An automatic real-time identification method for sea ice conditions based on shipborne acceleration measurement

Abstract

The application discloses a kind of sea ice ice condition automatic real-time identification method based on shipborne acceleration measurement, the identification method is realized by identification system, the identification system includes three modules: signal pre-processing module, time-frequency feature extraction module and sea ice ice condition inference module;1, by time-frequency analysis, acceleration data is converted into time-frequency image, short-time Fourier transform and Hough transform can be used;2.time-frequency image feature recognition to extract key time-frequency features, DETR model can be used;3.establish the mapping relationship between acceleration signal time-frequency feature and sea ice ice condition, machine learning methods such as support vector machine can be used;Compared with existing direct measurement technology, the present application uses reverse engineering method to real-time inversion sea ice ice condition, improves the identification accuracy, reduces the cost, and can provide assistance for polar safety operation and navigation decision.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of sea ice identification, and more particularly to a sea ice ice condition automatic real-time identification method based on shipborne acceleration measurement. BACKGROUND

[0002] Due to the diversity of polar maritime risks and the complexity of the interaction mechanism, real-time risk state monitoring of polar maritime navigation and operation needs more attention. Obviously, effective and real-time detection measures are crucial for minimizing operational risks and reducing the probability of maritime accidents, which requires a flexible sea ice ice condition automatic real-time identification method.

[0003] Current technology is difficult to measure the dynamic response generated by the polar environment conditions and the ship-ice contact, but the ice condition and ice load are important parameters that determine the safety and efficiency of polar ships; therefore, it is crucial to accurately extract the ship-ice contact information in the acceleration of polar ships and identify the sea ice ice condition through acceleration data.

[0004] Existing research mainly focuses on the acceleration frequency domain analysis of specific ship-ice contact events, but does not conduct in-depth research and quantitative analysis on ship-ice contact; in the case of given acceleration measurement signals, there is still a lack of effective method for automatic and rapid identification of sea ice ice condition, which makes it difficult for icebreakers to identify the current ice condition in real time, quickly and accurately during polar operation. SUMMARY

[0005] The present application provides a sea ice ice condition automatic real-time identification method based on shipborne acceleration measurement to solve the problem that in the case of given acceleration measurement signals, there is still a lack of effective method for automatic and rapid identification of sea ice ice condition in the prior art, which makes it difficult for icebreakers to identify the current ice condition in real time, quickly and accurately during polar operation.

[0006] To achieve the above purpose, the present application provides the following technical scheme:

[0007] A sea ice ice condition automatic real-time identification method based on shipborne acceleration measurement, comprising the following steps: the identification method is realized by an identification system, the identification system comprises three modules: a signal preprocessing module, a time-frequency feature extraction module and a sea ice ice condition inference module; the specific identification steps are as follows:

[0008] Step one: input the original data into the signal preprocessing module and convert it into a time-frequency graph through time-frequency analysis method;

[0009] Step two: automatically identify the ship-ice contact event and extract the acceleration time-frequency feature in the time-frequency feature extraction module;

[0010] Step three: establish the mapping relationship between the acceleration time-frequency feature and the sea ice ice condition.

[0011] Preferably, the step one comprises the following operations:

[0012] 1. Collecting raw data and performing noise reduction processing to extract ship-ice contact signals from acceleration data;

[0013] 2. Performing time-frequency analysis on the noise-reduced acceleration signals to generate time-frequency images;

[0014] 3. Processing the time-frequency images for subsequent recognition.

[0015] Preferably, the step two comprises the following operations:

[0016] 1. Manually annotating ship-ice contact time on the acceleration time-frequency images to generate annotation files;

[0017] 2. Training the image dataset using image recognition algorithms (such as DETR) to obtain a ship-ice contact automatic recognition model;

[0018] 3. Applying the ship-ice contact automatic recognition model to process the image dataset, recognizing ship-ice contact, and extracting typical acceleration time-frequency features such as ship-ice contact frequency, pulse width, and acceleration peak value.

[0019] Preferably, the step three comprises the following operations:

[0020] 1. Corresponding the acceleration time-frequency features with sea ice features (including thickness, density, and floe size) to establish a dataset;

[0021] 2. Applying machine learning methods such as support vector machines to establish a mapping model of acceleration time-frequency features and sea ice features;

[0022] 3. Integrating the established model with the signal preprocessing module in step one and the DETR model trained in step two to form an automatic sea ice condition recognition system.

[0023] Preferably, the noise reduction processing comprises low-pass filtering of acceleration to obtain only ship-ice contact acceleration data; and using short-time Fourier transform to visualize the extracted acceleration data to obtain ship-ice contact event time-frequency images.

[0024] Preferably, the ship-ice contact automatic recognition model takes acceleration time-frequency images as input and ship-ice contact events as recognition results.

[0025] Preferably, the acceleration time-frequency features include ship-ice contact frequency, acceleration response pulse width, and acceleration amplitude.

[0026] Preferably, the sea ice condition includes three aspects: sea ice concentration, sea ice growth stage, and sea ice morphology.

[0027] The principle and beneficial effects of this technical solution:

[0028] (1) This invention uses enhanced signal decomposition and data visualization modules to extract ship-ice acceleration features from the original acceleration data, extracting acceleration data that only includes ship-ice contact; it uses STFT and HF to visualize the ship-ice contact acceleration data from (ship-ice contact events); it constructs a ship-ice contact event dataset, and proposes the ICE-DETR model based on the characteristics of ship-ice contact event images, including the following three improvement methods: adding a CA attention mechanism, replacing the neck network with an FDPN feature diffusion focusing pyramid structure, and embedding a Dimension Aware Selective Integration (DASI) module in the neck network; this invention solves the interference of noise on acceleration data during the detection process through empirical mode decomposition, absolute median difference, short-time Fourier transform, and Hough transform, and achieves accurate extraction of ship-ice acceleration features. This invention solves the problem of inaccurate identification of ship-ice contact events with high density through the CA attention mechanism, FDPN feature diffusion focusing pyramid structure, and DASI dimensionality-aware selective integration module, and achieves accurate extraction of ship-ice contact event features.

[0029] (2) A deep learning method is used to train the neural network to obtain a ship-ice contact event recognition model. The trained model is validated using evaluation metrics such as accuracy, recall, and mean precision to obtain the optimal ship-ice contact event detection model. The test set is input into the target detection model to obtain the recognition results. Each ship-ice contact event is extracted as an acceleration feature n, the pulse width of the ship-ice contact event is extracted as an acceleration feature w, and the maximum amplitude of acceleration corresponding to the ship-ice contact event is extracted as A from the time-amplitude curve. This invention solves the problem of unquantifiable model parameters through evaluation metrics such as accuracy, recall, and mean precision, and realizes the visualization of the accuracy of the ICE-DETR model. This invention solves the feature extraction of ship-ice contact events through ICE-DETR, and realizes the quantification of ship-ice contact event features.

[0030] (3) The acceleration features (n, A, w) obtained from the first two modules are artificially matched with sea ice condition features (Ct, Sa, Fa) to construct a sea ice condition dataset. A GAPSO-SVM automatic ice condition inference model is proposed. The intelligent optimization algorithm (GAPSO) is used to optimize the support vector machine model to obtain the optimal parameters (C, g). Two single-mode functions and two multi-mode functions are selected to solve for the minimum value, and the optimization performance of GAPSO is verified. The performance of GA, PSO, and GAPSO is analyzed using the optimal value. The SVM is trained using machine methods to obtain the automatic ice condition inference model. Therefore, researchers can obtain a general understanding of the ice condition by judging the three major features when studying icebreakers with unknown locations, thus obtaining more accurate research results. The acceleration data, through the automatic ice condition inference model, can automatically infer the three parameters of the ice condition (Ct, Sa, Fa). Through these parameters, researchers can roughly understand the sea ice environment experienced by the icebreaker, thereby assisting polar engineering research in obtaining more detailed environmental information. Attached Figure Description

[0031] Figure 1 This is a flowchart of the algorithm of the present invention;

[0032] Figure 2 This is a schematic diagram of the noise reduction operation of the enhanced signal decomposition and data visualization module in the embodiment;

[0033] Figure 3 This is a schematic diagram illustrating the visualization operation of the enhanced signal decomposition and data visualization module in the embodiment;

[0034] Figure 4 This is a schematic diagram of the CA attention mechanism of the time-frequency feature extraction module based on ICE-DETR in the embodiment;

[0035] Figure 5 This is a schematic diagram of the FDPN feature focusing diffusion pyramid structure of the time-frequency feature extraction module based on ICE-DETR in the embodiment.

[0036] Figure 6 This is a schematic diagram of the activation function structure of the DASI dimension-aware selective ensemble module in the time-frequency feature extraction module based on ICE-DETR in the embodiment.

[0037] Figure 7 This is a schematic diagram of the recognition result of the time-frequency feature extraction module based on ICE-DETR in the embodiment;

[0038] Figure 8 This is a schematic diagram of the GAPSO-SVM structure of the ice condition parameter inference module in the embodiment;

[0039] Figure 9 The diagram shows the objective function of the ice condition parameter inference module in this embodiment. Detailed Implementation

[0040] The present invention will now be described in further detail with reference to the accompanying drawings and embodiments:

[0041] Example:

[0042] S1: Input the raw data into the enhanced signal decomposition and data visualization module;

[0043] When analyzing sea ice conditions, images should contain as little noise and interference as possible. This process requires multiple methods to ensure that the extracted sea ice features are accurate and representative.

[0044] The extraction of ship-ice acceleration data includes: collecting raw acceleration data and using Empirical Mode Decomposition (EMD) to reduce noise and obtain IMF1 containing the largest feature; using Median Absolute Deviation (MAD) to remove discrete data points in IMF1 and obtain acceleration data containing only ship-ice contact.

[0045] Because ice condition datasets require feature extraction from different dimensions, high-quality features available for training using a single time-amplitude model are limited. In such cases, visualization techniques can provide more features for analysis; for example... Figure 3 As shown, STFT is used to convert acceleration data from a time-amplitude image to a time-frequency image. Compared with the time-frequency function plotted without EMDMAD, this method effectively suppresses noise. The ship-ice contact features are extracted and the main components of the original signal are well preserved without information loss. Therefore, this method can effectively extract ship-ice contact acceleration data features and has good temporal resolution. In order to better provide the EMDMAD-STFT generated image for ICE-DETR learning, a line segment detection algorithm based on Hough transform is used to process the image. The impact pulse is extracted as line features, and these features can be detected by HF due to the color change.

[0046] S2: Input the ship-ice contact time into the ICE-DETR-based time-frequency feature extraction module.

[0047] Image annotation of ship-ice contact events was performed using the labelimg software, generating corresponding XML annotation files. The annotation files include bounding boxes, the location information of the bounding boxes, and the true category information of the content annotated by the bounding boxes. At the same time, the original dataset was augmented using the MixUp technique to obtain a dataset consisting of 8,000 ship-ice contact events. The dataset was then split into 80% training set and 20% test set.

[0048] The proposed ICE-DETR model includes the following three improvements: adding a CA attention mechanism, such as...Figure 4 As shown.

[0049] Replace the neck network with an FDPN feature diffusion focusing pyramid structure, such as Figure 5 As shown.

[0050] Implanting a dimension-aware selective integration (DASI) module into the neck network, such as Figure 6 As shown.

[0051] The CA module enhances the model's understanding of the importance of different locations in an image by applying attention weights to the feature map in spatial coordinates. The DASI module improves the detection accuracy and robustness for densely packed small targets by adaptively selecting and fusing features and dynamically adjusting the feature fusion strategy. The FDPN module enhances the performance of target detection and classification through feature focusing and diffusion mechanisms, enabling the model to achieve efficient and accurate target detection in complex scenes.

[0052] The dataset was input into ICE-DETR for training, resulting in a ship-ice contact event recognition model. Modified modules were combined and compared to determine the practical effectiveness of the proposed ICE-DETR.

[0053]

[0054] Using a trained ICE-DETR model, images of new ship-ice contact events are identified, such as... Figure 7 As shown;

[0055] Each ship-ice contact event is extracted as the acceleration feature n, the pulse width of the ship-ice contact event is extracted as the acceleration feature w, and the maximum amplitude of acceleration is extracted from the time-amplitude curve corresponding to the ship-ice contact event as A.

[0056] Step S3: Establish the relationship between the obtained acceleration features (n, w, A) and sea ice conditions (Ct, Sa and Fa) and input them into the ice condition parameter inference module;

[0057] An intelligent optimization algorithm combining genetic algorithm and particle swarm optimization is used to optimize (C,g) in SVM.

[0058] like Figure 9 As shown, two unimodal functions and two multimodal functions are selected to solve for the minimum value, and the optimization performance of GAPSO is verified. The obtained sea ice condition dataset is input into GAPSO-SVM, and a machine learning method is used to train the SVM to obtain an automatic ice condition inference model, as shown below. Figure 8 As shown, researchers can determine the approximate ice conditions and obtain more accurate research results by judging three main characteristics when studying icebreakers in unknown locations.

[0059] The specific usage and function of this embodiment are as follows:

[0060] like Figures 1 to 9 As shown, this invention provides an automatic real-time sea ice condition identification method based on shipborne acceleration measurement. The identification method is implemented through an identification system, which includes three modules: a signal preprocessing module, a time-frequency feature extraction module, and a sea ice condition inference module. The specific identification steps are as follows:

[0061] Step 1: Input the raw data into the signal preprocessing module, and convert it into a time-frequency graph using time-frequency analysis methods.

[0062] Step 2: Automatically identify ship-ice contact events and extract acceleration time-frequency features in the time-frequency feature extraction module;

[0063] Step 3: Establish the mapping relationship between the time-frequency characteristics of acceleration and sea ice conditions.

[0064] like Figure 1 As shown, step one includes the following operations:

[0065] 1. Collect raw data and perform noise reduction processing to extract the ship-ice contact signal from the acceleration data;

[0066] 2. Perform time-frequency analysis on the noise-reduced acceleration signal to generate a time-frequency image;

[0067] 3. The time-frequency images are processed for subsequent recognition.

[0068] like Figure 1 As shown, step two includes the following operations:

[0069] 1. Manually annotate the ship-ice contact time on the acceleration time-frequency images to generate XML annotation files, and use MixUp technology to enhance the dataset to obtain the image dataset;

[0070] 2. Add CA attention mechanism to the backbone network of ICE-DETR model, replace the original network with FDPN neck network with feature diffusion focusing, and implant dimension-aware selective ensemble module (DASI) into FDPN. Train on image dataset to obtain ship-ice contact automatic recognition model.

[0071] 3. Apply the ship-ice contact automatic identification model to the image dataset.

[0072] like Figure 1 As shown, step three includes the following operations:

[0073] 1. Map the time-frequency features of acceleration to sea ice features (including thickness, concentration, and ice floe size) to create a dataset;

[0074] 2. Apply machine learning methods such as support vector machines to establish a mapping model between acceleration time-frequency features and sea ice features;

[0075] 3. Connect the established model with the signal preprocessing module in step one and the DETR model trained in step two, and package them to form an automatic sea ice condition identification system.

[0076] like Figure 3 As shown, the noise reduction process includes low-pass filtering of the acceleration to obtain data containing only ship-ice contact acceleration; the extracted acceleration data is visualized using short-time Fourier transform to obtain a time-frequency image of the ship-ice contact event.

[0077] like Figure 4 As shown, the automatic ship-ice contact identification model takes the acceleration time-frequency diagram as input and the ship-ice contact event as the identification result.

[0078] like Figure 5 As shown, the time-frequency characteristics of acceleration include three components: ship-ice contact frequency, acceleration response pulse width, and acceleration amplitude.

[0079] like Figure 5 As shown, sea ice conditions include three aspects: sea ice concentration, sea ice growth stage, and sea ice morphology.

[0080] The above descriptions are merely embodiments of the present invention, and common knowledge such as specific technical solutions and / or characteristics are not described in detail here. It should be noted that those skilled in the art can make various modifications and improvements without departing from the technical solutions of the present invention, and these should also be considered within the scope of protection of the present invention. These modifications and improvements will not affect the effectiveness of the implementation of the present invention or the practicality of the patent. The scope of protection claimed in this application should be determined by the content of its claims, and the specific embodiments described in the specification can be used to interpret the content of the claims.

Claims

1. A method for automatic real-time identification of sea ice conditions based on shipborne acceleration measurement, characterized in that, Includes the following steps: The identification method is implemented through an identification system, which includes three modules: a signal preprocessing module, a time-frequency feature extraction module, and a sea ice condition inference module; the specific identification steps are as follows: Step 1: Input the raw data into the signal preprocessing module, and convert it into a time-frequency graph using time-frequency analysis methods; Step 2: Automatically identify ship-ice contact events and extract acceleration time-frequency features in the time-frequency feature extraction module; Step 3: Establish the mapping relationship between acceleration time-frequency characteristics and sea ice conditions; Step three includes the following operations:

1. Map the time-frequency features of acceleration to the features of sea ice and establish a dataset, where the features of sea ice include sea ice thickness, concentration and ice floe size; 2. Apply the support vector machine learning method to establish a mapping model between acceleration time-frequency features and sea ice features; 3. The established model is integrated with the signal preprocessing module in step one and the DETR model trained in step two, and packaged to form an automatic sea ice condition identification system. The system utilizes the intelligent optimization algorithm GAPSO to optimize the support vector machine model and obtain the optimal parameters C and g. Two single-mode functions and two multi-mode functions are selected to solve for the minimum value, verifying the optimization performance of GAPSO. The performance of GA, PSO, and GAPSO is analyzed using the optimal values. A machine learning approach is used to train the SVM to obtain the automatic sea ice condition inference model. The acceleration time-frequency characteristics include three items: ship-ice contact frequency, acceleration response pulse width, and acceleration amplitude.

2. The automatic real-time identification method for sea ice conditions based on shipborne acceleration measurement according to claim 1, characterized in that: Step one includes the following operations:

1. Collect raw data and perform noise reduction processing to extract the ship-ice contact signal from the acceleration data; 2. Perform time-frequency analysis on the noise-reduced acceleration signal to generate a time-frequency image; 3. The time-frequency images are processed for subsequent recognition.

3. The automatic real-time identification method for sea ice conditions based on shipborne acceleration measurement according to claim 2, characterized in that: Step two includes the following operations:

1. Manually annotate the ship-ice contact time on the acceleration time-frequency image and generate an annotation file; 2. The image dataset was trained using the DETR image recognition algorithm to obtain an automatic ship-ice contact recognition model; 3. The image dataset is processed using an automatic ship-ice contact recognition model to identify ship-ice contact and extract the time-frequency features of ship-ice contact acceleration, which include ship-ice contact frequency, pulse width and acceleration peak value.

4. The automatic real-time identification method for sea ice conditions based on shipborne acceleration measurement according to claim 3, characterized in that: The noise reduction process includes low-pass filtering of the acceleration to obtain data containing only ship-ice contact acceleration; and visualization of the extracted acceleration data using short-time Fourier transform to obtain a time-frequency image of the ship-ice contact event.

5. The automatic real-time identification method for sea ice conditions based on shipborne acceleration measurement according to claim 4, characterized in that: The automatic ship-ice contact identification model takes the acceleration time-frequency graph as input and the ship-ice contact event as the identification result.

6. The automatic real-time identification method for sea ice conditions based on shipborne acceleration measurement according to claim 5, characterized in that: The sea ice conditions include three aspects: sea ice concentration, sea ice growth stage, and sea ice morphology.

Images