding to the positive sample data set is greater than the number of negative samples corresponding to the negative sample data set, the data set with fewer samples is the negative sample data set (i.e. the to-be-enhanced sample data set).

[0076] In step S103, a synthetic minority over-sampling technique is used to perform sample balancing processing on the to-be-enhanced sample data set to obtain a synthetic sample data set.

[0077] Referring to Figure 2 , Figure 2 is a sample balancing process schematic diagram provided by an embodiment of the present application, as Figure 2 shown, in some embodiments, step S103 can include but is not limited to steps S201 to S204:

[0078] Step S201, based on the synthetic minority class oversampling technique, calculating the feature space distance of each to-be-enhanced sample data in the to-be-enhanced sample data set, and generating the nearest neighbor sample data corresponding to each to-be-enhanced sample data;

[0079] Step S202, using a random selection mechanism to determine a candidate nearest neighbor sample data set from the nearest neighbor sample data corresponding to each to-be-enhanced sample data;

[0080] Step S203, generating an initial synthetic data set according to the candidate nearest neighbor sample data set;

[0081] Step S204, quality detection is performed on the initial synthetic data set to determine the synthetic sample data set.

[0082] In a specific implementation, assuming that the multi-metal nodule abundance classification threshold is 8 kg / m 2 , samples higher than the multi-metal nodule abundance classification threshold (8 kg / m2) are divided into positive samples, and samples lower than the multi-metal nodule abundance classification threshold (8 kg / m2) are divided into negative samples, and then the SMOTE algorithm is used for sample balancing. Specifically, the Synthetic Minority Over-sampling Technique (SMOTE) is used to calculate N (such as 5) nearest neighbors of the minority class sample (such as the number of positive samples is less than the number of negative samples, then the minority class sample is the positive sample), and a new synthetic sample is generated by randomly selecting a neighbor from the N nearest neighbors, and then the generated synthetic sample is quality detected to ensure that the feature value of the new synthetic sample is within a reasonable range. By performing data enhancement on the minority class sample data higher / lower than the threshold, the bias caused by data imbalance can be reduced.

[0083] Step S104, constructing a training sample data set according to the target multi-source seabed observation data and the synthetic sample data set;

[0084] In a specific implementation, the target multi-source seabed observation data obtained after preprocessing and the synthetic sample data set are fused to obtain a training sample data set. It should be noted that the training sample data set also contains the measured station multi-metal nodule abundance data as label data for model training.

[0085] Step S105, input the training sample data set into the random forest classifier for model training to obtain a polymetallic nodule distribution prediction model;

[0086] In some embodiments, step S105 can include: inputting the training sample data set into the random forest classifier; training the random forest classifier according to the training sample data set using a preset cross-validation method to obtain the polymetallic nodule distribution prediction model.

[0087] In a specific implementation, the parameters set for constructing the random forest classifier include: using the total sample data composed of the original samples and the synthetic samples as the training data set for model training, and using K-fold cross-validation combined with grid search for parameter optimization. Specifically, first, randomly sample the training data set to form K groups of data sets, and the number of sample data in each fold is basically consistent; then define the range of random forest super parameters to be optimized: such as 100-200 trees for the number of decision trees; the maximum depth is not more than 12 layers; and the minimum number of nodes is 5. For each parameter combination, in turn, take the K-th fold as the validation set, and the remaining K-1 folds as the training set (k = 1, 2,..., K), calculate the accuracy rate of each fold validation set, select the parameter combination with the highest accuracy rate as the final model parameter, and obtain the polymetallic nodule distribution prediction model suitable for the region.

[0088] Step S106, input the to-be-predicted data set and the to-be-predicted nodule abundance classification threshold into the polymetallic nodule distribution prediction model to obtain a polymetallic nodule distribution prediction result of the study area.

[0089] In some embodiments, step S106 can include: inputting the to-be-predicted data set and the to-be-predicted nodule abundance classification threshold into the polymetallic nodule distribution prediction model; based on the to-be-predicted nodule abundance classification threshold, guiding the polymetallic nodule distribution prediction model to perform nodule abundance classification prediction according to the to-be-predicted data set, and obtaining a polymetallic nodule distribution prediction result of the study area corresponding to the to-be-predicted nodule abundance classification threshold.

[0090] The to-be-predicted data set refers to seabed topography data, echo intensity data, etc., but does not include measured station polymetallic nodule abundance data. Specifically, the to-be-predicted data set refers to the variables such as backscattering intensity, water depth, slope, vertical roughness metric, and geographical coordinates of the sample point extracted from the seabed topography data, echo intensity data, etc.

[0091] Optionally, the output result of the polymetallic nodule distribution prediction model is a binary classification (higher than the nodule abundance threshold \ lower than the nodule abundance threshold) result, and the binary classification result is output by setting an interactive threshold adjustment module in the polymetallic nodule distribution prediction model, which aims to meet the development economic needs of different study areas.

[0092] In some embodiments, after step S106, it can further include: generating a polymetallic nodule distribution prediction table according to the prediction result of the research area polymetallic nodule distribution, and displaying the polymetallic nodule distribution prediction table through a visual interface.

[0093] In a specific implementation, the related data of the research area is input into the prediction model to obtain the polymetallic nodule distribution of the area. Specifically, first, by inputting all variables (i.e. the to-be-predicted data set: backscatter intensity, water depth, slope, vertical roughness metric, and geographical coordinates of the sample point) of the research area into the input area of the trained and optimized polymetallic nodule distribution prediction model, the prediction result of each grid cell can be obtained; then a table data containing the prediction result is generated, which includes grid coordinate information, prediction category (enrichment area / non-enrichment area) and prediction probability value; finally, the polymetallic nodule distribution of the area can be obtained through visual processing, thereby providing certain guidance for further understanding of the nodule distribution area and continuity of the area.

[0094] The steps S101 to S106 shown in the embodiments of the present application obtain the target multi-source seabed observation data by acquiring the initial multi-source seabed observation data of the deep sea research area and pre-processing the initial multi-source seabed observation data; determine the to-be-enhanced sample data set by performing data analysis on the target multi-source seabed observation data; obtain the synthesized sample data set by performing sample balancing processing on the to-be-enhanced sample data set using the synthetic minority over-sampling technique; construct the training sample data set according to the target multi-source seabed observation data and the synthesized sample data set; input the training sample data set into the random forest classifier for model training to obtain the polymetallic nodule distribution prediction model; and input the to-be-predicted data set and the to-be-predicted nodule abundance classification threshold into the polymetallic nodule distribution prediction model to obtain the prediction result of the research area polymetallic nodule distribution. The embodiments of the present application can overcome the limitations of a single data source by constructing the training sample data set by acquiring the multi-source seabed observation data, significantly improve the spatial coverage and information dimension of the initial data, and solve the class imbalance problem that may exist in the polymetallic nodule distribution data by performing sample balancing processing on the minority class sample data set using the synthetic minority over-sampling technique, so that the number of different class samples is more balanced, avoiding the neglect of minority class samples in the model training process, effectively alleviating the modeling bias caused by the uneven distribution of nodule abundance extreme value samples, thereby improving the recognition ability and prediction accuracy of the model for each type of sample. In addition, the random forest classifier is used to learn the patterns and rules in the data for prediction and decision-making, reducing the influence of subjective factors, realizing the multi-scale fusion of geophysical surface data and point sampling data, and improving the spatial resolution of the prediction result.

[0095] In order to explain the principle of the technical scheme of the present application in detail, the overall process of the present application will be described below in combination with some specific embodiments. It is easy to understand that the following is an explanation of the technical principle of the present application and cannot be regarded as a limitation of the present application.

[0096] Please refer to Figure 3 , Figure 3 is a technical framework diagram of the multi-metal nodule distribution prediction method provided by the embodiments of the present application; as shown in Figure 3 , the technical framework of the multi-metal nodule distribution prediction method provided by the embodiments of the present application mainly includes four parts, which are data preparation part, sample processing part, model training part and prediction application part, and the specific process is to prepare multi-source data first, then process the samples of the multi-source data, then train the model using the training sample data set, and finally use the trained model to predict the distribution of multi-metal nodules. For the specific implementation process of the four parts, please refer to Figure 4 , Figure 4 is a flowchart of the multi-metal nodule distribution prediction method provided by the embodiments of the present application, as shown in Figure 4 , the specific implementation process of the multi-metal nodule distribution prediction method includes the following four steps:

[0097] First, data preparation.

[0098] In the specific implementation, multi-source data is first obtained, and then the multi-source data is preprocessed to determine the characteristic variables (seafloor topographic data, echo intensity data, etc.) and the target variables (measured station multi-metal nodule abundance data), and then the target multi-source seafloor observation data is constructed. The specific implementation process includes the following two steps: (1) collecting the water depth value (seafloor topographic data) from the multi-beam measurement system and the backscattering intensity value (echo intensity data) from the acoustic detection equipment, and the actual measurement value (measured station multi-metal nodule abundance data) obtained by seafloor sampling as label data of the training sample. (2) Uniformly process the spatial data. Specifically, first, the seafloor topographic data and the echo intensity data in the study area (such as a certain area of the ocean) are projected in Arcgis to convert the seafloor topographic data and the echo intensity data to a plane rectangular coordinate system; then, the natural neighborhood method is used to uniformly interpolate the data on the plane rectangular coordinate system to form regular grid data (such as 500m x 500m); then, the slope and topographic roughness are calculated according to the water depth data in the seafloor topographic data; finally, all the preprocessed data are cleaned (including but not limited to operations such as removing outliers and filling missing values), to obtain the target multi-source seafloor observation data. Among them, the basic investigation research such as the water depth echo information obtained by the multi-beam can cover a wide range of areas, so it is very important to fully exploit the existing sample data information and establish the relationship between the terrain attributes and the multi-metal nodule abundance.

[0099] Specifically, for each seabed sampling station, the topographic features (water depth, slope, roughness) of the grid cell where each station is located, the echo intensity value and the position information are extracted according to the latitude and longitude of the existing sampling station, and combined with the measured nodule abundance data of the station to form a training sample set. For example, assuming that there are 40 existing station positions, the water depth, slope, roughness and echo intensity data at each station can be extracted according to the 40 existing station positions. And the water depth, slope, roughness, echo intensity and the latitude and longitude information corresponding to the station position are used as the model input variables, and the station abundance data is used as the prediction output variable of the model to form a required data set.

[0100] It can be understood that the training data contains echo intensity data, seabed topographic data and station abundance data at stations in the study area. For example, if there are 50 stations, there are 50 data, each data includes echo intensity data, seabed topographic data, station abundance data and latitude and longitude information (position information). And the final prediction is the distribution of polymetallic nodules in the area that has not been sampled.

[0101] Second step, sample processing.

[0102] In a specific implementation, assuming that the polymetallic nodule abundance classification threshold is 8 kg / m2, samples higher than the polymetallic nodule abundance classification threshold (8 kg / m2) are divided into positive samples, and samples lower than the polymetallic nodule abundance classification threshold (8 kg / m2) are divided into negative samples, and then the SMOTE algorithm is used for sample balancing. Specifically, the SMOTE algorithm is used to calculate N nearest neighbors of the minority class samples, and a new synthetic sample is generated by randomly selecting a neighboring sample, and then the quality of the generated synthetic sample is detected to ensure that the feature value of the new synthetic sample is within a reasonable range. By data enhancement on the minority class sample data higher / lower than the threshold, the bias caused by data imbalance can be reduced.

[0103] It can be understood that, assuming that the polymetallic nodule abundance boundary value is determined to be 15 kg / m 2 , the sample data greater than the boundary value is the nodule abundance that needs to be identified, i.e. the outlier, and the sample data less than the boundary value is the background value, which needs to identify the outlier from the background value. Specifically, according to the size of the amount of data of the two categories, the synthetic minority over-sampling method is used to increase the amount of data of the kind of sample with less data, to form a data set for machine learning training and verification. Among them, the total sample data composed of the preprocessed original sample and the synthetic sample is used as the training data set for model training.

[0104] Third step, model training.

[0105] In a specific implementation, a prediction model is constructed and optimized using a random forest classification algorithm and a training data set. Specifically, a random forest classification method is used to train and validate the model using water depth, echo intensity, etc. as input variables and whether it is greater than the boundary value as the prediction variable. Among them, the parameters set when constructing the random forest classifier include: using the total sample data composed of original samples and synthetic samples as the training data set for model training, using K-fold cross-validation combined with grid search for parameter optimization. Specifically, first, randomly sample the training data set to form K data sets, and the number of sample data in each fold is basically consistent; then define the range of random forest super parameters to be optimized: such as 100-200 trees for decision tree number; the maximum depth does not exceed 12 layers; the minimum number of nodes is 5. For each parameter combination, in turn, the K-th fold is used as the validation set, and the remaining K-1 folds are used as the training set (k=1, 2,..., K), the accuracy of each fold validation set is calculated, and the parameter combination with the highest accuracy is selected as the final model parameter, obtaining a multi-metal nodule distribution prediction model suitable for the region.

[0106] Please refer to Figure 5 , Figure 5 is a confusion matrix diagram of the multi-metal nodule abundance prediction result provided by the embodiment of the application, as Figure 5 shown, the prediction result of 5-fold cross-validation in the embodiment of the application is shown in the confusion matrix, which is the most intuitive evaluation tool in the classification problem, and the model performance is shown through four key indicators, as Figure 5 shown, the following data can be obtained: ① true positive (TP): the number of "enriched areas" that are actually predicted correctly; ② false positive (FP): the number of "non-enriched areas" that are misjudged as enriched areas (false positives); ③ false negative (FN): the number of true enriched areas that are missed (false negatives); ④ true negative (TN): the number of non-enriched areas that are correctly identified. Therefore, from Figure 5 it can be seen that the recognition accuracy of the enriched area (i.e. greater than the threshold value) in the embodiment of the application reaches 82%, which indicates that the multi-metal nodule distribution prediction method provided by the embodiment of the application is basically suitable for nodule distribution prediction in the area.

[0107] Step 4, prediction application.

[0108] In a specific implementation, the relevant data of the study area is input into the prediction model to obtain the multi-metal nodule distribution in the region. Specifically, first, input all the variables of the study area into the input area of the multi-metal nodule distribution prediction model that has been trained and optimized to obtain the prediction result of each grid cell, then arrange the table data containing the prediction result, the table data includes grid coordinate information, prediction category (enriched area / non-enriched area) and prediction probability value, and finally the multi-metal nodule distribution in the area can be obtained through visualization processing.

[0109] See also Figure 6 , Figure 6 This is a schematic diagram of the interface of the polymetallic nodule abundance prediction model provided in the embodiment of the present application. Figure 6 As shown in the figure, a visual interface is provided. BS is the backscatter intensity, Depth is the water depth, Slope is the slope, VRM is the vertical roughness metric, Longitude is the geographic coordinate (longitude) of the sample point, and Latitude is the geographic coordinate (latitude) of the sample point. When the model is applied, these variables (i.e., the dataset to be predicted: backscatter intensity, water depth, slope, vertical roughness metric, and geographic coordinates of the sample point) need to be input to make predictions. In addition, different results can be obtained by inputting different abundance thresholds, such as Figure 6 As shown in the figure, assuming that the input abundance threshold is 10, the output result is greater than 10kg / m2 or less than 10kg / m 2 The data identified is that the abundance is 10kg / m 2 The embodiment of the present application proposes a dynamic threshold response mechanism, which realizes the adjustable output of exploration demand-oriented prediction results (threshold is continuously adjustable) and supports adaptive adjustment of exploration demand. It is worth mentioning that the visualization interface corresponding to the polymetallic nodule abundance prediction model provided by the embodiment of the present application can not only support single data prediction, but also support CSV (Comma-Separated Values, comma separated value file) batch prediction, where Figure 6 The "Predict Single Data" button allows users to manually input various parameters of a single sample (such as BS, Depth, Slope, VRM, Longitude, Latitude, and abundance threshold). After the user clicks the "Predict Single Data" button, the system will predict the polymetallic nodule abundance prediction result of the single sample based on these input parameters; Figure 6 The "Upload CSV Forecast" button means that users can upload a CSV file containing multiple rows of data (each row in the file represents a piece of data to be predicted, and the columns contain the required fields of BS, Depth, Slope, VRM, Longitude, Latitude, and abundance threshold). After the user clicks the "Upload CSV Forecast" button, the system will read the data in the CSV file, predict each item one by one, and generate the polymetallic nodule abundance prediction results.

[0110] See also Figure 7 , Figure 7 Schematic diagram of the prediction results of polymetallic nodule distribution provided in the embodiment of the present application. Figure 7 To obtain the distribution of polymetallic nodules in the study area through visualization.

[0111] See also Figure 8 ,Figure 8 is a comparison diagram of the multi-metal nodule distribution prediction method provided by the embodiment of the application, as shown in Figure 8 The key advantages of the multi-metal nodule distribution prediction method provided by the embodiment of the application are as follows: (1) the data enhancement algorithm is used to balance the deep-sea multi-metal nodule data, and the problems of subjective sampling and sample scarcity are solved; (2) the dynamic threshold decision mechanism is used to adjust the binary output design of the threshold, and the adaptive adjustment of the exploration demand is supported.

[0112] In the embodiment of the application, a multi-metal nodule distribution prediction method based on random forest and data enhancement is proposed. The patterns and rules in the data are learned by the random forest model to make prediction and decision, which reduces the influence of subjective factors, realizes the multi-scale fusion of geophysical surface data and point sampling data, and improves the spatial resolution of the prediction result. At the same time, the data enhancement algorithm (synthetic minority over-sampling technique, SMOTE) is combined to solve the problems of subjective sampling and sample scarcity, effectively alleviate the modeling bias caused by the uneven distribution of nodule abundance extreme value samples. In addition, the dynamic threshold response mechanism is proposed in the embodiment of the application, which realizes the adjustable output (continuous adjustable threshold) of the prediction result guided by the exploration demand, that is, the binary output design of the adjustable threshold, which supports the adaptive adjustment of the exploration demand. The multi-metal nodule distribution prediction method provided by the embodiment of the application can significantly optimize the resource calculation efficiency, support efficient regional prediction, and be suitable for real-time decision support systems for deep-sea mineral resource exploration and dynamic evaluation of multi-metal nodule resources and other application scenarios. By establishing a surface constraint transfer learning framework, the model knowledge obtained by training in the source area is transferred to the target area, so that the sample demand of the new area prediction is reduced.

[0113] It should be noted that the embodiment only briefly and schematically describes the general process of the multi-metal nodule distribution prediction method, and the detailed description of each step can refer to the related content in the foregoing embodiments, which will not be repeated here. It can be understood that the present application does not limit this.

[0114] The embodiment of the application obtains initial multi-source seafloor observation data of a deep sea research area, and pre-processes the initial multi-source seafloor observation data to obtain target multi-source seafloor observation data; performs data analysis on the target multi-source seafloor observation data to determine a to-be-enhanced sample data set; performs sample balancing processing on the to-be-enhanced sample data set by using a synthetic minority over-sampling technique to obtain a synthetic sample data set; constructs a training sample data set according to the target multi-source seafloor observation data and the synthetic sample data set; inputs the training sample data set into a random forest classifier to perform model training to obtain a polymetallic nodule distribution prediction model; and inputs a to-be-predicted data set and a to-be-predicted nodule abundance classification threshold into the polymetallic nodule distribution prediction model to obtain a polymetallic nodule distribution prediction result. The embodiment of the application can overcome the limitation of a single data source by obtaining multi-source seafloor observation data to construct a training sample data set, significantly improve the spatial coverage and information dimension of initial data, and solve the class imbalance problem that may exist in polymetallic nodule distribution data by performing sample balancing processing on a minority class sample data set by using a synthetic minority over-sampling technique, so that the number of samples of different classes is more balanced, the model ignores the minority class samples in the training process is avoided, the modeling bias caused by uneven distribution of nodule abundance extreme value samples is effectively alleviated, the recognition ability and prediction accuracy of the model for various samples are improved, and in addition, the random forest classifier is used to learn the patterns and rules in the data to make predictions and decisions, the influence of subjective factors is reduced, the multi-scale fusion of geophysical surface data and point sampling data is realized, and the spatial resolution of the prediction result is improved.

[0115] To sum up, the embodiment of the application provides a deep sea polymetallic nodule distribution prediction method based on data enhancement and random forest, which uses a synthetic minority over-sampling technique (SMOTE) and a random forest combination method based on seafloor topography, water depth, echo intensity and position information, increases nodule abundance extreme value samples, reduces prediction bias caused by data imbalance, constructs a deep sea polymetallic nodule prediction model, and predicts the distribution of polymetallic nodules in different abundance ranges. The method can improve the recognition ability of the model for polymetallic nodule enrichment areas, solve the problem of difficult deep sea polymetallic nodule exploration, realize regional prediction of deep sea polymetallic nodule distribution, and provide certain data reference for subsequent exploration. Moreover, the embodiment of the application can take into account multi-source data fusion, improve regional adaptability and engineering practicability.

[0116] For reference Figure 9 The embodiment of the application also provides a polymetallic nodule distribution prediction device 900, which can implement the above-mentioned polymetallic nodule distribution prediction method. The device includes the following modules:

[0117] The initial data preprocessing module 901 is configured to acquire initial multi-source seabed observation data of a deep-sea research area, and preprocess the initial multi-source seabed observation data to obtain target multi-source seabed observation data.

[0118] The target data analysis module 902 is configured to perform data analysis on the target multi-source seabed observation data to determine a sample data set to be enhanced.

[0119] The sample balancing processing module 903 is configured to perform sample balancing processing on the sample data set to be enhanced by using a synthetic minority over-sampling technique to obtain a synthetic sample data set.

[0120] The training data set construction module 904 is configured to construct a training sample data set according to the target multi-source seabed observation data and the synthetic sample data set.

[0121] The prediction model training module 905 is configured to input the training sample data set into a random forest classifier to perform model training, and obtain a polymetallic nodule distribution prediction model.

[0122] The polymetallic nodule distribution prediction module 906 is configured to input a data set to be predicted and a nodule abundance classification threshold to be predicted into the polymetallic nodule distribution prediction model to obtain a polymetallic nodule distribution prediction result of the research area.

[0123] It can be understood that the contents in the above method embodiments are all applicable to the device embodiments, the device embodiments specifically implement the functions of the above method embodiments, and achieve the same beneficial effects as the above method embodiments.

[0124] The embodiment of the application further provides an electronic device, which includes a memory and a processor, the memory stores a computer program, and the processor implements the above polymetallic nodule distribution prediction method when executing the computer program. The electronic device can be any intelligent terminal including a tablet computer, a vehicle-mounted computer, etc.

[0125] It can be understood that the contents in the above method embodiments are all applicable to the device embodiments, the device embodiments specifically implement the functions of the above method embodiments, and achieve the same beneficial effects as the above method embodiments.

[0126] Please refer to Figure 10 , Figure 10 The hardware structure of the electronic device of another embodiment is illustrated, and the electronic device includes:

[0127] The processor 1001 can be implemented by a general-purpose CPU (Central Processing Unit), a microprocessor, an ASIC (Application Specific Integrated Circuit), or one or more integrated circuits, and is configured to execute related programs to implement the technical solutions provided by the embodiments of the present application.

[0128] The memory 1002 can be implemented by a ROM (Read-Only Memory), a static storage device, a dynamic storage device, or a RAM (Random Access Memory), and the like. The memory 1002 can store an operating system and other application programs. When the technical solutions provided by the embodiments of the present application are implemented by software or firmware, the related program codes are stored in the memory 1002 and are called and executed by the processor 1001 to implement the multi-metal nodule distribution prediction method of the embodiments of the present application.

[0129] The input / output interface 1003 is configured to realize information input and output.

[0130] The communication interface 1004 is configured to realize the communication interaction between the device and other devices. The communication can be realized by a wired manner (for example, a USB, a network cable, or the like) or a wireless manner (for example, a mobile network, WIFI, Bluetooth, or the like).

[0131] The bus 1005 is configured to transmit information between various components (for example, the processor 1001, the memory 1002, the input / output interface 1003, and the communication interface 1004) of the device.

[0132] The processor 1001, the memory 1002, the input / output interface 1003, and the communication interface 1004 are connected to each other through the bus 1005 to realize the communication connection between the devices.

[0133] The embodiments of the present application further provide a computer readable storage medium, which stores a computer program. The computer program is executed by a processor to implement the multi-metal nodule distribution prediction method.

[0134] It can be understood that the contents of the above method embodiments are applicable to the present storage medium embodiments. The present storage medium embodiments specifically implement the functions of the above method embodiments, and achieve the same beneficial effects as the above method embodiments.

[0135] Memory, as a non-transitory computer-readable storage medium, can be used to store non-transitory software programs and non-transitory computer-executable programs. In addition, the memory can include high-speed random access memory and can also include non-transitory memory, such as at least one magnetic disk storage device, a flash memory device, or other non-transitory solid-state memory device. In some embodiments, the memory can optionally include memory that is remotely disposed relative to the processor, which can be connected to the processor through a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0136] The multi-metal nodule distribution prediction method, the multi-metal nodule distribution prediction device, the electronic equipment and the storage medium provided by the embodiments of the present application obtain initial multi-source seabed observation data of a deep sea research area, and pre-process the initial multi-source seabed observation data to obtain target multi-source seabed observation data; data analysis is performed on the target multi-source seabed observation data to determine a to-be-enhanced sample data set; a synthetic minority over-sampling technique is used to perform sample balancing processing on the to-be-enhanced sample data set to obtain a synthetic sample data set; a training sample data set is constructed according to the target multi-source seabed observation data and the synthetic sample data set; the training sample data set is input into a random forest classifier for model training to obtain a multi-metal nodule distribution prediction model; and the to-be-predicted data set and a to-be-predicted nodule abundance classification threshold are input into the multi-metal nodule distribution prediction model to obtain a multi-metal nodule distribution prediction result of the research area. The embodiments of the present application can overcome the limitations of a single data source, significantly improve the spatial coverage and information dimension of the initial data by obtaining multi-source seabed observation data to construct a training sample data set, and solve the class imbalance problem that may exist in the multi-metal nodule distribution data by using the synthetic minority over-sampling technique to perform sample balancing processing on the minority class sample data set, so that the number of samples of different classes is more balanced, the neglect of the minority class samples by the model in the training process is avoided, the modeling bias caused by the uneven distribution of nodule abundance extreme value samples is effectively alleviated, the recognition ability and prediction accuracy of the model for various samples are improved, and in addition, the patterns and rules in the learning data are learned based on the random forest classifier to make predictions and decisions, the influence of subjective factors is reduced, the multi-scale fusion of geophysical surface data and point sampling data is realized, and the spatial resolution of the prediction result is improved.

[0137] The embodiments described in the embodiments of the present application are used to more clearly illustrate the technical solutions of the embodiments of the present application, and do not constitute a limitation on the technical solutions provided by the embodiments of the present application. Those skilled in the art can know that, with the evolution of technology and the appearance of new application scenarios, the technical solutions provided by the embodiments of the present application are also applicable to similar technical problems.

[0138] Those skilled in the art can understand that the technical solutions shown in the figures do not constitute a limitation to the embodiments of the present application, and can include more or fewer steps than the figures, or combine certain steps, or different steps.

[0139] The device embodiments described above are merely illustrative, wherein the units described as separate components can or can not be physically separated, that is, can be located in one place, or can be distributed to multiple network units. Part or all of the modules can be selected according to actual needs to achieve the purpose of the embodiments.

[0140] Those skilled in the art can understand that all or some steps in the above disclosed method, functional modules / units in the system and device can be implemented as software, firmware, hardware and their appropriate combinations.

[0141] The terms "first", "second", "third", "fourth" and the like (if any) in the description and the above drawings of the present application are used to distinguish similar objects, and do not necessarily indicate a specific order or sequence. It should be understood that the data thus used can be interchanged under appropriate circumstances, so that the embodiments of the present application described herein can be implemented in an order other than that illustrated or described herein. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, system, product or device including a series of steps or units does not necessarily limit to those clearly listed steps or units, but can include other steps or units not clearly listed or inherent to these processes, methods, products or devices.

[0142] It should be understood that in the present application, "at least one" means one or more, and "multiple" means two or more. "And / or" is used to describe the association between the associated objects, which means that there can be three relationships, for example, "A and / or B" can mean that there are three cases of only A, only B and A and B at the same time, where A and B can be singular or plural. The character " / " generally represents an "or" relationship between the associated objects. "At least one of the following" or similar expressions means any combination of these items, including any combination of single or multiple items. For example, at least one of a, b or c can mean a, b, c, "a and b", "a and c", "b and c", or "a and b and c", where a, b, and c can be single or multiple.

[0143] In several embodiments provided in the present application, it should be understood that the disclosed apparatus and method can be implemented by other manners. For example, the apparatus embodiments described above are merely illustrative, for example, the division of the above units is merely a logical function division, and actual implementation can have another division manner, for example, a plurality of units or components can be combined or integrated into another system, or some features can be ignored or not executed. In addition, the coupling or direct coupling or communication connection between the units or components shown or discussed can be indirect coupling or communication connection through some interfaces, apparatuses or units, and can be electrical, mechanical or other forms.

[0144] The units described above as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, i.e., they can be located in one place or distributed on a plurality of network units. Some or all of the units can be selected according to actual needs to achieve the purpose of the embodiment.

[0145] In addition, the functional units in each embodiment of the present application can be integrated in one processing unit, or each unit can be physically present separately, or two or more units can be integrated in one unit. The integrated unit can be realized in the form of hardware or in the form of a software functional unit.

[0146] If the integrated unit is realized in the form of a software functional unit and sold or used as an independent product, it can be stored in a computer readable storage medium. Based on this understanding, the technical solutions of the present application essentially or the part of the prior art that makes a contribution or the whole or part of the technical solutions can be embodied in the form of a software product, which is stored in a storage medium and includes a plurality of instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the method of each embodiment of the present application. The foregoing storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk, and various program storage media.

[0147] The preferred embodiments of the embodiments of the present application are described above with reference to the accompanying drawings, but this does not limit the scope of the rights of the embodiments of the present application. Any modifications, equivalent replacements and improvements made by those skilled in the art without departing from the scope and essence of the embodiments of the present application shall be within the scope of the rights of the embodiments of the present application.

Claims

1. A method for predicting the distribution of polymetallic nodules, characterized in that: The method comprises the following steps: Acquiring initial multi-source seabed observation data of a deep-sea research area, and preprocessing the initial multi-source seabed observation data to obtain target multi-source seabed observation data; Performing data analysis on the target multi-source seabed observation data to determine a sample data set to be enhanced; Using a synthetic minority class oversampling technique to perform sample balancing processing on the sample data set to be enhanced to obtain a synthetic sample data set; constructing a training sample dataset based on the target multi-source seabed observation data and the synthetic sample dataset; Inputting the training sample data set into a random forest classifier for model training to obtain a polymetallic nodule distribution prediction model; The data set to be predicted and the classification threshold of the nodule abundance to be predicted are input into the polymetallic nodule distribution prediction model to obtain the prediction results of the polymetallic nodule distribution in the study area.

2. The method according to claim 1, characterized in that The initial multi-source seabed observation data includes initial seabed topography data, initial echo intensity data, and sample location information. The preprocessing of the initial multi-source seabed observation data to obtain target multi-source seabed observation data includes: Performing Mercator projection processing on the water depth data in the initial seabed topography data to obtain a first plane rectangular coordinate system; Performing Mercator projection processing on the initial echo intensity data to obtain a second plane rectangular coordinate system; Performing gridding processing on the first plane rectangular coordinate system to obtain first grid data; performing gridding processing on the second plane rectangular coordinate system to obtain second grid data; Calculating terrain slope data and terrain roughness data based on the water depth data in the first grid data; constructing candidate multi-source seabed observation data based on the water depth data in the first grid data, the second grid data, the terrain slope data, the terrain roughness data, and the sample position information; The candidate multi-source seabed observation data are cleaned to obtain the target multi-source seabed observation data.

3. The method according to claim 1, characterized in that The performing data analysis on the target multi-source seabed observation data to determine the sample data set to be enhanced includes: According to a preset test nodule abundance classification threshold, sample classification processing is performed on the target multi-source seabed observation data to obtain a positive sample data set and a negative sample data set; The sample data set to be enhanced is determined according to the number of positive samples corresponding to the positive sample data set and the number of negative samples corresponding to the negative sample data set.

4. The method according to claim 1, wherein The method of using the synthetic minority class oversampling technology to perform sample balancing processing on the sample dataset to be enhanced to obtain a synthetic sample dataset includes: Based on the synthetic minority class oversampling technology, feature space distance calculation is performed on each sample data to be enhanced in the sample data set to be enhanced, and the nearest neighbor sample data corresponding to each sample data to be enhanced is generated; Determine a candidate adjacent sample data set from the nearest adjacent sample data corresponding to each of the sample data to be enhanced by using a random selection mechanism; Generate an initial synthetic data set based on the candidate neighboring sample data set; A quality check is performed on the initial synthetic data set to determine the synthetic sample data set.

5. The method according to claim 1, wherein The step of inputting the training sample data set into a random forest classifier for model training to obtain a polymetallic nodule distribution prediction model comprises: Inputting the training sample data set into the random forest classifier; According to the training sample data set, the random forest classifier is trained using a preset cross-validation method to obtain the polymetallic nodule distribution prediction model.

6. The method according to claim 1, characterized in that The step of inputting the dataset to be predicted and the classification threshold of the nodule abundance to be predicted into the polymetallic nodule distribution prediction model to obtain the prediction result of the polymetallic nodule distribution in the study area includes: Inputting the to-be-predicted data set and the to-be-predicted nodule abundance classification threshold into the polymetallic nodule distribution prediction model; Based on the nodule abundance classification threshold to be predicted, the polymetallic nodule distribution prediction model is guided to perform nodule abundance classification prediction according to the data set to be predicted, and the polymetallic nodule distribution prediction result of the study area corresponding to the nodule abundance classification threshold to be predicted is obtained.

7. The method according to claim 1, characterized in that After inputting the data set to be predicted and the nodule abundance classification threshold to be predicted into the polymetallic nodule distribution prediction model to obtain the prediction result of the polymetallic nodule distribution in the study area, the method further includes: According to the prediction results of the polymetallic nodule distribution in the study area, a polymetallic nodule distribution prediction table is generated, and the polymetallic nodule distribution prediction table is displayed through a visual interface.

8. A device for predicting the distribution of polymetallic nodules, characterized in that: The device comprises the following modules: An initial data preprocessing module is used to obtain initial multi-source seabed observation data of the deep-sea research area and preprocess the initial multi-source seabed observation data to obtain target multi-source seabed observation data; A target data analysis module is used to analyze the target multi-source seabed observation data and determine a sample data set to be enhanced; A sample balancing processing module is used to perform sample balancing processing on the sample data set to be enhanced by using a synthetic minority class oversampling technology to obtain a synthetic sample data set; A training data set construction module, configured to construct a training sample data set based on the target multi-source seabed observation data and the synthetic sample data set; A prediction model training module is used to input the training sample data set into a random forest classifier for model training to obtain a polymetallic nodule distribution prediction model; The polymetallic nodule distribution prediction module is used to input the data set to be predicted and the classification threshold of the nodule abundance to be predicted into the polymetallic nodule distribution prediction model to obtain the prediction result of the polymetallic nodule distribution in the study area.

9. An electronic device, characterized in that: The electronic device includes a memory and a processor, the memory stores a computer program, and the processor implements the method according to any one of claims 1 to 7 when executing the computer program.

10. A computer-readable storage medium storing a computer program, characterized in that: When the computer program is executed by a processor, the method according to any one of claims 1 to 7 is implemented.

Images