Ship navigation method, device and equipment for sea ice area and storage medium

Abstract

The present application relates to the technical field of ship navigation, and provides a ship navigation method, device and equipment for a sea ice area and a storage medium, the method comprising: acquiring a remote sensing image of a sea ice area to be passed through by a target ship, and detecting feature information of each sea ice in the sea ice area based on the remote sensing image; the feature information comprises contour information and category information, and the category information is used to represent the size of the sea ice; clustering the sea ice according to the feature information to obtain sea ice clusters; planning a path for the target ship based on the sea ice clusters, and navigating the target ship according to the planned target path. By clustering the sea ice and planning a path for the ship according to the clustering result, the calculation amount of path planning is reduced, and the efficiency and accuracy of path planning are improved. Navigating the ship according to the planned target path is conducive to improving the safety of the ship passing through the sea ice area.

Description

TECHNICAL FIELD

[0001] The present application relates to the field of ship navigation technology, and in particular to a ship navigation method and device in a sea ice area, equipment and a storage medium. BACKGROUND

[0002] In the field of marine shipping, the real-time requirement for route planning of a ship is relatively high. In high-latitude areas, monsoon ocean currents and other natural factors cause sea ice to accumulate or disperse, which easily affects the normal driving of a ship. In an extreme environment of a large-area sea ice area, an icebreaker is generally used to open up a channel for the ship to plan a path, or an unmanned aerial vehicle or human observation is used to observe the situation of a channel in front of the ship to make a navigation decision. In the method of using an icebreaker to open up a channel, since the situation of sea ice is unknown, it is difficult to ensure the safety of the ship navigation along the planned driving path. In the method of using an unmanned aerial vehicle or human observation to make a navigation decision, the observation range of the channel in front of the ship is limited, and the accuracy of the observed channel situation is difficult to guarantee, which affects the accurate planning of the driving path of the ship and also makes it difficult to ensure the safety of the ship navigation. SUMMARY

[0003] The present application provides a ship navigation method and device in a sea ice area, equipment and a storage medium, to solve the defect that the existing ship path planning method in a sea ice area is difficult to ensure the safety of ship navigation, and to realize the accurate planning of the driving path of a ship in a sea ice area and improve the safety of ship navigation.

[0004] The present application provides a ship navigation method in a sea ice area, comprising the following steps:

[0005] Obtaining a remote sensing image of a sea ice area to be passed by a target ship, and detecting feature information of each sea ice in the sea ice area based on the remote sensing image; the feature information includes contour information and category information, and the category information is used to represent the size of the sea ice;

[0006] Clustering the sea ice according to the feature information to obtain a sea ice cluster;

[0007] Planning a path for the target ship based on the sea ice cluster, and navigating the target ship according to the planned target path.

[0008] According to the ship navigation method in a sea ice area provided by the present application, the category information at least includes large sea ice, medium sea ice and small sea ice; the clustering of the sea ice according to the feature information to obtain a sea ice cluster comprises:

[0009] Selecting a sea ice set in the sea ice area according to the category information; the sea ice set includes large sea ice and medium sea ice;

[0010] calculate a shortest distance between a first sea ice and each second sea ice based on the contour information; the first sea ice is any one of the set of sea ice, and the second sea ice is any one of the set of sea ice except the first sea ice;

[0011] perform a density-based clustering processing on the sea ice in the set of sea ice based on the shortest distance and a minimum turning radius of the target ship to obtain a sea ice cluster; wherein the shortest distance between adjacent sea ice in the same sea ice cluster is less than the minimum turning radius.

[0012] According to the ship navigation method for the sea ice area provided by the application, the target ship is path planned based on the sea ice cluster, which comprises:

[0013] obtain the maximum Feret diameter of the ice cluster area corresponding to each sea ice cluster, and classify the sea ice cluster into a medium sea ice cluster and a large sea ice cluster according to the maximum Feret diameter of the ice cluster area;

[0014] perform a globally optimal obstacle avoidance path planning on the large sea ice cluster and a locally optimal obstacle avoidance path planning on the medium sea ice cluster based on the start position and the end position of the target ship to obtain a path array; the path array is used to represent the path distance between key nodes, and the key nodes are selected from the edge pixel points of the medium sea ice cluster and the large sea ice cluster in the path planning;

[0015] perform an optimal path planning on the target ship by using a single-source shortest path algorithm based on the path array.

[0016] According to the ship navigation method for the sea ice area provided by the application, after the feature information of each sea ice in the sea ice area is detected based on the remote sensing image, the method further comprises:

[0017] obtain the maximum Feret diameter of each sea ice, and convert the maximum Feret diameter of the sea ice into first evidence information describing the sea ice contour;

[0018] convert the category information in the feature information into second evidence information describing the sea ice category; the category information contains the confidence degree of the sea ice belonging to each preset category;

[0019] fuse the first evidence information and the second evidence information based on a preset fusion algorithm to obtain a fusion feature;

[0020] determine the target category to which the sea ice belongs according to the fusion feature, and update the category information based on the target category.

[0021] The ship navigation method of the sea ice area provided by the application comprises the following steps:

[0022] The remote sensing image is input into a pre-trained target detection model to obtain feature information of each sea ice in the sea ice area output by the target detection model; wherein, contour information in the feature information is represented by a two-dimensional array, and the two-dimensional array is used to describe position coordinates of edge pixel points of the sea ice contour.

[0023] The target detection model comprises a backbone network and a head network, and the backbone network is built based on a spatial pyramid mechanism and a convolution attention mechanism; the backbone network comprises a first convolution layer, a residual feature extraction layer, a spatial pyramid layer and a convolution attention layer; and the head network comprises an up-sampling layer, a splicing layer, a non-residual feature extraction layer and a second convolution layer.

[0024] The ship navigation method of the sea ice area provided by the application further comprises the following steps before the feature information of each sea ice in the sea ice area is detected based on the remote sensing image:

[0025] A sample data set constructed based on a sea ice data set is obtained; the sample data set contains annotation information of a sea ice sample image, and the annotation information is used to annotate a sea ice category and a real region frame of sea ice; the sea ice data set is obtained by preprocessing a sea ice image, and the preprocessing comprises angle adjustment and pixel cropping;

[0026] A pre-set target detection model is iteratively trained based on the sample data set, and in each iteration process, a loss value of the target detection model is calculated.

[0027] Model parameters of the target detection model are updated according to the loss value until a maximum iteration number reaches a preset maximum iteration number or the loss value converges, so as to obtain the pre-trained target detection model.

[0028] The ship navigation method of the sea ice area provided by the application comprises the following steps in each iteration process:

[0029] In each iteration process, a predicted region frame of sea ice in the sea ice sample image output by the target detection model is obtained, and a shape intersection-over-union ratio of the predicted region frame and the real region frame is calculated.

[0030] A first transverse length, a first longitudinal length of the predicted region frame and a first position coordinate of a center point of the predicted region frame are obtained, and a second transverse length, a second longitudinal length of the real region frame and a second position coordinate of a center point of the real region frame are obtained.

[0031] According to the shape intersection-over-union ratio, the first transverse length, the first longitudinal length, the first position coordinate, the second transverse length, the second longitudinal length, and the second position coordinate, a regression coefficient of the predicted region frame is calculated; the regression coefficient is used to represent a regression direction of the predicted region frame relative to the real region frame;

[0032] Based on the shape intersection-over-union ratio and the regression coefficient, a loss value of the target detection model is calculated.

[0033] The present application also provides a ship navigation device for a sea ice region, comprising the following modules:

[0034] A sea ice detection module is configured to acquire a remote sensing image of a sea ice region to be passed by a target ship, and detect feature information of each sea ice in the sea ice region based on the remote sensing image; the feature information comprises contour information and category information, and the category information is used to represent a size of the sea ice.

[0035] A sea ice clustering module is configured to cluster the sea ice according to the feature information to obtain a sea ice cluster.

[0036] A path planning and navigation module is configured to plan a path for the target ship based on the sea ice cluster, and navigate the target ship according to the planned target path.

[0037] The present application also provides an electronic device comprising a memory, a processor, and a computer program stored in the memory and running on the processor, wherein the processor implements the ship navigation method for a sea ice region as described above.

[0038] The present application also provides a non-transitory computer readable storage medium having a computer program stored thereon, wherein the computer program is executed by a processor to implement the ship navigation method for a sea ice region as described above.

[0039] The present application also provides a computer program product comprising a computer program, wherein the computer program is executed by a processor to implement the ship navigation method for a sea ice region as described above.

[0040] The method, device, equipment and storage medium for ship navigation in a sea ice area provided by the application, by acquiring a remote sensing image of a sea ice area to be passed through by a target ship, detecting feature information of sea ice based on the remote sensing image of the sea ice area, clustering the sea ice according to the feature information to obtain sea ice clusters, planning a path for the target ship based on the sea ice clusters, and navigating the target ship according to the planned target path, the calculation amount of path planning is reduced, the efficiency and accuracy of path planning are improved, and the ship is navigated according to the planned target path, which is conducive to improving the safety of the ship passing through the sea ice area. BRIEF DESCRIPTION OF DRAWINGS

[0041] In order to more clearly illustrate the technical solutions in the application or prior art, the following will briefly introduce the drawings needed to be used in the embodiments or prior art description. Obviously, the drawings in the following description are some embodiments of the application, and other drawings can be obtained by those skilled in the art without creative labor.

[0042] Figure 1 is a flowchart of the method for ship navigation in a sea ice area provided by the application.

[0043] Figure 2 is a structural diagram of the convolution attention module provided by the application.

[0044] Figure 3 is a structural diagram of the device for ship navigation in a sea ice area provided by the application.

[0045] Figure 4 is a structural diagram of the electronic device provided by the application. DETAILED DESCRIPTION

[0046] In order to make the objects, technical solutions and advantages of the application clearer, the technical solutions in the application will be described clearly and completely below with reference to the drawings in the application. Obviously, the described embodiments are some embodiments of the application, not all embodiments. Based on the embodiments in the application, all other embodiments obtained by those skilled in the art without creative labor are within the protection scope of the application.

[0047] The embodiment of the application provides a method for ship navigation in a sea ice area, the size and contour of each sea ice in the sea ice area are identified based on a remote sensing image of the sea ice area, thereby the adjacent sea ice of a specific size that will hinder the navigation of a ship is clustered according to distance to obtain corresponding sea ice clusters, the navigation path of the ship in the sea ice area is planned according to the sea ice clusters, and the navigation of the ship in the sea ice area is navigated according to the planned path.

[0048] Specifically, referring to Figure 1 , Figure 1 is a flowchart of a ship navigation method for a sea ice region provided by an embodiment of the present application, as shown in Figure 1 , the ship navigation method for a sea ice region comprises the following steps:

[0049] Step 100, acquiring a remote sensing image of a sea ice region to be passed through by a target ship, and detecting feature information of each sea ice in the sea ice region based on the remote sensing image; the feature information comprises contour information and category information, and the category information is used to represent the size of the sea ice;

[0050] Step 200, clustering the sea ice according to the feature information to obtain sea ice clusters;

[0051] Step 300, planning a path for the target ship based on the sea ice clusters, and navigating the target ship according to the planned target path.

[0052] First, a remote sensing image of a sea ice region to be passed through by a target ship is acquired, and feature information of each sea ice in the sea ice region is detected based on the remote sensing image. The feature information of the sea ice comprises contour information and category information of the sea ice, wherein the category information is used to represent the size of the sea ice. Exemplarily, the category information of the sea ice comprises large, medium and small, which are used to represent the size of the sea ice, and the contour information of the sea ice is used to represent the shape of the sea ice.

[0053] The contour information of the sea ice can be represented by position coordinates of contour points of the sea ice edge, or the contour information of the sea ice is represented by position coordinates of contour key points of the sea ice edge.

[0054] According to the feature information of the sea ice, the sea ice is clustered to obtain sea ice clusters. In an embodiment, the sea ice is density-based clustered according to the feature information of the sea ice to obtain the sea ice clusters. Optionally, the distance between each two sea ices is determined according to the contour information in the feature information of the sea ice, and the sea ice is clustered according to the distance to obtain sea ice clusters with similar distances. Optionally, the distance between each two sea ices is determined according to the contour information in the feature information of the sea ice, and the sea ice is clustered according to the distance, and on this basis, further clustering is performed according to the category information in the feature information to obtain sea ice clusters with similar distances and same categories.

[0055] The target ship is path planned based on the obtained sea ice clusters, and is navigated according to the planned target path. For sea ice that is relatively close, the distance between the sea ice is not enough for the ship to pass through, but if path planning is performed based on each sea ice block, the amount of calculation required for path planning is large, which is not conducive to the real-time requirement of the ship for path planning and navigation. Therefore, by clustering the sea ice, the ship is path planned based on the obtained sea ice clusters, so as to reduce the amount of calculation during path planning, improve the path planning efficiency, and meet the real-time requirement of the ship for navigation.

[0056] Optionally, the path planning of the target ship is performed based on the position of the sea ice cluster under the condition that the starting position and the destination position of the ship relative to the sea ice region are known. The target path obtained by the planning has the starting position of the target ship relative to the sea ice region as the starting point, has the destination position of the target ship relative to the sea ice region as the ending point, and has the position of the sea ice cluster as an impassable obstacle region that needs to be avoided. The starting position and the destination position of the target ship relative to the sea ice region are determined based on the predetermined sailing route of the ship.

[0057] In this embodiment, the feature information of the sea ice is detected based on the remote sensing image of the sea ice region through which the target ship is to pass, the sea ice is clustered according to the feature information, the sea ice clusters are obtained, the target ship is path planned based on the sea ice clusters, and the target ship is navigated according to the planned target path. By clustering the sea ice and path planning the ship according to the clustering result, the amount of calculation for path planning is reduced, the efficiency and accuracy of path planning are improved, the ship is navigated according to the planned target path, and the safety of the ship passing through the sea ice region is improved.

[0058] In one embodiment, the category information in the feature information of the sea ice at least includes large sea ice, medium sea ice and small sea ice, and the clustering of the sea ice is based on the density of the sea ice. Therefore, step 200 further includes:

[0059] In step 201, a set of sea ice in the sea ice region is selected according to the category information; the set of sea ice includes large sea ice and medium sea ice;

[0060] In step 202, the shortest distance between a first sea ice and each second sea ice is calculated based on the contour information; the first sea ice is any one in the set of sea ice, and the second sea ice is any one in the set of sea ice except the first sea ice;

[0061] In step 203, the sea ice in the sea ice set is subjected to density-based clustering processing according to the shortest distance and the minimum turning radius of the target ship, to obtain a sea ice cluster; wherein the shortest distance between adjacent sea ice in the same sea ice cluster is less than the minimum turning radius.

[0062] According to the category information in the feature information of the sea ice, a sea ice set in the sea ice region is selected, the sea ice set including large sea ice and medium sea ice, and the shortest distance between a first sea ice and each second sea ice is calculated based on the contour information in the feature information of the sea ice; wherein the first sea ice is any one in the sea ice set, and the second sea ice is any one in the sea ice set except the first sea ice.

[0063] Further, the sea ice in the sea ice set is subjected to density-based clustering processing according to the shortest distance between any two sea ice in the sea ice set and the minimum turning radius of the target ship, to obtain a sea ice cluster; wherein the shortest distance between adjacent sea ice in the same sea ice cluster is less than the minimum turning radius of the ship.

[0064] For the sea ice in the sea ice region, the small sea ice does not affect the ship passage, therefore, only the large sea ice and the medium sea ice are taken as the obstacles affecting the ship passage. Further, the shortest distance between the large sea ice and the medium sea ice is compared with the minimum turning radius of the ship to determine whether the ship can pass, and for the shortest distance less than the minimum turning radius of the ship, the ship cannot pass between the sea ice, and the sea ice with the shortest distance less than the minimum turning radius of the ship is taken as a whole and clustered into the same sea ice cluster.

[0065] In one embodiment, the density clustering algorithm is adopted to group the sea ice, the similarity between the sea ice is measured based on the shortest distance between the sea ice, and the shortest distance between the sea ice is calculated based on the contour information in the feature information of the sea ice. Specifically, for the shortest distance between the first sea ice and the second sea ice, first, based on any edge pixel point in the contour information of the first sea ice, the distance between the edge pixel point and each edge pixel point in the contour information of the second sea ice is calculated, in this way, the distance between each edge pixel point in the contour information of the first sea ice and each edge pixel point in the contour information of the second sea ice is calculated, and the shortest distance is selected as the shortest distance between the first sea ice and the second sea ice.

[0066] Further, the shortest distance between the sea ice is compared with the minimum turning radius of the ship, so as to further measure the similarity between the sea ice, and determine whether the sea ice is clustered into the same group. Specifically, if the shortest distance between the sea ice is less than or equal to the minimum turning radius of the ship, it means that the shortest distance between the sea ice is not enough for the ship to pass normally, and the sea ice is clustered into the same group to obtain the corresponding sea ice cluster; if the shortest distance between the sea ice is greater than the minimum turning radius of the ship, it means that the shortest distance between the sea ice is enough for the ship to pass normally, and the sea ice does not belong to the same cluster group.

[0067] The sea ice clusters obtained by clustering are classified according to the maximum Feret diameter, so as to divide the sea ice clusters into large sea ice clusters and medium sea ice clusters, and different algorithms are used for path planning according to different sizes of the sea ice clusters. Based on this, in step 300, the path planning for the target ship based on the sea ice cluster further includes:

[0068] In step 301, the maximum Feret diameter of the ice cluster area corresponding to each sea ice cluster is obtained, and the sea ice cluster is classified into a medium sea ice cluster and a large sea ice cluster according to the maximum Feret diameter of the ice cluster area.

[0069] In step 302, the global optimal obstacle avoidance path planning is performed on the large sea ice cluster and the local optimal obstacle avoidance path planning is performed on the medium sea ice cluster based on the start position and the end position of the target ship, and a path array is obtained; the path array is used to represent the path distance between the key nodes, and the key nodes are selected from the edge pixel points of the medium sea ice cluster and the large sea ice cluster in the path planning.

[0070] In step 303, the single-source shortest path algorithm is used to perform optimal path planning on the target ship based on the path array.

[0071] The classification of the sea ice cluster is determined based on the maximum Feret diameter of the sea ice cluster. When the sea ice cluster is classified, the maximum Feret diameter of each sea ice cluster is first obtained, and the sea ice cluster is classified into a medium sea ice cluster and a large sea ice cluster according to the maximum Feret diameter of each sea ice cluster.

[0072] Optionally, the global optimal obstacle avoidance path planning is performed on the large sea ice cluster and the local optimal obstacle avoidance path planning is performed on the medium sea ice cluster based on the start position and the end position of the target ship, and a path array is obtained, which is used to identify the path distance between the key nodes, and the key nodes are selected from the edge pixel points of the medium sea ice cluster and the large sea ice cluster in the path planning. The single-source shortest path algorithm is used to perform optimal path planning on the target ship based on the path array obtained by planning.

[0073] In an embodiment, the area corresponding to the sea ice cluster is irregular, and the maximum Feret diameter of the irregular area occupied by all the sea ice in the sea ice cluster is greater than a first number of pixel points, and the sea ice cluster is a large sea ice cluster, and if the maximum Feret diameter of the irregular area occupied by all the sea ice in the sea ice cluster is greater than a second number of pixel points and less than the first number of pixel points, the sea ice cluster is a medium sea ice cluster. The first number of pixel points is greater than the second number of pixel points.

[0074] For the sea ice cluster obtained by clustering, a globally optimal obstacle avoidance path planning is performed on the large sea ice cluster. Specifically, for the large sea ice cluster, the starting point A and the ending point Z of the sea ice area are first connected, and then the tangent points B and C passing through the starting point A and the large sea ice cluster are found, and the distance AB between the starting point A and the tangent point B and the distance BC between the tangent points B and C are calculated. Finally, the tangent points B and C are taken as new starting points, and the above process is repeated to obtain a feature map of the path planning composed of nodes and edges.

[0075] For the sea ice cluster obtained by clustering, a locally optimal obstacle avoidance path planning is performed on the medium sea ice cluster. Specifically, when the ship encounters a medium sea ice cluster during navigation, the ship can slowly navigate around the medium sea ice cluster, and the contour points of the medium sea ice cluster are used to calculate the paths on both sides for local optimal path planning.

[0076] In this way, the key nodes of each sea ice cluster and the distance between any two adjacent key nodes can be obtained, and a path array for path planning can be obtained. A single-source shortest path algorithm is used to make a decision on the path array, and an optimal path is planned for the target ship to obtain a target path. Optionally, the single-source shortest path algorithm includes but is not limited to Dijkstra's algorithm, and the optimal path can be a distance optimal path, i.e., a shortest path.

[0077] In an embodiment, after detecting the feature information of the sea ice based on the remote sensing image of the sea ice area, the detected feature information is updated based on the maximum Feret diameter of the sea ice. Based on this, after step 100, the method can further include:

[0078] In step 110, the maximum Feret diameter of each sea ice is obtained, and the maximum Feret diameter of the sea ice is converted into first evidence information describing the contour of the sea ice.

[0079] In step 120, the category information in the feature information is converted into second evidence information describing the category of the sea ice. The category information includes the confidence of the sea ice belonging to each preset category.

[0080] In step 130, the first evidence information and the second evidence information are fused based on a preset fusion algorithm to obtain fused features.

[0081] Step 140, determining a target category to which the sea ice belongs according to the fusion feature, and updating the category information based on the target category.

[0082] After detecting the feature information of each sea ice, the maximum Feret diameter of each sea ice is obtained, and the maximum Feret diameter of the sea ice is converted into the first evidence information describing the profile of the sea ice. Optionally, the profile information in the feature information of the sea ice is represented by a two-dimensional array including the position coordinates of each edge pixel point of the sea ice, and the maximum Feret diameter of the sea ice is calculated according to the position coordinates of the edge pixel points of the sea ice.

[0083] The maximum width of the sea ice is a key factor affecting the navigation of the ship, and the maximum width of the sea ice is represented by the maximum Feret diameter, and the sea ice is clustered and path planned according to the maximum Feret diameter, to ensure the safety of the ship navigation. Optionally, for irregular sea ice, the maximum Feret diameter can be the length of the circumscribed rectangle of the sea ice, or the diagonal length of the circumscribed rectangle of the sea ice.

[0084] According to the maximum Feret diameter of the sea ice , the profile information of the sea ice is converted into the first evidence information describing the profile of the sea ice, and the conversion of the maximum Feret diameter can be calculated according to the following formula:

[0085] ; (1)

[0086] ; (2)

[0087] ; (3)

[0088] For the maximum Feret diameter of any sea ice , the probabilities of the sea ice belonging to large sea ice , medium sea ice and small sea ice , etc. are calculated according to the above formulas 1 to 3, respectively, as the first evidence information of the profile of the sea ice. , and are the evidence conversion functions of large sea ice, medium sea ice and small sea ice, is the gamma function.

[0089] Further, the category information in the feature information of the sea ice is converted into the second evidence information describing the category of the sea ice, and the category information of the sea ice includes the confidence of the sea ice belonging to each preset category. Optionally, the conversion of the category information can be converted according to the following formula 4:

[0090] , , ; (4)

[0091] confidence that the sea ice is small sea ice, confidence that the sea ice is medium sea ice, confidence that the sea ice is large sea ice. , and are the second evidence information converted from large sea ice, medium sea ice and small sea ice respectively.

[0092] Based on a preset fusion algorithm, the first evidence information and the second evidence information are fused to obtain a fusion feature, and a target category to which the sea ice belongs is determined according to the fusion feature, and the category information of the sea ice is updated based on the target category. Optionally, the category with the highest confidence in the category information of the sea ice is taken as an initial category of the sea ice, the target category of the sea ice is determined according to the fusion feature, and if the target category is different from the initial category of the sea ice, the category information of the sea ice is updated, and the target category of the sea ice is taken as a final category.

[0093] In one embodiment, the fusion algorithm for fusing the first evidence information and the second evidence information is as follows:

[0094] Definition ;

[0095] Let ;

[0096] ;

[0097] ;

[0098] .

[0099] Wherein, and are intermediate feature values, and the maximum value in the fusion feature ( ) corresponds to the target category of the sea ice, that is, the updated category of the sea ice.

[0100] Based on the feature information of the sea ice detected from the remote sensing image of the sea ice area, the feature information of each sea ice in the sea ice area is detected based on a pre-trained target detection model. Based on this, in step 100, the feature information of each sea ice in the sea ice area is detected based on the remote sensing image of the sea ice area, which can further include:

[0101] In step 101, the remote sensing image is input into the pre-trained target detection model to obtain feature information of each sea ice in the sea ice region output by the target detection model; wherein, the contour information in the feature information is represented by a two-dimensional array, and the two-dimensional array is used to describe the position coordinates of the edge pixel points of the sea ice contour.

[0102] The target detection model comprises a backbone network and a head network, the backbone network is built based on a spatial pyramid mechanism and a convolution attention mechanism; the backbone network comprises a first convolution layer, a residual feature extraction layer, a spatial pyramid layer and a convolution attention layer; the head network comprises an up-sampling layer, a splicing layer, a non-residual feature extraction layer and a second convolution layer.

[0103] The remote sensing image is input into the pre-trained target detection model to obtain feature information of each sea ice in the sea ice region output by the target detection model, wherein the contour information in the feature information of the sea ice is represented by a two-dimensional array, and the two-dimensional array is used to describe the position coordinates of the edge pixel points of the sea ice contour.

[0104] Further, the target detection model comprises a backbone network and a head network, the backbone network is built based on a spatial pyramid mechanism and a convolution attention mechanism, and the backbone network comprises a first convolution layer, a residual feature extraction layer, a spatial pyramid layer and a convolution attention layer, and the head network comprises an up-sampling layer, a splicing layer, a non-residual feature extraction layer and a second convolution layer.

[0105] Before the remote sensing image is input into the pre-trained target detection model, the target detection model needs to be trained, and the training process of the target detection model specifically comprises:

[0106] In step 001, a sample data set constructed based on a sea ice data set is obtained; the sample data set contains annotation information of a sea ice sample image, and the annotation information is used to annotate a sea ice category and a real region frame of the sea ice; the sea ice data set is obtained by preprocessing a sea ice image, and the preprocessing comprises angle adjustment and pixel cropping;

[0107] In step 002, a pre-set target detection model is iteratively trained based on the sample data set, and in each iteration process, a loss value of the target detection model is calculated;

[0108] In step 003, model parameters of the target detection model are updated according to the loss value until the number of iterations reaches a pre-set maximum number of iterations, or the loss value converges, and a pre-trained target detection model is obtained.

[0109] A sample data set constructed based on a sea ice data set is obtained, the sample data set containing annotation information of sea ice sample images, the annotation information of the sea ice sample images being used to annotate sea ice categories and real region frames of the sea ice. The sea ice data set is obtained by preprocessing sea ice images, and the preprocessing of the sea ice images includes but is not limited to angle adjustment and pixel cropping. The sea ice images are preprocessed to obtain corresponding sea ice sample images, and the sample data set is constructed according to the sea ice sample images and the annotation information thereof.

[0110] Further, based on the constructed sample data set, a preset target detection model is iteratively trained. In each iteration process, the loss value of the target detection model is calculated, and the model parameters of the target detection model are updated according to the calculated loss value, until the iteration number reaches a preset maximum iteration number, or the loss value converges, and a pre-trained target detection model is obtained.

[0111] The annotation information of the sea ice sample images is obtained by annotating the sea ice sample images. Specifically, a sea ice data set of satellite remote sensing images of sea ice regions is obtained, and the sea ice images in the sea ice data set are preprocessed and annotated to obtain a set of sea ice images based on visible light, i.e., a sample data set.

[0112] In one embodiment, the red, green and blue bands in the sea ice data set are extracted, and the three bands are fused at the pixel level to obtain high-resolution visible light sea ice images. Then, through preprocessing such as segmentation operation, the high-resolution visible light sea ice images are divided into image blocks of a certain size to obtain sea ice sample images. Then, the sea ice image blocks in the sea ice sample images are annotated. For example, a circumscribed rectangle of each sea ice image block is generated, and the sea ice in the circumscribed rectangle of the sea ice target is annotated as small sea ice if the longest side of the circumscribed rectangle is within 8 to 32 pixels, as medium sea ice if the longest side of the circumscribed rectangle is within 32 to 128 pixels, and as large sea ice if the longest side of the circumscribed rectangle is within 128 to 256 pixels.

[0113] Further, based on the sample data set constructed based on the annotated sea ice sample images, the target detection model is trained. In one embodiment, the target detection model is built by using a YOLOv8 network model, and the training process of the target detection model mainly includes:

[0114] Step 1, configure the environment required for model training;

[0115] Step 2, build an improved YOLOv8 network model as the target detection model, such as Figure 2As shown, in the target detection model, a convolutional block attention module (CBAM) is added to the 10th layer of the backbone network of YOLOv8. On this basis, the architecture of the head network of YOLOv8 is improved; Shape-IoU is used as the loss function of the bounding box regression, and then an evidence fusion module is integrated into the non-maximum suppression module of the target detection model. The convolutional block attention module is constructed based on the convolutional attention mechanism.

[0116] Step 3, model training and verification: the improved YOLOv8 network model is trained and verified using the sample data set constructed from the sea ice image.

[0117] The last module in the original YOLOv8 backbone network is the improved spatial pyramid pooling module (SPPF), and the CBAM is arranged after the SPPF module.

[0118] The method for CBAM to extract features is as follows:

[0119] Step 1: the first sea ice feature map F1 output by the 9th layer (i.e. the spatial pooling layer) of the target detection model is taken as the input of the convolutional block attention module. The convolutional block attention module includes a channel attention module and a spatial attention module.

[0120] Step 2: the channel attention module performs one-dimensional convolution operation on the first sea ice feature map F1 to obtain one-dimensional convolution result F1. Then, the one-dimensional convolution result is multiplied by the first sea ice feature map to obtain the second sea ice feature map F2. .

[0121] Step 3: in the spatial attention module, the second sea ice feature map F2 is taken as the input, and two-dimensional convolution operation is performed on the second sea ice feature map F2 to obtain two-dimensional convolution result F2. Then, the two-dimensional convolution result is multiplied by the second sea ice feature map to obtain the third sea ice feature map F3. .

[0122] The head network of YOLOv8 has three branches, which are respectively used for detecting small targets, medium targets and large targets. The branches for detecting medium targets and large targets are connected after the convolutional block attention module, and the branch for detecting small targets is still connected after the SPPF module.

[0123] Based on this, the process of the target detection model detecting the feature information of sea ice includes the following steps:

[0124] input the first sea ice feature map into the channel attention module to obtain a second sea ice feature map output by the channel attention module; the first sea ice feature map is output by a spatial pooling layer before the convolution attention module in the backbone network;

[0125] input the first sea ice feature map into the channel attention module to obtain a second sea ice feature map output by the channel attention module, and the method further comprises:

[0126] perform one-dimensional convolution on the first sea ice feature map; and determine a second sea ice feature map based on a one-dimensional convolution result and the first sea ice feature map.

[0127] Further, input the second sea ice feature map into the spatial attention module to obtain a third sea ice feature map output by the spatial attention module;

[0128] input the second sea ice feature map into the spatial attention module to obtain a third sea ice feature map output by the spatial attention module, and the method further comprises:

[0129] perform two-dimensional convolution on the second sea ice feature map; and determine a third sea ice feature map based on a two-dimensional convolution result and the second sea ice feature map.

[0130] Specifically, the explanation of the above is shown in formula 5. The input first sea ice feature map is subjected to two parallel average pooling AvgPool and maximum pooling MaxPool, and the shape of the feature map is 1x1xC at this time. The feature map is input into a multi-layer perception module MLP, which is also called an artificial neural network and includes an input layer, a hidden layer and an output layer. The elements of the two outputs are added one by one, and then subjected to a sigmoid (activation function) to obtain the output result of the channel attention module, i.e. the second sea ice feature map in this embodiment.

[0131] ; (5)

[0132] In this embodiment, the channel attention module in the convolution attention module performs one-dimensional convolution operation on the first sea ice feature map.

[0133] The explanation of the above is shown in formula 6. The input second sea ice feature map is subjected to two parallel average pooling AvgPool and maximum pooling MaxPool, and the shape of the feature map is HxWx1 at this time. Then, the two feature maps are spliced using a splicing operation. A 7x7 convolution is used to change the feature map into a 1-channel feature map. A sigmoid operation is used on the 1-channel feature map to obtain the output result of the spatial attention module, i.e. the third sea ice feature map in this embodiment.

[0134] ; (6)

[0135] The second sea ice feature map is subjected to a two-dimensional convolution operation by a spatial attention module in the convolution attention module.

[0136] In the training process of the target detection model, a shape intersection over union (Shape-IoU) is used to construct a loss function of the target detection model, which is determined based on a predicted region frame of the target detection model and a real region frame of a labeled sea ice sample image.

[0137] Specifically, the training method of the improved YOLOv8 model is as follows: the labeled sea ice sample images are divided into a training set and a validation set according to a certain proportion (for example, 7:3), a training coefficient (for example, 300) is set, a certain number of sea ice sample images (for example, 16) are input each time, and the weight parameters of the model after training are saved after training.

[0138] As described above, in the training process of the target detection model, the intersection over union of the predicted region frame and the real region frame of the sea ice sample image is used as the loss function to calculate the loss value of the target detection model. Based on this, the calculation process of the loss value in the training process of the target detection model mainly includes the following steps:

[0139] Step 10, determining first size data of the predicted region frame; the first size data includes a first center point and a first size;

[0140] Step 20, determining second size data of the real region frame; the second size data includes a second center point and a second size;

[0141] Step 30, determining third size data of an external rectangle of the predicted region frame and the real region frame; the third size data includes a horizontal length, a vertical length, and a scale parameter; the external rectangle is a maximum external matrix of the predicted region frame and the real region frame;

[0142] Step 40, calculating the loss value of the target detection model based on the first size data, the second size data, and the third size data.

[0143] Specifically, the loss function of the target detection model is constructed based on the shape intersection over union Shape-IoU, and the loss value is calculated based on the loss function, as follows:

[0144] The first center point of the predicted region frame Anchor is defined as (x, y), the horizontal length is , and the vertical length is , that is, the first size data in this embodiment.

[0145] The second center point of the ground truth region frame Ground_Truth is defined as (x, y) , the horizontal length is , and the vertical length is , that is, the second size data in the embodiment.

[0146] The horizontal length of the maximum circumscribed rectangle of the two edge frames of the predicted region frame and the ground truth region frame is defined as W, the vertical length is H, and the scale parameter scale can be adjusted according to actual conditions, and the value range is: scale≥1, to obtain the third size data in the embodiment.

[0147] The loss function constructed based on the shape IoU is beneficial to accelerate the convergence speed of the target detection model, thereby improving the detection accuracy of the target detection model.

[0148] Further, according to the first size data, the second size data and the third size data, the process of calculating the loss value of the target detection model is as shown in the following formulas 7 to 12.

[0149] ; (7)

[0150] ; (8)

[0151] ; (9)

[0152] ; (10)

[0153] ; (11)

[0154] ; (12)

[0155] wherein, represents the predicted region frame, represents the ground truth region frame, is a loss function for calculating the loss value.

[0156] The shape IoU is a loss function that comprehensively considers the relative position difference between the center points of the predicted region frame and the ground truth region frame, and the relative shape difference between the predicted region frame and the ground truth region frame. In addition, the shape IoU calculates a regression coefficient from the ground truth region frame, which is used to refine the regression process. The specific regression direction of the predicted region frame is from the short side direction of the ground truth region frame or from the long side direction of the ground truth region frame. The loss function constructed by the shape IoU improves the detection accuracy of the target detection model.

[0157] The first size data includes a first horizontal length and a first vertical length of the predicted region frame, the second size data includes a second horizontal length and a second vertical length of the real region frame, a first center point of the predicted region frame is represented by first position coordinates, and a second center point of the real region frame is represented by second position coordinates. Based on this, in step 002, in each iteration process, the loss value of the target detection model is calculated, specifically including:

[0158] In step 021, in each iteration process, the predicted region frame of the sea ice region in the sea ice sample image output by the target detection model is obtained, and the shape intersection-over-union of the predicted region frame and the real region frame is calculated.

[0159] In step 022, the first horizontal length, the first vertical length of the predicted region frame, and the first position coordinates of the center point of the predicted region frame are obtained, and the second horizontal length, the second vertical length of the real region frame, and the second position coordinates of the center point of the real region frame are obtained.

[0160] In step 023, according to the shape intersection-over-union, the first horizontal length, the first vertical length, the first position coordinates, the second horizontal length, the second vertical length, and the second position coordinates, the regression coefficient of the predicted region frame is calculated; the regression coefficient is used to represent the regression direction of the predicted region frame relative to the real region frame.

[0161] In step 024, based on the shape intersection-over-union and the regression coefficient, the loss value of the target detection model is calculated.

[0162] In each iteration process of model training, the predicted region frame of the sea ice in the sea ice sample image output by the target detection model is obtained, and the shape intersection-over-union of the predicted region frame and the real region frame labeled by the sea ice sample image is calculated.

[0163] Further, the first size data of the predicted region frame and the first position coordinates of the center point thereof are obtained, and the second size data of the real region frame and the second position coordinates of the center point thereof are obtained. The first size data includes a first horizontal length and a first vertical length of the predicted region frame, and the second size data includes a second horizontal length and a second vertical length of the real region frame.

[0164] According to the calculated shape intersection-over-union, the first size data, the second size data, the first position coordinates, and the second position coordinates, the regression coefficient of the predicted region frame is calculated, which is used to represent the regression direction of the predicted region frame relative to the real region frame. Based on the shape intersection-over-union and the regression coefficient of the target detection model, the loss value of the target detection model is calculated.

[0165] In one embodiment, during the training process of the target detection model, non-maximum suppression is used to update the model parameters, specifically including the following steps:

[0166] Step 50, input the longest side of the predicted region frame into the non-maximum suppression module, and determine a first uncertainty evidence and a second uncertainty evidence based on the longest side of the predicted region frame; the first uncertainty evidence is an evidence for describing the uncertainty of the bounding box of sea ice; the second uncertainty evidence is an evidence for describing the class uncertainty of sea ice;

[0167] Step 60, perform fusion reasoning on the first uncertainty evidence and the second uncertainty evidence to obtain a new predicted class of the predicted region frame;

[0168] Step 70, update the model parameters of the target detection model according to the new predicted class of the predicted region frame.

[0169] Specifically, the present application integrates evidence fusion into the non-maximum suppression of YOLOv8, adjusts the predicted sea ice size class information, and improves the non-maximum suppression module of the original YOLOv8. The evidence fusion module is added in the non-maximum suppression module. The non-maximum suppression module of the original YOLOv8 is used to eliminate redundant detection frames in the prediction result, and then the final prediction result is obtained.

[0170] Further, on this basis, the present application extracts the prediction result obtained by the original YOLOv8 in the non-maximum suppression module, including the position of the predicted region frame, the predicted class, and processes the position and class of the predicted region frame. For the position information of the predicted region frame, the longest side of the predicted region frame is determined l , which is converted into two types of evidence for describing uncertainty, a and b.

[0171] The above two types of evidence are fused according to the improved DSmT (Dezert-Smarandache Theory, an algorithm for processing information fusion) fusion reasoning algorithm to obtain a new predicted class of sea ice size. By integrating the evidence fusion module into the non-maximum suppression module, the sea ice size prediction class is adjusted, and the detection accuracy of sea ice size is improved in the model training stage and the model application stage, thereby further improving the path planning and navigation accuracy of the ship in the sea ice area.

[0172] By adding a convolution attention module to the target detection model and correspondingly improving the construction of the head network of the target detection model, the target detection model is guided to extract the features of targets of different sizes and types evenly; in the model loss, a shape intersection over union boundary regression loss function is introduced to reduce the error of the target detection model in detecting the shape and position of the target, thereby improving the accuracy of ship navigation in the sea ice area.

[0173] The pre-trained target detection model is used to detect the contour information and size category of the sea ice in the remote sensing image of the sea ice area to be passed through by the ship after the remote sensing image of the sea ice area to be passed through by the ship is obtained. The remote sensing image of the sea ice area obtained includes one or more, and when multiple, the remote sensing images of the multiple sea ice areas form a sea ice image set. The remote sensing image in the sea ice image set is a visible light-based sea ice image obtained from satellite remote sensing data of the sea ice in a specified area, and the sea ice image contains area positioning information, for example, four vertices of a rectangular sea ice image correspond to four latitude and longitude information. Through an image-to-real-world distance conversion algorithm, the real distance of a certain sea ice in the sea ice image from the image vertex can be obtained, and further based on the latitude and longitude of the image vertex, the coordinates of the sea ice in the real world in the image can be calculated.

[0174] The sea ice size can be divided into small, medium and large, and the medium and large sizes are considered as the sea ice sizes affecting the navigation of the ship. Therefore, by calculating the positions of the medium and large sea ice in the sea ice image, the navigation path and navigation information of the ship are planned, the navigation information includes the position information of the medium sea ice and the position information of the large sea ice, and the safe driving path of the ship is planned through the acquisition of the sea ice position.

[0175] In the embodiment, by obtaining the remote sensing image of the sea ice and performing preprocessing such as labeling, a sample data set is constructed, the target detection model is trained, the pre-trained target detection model is used to detect the remote sensing image of the sea ice area to be passed through by the ship, the detection result of the sea ice is obtained, including the contour of the sea ice and the category of the size of the sea ice, and based on this, the path of the ship is planned and the navigation is performed, the efficiency of the ship sailing in the sea ice area is improved, and the safety of the ship sailing is ensured.

[0176] Further, when the path of the ship is planned, a density clustering algorithm is used to cluster the medium and large size sea ice affecting the navigation of the ship according to the passable distance of the ship, so as to reduce the calculation amount when the path is planned, improve the efficiency of the path planning, and meet the real-time demand of the navigation of the ship. Moreover, for the planning of the path of the ship, a combination of global optimization and local optimization is adopted to ensure that the ship can pass through the sea ice area at the shortest distance.

[0177] The ship navigation device in the sea ice area provided by the present application is described below, and the ship navigation device in the sea ice area described below can be referred to each other in the description of the ship navigation method in the sea ice area described above.

[0178] Reference Figure 3 The ship navigation device in the sea ice area provided by the embodiment of the present application comprises:

[0179] The sea ice detection module 10 is configured to acquire a remote sensing image of a sea ice area to be passed by a target ship, and detect feature information of each sea ice in the sea ice area based on the remote sensing image; the feature information includes contour information and category information, and the category information is used to represent a size of the sea ice.

[0180] The sea ice clustering module 20 is configured to cluster the sea ice according to the feature information, to obtain a sea ice cluster.

[0181] The path planning and navigation module 30 is configured to perform path planning for the target ship based on the sea ice cluster, and perform navigation for the target ship according to a planned target path.

[0182] In an embodiment, the category information at least includes large sea ice, medium sea ice and small sea ice; and the sea ice clustering module 20 is further configured to:

[0183] select a sea ice set in the sea ice area according to the category information; the sea ice set includes large sea ice and medium sea ice;

[0184] calculate a shortest distance between a first sea ice and each second sea ice based on the contour information; the first sea ice is any one in the sea ice set, and the second sea ice is any one in the sea ice set except the first sea ice;

[0185] perform density-based clustering processing on the sea ice in the sea ice set according to the shortest distance and a minimum turning radius of the target ship, to obtain a sea ice cluster; wherein a shortest distance between adjacent sea ice in the same sea ice cluster is less than the minimum turning radius.

[0186] In an embodiment, the path planning and navigation module 30 is further configured to:

[0187] acquire a maximum Feret diameter of an ice cluster area corresponding to each sea ice cluster, and classify the sea ice cluster into a medium sea ice cluster and a large sea ice cluster according to the maximum Feret diameter of the ice cluster area;

[0188] perform globally optimal obstacle avoidance path planning on the large sea ice cluster and locally optimal obstacle avoidance path planning on the medium sea ice cluster based on a start position and an end position of the target ship, to obtain a path array; the path array is used to represent a path distance between key nodes, and the key nodes are selected from edge pixel points of the medium sea ice cluster and the large sea ice cluster in path planning;

[0189] perform optimal path planning on the target ship by using a single-source shortest path algorithm based on the path array.

[0190] In one embodiment, the sea ice detection module 10 is further configured to:

[0191] obtain a maximum Feret diameter of each sea ice, and convert the maximum Feret diameter of the sea ice into first evidence information describing a sea ice profile;

[0192] convert the category information in the feature information into second evidence information describing a sea ice category; the category information includes a confidence level of the sea ice belonging to each preset category;

[0193] fuse the first evidence information and the second evidence information based on a preset fusion algorithm to obtain a fusion feature;

[0194] determine a target category to which the sea ice belongs according to the fusion feature, and update the category information based on the target category.

[0195] In one embodiment, the sea ice detection module 10 is further configured to:

[0196] input the remote sensing image into a pre-trained target detection model to obtain feature information of each sea ice in the sea ice region output by the target detection model; wherein the profile information in the feature information is represented by a two-dimensional array, and the two-dimensional array is used to describe the position coordinates of edge pixel points of the sea ice profile;

[0197] The target detection model includes a backbone network and a head network, and the backbone network is built based on a spatial pyramid mechanism and a convolution attention mechanism; the backbone network includes a first convolution layer, a residual feature extraction layer, a spatial pyramid layer, and a convolution attention layer; and the head network includes an up-sampling layer, a concatenation layer, a non-residual feature extraction layer, and a second convolution layer.

[0198] In one embodiment, the ship navigation device of the sea ice region further includes a model training module configured to:

[0199] obtain a sample data set constructed based on a sea ice data set; the sample data set includes annotation information of a sea ice sample image, and the annotation information is used to annotate a sea ice category and a real region frame of the sea ice; the sea ice data set is obtained by preprocessing a sea ice image, and the preprocessing includes angle adjustment and pixel cropping;

[0200] iteratively train a preset target detection model based on the sample data set, and calculate a loss value of the target detection model in each iteration process;

[0201] update model parameters of the target detection model according to the loss value until a preset maximum number of iterations is reached, or the loss value converges, to obtain a pre-trained target detection model.

[0202] In one embodiment, the model training module is further configured to:

[0203] In each iteration, the predicted region bounding box of the sea ice in the sea ice sample image output by the target detection model is obtained, and the shape intersection-union ratio of the predicted region bounding box and the real region bounding box is calculated.

[0204] Obtain the first horizontal length, the first vertical length, and the first position coordinates of the center point of the predicted region box; and obtain the second horizontal length, the second vertical length, and the second position coordinates of the center point of the real region box.

[0205] The regression coefficients of the predicted region box are calculated based on the shape intersection-union ratio, the first horizontal length, the first vertical length, the first position coordinates, the second horizontal length, the second vertical length, and the second position coordinates; the regression coefficients are used to represent the regression direction of the predicted region box relative to the true region box.

[0206] The loss value of the target detection model is calculated based on the shape intersection-union ratio and the regression coefficient.

[0207] Figure 4 An example is a schematic diagram of the physical structure of an electronic device, such as... Figure 4 As shown, the electronic device may include: a processor 410, a communications interface 420, a memory 430, and a communication bus 440, wherein the processor 410, the communications interface 420, and the memory 430 communicate with each other via the communication bus 440. The processor 410 can call logical instructions from the memory 430 to execute a ship navigation method in sea ice areas, the method including:

[0208] The system acquires remote sensing images of the sea ice area through which the target vessel intends to pass, and detects feature information of each sea ice in the sea ice area based on the remote sensing images; the feature information includes contour information and category information, and the category information is used to characterize the size of the sea ice.

[0209] The sea ice is clustered based on the aforementioned feature information to obtain sea ice clusters;

[0210] The target vessel is route-planned based on the sea ice clusters, and then navigated according to the planned target route.

[0211] Moreover, the logic instructions in the memory 430 described above can be implemented in the form of software functional units and sold or used as independent products, and can be stored in a computer readable storage medium. Based on such understanding, the technical solutions of the present application essentially or the parts that contribute to the prior art or parts of the technical solutions can be embodied in the form of a software product. The computer software product is stored in a storage medium, and includes several instructions for causing a computer device (which can be a personal computer, a server, or a network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of the present application. The aforementioned storage medium includes: a U disk, a mobile hard disk, a read-only memory (ROM, Read-Only Memory), a random access memory (RAM, Random Access Memory), a magnetic disk or an optical disk, and various media that can store program codes.

[0212] In another aspect, the present application also provides a computer program product, which comprises a computer program, the computer program can be stored on a non-transitory computer readable storage medium, and the computer program can be executed by a processor to enable a computer to perform the ship navigation method in the sea ice area provided by the above-mentioned methods, and the method comprises:

[0213] obtaining a remote sensing image of a sea ice area to be passed by a target ship, and detecting feature information of each sea ice in the sea ice area based on the remote sensing image; the feature information comprises contour information and category information, and the category information is used to represent the size of the sea ice;

[0214] clustering the sea ice according to the feature information to obtain a sea ice cluster;

[0215] planning a path for the target ship based on the sea ice cluster, and navigating the target ship according to the planned target path.

[0216] In another aspect, the present application also provides a non-transitory computer readable storage medium, which stores a computer program, and the computer program is executed by a processor to implement the ship navigation method in the sea ice area provided by the above-mentioned methods, and the method comprises:

[0217] obtaining a remote sensing image of a sea ice area to be passed by a target ship, and detecting feature information of each sea ice in the sea ice area based on the remote sensing image; the feature information comprises contour information and category information, and the category information is used to represent the size of the sea ice;

[0218] clustering the sea ice according to the feature information to obtain a sea ice cluster;

[0219] Path planning is performed on the target ship based on the sea ice cluster, and the target ship is navigated according to the planned target path.

[0220] The device embodiments described above are merely illustrative, wherein the units described as separate components can or can not be physically separated, and the components displayed as units can or can not be physical units, i.e., can be located in one place or distributed on multiple network units. Part or all of the modules can be selected to achieve the purpose of the embodiment scheme according to actual needs. Those skilled in the art can understand and implement without creative labor.

[0221] Through the description of the above embodiments, those skilled in the art can clearly understand that the embodiments can be realized by means of software and the necessary general hardware platform, and of course can also be realized by hardware. Based on such understanding, the above technical solutions can be embodied in the form of a software product, which can be stored in a computer readable storage medium, such as a ROM / RAM, a magnetic disk, an optical disk, etc., and includes a plurality of instructions to make a computer device (which can be a personal computer, a server, or a network device, etc.) execute the methods described in each embodiment or some parts of the embodiments.

[0222] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, and not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: it can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement for part of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the spirit and scope of the technical solutions of the embodiments of the present application.

Claims

1. A method of navigation of a ship in a sea ice region, characterized in that, The method comprises the following steps: acquiring a remote sensing image of a sea ice area to be passed by a target ship, and detecting feature information of each sea ice in the sea ice area based on the remote sensing image; the feature information comprises contour information and category information, and the category information is used to represent the size of the sea ice; clustering the sea ice according to the feature information to obtain sea ice clusters; planning a path for the target ship based on the sea ice clusters, and navigating the target ship according to the planned target path; the path planning for the target ship based on the sea ice clusters comprises: acquiring the maximum Feret diameter of an ice cluster area corresponding to each sea ice cluster, and classifying the sea ice clusters into medium-sized sea ice clusters and large-sized sea ice clusters according to the maximum Feret diameter of the ice cluster area; performing globally optimal obstacle avoidance path planning for the large-sized sea ice clusters and locally optimal obstacle avoidance path planning for the medium-sized sea ice clusters based on the start position and the end position of the target ship to obtain a path array; the path array is used to represent the path distance between key nodes, and the key nodes are selected from edge pixel points of the medium-sized sea ice clusters and the large-sized sea ice clusters in the path planning; performing optimal path planning for the target ship based on the path array by using a single-source shortest path algorithm.

2. The method of claim 1, wherein, The category information at least comprises large-sized sea ice, medium-sized sea ice and small-sized sea ice; and the clustering of the sea ice according to the feature information to obtain sea ice clusters comprises: selecting a sea ice set in the sea ice area according to the category information; the sea ice set comprises large-sized sea ice and medium-sized sea ice; calculating the shortest distance between a first sea ice and each second sea ice based on the contour information; the first sea ice is any one in the sea ice set, and the second sea ice is any one in the sea ice set except the first sea ice; performing density-based clustering processing on the sea ice in the sea ice set according to the shortest distance and the minimum turning radius of the target ship to obtain sea ice clusters; wherein the shortest distance between adjacent sea ice in the same sea ice cluster is less than the minimum turning radius.

3. The method of claim 1, wherein, After the feature information of each sea ice in the sea ice area is detected based on the remote sensing image, the method further comprises: acquiring the maximum Feret diameter of each sea ice, and converting the maximum Feret diameter of the sea ice into first evidence information describing the contour of the sea ice; converting the category information in the feature information into second evidence information describing the category of the sea ice; the category information comprises a confidence degree of the sea ice belonging to each preset category; fusing the first evidence information and the second evidence information based on a preset fusion algorithm to obtain fused features; determining a target category to which the sea ice belongs according to the fused features, and updating the category information based on the target category.

4. The method of claim 1, wherein, The detection of the feature information of each sea ice in the sea ice area based on the remote sensing image comprises: inputting the remote sensing image into a pre-trained target detection model to obtain feature information of each sea ice in the sea ice region output by the target detection model; wherein, contour information in the feature information is represented by a two-dimensional array, and the two-dimensional array is used to describe position coordinates of edge pixel points of the sea ice contour; the target detection model comprises a backbone network and a head network, and the backbone network is built based on a spatial pyramid mechanism and a convolution attention mechanism; the backbone network comprises a first convolutional layer, a residual feature extraction layer, a spatial pyramid layer, and a convolution attention layer; and the head network comprises an up-sampling layer, a concatenation layer, a non-residual feature extraction layer, and a second convolutional layer.

5. The method of claim 4, wherein, Before detecting the feature information of each sea ice in the sea ice region based on the remote sensing image, the method further comprises: obtaining a sample data set constructed based on a sea ice data set; the sample data set contains annotation information of a sea ice sample image, and the annotation information is used to annotate a sea ice category and a real region frame of sea ice; the sea ice data set is obtained by preprocessing a sea ice image, and the preprocessing comprises angle adjustment and pixel cropping; iteratively training a pre-set target detection model based on the sample data set; in each iteration process, a loss value of the target detection model is calculated; updating model parameters of the target detection model according to the loss value until a maximum iteration number is reached or the loss value converges, to obtain a pre-trained target detection model.

6. The method of claim 5, wherein, In each iteration process, the loss value of the target detection model is calculated, comprising: in each iteration process, a predicted region frame of sea ice in the sea ice sample image output by the target detection model is obtained, and a shape intersection-over-union ratio of the predicted region frame and the real region frame is calculated; a first horizontal length and a first vertical length of the predicted region frame and a first position coordinate of a center point of the predicted region frame are obtained, and a second horizontal length and a second vertical length of the real region frame and a second position coordinate of a center point of the real region frame are obtained; a regression coefficient of the predicted region frame is calculated according to the shape intersection-over-union ratio, the first horizontal length, the first vertical length, the first position coordinate, the second horizontal length, the second vertical length, and the second position coordinate; the regression coefficient is used to represent a regression direction of the predicted region frame relative to the real region frame; the loss value of the target detection model is calculated based on the shape intersection-over-union ratio and the regression coefficient.

7. A ship navigation device for a sea ice region, characterized by comprising: a sea ice detection module, configured to obtain a remote sensing image of a sea ice region to be passed by a target ship, and detect feature information of each sea ice in the sea ice region based on the remote sensing image; the feature information comprises contour information and category information, and the category information is used to represent a size of the sea ice; a sea ice clustering module, configured to cluster the sea ice according to the feature information to obtain a sea ice cluster; a path planning and navigation module, configured to plan a path for the target ship based on the sea ice cluster, and navigate the target ship according to the planned target path. The path planning and navigation module is further configured to acquire a maximum Feret diameter of an ice cluster area corresponding to each of the sea ice clusters, and classify the sea ice clusters into medium-sized sea ice clusters and large-sized sea ice clusters according to the maximum Feret diameter of the ice cluster area; perform globally optimal obstacle avoidance path planning on the large-sized sea ice clusters and locally optimal obstacle avoidance path planning on the medium-sized sea ice clusters based on a start position and an end position of the target ship, to obtain a path array; the path array is used to represent path distances between key nodes, the key nodes being selected from edge pixel points of the medium-sized sea ice clusters and the large-sized sea ice clusters in path planning; and perform optimal path planning on the target ship by using a single-source shortest path algorithm based on the path array.

8. An electronic device comprising a memory, a processor, and a computer program stored on the memory and running on the processor, characterized in that, The computer program, when executed by the processor, implements the ship navigation method for a sea ice area as claimed in any one of claims 1 to 6. 9.A non-transitory computer-readable storage medium having stored thereon a computer program, characterized in that, The computer program, when executed by the processor, implements the ship navigation method for a sea ice area as claimed in any one of claims 1 to 6.

Images