Electronic chart data multi-scale hierarchical organization method

Abstract

The invention discloses a multi-scale hierarchical organization method for electronic chart data, and belongs to the technical field of electronic chart processing. The problem that a traditional multi-scale layered display method generally depends on predefined scale data and is difficult to dynamically adjust the resolution to adapt to different navigation requirements is solved, a weight distribution scheme is preset according to different navigation scenes, and weights of different features are dynamically adjusted according to specific requirements of navigation tasks; it is ensured that the convolutional neural network model can automatically adjust the importance of the features in different navigation scenes, so that the requirements of navigation tasks are better met; a priority mechanism is established according to the weights allocated to the coastline features, the water depth features and the navigation obstacle features, so that the method can process complex navigation scenes more flexibly, and the safety and efficiency of navigation can be remarkably improved.

Description

Technical Field

[0001] The present invention relates to the technical field of electronic chart processing, and specifically provides a multi-scale hierarchical organization method for electronic chart data. Background Art

[0002] Electronic chart data is digital chart information used for navigation and related applications. It is stored and displayed in electronic form, providing detailed data on the marine environment, navigation conditions, and geographical information for mariners. As an important tool in the navigation field, the multi-scale hierarchical organization of electronic charts is an important technology in navigation, which has an important impact on navigation safety and efficiency.

[0003] Traditional multi-scale hierarchical display methods usually rely on predefined scale data and are difficult to dynamically adjust the resolution to meet different navigation requirements.

[0004] Therefore, it does not meet the existing requirements, and for this reason, we propose a multi-scale hierarchical organization method for electronic chart data. Summary of the Invention

[0005] The purpose of the present invention is to provide a multi-scale hierarchical organization method for electronic chart data. By presetting a weight allocation scheme according to different navigation scenarios and dynamically adjusting the weights of different features according to the specific requirements of navigation tasks, it is ensured that in different navigation scenarios, the convolutional neural network model can automatically adjust the importance of features, thereby better meeting the requirements of navigation tasks. By establishing a priority mechanism for the weights assigned to coastline features, water depth features, and navigation obstacle features, the method can more flexibly handle complex navigation scenarios, thereby significantly improving the safety and efficiency of navigation, improving the efficiency of electronic chart processing, and solving the problems raised in the above background art.

[0006] To achieve the above purpose, the present invention provides the following technical solutions: A multi-scale hierarchical organization method for electronic chart data, the method comprising the following steps: Extracting feature information from historical electronic chart data based on a convolutional neural network model, including: coastline features, water depth, and navigation obstacles; Performing multi-scale fusion on the extracted feature information to simultaneously capture obstacle details at different precisions; Presetting a weight allocation scheme according to different navigation scenarios, and dynamically adjusting the weights of different features according to the specific requirements of navigation tasks. The process of adjusting weights includes: dynamically adjusting feature weights by monitoring the yaw situation of the ship and changes in the attribute information of the current position, as well as the change amplitudes of coastline data, water depth, navigation obstacles, and weather information at the current position; The reward function is used to divide the historical electronic nautical chart data into multiple layers of skeleton data according to their uses, and the resolution under different features is dynamically adjusted based on the feature weights. The process of adjusting the resolution includes: by real-time monitoring the number of emergency events in the navigation scene, when the threshold is exceeded, the weight ratio coefficient is calculated by combining the current feature weight and the historical feature weight, and the resolution of the chart data at different levels in the multi-layer skeleton data is adjusted accordingly; The verified convolutional neural network model is applied to actual navigation.

[0007] Furthermore, feature information from historical electronic chart data is extracted based on the convolutional neural network model, including: Collect historical electronic nautical chart data, including: high-resolution remote sensing images and related annotation data; Preprocess historical electronic chart data for training convolutional neural network models; Construct a multi-scale convolutional neural network model, and use convolution kernels of different scales in the convolution layer and pooling layer to automatically identify coastline features, water depth features, and navigation obstacle features of different scales in historical electronic nautical chart data; Different types of coastline features, water depth features, and navigation obstacle features are adaptively classified and located through the fully connected layer.

[0008] Furthermore, the extracted feature information is subjected to multi-scale fusion, including: The extracted coastline features, water depth features and navigation obstacle features of different scales are spliced ​​to form a multi-scale feature map; Through upsampling and downsampling operations, feature maps of different scales are adjusted to the same size and then fused for subsequent classification and segmentation tasks.

[0009] Furthermore, a weight distribution scheme is pre-set according to different navigation scenarios, and the weights of different features are dynamically adjusted according to the specific needs of the navigation mission, including: Assigning different navigation scenarios to the extracted coastline features, water depth features and navigation obstacle features; Different weights are assigned based on the importance of the navigation scenarios corresponding to the coastline characteristics, water depth characteristics and navigation obstacle characteristics; Establish a priority mechanism for the weights assigned to coastline characteristics, water depth characteristics, and navigation obstacle characteristics based on maritime navigation needs; Combined with the real-time feedback mechanism, the feature weights are dynamically adjusted according to the execution of the navigation mission.

[0010] Furthermore, combined with the real-time feedback mechanism, the feature weights are dynamically adjusted according to the execution of the navigation mission, including: collecting first feedback information in real time, wherein the first feedback information includes real-time positioning of the current position information of the ship during navigation, and comparing the current position information of the ship with a preset navigation route to determine whether there is deviation and the degree of deviation; When the ship deviates and the degree of deviation is greater than a preset deviation threshold, it is determined that the attribute information of the current position of the ship has changed, and the attribute information of the ship's position includes features of being close to a coastline, close to a shallow water area, close to a port, or close to a deep water area. The first feature weight is adjusted based on the change in the attribute information of the ship's position; In the process of adjusting the weight of the first feature, the corresponding weight assignment interval is retrieved according to the preset attribute information of the ship; collecting second feedback information in real time, wherein the second feedback information includes coastline data information, water depth information, navigation obstacle information and weather information under the current position information of the ship, At the same time, the coastline data information, water depth information, navigation obstacle information and weather information under the current location information are compared with the reference coastline data information, reference water depth information and reference navigation obstacle information when formulating the preset navigation route, and the change amplitude values ​​are compared. Based on the change amplitude value, the second feature weight is adjusted within the weight assignment range to obtain the adjusted feature weight value.

[0011] Furthermore, according to the needs of maritime navigation, a priority mechanism is established for the weights assigned to the coastline characteristics, water depth characteristics and navigation obstacle characteristics, including: Analyze the content of the current navigation task, prioritize the weights according to the content, and dynamically adjust the feature weights; Analyze the changes in the current navigation scene and divide the weight priorities according to the changes in the navigation scene.

[0012] Furthermore, a reward function is used to divide the historical electronic chart data into multiple layers of skeleton data according to their usage, including: By introducing a reward function into the convolutional neural network model, it is evaluated whether the behavior of the convolutional neural network model meets the requirements of the current navigation task; According to the current navigation scenario, analyze the importance of different feature information as the basis for dividing weight priorities; The reward function is used to divide the historical electronic chart data into multiple layers of skeleton data according to the current navigation scene and navigation mission purpose.

[0013] Furthermore, based on the feature weights, the resolution of different features is dynamically adjusted, including: Based on the dynamically adjusted feature weights and the specific needs of the current navigation mission, the resolution of the chart data at different levels in the multi-layer skeleton data is dynamically adjusted; According to the specific requirements of the navigation mission, dynamically adjust the resolution of chart data at different scales; Based on the adaptability of the mission to the adjusted resolution of chart data, issue a reward signal according to the reward function, and the convolutional neural network model updates and adjusts the strategy according to the reward signal.

[0014] Furthermore, after dynamically adjusting the resolution of chart data at different levels in the multi-layer skeleton data, real-time monitor the number of emergency events in different navigation scenarios, and compensate according to the resolution dynamically adjusted by the number of emergency events, including: Real-time monitor the number of emergency events in different navigation scenarios; wherein, the emergency events include coastline mutation events, sharp water depth change events, sudden appearance of navigation obstacles, and emergency avoidance events caused by incorrect position information of navigation obstacles; and, the coastline mutation events are related to the coastline features; the sharp water depth change events are related to the water depth features; the sudden appearance of navigation obstacles and the emergency avoidance events caused by incorrect position information of navigation obstacles are related to the navigation obstacle features; When the number of emergency events exceeds the preset emergency event number threshold, then real-time retrieve the current feature weights corresponding to each feature; Obtain the real-time weight sum value corresponding to all features according to the current feature weights corresponding to each feature; Perform a ratio process on the current feature weights corresponding to each feature and the real-time weight sum value corresponding to all features to obtain the first weight ratio coefficient corresponding to each feature; Extract the historical feature weights corresponding to each feature; Use the historical feature weights corresponding to each feature to obtain the weight standard deviation and the weight average value; Perform a ratio process on the weight standard deviation and the weight average value to obtain the second weight ratio coefficient; Use the first weight ratio coefficient and the second weight ratio coefficient corresponding to each feature to adjust the resolution of chart data at the level of the multi-layer skeleton data where the feature is located, and obtain the adjusted resolution of chart data.

[0015] Furthermore, the convolutional neural network model updates and adjusts the strategy according to the reward signal, including: Quantify the reward signal through the sign function and feedback it into the convolutional neural network model; The convolutional neural network model dynamically adjusts the weights of different features according to the reward signal; Based on the adjustment of the feature weights, the convolutional neural network model further dynamically adjusts the resolution of chart data under different features.

[0016] Further, the convolutional neural network model updates and adjusts the policy according to the reward signal, and further includes: Manually correct the decision of the convolutional neural network model through the human-machine collaborative intervention interface; The convolutional neural network model updates and adjusts the policy according to the modified intervention behavior.

[0017] Further, preprocess the historical electronic chart data, including: Convert the historical electronic chart data to a unified coordinate system and rasterize it into a standardized image of 512×512 pixels; Normalize the historical electronic chart data and adjust the pixel values to an appropriate range; Fuse the radar image data, achieve spatial registration with the electronic chart through affine transformation, and generate a multi-modal training set with annotations; Divide the multi-modal training set, including: a training set and a test set, train the convolutional neural network model with the training set, and verify the performance of the convolutional neural network model with the test set.

[0018] Compared with the prior art, the beneficial effects of the present invention are: 1. In the present invention, by presetting a weight allocation scheme according to different navigation scenarios and dynamically adjusting the weights of different features according to the specific requirements of the navigation task; it is ensured that in different navigation scenarios, the convolutional neural network model can automatically adjust the importance of features, so as to better meet the needs of the navigation task.

[0019] 2. In the present invention, by establishing a priority mechanism for the weights assigned to the coastline feature, water depth feature, and navigation obstacle feature, the method can more flexibly handle complex navigation scenarios, thereby significantly improving the safety and efficiency of marine navigation. BRIEF DESCRIPTION OF THE DRAWINGS

[0020] Figure 1 It is a flowchart of the multi-scale hierarchical organization method for electronic chart data of the present invention. DETAILED DESCRIPTION OF THE EMBODIMENTS

[0021] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the accompanying drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those of ordinary skill in the art without creative efforts shall fall within the protection scope of the present invention.

[0022] To solve the technical problem in the prior art that the traditional multi-scale hierarchical display method usually relies on predefined scale data and is difficult to dynamically adjust the resolution to meet different navigation requirements, please refer toFigure 1 , this embodiment provides the following technical solutions: A multi-scale hierarchical organization method for electronic chart data, the method comprising the following steps: Extract feature information from historical electronic chart data based on a convolutional neural network model, including: coastline features, water depth, and navigation obstacles, including: Collect historical electronic chart data, including: high-resolution remote sensing images and related annotation data, for providing detailed coastline features, water depth, and navigation obstacle information; preprocess the historical electronic chart data for training the convolutional neural network model, including: converting the historical electronic chart data to the unified WGS-84 coordinate system and rasterizing it into a standardized image of 512×512 pixels; Channel 1 (water depth): The water depth data represents the water depth information at different positions on the chart, usually presented in the form of isobaths or point water depth annotations; The purpose is to help mariners evaluate the water depth of the navigation area, ensure that the ship sails within a safe water depth, and avoid stranding; Normalize it to the 0-1 interval and truncate extreme values (>100m); Channel 2 (shoreline): The coastline is the boundary between land and sea and is one of the most important geographical information in the electronic chart; The purpose is to help mariners determine the position of the ship relative to the land and avoid stranding and collisions; Binarize it, mark the shoreline area as 1, and the others as 0; Channel 3 (obstacles): Navigation obstacles include reefs, shipwrecks, shoals, buoys, etc. that may affect the safe navigation of ships; The purpose is to help mariners identify potential navigation risks and take avoidance measures in advance; Mark the coordinates of navigation obstacles based on the IHOS-57 standard; Normalize the historical electronic chart data, adjust the pixel values to a suitable range to improve the training efficiency of the convolutional neural network model; Integrate radar image data, achieve spatial registration with the electronic chart through affine transformation, and generate a multi-modal training set with annotations; Divide the multi-modal training set, including: a training set and a test set, train the convolutional neural network model with the training set, and use the test set to verify the performance of the convolutional neural network model to ensure the generalization ability and reliability of the convolutional neural network model.

[0023] A multi-scale convolutional neural network model is constructed, and convolution kernels of different scales in the convolution layer and pooling layer are used to automatically identify coastline features, water depth features and navigation obstacle features of different scales in historical electronic chart data. Feature information of different scales is crucial for navigation safety and navigation. For example, high-resolution coastline details are very important for near-shore navigation, while low-resolution water depth and obstacle information are suitable for ocean navigation. Different types of coastline features, water depth features and navigation obstacle features are adaptively classified and located through the fully connected layer, which can significantly improve the safety and efficiency of navigation. Multi-scale fusion of the extracted feature information can simultaneously capture obstacle details of different precisions, including: splicing the extracted coastline features, water depth features and navigation obstacle features of different scales to form a multi-scale feature map. In this way, global information and local details can be captured simultaneously, significantly improving the richness and accuracy of feature representation. Feature maps of different scales are adjusted to the same size through upsampling and downsampling operations, and then fused for subsequent classification and segmentation tasks, ensuring the spatial consistency between feature maps and further optimizing the effect of feature fusion.

[0024] The weight distribution scheme is pre-set according to different navigation scenarios, and the weights of different features are dynamically adjusted according to the specific needs of the navigation mission. The process of adjusting the weight includes: dynamically adjusting the feature weight by monitoring the ship's yaw and the change of the attribute information of the current location, as well as the change range of the coastline data, water depth, navigation obstacles and weather information at the current location; specifically including: Assign different navigation scenarios to the extracted coastline features, water depth features and navigation obstacle features; assign different weights based on the importance of the navigation scenarios corresponding to the coastline features, water depth features and navigation obstacle features; for example, increase the weight of obstacle features in emergency collision avoidance tasks; establish a priority mechanism for the weights assigned to coastline features, water depth features and navigation obstacle features according to maritime navigation needs, including: analyzing the content of the current navigation task, dividing the weight priority according to the content, and dynamically adjusting the feature weights; for example, in the case of low visibility, increase the weight of obstacle features in electronic chart data; Analyze the changes in the current navigation scene and divide the weight priority according to the changes in the navigation scene; for example: in the port area, the weight priority of the coastline and obstacle features is higher; when the ship enters the ocean area from the nearshore area, the weight of the water depth feature should be gradually increased; combined with the real-time feedback mechanism, dynamically adjust the feature weight according to the execution of the navigation mission; for example: if it is found that the weight setting of the obstacle feature is unreasonable during a certain voyage, it can be adjusted in time through the feedback mechanism to optimize the effect of feature fusion, ensure that the convolutional neural network model can maintain the best performance in different scenarios, and improve the safety and efficiency of navigation.

[0025] Beneficial effects achieved by the above content: By dynamically adjusting feature weights, the convolutional neural network model can more accurately identify and respond to potential risks in navigation; by dynamically adjusting weights according to different navigation scenarios, the convolutional neural network model can better adapt to complex navigation environments and improve its robustness and reliability; combining a real-time feedback mechanism to ensure that the dynamic adjustment of feature weights can timely reflect the actual needs of navigation tasks, thereby optimizing the effect of feature fusion; through the above operations, the electronic chart data processing method can flexibly adjust feature weights according to different navigation scenarios and task requirements, significantly enhancing the safety and efficiency of navigation.

[0026] Among them, combining a real-time feedback mechanism, according to the execution situation of the navigation task, dynamically adjusting feature weights includes: Real-time collection of the first feedback information, the first feedback information includes real-time positioning of the current position information of the ship during navigation, and comparing the current position information of the ship with the preset navigation route to determine whether there is a deviation and the degree of deviation; When the ship has a deviation and the degree of deviation is greater than the preset deviation threshold, it is determined that the attribute information of the current position of the ship has changed. The attribute information of the ship's position includes characteristics of approaching the coastline, approaching shallow water areas, approaching ports or approaching deep water areas, and performing the first feature weight adjustment based on the change in the attribute information of the ship's position; Among them, in the process of the first feature weight adjustment, the corresponding weight assignment interval is retrieved according to the preset attribute information of the ship; Real-time collection of the second feedback information, the second feedback information includes coastline data information, water depth information, navigation obstacle information, and weather information under the current position information of the ship, At the same time, comparing the coastline data information, water depth information, navigation obstacle information, and weather information under the current position information with the reference coastline data information, reference water depth information, and reference navigation obstacle information when formulating the preset navigation route, comparing the change amplitude value, and performing the second feature weight adjustment within the weight assignment interval based on the change amplitude value to obtain the adjusted feature weight value.

[0027] The above technical solution is as follows: During the execution of a navigation mission, the navigation route may deviate from the preset route due to changes in the environment or weather. Therefore, after setting the preset navigation route, a feedback mechanism needs to be set up. The first feedback information is the current position information of the ship. The change in the position information of the ship is the most important factor in navigation. There are many situations where the position deviates from the preset route during navigation, some are active deviations, some are passive or accidental deviations. In either case, it will cause deviations or problems in the weight distribution of the features set based on the preset route. Therefore, when the current position information of the ship deviates from the preset route, it is first necessary to determine whether the attribute information of the ship has changed. The attribute information of the ship generally includes features such as being close to the coastline, close to shallow water areas, close to ports or close to deep water areas, etc. Different attribute information of the ship will cause changes in the weight values of the preset features. For example, if sailing close to the coastline, the weight of the coastline feature will increase. If close to deep water areas, the weights of the water depth feature and the obstacle feature will increase. The attribute information of the ship and the general range of the weights of the set features are preset. If the attribute information of the ship changes, first perform the first feature weight adjustment. The first feature weight adjustment is to first adjust the general weight focus direction of each feature based on the attribute information of the ship. At this time, the weight assignment interval for each feature can be obtained. That is to say, the attribute information of the ship determines the general weight focus direction of each feature, but does not determine the specific weight value of each feature, only gives a weight assignment interval. The adjustment of the specific weight value of each feature needs to consider more factors.

[0028] On this premise, the second feedback information is collected in real time, including the coastline data information, water depth information, navigation obstacle information, weather information, etc. under the current position information of the ship. If the changes in these information are large compared to the reference coastline data information, reference water depth information, and reference navigation obstacle information when formulating the preset navigation route, it is necessary to combine the magnitude of the changes and perform the second feature weight adjustment within the weight assignment interval to obtain the adjusted feature weight value. The adjusted feature weight value is the result value of the dynamic adjustment. The above technical solution can monitor the preset navigation route in real time through the feedback mechanism, dynamically adjust the weight values of the features in combination with the actual situation of navigation, adapt to the changes in the environment and tasks, and has good flexibility.

[0029] The historical electronic chart data is divided into multiple layers of skeleton data according to its uses by using a reward function. Based on the feature weights, the resolution under different features is dynamically adjusted. The process of adjusting the resolution includes: by real-time monitoring the number of emergency events in the navigation scenario, when it exceeds the threshold, combining the current feature weight and the historical feature weight, calculating the weight ratio coefficient, and accordingly adjusting the resolution of the chart data at different levels in the multiple layers of skeleton data. Specifically, it includes: By introducing a reward function into the convolutional neural network model, evaluate whether the behavior of the convolutional neural network model meets the requirements of the current navigation task; for example: for key channel obstacles, the reward function can be designed to give a positive reward when the obstacle is promoted to a higher precision level, and a negative reward otherwise; according to the current navigation scenario, analyze the importance of different feature information as the basis for dividing the weight priority; for example: when navigating near the shore, the weights of the coastline and obstacle features should be higher; while when navigating in the open ocean, the weight of the water depth feature should be higher; use the reward function to divide the historical electronic chart data into multiple layers of skeleton data according to the current navigation scenario and the purpose of the navigation task; for example: high-precision coastline data can be used as one layer, medium-precision water depth data as another layer, and low-precision obstacle data as the third layer; based on the dynamically adjusted feature weights, combined with the specific requirements of the current navigation task, dynamically adjust the resolution of the chart data at different levels in the multiple layers of skeleton data; for example: adjust the resolution from 1:500,000 to 1:10,000 to adapt to different navigation scenarios; then according to the specific requirements of the navigation task, dynamically adjust the resolution of the chart data at different scales; for example: when navigating near the shore, increase the resolution of the coastline data; while when navigating in the open ocean, the model can reduce the resolution of the coastline data and at the same time increase the resolution of the water depth data; according to the adaptability of the adjusted chart data resolution to the task, send a reward signal based on the reward function, and the convolutional neural network model updates the adjustment strategy according to the reward signal, including: quantifying the reward signal through a sigmoid function and feeding it back into the convolutional neural network model; for example: if the adjusted chart data resolution is closer to the task requirements, give a positive reward; if the adjusted chart data resolution deviates from the task requirements, give a negative reward; the convolutional neural network model dynamically adjusts different features according to the reward signal, such as: the weights of the coastline feature, water depth, and navigation obstacles; for example: if the convolutional neural network model finds that increasing the weight of a certain feature can obtain a higher reward, it will automatically adjust the weight of that feature; based on the adjustment of the feature weights, the convolutional neural network model further dynamically adjusts the resolution of the chart data under different features; for example: if the reward signal indicates that increasing the resolution of key channel obstacles can better adapt to the task requirements, the convolutional neural network model will automatically increase the resolution of that feature.

[0030] It also includes: Manually correct the decisions of the convolutional neural network model through the human-machine collaborative intervention interface; for example, if a mariner finds that the model's resolution adjustment for a certain key obstacle is inappropriate, the mariner can manually make corrections; the convolutional neural network model updates the adjustment strategy based on the modified intervention behavior to improve future decision-making accuracy; based on this, not only the adaptability and flexibility of the convolutional neural network model are improved, but also the mariner's trust and understanding ability of the model's decisions are enhanced; at the same time, through real-time feedback and manual correction, the convolutional neural network model can continuously optimize its decision-making process to ensure reliability and safety in complex navigation environments.

[0031] Apply the verified convolutional neural network model to actual navigation.

[0032] The beneficial effects achieved by the above content: By presetting the weight allocation scheme according to different navigation scenarios and dynamically adjusting the weights of different features according to the specific requirements of navigation tasks; ensuring that in different navigation scenarios, the convolutional neural network model can automatically adjust the importance of features, thus better adapting to the needs of navigation tasks; also by establishing a priority mechanism for the weights assigned to coastline features, water depth features, and navigation obstacle features, the method can more flexibly handle complex navigation scenarios, thereby significantly improving the safety and efficiency of navigation.

[0033] Working principle: Automatically identify and splice the coastline features, water depth features, and navigation obstacle features in the historical electronic chart data by constructing a multi-scale convolutional neural network model to form a multi-scale feature map; preset the weight allocation scheme according to different navigation scenarios and assign different weights to the coastline features, water depth features, and navigation obstacle features; dynamically adjust the weights of different features according to the specific requirements of navigation tasks; and dynamically adjust the chart data resolution of different levels in the multi-layer skeleton data in combination with the specific requirements of the current navigation task; thereby ensuring that the method can effectively adapt to different navigation scenarios and task requirements and significantly improve the safety and efficiency of navigation.

[0034] It also includes: After dynamically adjusting the chart data resolution of different levels in the multi-layer skeleton data, real-time monitor the number of emergency events occurring in different navigation scenarios, and make compensations according to the resolution dynamically adjusted by the number of emergency events, including: Real-time monitor the number of emergency events occurring in different navigation scenarios; among them, the emergency events include coastline mutation events, sharp water depth change events, sudden appearance of navigation obstacles events, and emergency avoidance events caused by incorrect position information of navigation obstacles; and, the coastline mutation events are related to the coastline features; the sharp water depth change events are related to the water depth features; the sudden appearance of navigation obstacles events and the emergency avoidance events caused by incorrect position information of navigation obstacles are related to the navigation obstacle features; When the number of emergency events exceeds the preset emergency event quantity threshold, the current feature weight corresponding to each feature is retrieved in real time; The real-time weight sum value corresponding to all features is obtained according to the current feature weight corresponding to each feature; The current feature weight corresponding to each feature and the real-time weight sum value corresponding to all features are subjected to a ratio process to obtain a first weight ratio coefficient corresponding to each feature; The historical feature weight corresponding to each feature is extracted; The weight standard deviation and the weight average value are obtained by using the historical feature weight corresponding to each feature; The weight standard deviation and the weight average value are subjected to a ratio process to obtain a second weight ratio coefficient; The chart data resolution at the multi-layer skeleton data level where the feature is located is adjusted by using the first weight ratio coefficient and the second weight ratio coefficient corresponding to each feature to obtain the adjusted chart data resolution; Among them, the adjusted chart data resolution is obtained through the following formula:

[0035] Among them, R represents the adjusted chart data resolution; R b represents the chart data resolution before adjustment; n represents the number of features included in each layer of the skeleton data level; W 01i represents the first weight ratio coefficient corresponding to the i-th feature; W 02i represents the second weight ratio coefficient corresponding to the i-th feature; a represents the resolution sensitivity coefficient, and the value range of the resolution sensitivity coefficient is 0.6 - 1.2; b represents the emergency event gain coefficient, and the value range of the emergency event gain coefficient is 0.38 - 0.53; S i represents the number of times the i-th feature participates during the occurrence of all emergency events; S z represents the total number of times all emergency events occur. This part may have a greater effect on the resolution adjustment. Overall, this part represents the additional gain for the resolution adjustment considering emergency events and feature-related weight factors. This calculation step is to perform the a-th power operation on the first weight ratio coefficient Through the power operation, the influence degree of the first weight ratio coefficient on the final chart data resolution adjustment is amplified or reduced. The formula comprehensively considers the current weight of the feature (through W 01i ), the historical dispersion of the feature weight (through ), the participation of the feature in emergency events (through S i and S z), as well as various factors such as the sensitivity coefficient a and the gain coefficient b. The compensation adjustment of the resolution is not based on a single factor, but comprehensively weighs the impacts of various factors on navigation tasks, so as to achieve a more reasonable and practical adjustment. By setting the value ranges of the resolution sensitivity coefficient a and the emergency event gain coefficient b, and using different weight coefficients and the number of feature participations for relevant calculations, the adjustment has flexibility. Parameters can be reasonably set according to different navigation scenarios, task requirements, and expectations for resolution adjustment to effectively control the degree of resolution compensation adjustment and meet diverse application requirements.

[0036] The technical effects of the above technical solution are as follows: During navigation, first complete the dynamic adjustment of the resolution of chart data at different levels in the multi-layer skeleton data. This is the basic adjustment to set an appropriate resolution for dealing with general navigation situations. Continuously and real-time monitor the number of emergency events in different navigation scenarios, covering events such as sudden changes in the coastline, sharp changes in water depth, sudden appearance of navigation obstacles, and emergency avoidance due to incorrect position information of navigation obstacles. These emergency events will have a significant impact on navigation safety and require the system to respond in a timely manner. Compare the number of monitored emergency events with a preset threshold of the number of emergency events. If the number of emergency events does not exceed the threshold, the system maintains the current state such as the resolution; if it exceeds the threshold, it indicates that the current navigation situation is complex and a compensation adjustment to the resolution is required, and then enter the subsequent adjustment process. Calculation of the first weight ratio coefficient: Real-time retrieve the current feature weight corresponding to each feature, calculate the real-time weight sum value of all features, and then calculate the ratio of the current weight of each feature to the real-time weight sum value to obtain the first weight ratio coefficient corresponding to each feature. This coefficient reflects the relative proportion relationship of the current weights of each feature in the overall weight and reflects the importance of the current feature. Extract the historical feature weight corresponding to each feature, calculate the weight standard deviation and weight average value based on the historical weights, and then calculate the ratio of the weight standard deviation to the weight average value to obtain the second weight ratio coefficient. This coefficient reflects the degree of dispersion of the historical weights of the features and reflects the stability of the feature weights, etc. Use the obtained first weight ratio coefficient and second weight ratio coefficient to adjust the resolution of the chart data at the level of the multi-layer skeleton data where the feature is located. Considering factors such as the current importance of the feature and the stability of the historical weights, obtain the adjusted resolution of the chart data, so that the chart data can be presented at a more appropriate resolution to assist navigation decision-making and ensure navigation safety. Among them, the sudden change in the coastline event in the above emergency events refers to the sudden change in the shape of the coastline during navigation due to natural disasters such as earthquakes, tsunamis, and landslides, such as the collapse of the coastline, the formation of new islands, or the disappearance of original islands. This event is directly related to the coastline feature, and this mutation may cause a previously safe waterway to become dangerous; the sharp change in water depth event refers to the sharp change in water depth within a short period of time in a specific sea area due to changes in the seabed topography (such as underwater volcanic eruptions, mudslides, etc.) or abnormal tides, such as a sudden deepening of a shallow water area or a sudden shallowing of a deep water area. This type of event is directly related to the water depth feature. The change in water depth may directly affect the navigation safety of ships, such as an increased risk of grounding or a change in navigation resistance, etc. The sudden appearance of a navigation obstacle event refers to the sudden discovery of a navigation obstacle that did not exist before during navigation, such as a sunken ship, floating object, unmarked reef or rock, etc., which is directly related to the navigation obstacle feature. The appearance of these obstacles may seriously threaten the navigation safety of ships and immediate avoidance measures need to be taken. The emergency avoidance event caused by incorrect position information of navigation obstacles refers to the inaccurate position information of navigation obstacles due to incorrect electronic chart data or untimely updates.The ship discovers that the actual situation does not match the nautical chart only when approaching these obstacles, and it needs to make an emergency avoidance, which is directly related to the characteristics of navigation obstacles. This emergency requires the ship to respond quickly and may need to re-plan the route according to the characteristics of the coastline and water depth.

[0037] The above technical solution can dynamically adjust the resolution of nautical chart data according to whether the number of emergency events exceeds the threshold and in combination with the feature weights. When an emergency event occurs, considering the participation of different features in the emergency event, the system can be adjusted according to the specific requirements of the navigation task, improving the system's adaptability to complex and changeable navigation scenarios and better meeting the actual task requirements. By calculating various ratio coefficients of the feature weights (the first weight ratio coefficient, the second weight ratio coefficient), the resolution is adjusted considering multiple factors. Compared with the fixed resolution or simple adjustment method, the resolution of nautical chart data can be adjusted more accurately according to the importance of the features and the situation of emergency events, making the nautical chart data more in line with the actual task and improving the accuracy of data use. The feature weights are retrieved in real time and adjusted based on this to ensure that the system can respond in a timely manner to the changes in emergency events and feature weights, quickly adjust the resolution of nautical chart data, meet the real-time requirements of navigation tasks, and enable navigation decisions to be based on more timely and accurate data. The resolution is compensated by using the emergency events in combination with the dynamic weight coefficients corresponding to the features. According to the participation of different features in the emergency event (reflected by S i and S z ), as well as the relevant calculations of the feature weights (the first and second weight ratio coefficients), the compensation adjustment of the resolution is made more reasonable and scientific, avoiding excessive or insufficient adjustment. The resolution of nautical chart data at different levels in the multi-layer skeleton data is adjusted, considering the hierarchical structure characteristics of the data. In combination with the feature weights, the resolution adjustment of each level is made more refined, ensuring that the data at different levels can be presented at an appropriate resolution in case of an emergency event, and improving the overall data display and use effect.

[0038] It should be noted that in this article, relational terms such as first and second are only used to distinguish one entity or operation from another entity or operation, and do not necessarily require or imply any actual relationship or order between these entities or operations. Moreover, the term "comprising", "having" or any other variant thereof is intended to cover non-exclusive inclusion, so that a process, method, article or device comprising a series of elements includes not only those elements but also other elements not expressly listed, or elements inherent to such process, method, article or device.

[0039] Although embodiments of the present invention have been shown and described, it will be understood by those of ordinary skill in the art that various changes, modifications, substitutions and variations can be made therein without departing from the principles and spirit of the present invention, and the scope of the present invention is defined by the appended claims and their equivalents.

Claims

1. A multi-scale hierarchical organization method for electronic chart data, characterized in that: The method comprises the following steps: Extract feature information from historical electronic nautical chart data based on a convolutional neural network model, including coastline features, water depth, and navigation obstacles; The extracted feature information is fused at multiple scales to capture obstacle details of different precisions at the same time; Preset weight distribution schemes according to different navigation scenarios, and dynamically adjust the weights of different features according to the specific needs of the navigation mission; the process of adjusting weights includes: dynamically adjusting feature weights by monitoring the ship's yaw and changes in attribute information at the current location, as well as changes in coastline data, water depth, navigation obstacles and weather information at the current location; The reward function is used to divide the historical electronic chart data into multiple layers of skeleton data according to their uses, and the resolution under different features is dynamically adjusted based on the feature weights. The process of adjusting the resolution includes: by real-time monitoring the number of emergency events in the navigation scene, when the threshold is exceeded, the weight ratio coefficient is calculated by combining the current feature weight and the historical feature weight, and the resolution of the chart data at different levels in the multi-layer skeleton data is adjusted accordingly; The verified convolutional neural network model is applied to actual navigation.

2. The method for multi-scale hierarchical organization of electronic nautical chart data according to claim 1, characterized in that: Based on the convolutional neural network model, feature information is extracted from historical electronic chart data, including: Collect historical electronic nautical chart data, including: high-resolution remote sensing images and related annotation data; Preprocess historical electronic chart data for training convolutional neural network models; including: Convert historical electronic chart data into a unified coordinate system and rasterize them into standardized images of 512×512 pixels; Normalize historical electronic chart data and adjust pixel values ​​to an appropriate range; Fuse radar image data, achieve spatial registration with electronic charts through affine transformation, and generate annotated multimodal training sets; Divide the multimodal training set into a training set and a test set. Use the training set to train the convolutional neural network model, and use the test set to verify the performance of the convolutional neural network model. Construct a multi-scale convolutional neural network model, and use convolution kernels of different scales in the convolution layer and pooling layer to automatically identify coastline features, water depth features, and navigation obstacle features of different scales in historical electronic nautical chart data; Different types of coastline features, water depth features, and navigation obstacle features are adaptively classified and located through the fully connected layer.

3. The method for multi-scale hierarchical organization of electronic chart data according to claim 2, characterized in that: The extracted feature information is fused at multiple scales, including: The extracted coastline features, water depth features and navigation obstacle features of different scales are spliced ​​to form a multi-scale feature map; Through upsampling and downsampling operations, feature maps of different scales are adjusted to the same size and then fused for subsequent classification and segmentation tasks.

4. The method for multi-scale hierarchical organization of electronic chart data according to claim 3, characterized in that: Preset weight distribution schemes according to different navigation scenarios, and dynamically adjust the weights of different features according to the specific needs of the navigation mission, including: Assigning different navigation scenarios to the extracted coastline features, water depth features and navigation obstacle features; Different weights are assigned based on the importance of the navigation scenarios corresponding to the coastline characteristics, water depth characteristics and navigation obstacle characteristics; Establish a priority mechanism for the weights assigned to coastline characteristics, water depth characteristics, and navigation obstacle characteristics based on maritime navigation needs; Combined with the real-time feedback mechanism, the feature weights are dynamically adjusted according to the execution of the navigation mission.

5. The method for multi-scale hierarchical organization of electronic chart data according to claim 4, characterized in that: Combined with the real-time feedback mechanism, the feature weights are dynamically adjusted according to the execution of the navigation mission, including: collecting first feedback information in real time, wherein the first feedback information includes real-time positioning of the current position information of the ship during navigation, and comparing the current position information of the ship with a preset navigation route to determine whether there is deviation and the degree of deviation; When the ship deviates and the degree of deviation is greater than a preset deviation threshold, it is determined that the attribute information of the current position of the ship has changed, and the attribute information of the ship's position includes features of being close to a coastline, close to a shallow water area, close to a port, or close to a deep water area. The first feature weight is adjusted based on the change in the attribute information of the ship's position; In the process of adjusting the weight of the first feature, the corresponding weight assignment interval is retrieved according to the preset attribute information of the ship; collecting second feedback information in real time, wherein the second feedback information includes coastline data information, water depth information, navigation obstacle information and weather information under the current position information of the ship; At the same time, the coastline data information, water depth information, navigation obstacle information and weather information under the current location information are compared with the reference coastline data information, reference water depth information and reference navigation obstacle information when formulating the preset navigation route, and the change amplitude values ​​are compared. Based on the change amplitude value, the second feature weight is adjusted within the weight assignment range to obtain the adjusted feature weight value.

6. The method for multi-scale hierarchical organization of electronic nautical chart data according to claim 4, characterized in that: According to the needs of maritime navigation, a priority mechanism is established to assign weights to the coastline characteristics, water depth characteristics and navigation obstacle characteristics, including: Analyze the content of the current navigation task, prioritize the weights according to the content, and dynamically adjust the feature weights; Analyze the changes in the current navigation scene and divide the weight priorities according to the changes in the navigation scene.

7. The method for multi-scale hierarchical organization of electronic chart data according to claim 4, characterized in that: The reward function is used to divide the historical electronic chart data into multiple layers of skeleton data according to their usage, including: By introducing a reward function into the convolutional neural network model, it is evaluated whether the behavior of the convolutional neural network model meets the requirements of the current navigation task; According to the current navigation scenario, analyze the importance of different feature information as the basis for dividing weight priorities; The reward function is used to divide the historical electronic chart data into multiple layers of skeleton data according to the current navigation scene and navigation mission purpose.

8. The method for multi-scale hierarchical organization of electronic nautical chart data according to claim 7, characterized in that: Based on feature weights, the resolution of different features is dynamically adjusted, including: Based on the dynamically adjusted feature weights and the specific needs of the current navigation mission, the resolution of the chart data at different levels in the multi-layer skeleton data is dynamically adjusted; Then, according to the specific needs of the navigation mission, the resolution of the chart data at different scales is dynamically adjusted; According to the adaptability of the adjusted chart data resolution to the task, a reward signal is sent based on the reward function, and the convolutional neural network model updates the adjustment strategy according to the reward signal.

9. The method for multi-scale hierarchical organization of electronic nautical chart data according to claim 8, characterized in that: After dynamically adjusting the resolution of different levels of chart data in the multi-layer skeleton data, the number of emergencies in different navigation scenarios is monitored in real time, and compensation is made according to the dynamically adjusted resolution according to the number of emergencies, including: Real-time monitoring of the number of emergency events in different navigation scenarios; wherein the emergency events include coastline mutation events, water depth drastic change events, navigation obstacle sudden appearance events and emergency avoidance events caused by navigation obstacle location information errors; and the coastline mutation events are related to the coastline characteristics; the water depth drastic change events are related to the water depth characteristics; the navigation obstacle sudden appearance events and the emergency avoidance events caused by navigation obstacle location information errors are related to the navigation obstacle characteristics; When the number of emergency events exceeds the preset emergency event number threshold, the current feature weight corresponding to each feature is retrieved in real time; Obtaining real-time weights and values ​​corresponding to all features according to the current feature weight corresponding to each feature; Performing ratio processing on the current feature weight corresponding to each feature and the real-time weight and value corresponding to all features to obtain a first weight ratio coefficient corresponding to each feature; Extract the historical feature weights corresponding to each feature; Obtaining a weight standard deviation and a weight average using the historical feature weights corresponding to each feature; Performing ratio processing using the weight standard deviation and the weight average value to obtain a second weight ratio coefficient; The first weight ratio coefficient and the second weight ratio coefficient corresponding to each feature are used to adjust the resolution of the nautical chart data at the multi-layer skeleton data level where the feature is located, so as to obtain the adjusted nautical chart data resolution.

10. The method for multi-scale hierarchical organization of electronic nautical chart data according to claim 8, characterized in that: The convolutional neural network model updates and adjusts the strategy based on the reward signal, including: The reward signal is quantified through a symbolic function and fed back into the convolutional neural network model; The convolutional neural network model dynamically adjusts the weights of different features based on the reward signal; Based on the adjustment of feature weights, the convolutional neural network model further dynamically adjusts the resolution of chart data under different features; Also includes: Manually correct the decision of the convolutional neural network model through the human-computer collaborative intervention interface; The convolutional neural network model updates and adjusts the strategy based on the modified intervention behavior.

Images