Dynamic monitoring method and system for marine sediment stratification collection based on the Internet of Things

Abstract

The present invention provides a method and system for dynamic monitoring of marine sediment layered collection based on the Internet of Things. The method and system acquire original marine environmental parameters collected in real time; calculate an adaptive forgetting factor based on the significance of the original marine environmental parameters indicating the transition of sediment interlayer interfaces; use the adaptive forgetting factor to perform Kalman filtering on the non-image original marine environmental parameters to obtain filtered environmental parameters; select a feature subset from a feature set containing features with preset spatial correlation with a current image area, determine partitioning features from the feature subset, and extract sediment morphological features using a random uniform forest; fuse the filtered environmental parameters and the sediment morphological features to obtain a comprehensive state feature vector; and calculate an adjusted collection depth and collection speed of a marine sediment collection device based on the comprehensive state feature vector and using a preset layered collection model, and send the adjusted collection depth and collection speed to a control system of the collection device.

Description

Technical Field

[0001] The present application relates to the field of marine sediment monitoring, and in particular to a method and system for dynamic monitoring of marine sediment stratification collection based on the Internet of Things. Background Art

[0002] Marine sediments record key information such as paleoclimate, paleoenvironment, paleoocean, and tectonic evolution throughout geological history. Accurately acquiring marine sediment samples with a clear stratified structure is crucial for a deeper understanding of Earth's evolutionary processes, assessing the potential of marine resources, and predicting future environmental changes. Traditional marine sediment collection methods, such as gravity column sampling and piston column sampling, often rely on empirical judgment during the collection process and are unable to perceive and adapt to subtle changes in sediment stratification in real time. Especially in cases where the transition between sediment layers is unclear, the strata are complex and variable, or thin interbeds exist, traditional methods are prone to loss of stratification information, sample sequence confusion, or confusion between samples from different layers, seriously affecting the accuracy and scientific value of subsequent analysis results. Although some modern collection equipment integrates some basic sensors to monitor equipment status, this information is often only used as a rough reference and lacks the perception of the real-time characteristics of the sediment during the collection process.

[0003] The rapid development of IoT, sensor technology, and artificial intelligence algorithms has provided new directions for the intelligent upgrade of ocean exploration and sampling equipment. Integrating multiple sensors at the front end of acquisition equipment and transmitting real-time data to a control center via low-power wide-area IoT (such as LoRa and NB-IoT) or underwater acoustic / optical communication technologies lays the data foundation for dynamic monitoring. However, the marine environment is complex and ever-changing, and sensor data often contains a large amount of noise and uncertainty. On the other hand, the visual morphological characteristics of sediments contain rich geological information and are an important basis for determining stratification. Machine learning algorithms such as random forests have demonstrated good performance in image feature extraction and classification, but traditional random forest algorithms are insensitive to key morphological differences, which affects the accuracy of stratification identification. Summary of the Invention

[0004] In order to solve the problems pointed out in the above background technology, this application proposes a method for dynamic monitoring of marine sediment layer collection based on the Internet of Things, including:

[0005] Step 1: obtaining raw ocean environmental parameters collected in real time by multiple sensor nodes deployed in the ocean sediment collection area and transmitted through the Internet of Things communication module;

[0006] Step 2: calculating an adaptive forgetting factor of a Kalman filter algorithm based on the significance of the interface transition between sediment layers indicated in the original ocean environmental parameters, and performing a Kalman filter on the original ocean environmental parameters of the non-image using the adaptive forgetting factor to obtain filtered environmental parameters;

[0007] Step 3: Select a feature subset from a feature set having a preset spatial correlation with the current image area, determine a partitioning feature from the feature subset, extract sediment morphological features using a random uniform forest, and determine the selected nodes according to the partitioning feature when constructing a decision tree; and fuse the filtered environmental parameters and the sediment morphological features to obtain a comprehensive state feature vector.

[0008] Step 4: Based on the comprehensive state feature vector and using a preset hierarchical acquisition model, the adjusted acquisition depth and acquisition speed of the marine sediment acquisition equipment are calculated, and the adjusted acquisition depth and acquisition speed are sent to the control system of the acquisition equipment through the Internet of Things communication module.

[0009] Optionally, the step of calculating the adaptive forgetting factor of the Kalman filter algorithm based on the significance of the original ocean environment parameter indicating the interface transition between sediment layers comprises:

[0010] receiving the original ocean environment parameters and extracting a plurality of indicators related to the interface transition between sediment layers, wherein the plurality of indicators include a pressure change rate, a depth change rate, an image texture mutation degree, and a temperature gradient;

[0011] Setting weight coefficients for the respective multiple indicators, and performing weighted fusion on the multiple indicators according to the weight coefficients to obtain an initial transition significance score;

[0012] Acquiring historical sediment stratification data and current acquisition target layer information, and dynamically adjusting the threshold of the initial transition significance score according to the historical sediment stratification data and the current acquisition target layer information;

[0013] When the initial transition significance score exceeds the dynamically adjusted threshold, the adaptive forgetting factor adjustment amplitude is set to a first preset adjustment amplitude value; otherwise, the adaptive forgetting factor adjustment amplitude is set to a second preset adjustment amplitude value, where the first preset adjustment amplitude value is greater than the second preset adjustment amplitude value;

[0014] The final adaptive forgetting factor is calculated according to the adaptive forgetting factor adjustment amplitude and a preset reference forgetting factor.

[0015] Optionally, the adopting the adaptive forgetting factor to perform Kalman filtering on the original ocean environment parameters of the non-image to obtain filtered environment parameters includes:

[0016] Initializing a state vector and a covariance matrix of a Kalman filter, wherein the state vector includes non-image ocean environment parameters to be estimated;

[0017] Based on the system dynamics model, a one-step prediction is performed on the state vector and covariance matrix at the current moment to obtain a predicted state vector and a predicted covariance matrix;

[0018] Obtain the current non-image original ocean environment parameters as observation values;

[0019] Calculating a Kalman gain using the predicted state vector, the predicted covariance matrix, the observation value, and the adaptive forgetting factor;

[0020] Combining the predicted state vector, the observed value, and the Kalman gain, updating the state vector to obtain the filtered environmental parameters at the current moment;

[0021] The covariance matrix is updated for filtering at the next moment.

[0022] Optionally, selecting a feature subset from a feature set having a preset spatial correlation with the current image region includes:

[0023] Preprocessing the image data in the original ocean environment parameters to extract underlying visual features, and preliminarily grouping the underlying visual features according to physical meaning and spatial scale to form an initial feature set;

[0024] Acquire macro features of the current image area to be processed, wherein the macro features include average depth information and color information of the area;

[0025] Determine the type of sediment morphological characteristics that need to be distinguished based on the macroscopic characteristics;

[0026] Based on the determined sediment morphological feature type, adjusting the prior probability of selecting a feature group in the initial feature set that is related to the determined sediment morphological feature type and has a strong spatial correlation with the current image area, to obtain an adjusted feature set;

[0027] A specified number of features are randomly extracted from the adjusted feature set according to the adjusted prior probability to form the feature subset.

[0028] Optionally, the random uniform forest is used to extract sediment morphological features, and the random uniform forest determines the selected nodes according to the partitioning features when constructing a decision tree, including:

[0029] Step 3.1: Initialize the random uniform forest, set the number of decision trees and the maximum depth of each tree;

[0030] Step 3.2: For each decision tree to be constructed, perform bootstrap sampling with replacement on the training samples to form a training subset for the current decision tree;

[0031] Step 3.3: Recursively construct a decision tree starting from the root node. For the current node to be split, obtain the feature subset for the samples of the image area represented by the current node; randomly select a feature from the feature subset as a partitioning feature and randomly generate a splitting point; divide the samples of the current node into child nodes based on the partitioning feature and the splitting point; if the stop splitting condition is met, mark the current node as a leaf node and assign a category label; otherwise, repeat step 3.3 for the child nodes;

[0032] Step 3.4: Combine all constructed decision trees into a random uniform forest model;

[0033] Step 3.5: Input the image data to be extracted for morphological features into the random uniform forest model, and obtain the final sediment morphological features by majority voting or average.

[0034] This application also proposes a marine sediment stratification collection dynamic monitoring system based on the Internet of Things, including:

[0035] A data acquisition unit, used to obtain raw ocean environmental parameters transmitted through an Internet of Things communication module and collected in real time by multiple sensor nodes deployed in the ocean sediment collection area;

[0036] a filtering unit, configured to calculate an adaptive forgetting factor of a Kalman filter algorithm based on the significance of the sediment layer interface transition indicated in the original ocean environmental parameters, and perform a Kalman filter on the non-image original ocean environmental parameters using the adaptive forgetting factor to obtain filtered environmental parameters;

[0037] a feature extraction unit configured to select a feature subset from a feature set having a preset spatial correlation with a current image region, determine a partitioning feature from the feature subset, extract sediment morphological features using a random uniform forest, wherein the random uniform forest determines selected nodes based on the partitioning feature when constructing a decision tree; and obtain a comprehensive state feature vector by fusing the filtered environmental parameters and the sediment morphological features;

[0038] The monitoring unit is used to calculate the adjusted collection depth and collection speed of the marine sediment collection equipment based on the comprehensive state feature vector and using a preset hierarchical collection model, and send the adjusted collection depth and collection speed to the control system of the collection equipment through the Internet of Things communication module.

[0039] Optionally, the step of calculating the adaptive forgetting factor of the Kalman filter algorithm based on the significance of the original ocean environment parameter indicating the interface transition between sediment layers comprises:

[0040] receiving the original ocean environment parameters and extracting a plurality of indicators related to the interface transition between sediment layers, wherein the plurality of indicators include a pressure change rate, a depth change rate, an image texture mutation degree, and a temperature gradient;

[0041] Setting weight coefficients for the respective multiple indicators, and performing weighted fusion on the multiple indicators according to the weight coefficients to obtain an initial transition significance score;

[0042] Acquiring historical sediment stratification data and current acquisition target layer information, and dynamically adjusting the threshold of the initial transition significance score according to the historical sediment stratification data and the current acquisition target layer information;

[0043] When the initial transition significance score exceeds the dynamically adjusted threshold, the adaptive forgetting factor adjustment amplitude is set to a first preset adjustment amplitude value; otherwise, the adaptive forgetting factor adjustment amplitude is set to a second preset adjustment amplitude value, where the first preset adjustment amplitude value is greater than the second preset adjustment amplitude value;

[0044] The final adaptive forgetting factor is calculated according to the adaptive forgetting factor adjustment amplitude and the preset reference forgetting factor.

[0045] Optionally, the adopting the adaptive forgetting factor to perform Kalman filtering on the original ocean environment parameters of the non-image to obtain filtered environment parameters includes:

[0046] Initializing a state vector and a covariance matrix of a Kalman filter, wherein the state vector includes non-image ocean environment parameters to be estimated;

[0047] Based on the system dynamics model, a one-step prediction is performed on the state vector and covariance matrix at the current moment to obtain a predicted state vector and a predicted covariance matrix;

[0048] Obtain the current non-image original ocean environment parameters as observation values;

[0049] Calculating a Kalman gain using the predicted state vector, the predicted covariance matrix, the observation value, and the adaptive forgetting factor;

[0050] Combining the predicted state vector, the observed value, and the Kalman gain, updating the state vector to obtain the filtered environmental parameters at the current moment;

[0051] The covariance matrix is updated for filtering at the next moment.

[0052] Optionally, selecting a feature subset from a feature set having a preset spatial correlation with the current image region includes:

[0053] Preprocessing the image data in the original ocean environment parameters to extract underlying visual features, and preliminarily grouping the underlying visual features according to physical meaning and spatial scale to form an initial feature set;

[0054] Acquire macro features of the current image area to be processed, wherein the macro features include average depth information and color information of the area;

[0055] Determine the type of sediment morphological characteristics that need to be distinguished based on the macroscopic characteristics;

[0056] Based on the determined sediment morphological feature type, adjusting the prior probability of selecting a feature group in the initial feature set that is related to the determined sediment morphological feature type and has a strong spatial correlation with the current image area, to obtain an adjusted feature set;

[0057] A specified number of features are randomly extracted from the adjusted feature set according to the adjusted prior probability to form the feature subset.

[0058] Optionally, the random uniform forest is used to extract sediment morphological features, and the random uniform forest determines the selected nodes according to the partitioning features when constructing a decision tree, including:

[0059] Step 3.1: Initialize the random uniform forest, set the number of decision trees and the maximum depth of each tree;

[0060] Step 3.2: For each decision tree to be constructed, perform bootstrap sampling with replacement on the training samples to form a training subset for the current decision tree;

[0061] Step 3.3: Recursively construct a decision tree starting from the root node. For the current node to be split, obtain the feature subset for the samples of the image area represented by the current node; randomly select a feature from the feature subset as a partitioning feature and randomly generate a splitting point; divide the samples of the current node into child nodes based on the partitioning feature and the splitting point; if the stop splitting condition is met, mark the current node as a leaf node and assign a category label; otherwise, repeat step 3.3 for the child nodes;

[0062] Step 3.4: Combine all constructed decision trees into a random uniform forest model;

[0063] Step 3.5: Input the image data to be extracted for morphological features into the random uniform forest model, and obtain the final sediment morphological features by majority voting or average.

[0064] Compared with the prior art, this application has the following beneficial effects:

[0065] An adaptive forgetting factor Kalman filter, sensitive to interlayer transitions, can capture changes between sediment layers, quickly adapt to the characteristics of new layers, and accurately locate stratification interfaces. Furthermore, a spatial context-guided random uniform forest algorithm can more intelligently extract key morphological features from images, providing a reliable visual basis for stratification judgment. This enables more precise and adaptive dynamic monitoring and control of the marine sediment stratification collection process, resulting in higher-quality sediment samples and more valuable data for marine scientific research. BRIEF DESCRIPTION OF THE DRAWINGS

[0066] Figure 1 This is a flow chart of Example 1;

[0067] Figure 2 is the calculation process of the comprehensive state eigenvector;

[0068] Figure 3 Flowchart for extracting sediment morphological features for random uniform forests;

[0069] Figure 4 A schematic diagram of a single decision tree. DETAILED DESCRIPTION

[0070] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the embodiments described are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without making creative efforts are within the scope of protection of this application.

[0071] The terms "first", "second" and corresponding terminology numbers in the specification and claims of this application and the above-mentioned drawings are used to distinguish similar objects, and are not necessarily used to describe a specific order or sequence. It should be understood that the terms used in this way can be interchanged under appropriate circumstances. This is merely a way of distinguishing when describing objects with the same properties in the embodiments of the present application. In addition, the terms "including" and "having" and any variations thereof are intended to cover non-exclusive inclusions, so that a process, method, system, product or apparatus that includes a series of units is not necessarily limited to those units, but may include other units that are not explicitly listed or inherent to these processes, methods, products or apparatuses.

[0072] In addition, in the description of this application, unless otherwise specified, "plurality" means two or more. The term "and / or" or the character " / " in this application is merely a description of the association relationship between related objects, indicating that three relationships can exist. For example, A and / or B, or A / B, can mean: A exists alone, A and B exist at the same time, or B exists alone.

[0073] Specific embodiment, a method for dynamic monitoring of marine sediment layer collection based on the Internet of Things, such as Figure 1 Shown, including:

[0074] Step 1: obtaining raw ocean environmental parameters collected in real time by multiple sensor nodes deployed in the ocean sediment collection area and transmitted through the Internet of Things communication module;

[0075] A pre-set IoT communication module receives raw ocean environmental parameters collected and transmitted in real time by multiple sensor nodes deployed in the marine sediment collection area. This IoT communication module utilizes various wireless or wired communication technologies, such as low-power wide area network (LPWAN) technologies like LoRa, NB-IoT, and Sigfox. The sensor nodes are integrated into the marine sediment collection equipment and include, but are not limited to, depth sensors, pressure sensors, temperature sensors, and image sensors. These sensors work together to provide real-time perception of key environmental parameters and sediment characteristics during the collection process from different dimensions.

[0076] Step 2: calculating an adaptive forgetting factor of a Kalman filter algorithm based on the significance of the interface transition between sediment layers indicated in the original ocean environmental parameters, and performing a Kalman filter on the original ocean environmental parameters of the non-image using the adaptive forgetting factor to obtain filtered environmental parameters;

[0077] Analyze received raw ocean environmental parameters, particularly those that indicate an impending or ongoing transition at the sediment interface. Determine significance based on the trend, rate of change, or presence of specific patterns in single or multiple parameters. Examples include rapid changes in pressure, stable or sudden changes in depth, abnormal temperature gradients, or macroscopic texture changes initially extracted from image data.

[0078] Based on the results of the transition salience assessment, the adaptive forgetting factor used in the Kalman filter algorithm is dynamically calculated or adjusted. The adaptive forgetting factor controls the weighting of historical data and current observations during state estimation. A properly adjusted forgetting factor allows the filter to converge to the new true state more quickly during sudden changes, such as transitions between layers. This calculation can be performed in a variety of ways. For example, when the transition salience is judged to be increasing, a pre-defined strategy is used to increase the forgetting effect, assigning greater weight to the current observation and faster forgetting of the previous state. Conversely, when the transition salience is increasing, the forgetting effect is reduced.

[0079] Using a dynamically calculated or adjusted adaptive forgetting factor, a Kalman filter is applied to non-image data in the raw ocean environmental parameters, such as time series data from depth sensors, pressure sensors, and temperature sensors. Kalman filtering can estimate the state of a dynamic system from a series of incomplete and noisy measurements. This filtering effectively removes random noise from the sensor data, improving parameter accuracy and stability, ultimately yielding filtered environmental parameters.

[0080] Step 3: Select a feature subset from a feature set having a preset spatial correlation with the current image area, determine a partitioning feature from the feature subset, extract sediment morphological features using a random uniform forest, and determine the selected nodes according to the partitioning feature when constructing a decision tree; and fuse the filtered environmental parameters and the sediment morphological features to obtain a comprehensive state feature vector.

[0081] When extracting sediment morphological features from images, feature selection considers the spatial correlation between features within the image region. Rather than simply viewing each feature in isolation, the team also focuses on spatially adjacent or co-characteristic combinations of geological structures, such as bedding, nodules, and bioturbation traces. In one embodiment, a set of spatially correlated features associated with a specific sediment morphology is predefined or learned.

[0082] Then, one or more feature subsets are selected from this feature set with preset spatial correlation, and the partitioning features used for subsequent model construction are determined from these subsets. The selection method includes but is not limited to feature importance evaluation, mutual information between features, etc. The random uniform forest algorithm is used to extract the morphological characteristics of the sediment. In the process of constructing each decision tree, when it is necessary to select a node for splitting, the selection and splitting of the node will be guided by the previously determined partitioning features. The growth of the decision tree is not completely random, but is guided by feature information related to sediment morphology and spatial correlation. Through integrated learning of the entire forest, the final output is morphological features that can characterize information such as sediment color, texture, particle size, and bedding structure. The obtained filtered (non-image) environmental parameters are fused with the extracted sediment morphological features in the data layer or feature layer. The information from different sensors that describes different aspects is integrated to form a comprehensive state feature vector that can more comprehensively and accurately reflect the current state of marine sediment collection. Fusion methods include but are not limited to simple feature splicing, weighted averaging, or fusion using neural networks or Bayesian networks. Figure 2 The calculation process of the comprehensive state feature vector is shown.

[0083] Step 4: Based on the comprehensive state feature vector and using a preset hierarchical acquisition model, the adjusted acquisition depth and acquisition speed of the marine sediment acquisition equipment are calculated, and the adjusted acquisition depth and acquisition speed are sent to the control system of the acquisition equipment through the Internet of Things communication module.

[0084] The obtained comprehensive state feature vector is used as input and combined with a preset layered collection model to calculate the required collection depth and speed for the marine sediment collection equipment. Preferably, the layered collection model is a rule-based expert system or a machine learning model trained through historical data, such as a regression model or reinforcement learning model. Based on the real-time perception of the sediment state represented by the comprehensive state feature vector, the layered collection model determines whether the collection strategy needs to be adjusted and how to adjust it to optimize the collection of samples from the target layer while avoiding inter-layer confusion. The calculated adjusted collection depth and speed parameters are sent back to the control system of the marine sediment collection equipment in real time via the Internet of Things communication module. After receiving these parameters, the control system of the collection equipment will adjust the operation of actuators such as the propulsion system accordingly, thereby achieving closed-loop dynamic control of the sediment layered collection process. For example, when it is determined that a new sediment layer is about to be reached, the collection speed will be slowed down and the collection depth will be fine-tuned to ensure accurate sampling.

[0085] In an optional embodiment, the adaptive forgetting factor of the Kalman filter algorithm is calculated based on the significance of the original ocean environment parameters indicating the interface transition between sediment layers, including:

[0086] receiving the original ocean environment parameters and extracting a plurality of indicators related to the interface transition between sediment layers, wherein the plurality of indicators include a pressure change rate, a depth change rate, an image texture mutation degree, and a temperature gradient;

[0087] Setting weight coefficients for the respective multiple indicators, and performing weighted fusion on the multiple indicators according to the weight coefficients to obtain an initial transition significance score;

[0088] Acquiring historical sediment stratification data and current acquisition target layer information, and dynamically adjusting the threshold of the initial transition significance score according to the historical sediment stratification data and the current acquisition target layer information;

[0089] When the initial transition significance score exceeds the dynamically adjusted threshold, the adaptive forgetting factor adjustment amplitude is set to a first preset adjustment amplitude value; otherwise, the adaptive forgetting factor adjustment amplitude is set to a second preset adjustment amplitude value, where the first preset adjustment amplitude value is greater than the second preset adjustment amplitude value;

[0090] The final adaptive forgetting factor is calculated according to the adaptive forgetting factor adjustment amplitude and a preset reference forgetting factor.

[0091] Specifically, the image data of the acquired raw ocean environment parameters is preprocessed, including image correction, region of interest (ROI) extraction, image enhancement, etc. Low-level visual features are extracted from the preprocessed image, including but not limited to Gabor features, local binary patterns, gray-level co-occurrence matrix features, color histograms, etc. In one embodiment, shape features are also included, such as the shape of the particle after the particle outline is extracted.

[0092] The extracted underlying visual features are preliminarily grouped according to the geological or morphological meanings they describe, such as the Gabor feature group related to bedding direction, the LBP or GLCM feature group related to grain roughness, and the color feature group related to color change, and the spatial scale to form a structured initial feature set.

[0093] In the construction process of each decision tree in the random uniform forest, when an internal node is selected , this node represents a part of the training sample, that is, a part of the image area or pixel, when splitting:

[0094] Calculate the macro-context features of the sample represented by the node, including the average depth value of the node sample area obtained by associating the depth sensor data synchronized with the image, the average color obtained by clustering the pixel colors in the area or calculating the average RGB value, the operating status of the current acquisition device, etc. If the average color of the node's sample area is dark and the average depth value is large, the surface is currently in dark muddy sediments. Based on the macro-context features and combined with the preset geological knowledge rule base or auxiliary classifier, dynamically determine what type of sediment morphological features are most likely to be further distinguished and identified in the sample of the current node. For example, based on the above example context, determine whether there is fine stratification or whether there are signs of bioturbation, rather than macroscopic color differences.

[0095] Based on the sediment morphological feature type that is judged to be most needed to be distinguished at the current node, such as fine bedding, the prior probability of each feature or feature group in the initial feature set being selected into the candidate feature subset of the current node is adjusted. For example, if it is judged that fine bedding needs to be distinguished, the basic probability of the Gabor feature group that describes horizontal or near-horizontal textures or the LBP feature group that can reflect subtle grayscale changes in the initial feature set being selected will be temporarily increased. For example, the sampling weights of these related features are changed from Increase to , where c>1 is an amplification factor, and the weights of other features are reduced accordingly to ensure that the total probability sum is 1. According to the prior probability distribution dynamically adjusted by context information, from the total From the image features, m features are randomly extracted with weights. The larger the weight, the greater the probability of being selected, forming a candidate feature subset for the current node split , where m is much smaller than , preferably From the constructed candidate feature subset with context-guided bias In the standard way of random uniform forest algorithm, the final method for the current node is determined Division characteristics of the split and split point threshold In one embodiment, from Randomly select a feature from , then Randomly generate a split point within its value range ,Right now and . Based on the selected partition features and split points, the samples of the current node are divided into its left and right child nodes.

[0096] Repeat context analysis, guided feature subset construction, and node splitting for the newly generated child nodes until the conditions for the decision tree to stop growing are met, for example, the node reaches the preset maximum depth, the number of samples in the node is too small to continue splitting, etc.

[0097] The above steps are repeated multiple times, and each tree is independently constructed based on a different training subset obtained by bootstrapping with replacement the original training samples of labeled sediment image areas and their corresponding morphological categories. All these independently constructed decision trees together constitute a random uniform forest model.

[0098] When new, unlabeled sediment image data is input, the image data or its extracted underlying features are fed into each decision tree in the forest. Each tree independently generates a classification result for the input data, for example, determining whether the image area belongs to a predefined category such as homogeneous mud, parallel-bedding sand, or cross-bedding silt sand, or a regression value. The outputs of all decision trees in the forest are integrated, for example, using majority voting for classification tasks or taking the average of all tree outputs for regression tasks, to obtain a final description of the sediment morphological characteristics or classification result for the input image data.

[0099] In an optional embodiment, the step of performing Kalman filtering on the original ocean environment parameters of the non-image image using the adaptive forgetting factor to obtain filtered environment parameters includes:

[0100] Initializing a state vector and a covariance matrix of a Kalman filter, wherein the state vector includes non-image ocean environment parameters to be estimated;

[0101] Based on the system dynamics model, a one-step prediction is performed on the state vector and covariance matrix at the current moment to obtain a predicted state vector and a predicted covariance matrix;

[0102] Obtain the current non-image original ocean environment parameters as observation values;

[0103] Calculating a Kalman gain using the predicted state vector, the predicted covariance matrix, the observation value, and the adaptive forgetting factor;

[0104] Combining the predicted state vector, the observed value, and the Kalman gain, updating the state vector to obtain the filtered environmental parameters at the current moment;

[0105] The covariance matrix is updated for filtering at the next moment.

[0106] Specifically, the state vector of the system is initialized. The state vector contains the non-image ocean environment parameters to be estimated. For example, the state vector includes the current depth, speed and / or salinity of the acquisition device. The error covariance matrix of the state vector is also initialized. The covariance matrix characterizes the degree of uncertainty of the initial state estimate. Based on the system dynamics model, such as uniform motion, the state and error covariance at the current moment are predicted according to the posterior optimal estimated state and posterior error covariance at the previous moment. The non-image original ocean environment parameters collected in real time by the sensor nodes and transmitted through the Internet of Things communication module at the current moment are obtained as the observation values at the current moment, such as a vector composed of data such as the actual speed, water temperature and / or salinity measured by the sensor at this moment.

[0107] The parameters in the Kalman filter equation are adjusted based on the calculated adaptive forgetting factor, for example by influencing the effective values of the process noise covariance and the observation noise covariance, or by directly adjusting the update weights of the prior error covariance. When a significant interface transition between sediment layers is detected, the forgetting factor is adjusted so that the filter places more emphasis on the current new observation data, thereby quickly responding to parameter changes. Conversely, when the parameters are stable, it places more emphasis on historical data to smooth out the noise. Using the calculated Kalman gain, combined with the prior state estimate and the current observation value, the state vector is corrected to obtain the optimal state estimate at the current moment, namely the filtered environmental parameters. Finally, the error covariance matrix is updated to reflect the degree of uncertainty in the posterior state estimate. This updated covariance matrix will be used in the prediction phase at the next moment.

[0108] Among them, the Kalman gain is calculated using the predicted state vector, the predicted covariance matrix, the observation value and the adaptive forgetting factor. In one embodiment, the predicted covariance matrix is weighted using the adaptive forgetting factor, and then the Kalman gain is calculated using the weighted predicted covariance matrix, the predicted state vector and the observation value.

[0109] In an optional embodiment, selecting a feature subset from a feature set having a preset spatial correlation with the current image region includes:

[0110] Preprocessing the image data in the original ocean environment parameters to extract underlying visual features, and preliminarily grouping the underlying visual features according to physical meaning and spatial scale to form an initial feature set;

[0111] Acquire macro features of the current image area to be processed, wherein the macro features include average depth information and color information of the area;

[0112] Determine the type of sediment morphological characteristics that need to be distinguished based on the macroscopic characteristics;

[0113] Based on the determined sediment morphological feature type, adjusting the prior probability of selecting a feature group in the initial feature set that is related to the determined sediment morphological feature type and has a strong spatial correlation with the current image area, to obtain an adjusted feature set;

[0114] A specified number of features are randomly extracted from the adjusted feature set according to the adjusted prior probability to form the feature subset.

[0115] Specifically, after image preprocessing, features that describe the basic visual content of the image are extracted, including but not limited to texture and shape features. The various underlying visual features extracted are then categorized and organized according to their physical meaning, such as color, texture, and shape, and spatial scale, such as features describing large-scale variations and features describing detailed textures. For example, all color-related features are grouped into the color feature group, and all features describing local texture are grouped into the texture feature group. These feature groups together constitute the initial feature set.

[0116] Additional information is obtained from other sensors, such as depth sensors. This information includes the absolute water depth at the current image acquisition location (for example, the average depth of the current area is 1500 meters) and average color information. Macro features provide context for subsequent sediment type determination and feature selection strategy adjustment. For example, sediment types and optical properties vary significantly in marine environments at different depths.

[0117] First, a preliminary judgment is made to predict the most likely sediment morphology in the current image area, or the sediment type that needs to be distinguished. Preferably, a priori knowledge base or pre-trained model is used, which contains common sediment types and their distinguishing features at different depths and colors. For example, if the average depth is shallow and the average color is light yellow, the main types to be distinguished are fine sand, coarse sand, or shell fragments. If the average depth is deep and the average color is dark brown or gray-black, the current focus is on different types of deep-sea mud, clay, or the early forms of manganese nodules.

[0118] The sediment morphological feature types most likely to be distinguished are associated with specific feature groups in the initial feature set. If fine sand and coarse sand are to be distinguished, the texture feature group and some edge feature groups in the initial feature set are considered highly relevant. The degree of match between the features and the content of the current image region is also considered. For example, if the current image region appears visually very smooth, features describing rough textures, while belonging to the texture feature group, may have a lower strong spatial correlation with the current region. Features deemed highly relevant to the target sediment type and highly compatible with the current image region have their prior probabilities, or selection weights, increased for inclusion in the final feature subset. Conversely, the prior probabilities of irrelevant or redundant features are decreased. Continuing with the above example, if fine sand and coarse sand are to be distinguished, the prior probabilities of features describing grain size in the texture feature group are significantly increased, while the prior probabilities of global color features in the color feature group are appropriately decreased. The adjusted feature set assigns an updated selection probability to each feature in the initial feature set. While the set itself does not reduce the number of elements, the importance of each element, i.e., the chance of selection, is altered. The feature subset is formed by using a weighted random sampling method.

[0119] In an optional embodiment, if Figure 3 As shown, the random uniform forest is used to extract sediment morphological features. The random uniform forest determines the selected nodes according to the partitioning features when constructing a decision tree, including:

[0120] Step 3.1: Initialize the random uniform forest, set the number of decision trees and the maximum depth of each tree;

[0121] Step 3.2: For each decision tree to be constructed, perform bootstrap sampling with replacement on the training samples to form a training subset for the current decision tree;

[0122] Step 3.3: Recursively construct a decision tree starting from the root node. For the current node to be split, obtain the feature subset for the samples of the image area represented by the current node; randomly select a feature from the feature subset as a partitioning feature and randomly generate a splitting point; divide the samples of the current node into child nodes based on the partitioning feature and the splitting point; if the stop splitting condition is met, mark the current node as a leaf node and assign a category label; otherwise, repeat step 3.3 for the child nodes;

[0123] Step 3.4: Combine all constructed decision trees into a random uniform forest model;

[0124] Step 3.5: Input the image data to be extracted for morphological features into the random uniform forest model, and obtain the final sediment morphological features by majority voting or average.

[0125] Specifically, the basic parameters of the forest are set, including the total number of decision trees in the forest (e.g., 100 trees) and the maximum depth allowed for each decision tree (e.g., 20 layers). For each decision tree to be constructed in the forest, a bootstrap sampling method with replacement is used. Specifically, samples are randomly selected from the total training set (e.g., samples containing multiple sediment image regions with known morphological features) to form a training subset specific to that decision tree.

[0126] Each decision tree starts from its root node and is constructed by recursive splitting. For the node to be split, the pre-selected feature subset corresponding to the image area sample represented by the node is first obtained, and a feature is randomly selected from the feature subset as the partitioning feature for the current split. For example, if the feature subset contains {average particle size, porosity}, the average particle size may be randomly selected as the partitioning feature. Then, a splitting point is randomly generated for this selected partitioning feature. For example, if the partitioning feature is the average particle size, and its value range in the node sample is 0.01mm to 2mm, a splitting point such as 0.5mm may be randomly generated. Based on the partitioning feature and the splitting point, the samples in the current node are assigned to two child nodes. For example, samples with an average particle size less than 0.5mm enter the left child node, while samples with an average particle size greater than or equal to 0.5mm enter the right child node. The splitting process will continue until the preset stop condition is met.

[0127] After all the preset number of decision trees, for example 100, are independently constructed according to step 3.3, these decision trees are organically combined to form the final random uniform forest model. Figure 4 Figure 1 shows a schematic diagram of a single decision tree. When morphological features need to be extracted from new, unknown sediment image data, the image data is input into a trained random uniform forest model. Each decision tree in the model independently analyzes the input data and outputs a prediction (such as a class label). Ultimately, a majority vote is taken on the outputs of all decision trees. For example, if 70 out of 100 trees identify an area as sandy sediment, the final morphological features for that area are either sandy or calculated as an average, yielding the sediment morphological features corresponding to the input image data.

[0128] In the several embodiments provided in this application, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are merely schematic. For example, the division of the units is merely a logical function division. In actual implementation, there may be other division methods, such as multiple units or components can be combined or integrated into another system, or some feature data can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be an indirect coupling or communication connection through some interfaces, devices or units, which can be electrical, mechanical or other forms.

[0129] The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed across multiple network units. Some or all of these units may be selected to achieve the purpose of this embodiment according to actual needs.

[0130] In addition, the functional units in the various embodiments of the present application may be integrated into a single processing unit, or each unit may exist physically separately, or two or more units may be integrated into a single unit. The aforementioned integrated units may be implemented in the form of hardware or software functional units.

[0131] If the integrated unit is implemented 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 solution of the present application is essentially or the part that contributes to the prior art or all or part of the technical solution can be embodied in the form of a software product, and the computer software product is stored in a storage medium, including a number of instructions for enabling 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 described in each embodiment of the present application. The aforementioned storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

Claims

1. A method for dynamic monitoring of marine sediment layer collection based on the Internet of Things, characterized in that: include: Step 1: obtaining raw ocean environmental parameters collected in real time by multiple sensor nodes deployed in the ocean sediment collection area and transmitted through the Internet of Things communication module; Step 2: calculating an adaptive forgetting factor of a Kalman filter algorithm based on the significance of the interface transition between sediment layers indicated in the original ocean environmental parameters, and performing a Kalman filter on the original ocean environmental parameters of the non-image using the adaptive forgetting factor to obtain filtered environmental parameters; Step 3: Select a feature subset from a feature set having a preset spatial correlation with the current image area, determine a partitioning feature from the feature subset, extract sediment morphological features using a random uniform forest, and determine the selected nodes according to the partitioning feature when constructing a decision tree; and fuse the filtered environmental parameters and the sediment morphological features to obtain a comprehensive state feature vector. Step 4: Based on the comprehensive state feature vector and using a preset hierarchical acquisition model, the adjusted acquisition depth and acquisition speed of the marine sediment acquisition equipment are calculated, and the adjusted acquisition depth and acquisition speed are sent to the control system of the acquisition equipment through the Internet of Things communication module.

2. The method according to claim 1, characterized in that The adaptive forgetting factor of the Kalman filter algorithm is calculated based on the significance of the interface transition between sediment layers indicated in the original ocean environment parameters, including: receiving the original ocean environment parameters and extracting a plurality of indicators related to the interface transition between sediment layers, wherein the plurality of indicators include a pressure change rate, a depth change rate, an image texture mutation degree, and a temperature gradient; Setting weight coefficients for the respective multiple indicators, and performing weighted fusion on the multiple indicators according to the weight coefficients to obtain an initial transition significance score; Acquiring historical sediment stratification data and current acquisition target layer information, and dynamically adjusting the threshold of the initial transition significance score according to the historical sediment stratification data and the current acquisition target layer information; When the initial transition significance score exceeds the dynamically adjusted threshold, the adaptive forgetting factor adjustment amplitude is set to a first preset adjustment amplitude value; otherwise, the adaptive forgetting factor adjustment amplitude is set to a second preset adjustment amplitude value, where the first preset adjustment amplitude value is greater than the second preset adjustment amplitude value; The final adaptive forgetting factor is calculated according to the adaptive forgetting factor adjustment amplitude and the preset reference forgetting factor.

3. The method according to claim 1, characterized in that The step of using the adaptive forgetting factor to perform Kalman filtering on the original ocean environment parameters of the non-image to obtain filtered environment parameters includes: Initializing a state vector and a covariance matrix of a Kalman filter, wherein the state vector includes non-image ocean environment parameters to be estimated; Based on the system dynamics model, a one-step prediction is performed on the state vector and covariance matrix at the current moment to obtain a predicted state vector and a predicted covariance matrix; Obtain the current non-image original ocean environment parameters as observation values; Calculating a Kalman gain using the predicted state vector, the predicted covariance matrix, the observation value, and the adaptive forgetting factor; Combining the predicted state vector, the observed value, and the Kalman gain, updating the state vector to obtain the filtered environmental parameters at the current moment; The covariance matrix is updated for filtering at the next moment.

4. The method according to claim 1, wherein The selecting of a feature subset from a feature set having a preset spatial correlation with the current image region comprises: Preprocessing the image data in the original ocean environment parameters to extract underlying visual features, and preliminarily grouping the underlying visual features according to physical meaning and spatial scale to form an initial feature set; Acquire macro features of the current image area to be processed, wherein the macro features include average depth information and color information of the area; Determine the type of sediment morphological characteristics that need to be distinguished based on the macroscopic characteristics; Based on the determined sediment morphological feature type, adjusting the prior probability of selecting a feature group in the initial feature set that is related to the determined sediment morphological feature type and has a strong spatial correlation with the current image area, to obtain an adjusted feature set; A specified number of features are randomly extracted from the adjusted feature set according to the adjusted prior probability to form the feature subset.

5. The method according to claim 1, wherein The random uniform forest is used to extract sediment morphological features, and the random uniform forest determines the selected nodes according to the partitioning features when constructing a decision tree, including: Step 3.1: Initialize the random uniform forest, set the number of decision trees and the maximum depth of each tree; Step 3.2: For each decision tree to be constructed, perform bootstrap sampling with replacement on the training samples to form a training subset for the current decision tree; Step 3.3: Recursively construct a decision tree starting from the root node. For the current node to be split, obtain the feature subset for the samples of the image area represented by the current node; randomly select a feature from the feature subset as a partitioning feature and randomly generate a splitting point; divide the samples of the current node into child nodes based on the partitioning feature and the splitting point; if the stop splitting condition is met, mark the current node as a leaf node and assign a category label; otherwise, repeat step 3.3 for the child nodes; Step 3.4: Combine all constructed decision trees into a random uniform forest model; Step 3.5: Input the image data to be extracted for morphological features into the random uniform forest model, and obtain the final sediment morphological features by majority voting or average.

6. A marine sediment stratification acquisition dynamic monitoring system based on the Internet of Things, characterized in that: include: A data acquisition unit, used to obtain raw ocean environmental parameters transmitted through an Internet of Things communication module and collected in real time by multiple sensor nodes deployed in the ocean sediment collection area; a filtering unit, configured to calculate an adaptive forgetting factor of a Kalman filter algorithm based on the significance of the sediment layer interface transition indicated in the original ocean environmental parameters, and perform a Kalman filter on the non-image original ocean environmental parameters using the adaptive forgetting factor to obtain filtered environmental parameters; a feature extraction unit configured to select a feature subset from a feature set having a preset spatial correlation with a current image region, determine a partitioning feature from the feature subset, extract sediment morphological features using a random uniform forest, wherein the random uniform forest determines selected nodes based on the partitioning feature when constructing a decision tree; and obtain a comprehensive state feature vector by fusing the filtered environmental parameters and the sediment morphological features; The monitoring unit is used to calculate the adjusted collection depth and collection speed of the marine sediment collection equipment based on the comprehensive state feature vector and using a preset hierarchical collection model, and send the adjusted collection depth and collection speed to the control system of the collection equipment through the Internet of Things communication module.

7. The system according to claim 6, characterized in that The adaptive forgetting factor of the Kalman filter algorithm is calculated based on the significance of the interface transition between sediment layers indicated in the original ocean environment parameters, including: receiving the original ocean environment parameters and extracting a plurality of indicators related to the interface transition between sediment layers, wherein the plurality of indicators include a pressure change rate, a depth change rate, an image texture mutation degree, and a temperature gradient; Setting weight coefficients for the respective multiple indicators, and performing weighted fusion on the multiple indicators according to the weight coefficients to obtain an initial transition significance score; Acquiring historical sediment stratification data and current acquisition target layer information, and dynamically adjusting the threshold of the initial transition significance score according to the historical sediment stratification data and the current acquisition target layer information; When the initial transition significance score exceeds the dynamically adjusted threshold, the adaptive forgetting factor adjustment amplitude is set to a first preset adjustment amplitude value; otherwise, the adaptive forgetting factor adjustment amplitude is set to a second preset adjustment amplitude value, where the first preset adjustment amplitude value is greater than the second preset adjustment amplitude value; The final adaptive forgetting factor is calculated according to the adaptive forgetting factor adjustment amplitude and a preset reference forgetting factor.

8. The system according to claim 6, characterized in that The step of using the adaptive forgetting factor to perform Kalman filtering on the original ocean environment parameters of the non-image to obtain filtered environment parameters includes: Initializing a state vector and a covariance matrix of a Kalman filter, wherein the state vector includes non-image ocean environment parameters to be estimated; Based on the system dynamics model, a one-step prediction is performed on the state vector and covariance matrix at the current moment to obtain a predicted state vector and a predicted covariance matrix; Obtain the current non-image original ocean environment parameters as observation values; Calculating a Kalman gain using the predicted state vector, the predicted covariance matrix, the observation value, and the adaptive forgetting factor; Combining the predicted state vector, the observed value, and the Kalman gain, updating the state vector to obtain the filtered environmental parameters at the current moment; The covariance matrix is updated for filtering at the next moment.

9. The system according to claim 6, wherein: The selecting of a feature subset from a feature set having a preset spatial correlation with the current image region comprises: Preprocessing the image data in the original ocean environment parameters to extract underlying visual features, and preliminarily grouping the underlying visual features according to physical meaning and spatial scale to form an initial feature set; Acquire macro features of the current image area to be processed, wherein the macro features include average depth information and color information of the area; Determine the type of sediment morphological characteristics that need to be distinguished based on the macroscopic characteristics; Based on the determined sediment morphological feature type, adjusting the prior probability of selecting a feature group in the initial feature set that is related to the determined sediment morphological feature type and has a strong spatial correlation with the current image area, to obtain an adjusted feature set; A specified number of features are randomly extracted from the adjusted feature set according to the adjusted prior probability to form the feature subset.

10. The system according to claim 6, wherein: The random uniform forest is used to extract sediment morphological features, and the random uniform forest determines the selected nodes according to the partitioning features when constructing a decision tree, including: Step 3.1: Initialize the random uniform forest, set the number of decision trees and the maximum depth of each tree; Step 3.2: For each decision tree to be constructed, perform bootstrap sampling with replacement on the training samples to form a training subset for the current decision tree; Step 3.3: Recursively construct a decision tree starting from the root node. For the current node to be split, obtain the feature subset for the samples of the image area represented by the current node; randomly select a feature from the feature subset as a partitioning feature and randomly generate a splitting point; divide the samples of the current node into child nodes based on the partitioning feature and the splitting point; if the stop splitting condition is met, mark the current node as a leaf node and assign a category label; otherwise, repeat step 3.3 for the child nodes; Step 3.4: Combine all constructed decision trees into a random uniform forest model; Step 3.5: Input the image data to be extracted for morphological features into the random uniform forest model, and obtain the final sediment morphological features by majority voting or average.

Images