Ship body identification method and electronic device

Abstract

The application discloses a ship body identification method and an electronic device, including: acquiring reference point cloud data of a target area, and determining a plurality of pixel grids in an overhead view based on the reference point cloud data; counting point cloud feature values of each pixel grid in a plurality of feature dimensions, the point cloud feature values being used to represent features of point clouds in the pixel grid; according to the point cloud feature values of each pixel grid in the plurality of feature dimensions and whether the reference point cloud data in each pixel grid contains ship body point clouds, a target feature dimension is screened from the plurality of feature dimensions; and based on the target feature dimension, a ship body in the target area is identified. The scheme filters a plurality of feature dimensions existing in point cloud data according to different target areas, identifies ship body point clouds in the target area based on the filtered feature dimensions, and improves the accuracy of ship body identification.

Description

TECHNICAL FIELD

[0001] The present application belongs to the technical field of ship body identification, and particularly relates to a ship body identification method and an electronic device. BACKGROUND

[0002] There are mainly two methods for object detection based on laser radar point cloud.

[0003] One is to divide the point cloud space into a voxel grid, and then directly extract features on the voxel using a convolutional neural network. The advantage of this method is to fully utilize the three-dimensional spatial geometric information of the point cloud. However, the computational complexity of this method is very high, and it is very difficult to apply in practice. The second method is to project the point cloud into a depth dimension to generate a depth feature map. This method can produce regular and ordered depth image data, but due to the perspective changes of objects at different distances, the size of the object on the depth image changes in a large range, which makes it difficult for CNN methods to detect objects at a distance. It can be seen that the above two schemes are not suitable for object detection, especially for ship body identification scenarios with high precision requirements.

[0004] Therefore, how to improve the accuracy of ship body identification is a technical problem to be solved. SUMMARY

[0005] Embodiments of the present application provide a ship body identification method and an electronic device. For different target areas, the accuracy of ship body identification is improved by screening a plurality of feature dimensions existing in the point cloud data, and identifying the ship body point cloud in the target area based on the screened feature dimensions.

[0006] Other characteristics and advantages of the present application will become apparent from the following detailed description, or will be learned by practice of the present application.

[0007] According to a first aspect of an embodiment of the present application, a ship body identification method is provided, the method comprising:

[0008] acquiring reference point cloud data of a target area, and determining a plurality of pixel grids in an overhead view based on the reference point cloud data;

[0009] counting point cloud feature values of each pixel grid in a plurality of feature dimensions, the point cloud feature values being used to represent the features of the point cloud in the pixel grid;

[0010] screening a target feature dimension from the plurality of feature dimensions according to the point cloud feature values of each pixel grid in the plurality of feature dimensions, and whether the reference point cloud data in each pixel grid contains ship body point cloud;

[0011] identifying a ship body in the target area based on the target feature dimension.

[0012] In some embodiments of this application, the step of filtering target feature dimensions from the multiple feature dimensions based on the point cloud feature values ​​of each pixel grid across the multiple feature dimensions and whether the reference point cloud data in each pixel grid contains hull point clouds includes:

[0013] Define the plurality of feature dimensions as candidate feature dimensions;

[0014] The target feature dimension is determined from the candidate feature dimension based on the point cloud feature values ​​of each pixel grid in the candidate feature dimension and whether the reference point cloud data in each pixel grid contains the hull point cloud.

[0015] If the number of dimensions of the target feature dimension exceeds the set number of dimensions, then the target feature dimension is used as a new candidate feature dimension, and the process returns to iteratively executing the step of determining the target feature dimension from the candidate feature dimensions based on the point cloud feature values ​​of each pixel grid on the candidate feature dimension and whether the reference point cloud data in each pixel grid contains the hull point cloud, until the number of dimensions of the target feature dimension meets the set number of dimensions.

[0016] In some embodiments of this application, determining the target feature dimension from the candidate feature dimensions based on the point cloud feature values ​​of each pixel grid in the candidate feature dimension, and whether the reference point cloud data in each pixel grid contains hull point clouds, includes:

[0017] The point cloud feature matrix is ​​determined based on the point cloud feature values ​​of each pixel grid in the candidate feature dimension.

[0018] Each pixel is labeled based on whether the reference point cloud data in each pixel contains the hull point cloud, so as to determine the label vector;

[0019] The point cloud feature matrix and the label vector are fitted to obtain the regression coefficients of each feature dimension. The absolute value of the regression coefficient is used to characterize the contribution of the corresponding feature dimension to whether the reference point cloud data in the pixel grid contains the hull point cloud.

[0020] The target feature dimension is determined from the candidate feature dimensions based on the regression coefficients of each feature dimension.

[0021] In some embodiments of this application, determining the target feature dimension from the candidate feature dimensions based on the regression coefficients of each feature dimension includes:

[0022] Based on the number of dimensions of the multiple feature dimensions, determine the number of target feature dimensions to retain;

[0023] Based on the number of target feature dimensions retained, and using the regression coefficients of each feature dimension, the target feature dimension is determined from the candidate feature dimensions.

[0024] In some embodiments of this application, determining the number of target feature dimensions to retain based on the number of dimensions of the plurality of feature dimensions includes:

[0025] If the number of dimensions of the multiple feature dimensions is less than or equal to the number of dimensions threshold, then the number of target feature dimensions to be retained is determined according to the set number of feature dimensions to be removed.

[0026] If the total number of feature dimensions exceeds the threshold, the number of feature dimensions to be removed is calculated using an exponential decay method. Based on the calculated number of feature dimensions to be removed, the number of target feature dimensions to be retained is determined.

[0027] In some embodiments of this application, determining the target feature dimension from the candidate feature dimensions based on the regression coefficients of each feature dimension according to the number of retained target feature dimensions includes:

[0028] Sort the regression coefficients of each feature dimension in descending order of absolute value;

[0029] The candidate feature dimensions with the highest number of retained items are determined as the target feature dimensions.

[0030] In some embodiments of this application, identifying the ship hull in the target region based on the target feature dimension includes:

[0031] Acquire the point cloud data to be identified in the target area;

[0032] The ship hull is identified based on the point cloud data to be identified and the target feature dimensions with the set number of dimensions.

[0033] In some embodiments of this application, the number of dimensions of the target feature dimension is three, and the identification of the ship hull based on the point cloud data to be identified and the target feature dimension includes:

[0034] Obtain the point cloud feature values ​​of each pixel grid corresponding to the point cloud data to be identified in the three target feature dimensions;

[0035] The point cloud feature values ​​in the three target feature dimensions are converted into RGB three-channel color information.

[0036] Based on the RGB three-channel color information, an image to be identified is generated, and the ship hull of the target area is identified using the image to be identified.

[0037] According to a second aspect of the embodiments of this application, a ship hull identification method is provided, the method comprising:

[0038] The method described in any of the first aspects above determines multiple sets of target feature dimensions, wherein each set of target feature dimensions includes at least one target feature dimension;

[0039] The target feature dimensions of each group are validated based on the validation data, and the root mean square error corresponding to each target feature dimension is calculated.

[0040] The ship hull in the target region is identified based on the set of target feature dimensions that minimize the root mean square error.

[0041] According to a third aspect of the present application, a hull identification device is provided, applied to the method described in any of the above embodiments, the device comprising:

[0042] The acquisition unit is used to acquire reference point cloud data of the target area and point cloud data to be identified in the target area.

[0043] A data processing unit is used to statistically analyze the point cloud feature values ​​of each pixel grid in multiple feature dimensions, wherein the point cloud feature values ​​are used to characterize the features of the point cloud within the pixel grid.

[0044] An iterative unit is used to filter target feature dimensions from the multiple feature dimensions based on the point cloud feature values ​​of each pixel grid in the multiple feature dimensions, and whether the reference point cloud data in each pixel grid contains hull point cloud.

[0045] The identification unit is used to identify the ship hull in the target region based on the target feature dimension.

[0046] According to a fourth aspect of the present application, a hull identification device is provided, applied to the method described in any of the above embodiments, the device comprising:

[0047] The determining unit is configured to determine multiple sets of target feature dimensions according to the method described in any of the above embodiments;

[0048] The analysis unit verifies the target feature dimensions of each group based on the validation data and calculates the root mean square error corresponding to each target feature dimension.

[0049] The identification unit is used to identify the ship hull in the target region based on the set of target feature dimensions that minimize the root mean square error.

[0050] According to a fifth aspect of the present application, a ship identification electronic device is provided, the electronic device including one or more processors and one or more memories, the one or more memories storing at least one piece of program code, the at least one piece of program code being loaded and executed by the one or more processors to implement the method described in any of the above embodiments.

[0051] By using some of the embodiments proposed in this application, the point cloud feature values ​​of each pixel grid on the multiple feature dimensions, and whether the reference point cloud data in each pixel grid contains the hull point cloud, the inherent feature dimensions of the pixel grid corresponding to the point cloud data are filtered out. This can filter out the target feature dimensions related to the accuracy of hull recognition, thereby reducing the influence of human subjective judgment on hull recognition during the process of recognizing the hull in the target area, and thus improving the accuracy of hull recognition to a certain extent.

[0052] The beneficial effects of the embodiments of the second to fifth aspects described above can be referred to the beneficial effects of the first aspect and the embodiments of the first aspect described above, and will not be repeated here.

[0053] It should be understood that the above general description and the following detailed description are exemplary and explanatory only, and do not limit this application. Attached Figure Description

[0054] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application. It is obvious that the drawings described below are merely some embodiments of this application, and those skilled in the art can obtain other drawings based on these drawings without any inventive effort. In the drawings:

[0055] Figure 1 A detailed flowchart of a ship hull identification method according to an embodiment of this application is shown;

[0056] Figure 2 This is a top view schematic diagram of the reference hull point cloud data within the target area in an embodiment of this application;

[0057] Figure 3 The following is a detailed flowchart illustrating how, in an embodiment of this application, a target feature dimension is selected from multiple feature dimensions based on the point cloud feature values ​​of each pixel grid in multiple feature dimensions, and whether the reference point cloud data in each pixel grid contains the hull point cloud.

[0058] Figure 4The following is a detailed flowchart illustrating how, in an embodiment of this application, the target feature dimension is determined from the candidate feature dimension based on the point cloud feature values ​​of each pixel grid in the candidate feature dimension, and whether the reference point cloud data in each pixel grid contains the hull point cloud.

[0059] Figure 5 The flowchart illustrating the detailed process of determining the target feature dimension from candidate feature dimensions based on regression coefficients of each feature dimension in an embodiment of this application is shown.

[0060] Figure 6 The flowchart illustrates the details of determining the number of target feature dimensions to be retained based on the number of dimensions of multiple feature dimensions in an embodiment of this application.

[0061] Figure 7 The flowchart illustrates the detailed process in this application embodiment of determining the target feature dimension from candidate feature dimensions based on the regression coefficients of each feature dimension, according to the number of retained items of the target feature dimension.

[0062] Figure 8 A detailed flowchart illustrating the identification of ship hulls in a target region based on target feature dimensions is shown in an embodiment of this application.

[0063] Figure 9 This document illustrates a detailed flowchart of how the ship hull is identified based on the point cloud data to be identified and the target feature dimensions in an embodiment of this application.

[0064] Figure 10 A detailed flowchart of a ship hull identification method according to an embodiment of this application is shown;

[0065] Figure 11 A block diagram of a ship identification device according to an embodiment of this application is shown;

[0066] Figure 12 A block diagram of another hull identification device in an embodiment of this application is shown;

[0067] Figure 13 A schematic diagram of the structure of the ship identification electronic device in an embodiment of this application is shown. Detailed Implementation

[0068] The technical solutions of the embodiments of this application will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of this application, and not all embodiments. Based on the embodiments of this application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of this application.

[0069] Furthermore, the described features, structures, or characteristics can be combined in any suitable manner in one or more embodiments. Numerous specific details are provided in the following description to give a thorough understanding of embodiments of this application. However, those skilled in the art will recognize that the technical solutions of this application can be practiced without one or more of the specific details, or other methods, components, apparatuses, steps, etc., can be employed. In other instances, well-known methods, apparatuses, implementations, or operations are not shown or described in detail to avoid obscuring various aspects of this application.

[0070] The block diagrams shown in the accompanying drawings are merely functional entities and do not necessarily correspond to physically independent entities. That is, these functional entities can be implemented in software, in one or more hardware modules or integrated circuits, or in different network and / or processor devices and / or microcontroller devices.

[0071] The flowcharts shown in the accompanying drawings are merely illustrative and do not necessarily include all content and operations / steps, nor do they necessarily have to be performed in the described order. For example, some operations / steps can be broken down, while others can be combined or partially combined; therefore, the actual execution order may change depending on the specific circumstances.

[0072] In the description of this application, it should be understood that the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance or implicitly specifying the number of indicated technical features. Therefore, a feature defined as "first" or "second" may explicitly or implicitly include one or more of that feature. In the description of this application, unless otherwise stated, "multiple" means two or more.

[0073] The implementation details of the technical solutions in the embodiments of this application are described in detail below:

[0074] In a first aspect, embodiments of this application disclose a method for identifying ship hulls.

[0075] Reference Figure 1 The diagram shows a detailed flowchart of a hull identification method according to an embodiment of this application.

[0076] Reference Figure 1 As shown, the hull identification method disclosed in the embodiments of this application can be executed by a device with computing processing capabilities. The hull identification method includes at least steps S1 to S4, which are described in detail below:

[0077] In S1, reference point cloud data of the target area is acquired, and multiple pixel grids are determined in the top view based on the reference point cloud data.

[0078] To enable those skilled in the art to better understand the point cloud data processing process employed by the inventors in proposing this invention, the following will be combined with... Figure 2 A brief explanation is provided:

[0079] Reference Figure 2 As shown, a top-view schematic diagram of the reference hull point cloud data in the target area of ​​this application embodiment is shown. In the process of dividing the reference hull point cloud data into pixel grids B, the distribution range of all point clouds in each pixel grid is actually a three-dimensional space. In the top-view schematic diagram, each pixel grid that appears as a rectangle may contain not only several point clouds A, but also several point clouds directly below each point cloud. Some pixel grids may not contain any point clouds.

[0080] It should be noted that the pixel grid mentioned in step S1 refers to the pixel grid of point cloud data. Furthermore, multiple pixel grids can be a set of pixel grids containing point cloud data in a single frame of point cloud data, or a set of pixel grids from multiple frames of point cloud data collected for the same target area.

[0081] In this application, the number of pixel grids in step S1 can be represented as n.

[0082] Continue to refer to Figure 1 In S2, the point cloud feature values ​​of each pixel grid are statistically analyzed in multiple feature dimensions. The point cloud feature values ​​are used to characterize the features of the point cloud within the pixel grid.

[0083] It should be understood that the feature dimensions mentioned in step S2 may include the maximum height value, the intensity value corresponding to the maximum height value, the minimum height value, the intensity value corresponding to the minimum height value, the average height value, the average intensity value, the number of points, the median of height values, the variance of height values, and other feature dimensions of all point clouds in each pixel grid along the direction perpendicular to the top view plane. The point cloud feature value is the value of the point cloud data corresponding to each of the above feature dimensions.

[0084] It should be noted that the number of multiple feature dimensions in step S2 can be represented as m.

[0085] Continue to refer to Figure 1 In S3, the target feature dimension is selected from multiple feature dimensions based on the point cloud feature values ​​of each pixel grid in multiple feature dimensions, and whether the reference point cloud data in each pixel grid contains the hull point cloud.

[0086] It should be understood that the target feature dimension mentioned in step S3 can be understood as several feature dimensions or one feature dimension that are closely related to whether the reference point cloud data of this target area contains the hull point cloud.

[0087] It is important to understand that by selecting the target feature dimension based on the point cloud feature values ​​of each pixel grid across multiple feature dimensions, and whether the reference point cloud data in each pixel grid contains the hull point cloud, the target feature dimension can be selected. This allows for the selection of the target feature dimension with the highest correlation to the accuracy of hull recognition, thereby reducing the impact of human subjective judgment on hull recognition and thus improving the accuracy of hull recognition to a certain extent.

[0088] In some embodiments of this application, target feature dimensions are filtered from multiple feature dimensions based on the point cloud feature values ​​of each pixel grid across multiple feature dimensions, and whether the reference point cloud data in each pixel grid contains hull point clouds. This can be done as follows: Figure 3 Perform the steps shown.

[0089] Reference Figure 3 The diagram illustrates a detailed flowchart of how, in an embodiment of this application, a target feature dimension is selected from multiple feature dimensions based on the point cloud feature values ​​of each pixel grid across multiple feature dimensions, and whether the reference point cloud data in each pixel grid contains the hull point cloud. Specifically, steps S31 to S33 are included.

[0090] S31. Define multiple feature dimensions as candidate feature dimensions.

[0091] S32. Determine the target feature dimension from the candidate feature dimensions based on the point cloud feature values ​​of each pixel grid in the candidate feature dimension, and whether the reference point cloud data in each pixel grid contains the hull point cloud.

[0092] It should be understood that the candidate feature dimension mentioned in step S31 is the feature dimension selected in each filtering cycle, and the target feature dimension mentioned in step S32 is the result of the selected feature dimension in each filtering cycle.

[0093] In some embodiments of this application, target feature dimensions are filtered from multiple feature dimensions based on the point cloud feature values ​​of each pixel grid across multiple feature dimensions, and whether the reference point cloud data in each pixel grid contains hull point clouds. This can be done as follows: Figure 4 Perform the steps shown.

[0094] Reference Figure 4 The diagram illustrates a flowchart of a process in this application embodiment for filtering target feature dimensions based on point cloud feature values ​​of each pixel grid across multiple feature dimensions, and whether the reference point cloud data in each pixel grid contains hull point clouds. The flowchart specifically includes steps S321 to S324:

[0095] S321. Determine the point cloud feature matrix based on the point cloud feature values ​​of each pixel grid in the candidate feature dimension.

[0096] S322. Based on whether the reference point cloud data in each pixel cell contains the hull point cloud, mark each pixel cell to determine the marking vector.

[0097] S323. Fit the point cloud feature matrix and the label vector to obtain the regression coefficients of each feature dimension. The absolute value of the regression coefficient is used to characterize the contribution of the corresponding feature dimension to whether the reference point cloud data in the pixel grid contains the hull point cloud, that is, to characterize the correlation between the corresponding feature dimension and whether the reference point cloud data in the pixel grid contains the hull point cloud.

[0098] S324. Based on the regression coefficients of each feature dimension, determine the target feature dimension from the candidate feature dimensions.

[0099] It should be noted that the point cloud feature matrix described in step S321 can be represented as m×n.

[0100] It should be noted that in step S322, each pixel is marked according to whether the reference point cloud data in each pixel contains the hull point cloud. Specifically, pixel pixels containing the hull point cloud can be marked as 1, and pixel pixels without the hull point cloud can be marked as 0. Furthermore, the marking vector mentioned in step S322 can be represented as n×1.

[0101] In this application, step S323 can involve performing least-squares fitting on the point cloud feature matrix and the label vector. This has the advantage of ensuring the accuracy of the regression coefficients for each feature dimension. Furthermore, the regression coefficients described in step S323 can be labeled as m×1.

[0102] In some embodiments of this application, the target feature dimension is determined from the candidate feature dimensions based on the regression coefficients of each feature dimension, which can be done as follows: Figure 5 Perform the steps shown.

[0103] Reference Figure 5 The diagram illustrates a flowchart of determining the target feature dimension from candidate feature dimensions based on regression coefficients of various feature dimensions in an embodiment of this application. Specifically, it includes steps S3241 to S3242:

[0104] S3241. Based on the number of dimensions of multiple feature dimensions, determine the number of target feature dimensions to retain.

[0105] S3242. Based on the number of target feature dimensions to be retained, determine the target feature dimension from the candidate feature dimensions according to the regression coefficients of each feature dimension.

[0106] In some embodiments of this application, the number of target feature dimensions to be retained is determined based on the number of dimensions of multiple feature dimensions, which can be done as follows: Figure 6Perform the steps shown.

[0107] Reference Figure 6 The diagram illustrates a detailed flowchart of determining the number of target feature dimensions to retain based on the number of dimensions in multiple feature dimensions, as shown in this embodiment of the application. Specifically, it includes steps S32421 to S32422:

[0108] S32411. If the number of dimensions of multiple feature dimensions is less than or equal to the number of dimensions threshold, then the number of target feature dimensions to be retained is determined by eliminating the set number of feature dimensions.

[0109] S32412. If the number of dimensions of multiple feature dimensions is greater than the threshold for the number of dimensions, the number of feature dimensions to be removed is calculated by exponential decay. Based on the calculated number of feature dimensions to be removed, the number of target feature dimensions to be retained is determined.

[0110] It should be understood that the feature dimension set in step S32411 can be one feature dimension or three feature dimensions. The threshold for the number of dimensions in step S32411 can be 10. The purpose of the threshold for the number of dimensions is to accommodate different numbers of m feature dimensions set for different target regions. It is divided into cases where the number of feature dimensions m is large and cases where the number of feature dimensions m is small. When the number of feature dimensions is small, one or two are eliminated in one iteration (a small number of eliminations), which can make the analysis more accurate.

[0111] In this application, the exponential decay method is as follows: First, all feature dimensions are used for iteration to determine the... i The number of dimensions used for each specific feature dimension is determined according to the following formula:

[0112]

[0113]

[0114] in, N Characterizes the number of iterations. k Characterizing the parameters of the exponential decay function, The integer part represents the first i The number of dimensions of the feature dimension is retained in the next iteration.

[0115] It is important to understand that using an exponential decay method to calculate the number of feature dimensions to be removed can quickly remove a large number of feature dimensions in the early stages of iteration, while removing fewer feature dimensions each time in the later stages of iteration. Using this method to remove the number of feature dimensions can make the analysis of the correlation between target feature dimensions and ship identification more accurate.

[0116] In some embodiments of this application, the target feature dimension is determined from the candidate feature dimensions based on the regression coefficients of each feature dimension, according to the number of target feature dimensions retained. This can be done as follows: Figure 7 Perform the steps shown.

[0117] Reference Figure 7 The diagram illustrates a detailed flowchart of how, in an embodiment of this application, the target feature dimension is determined from candidate feature dimensions based on the regression coefficients of each feature dimension and the number of retained features. Specifically, steps S32421 to S32422 are included.

[0118] S32421. Sort the absolute values ​​of the regression coefficients for each feature dimension in descending order.

[0119] S32422. Determine the candidate feature dimension with the highest number of retained items as the target feature dimension.

[0120] It should be noted that the target feature dimension mentioned in step S32422 is the feature dimension obtained after each retention and filtering.

[0121] It is important to understand that the larger the absolute value of the regression coefficient, the higher the correlation between the corresponding feature dimension and whether the reference point cloud data in each pixel grid of the target area contains the hull point cloud. The feature dimension with the larger absolute value is more correlated with whether the reference point cloud data in the pixel grid contains the hull point cloud.

[0122] Continue to refer to Figure 3 In S33, if the number of dimensions of the target feature dimension exceeds the set number of dimensions, the target feature dimension is taken as a new candidate feature dimension, and the iterative execution is returned to determine the target feature dimension from the candidate feature dimensions based on the point cloud feature values ​​of each pixel grid on the candidate feature dimension and whether the reference point cloud data in each pixel grid contains the hull point cloud, until the number of dimensions of the target feature dimension meets the set number of dimensions.

[0123] It should be understood that the definition of the number of dimensions in step S33 refers to the number of the most relevant feature dimensions selected in the final step. This number of dimensions can be a number that is defined manually. Here, in order to correspond to the image detection algorithm used in the ship hull recognition process below, we can set the number of dimensions to 3.

[0124] To enable those skilled in the art to better understand the above embodiments, a specific example will be used below to illustrate them. Figure 1 Step S3 is explained below:

[0125] A 5×4 (n×m) point cloud feature matrix is ​​as follows:

[0126]

[0127] The 5×1 (n×1) label vector is as follows:

[0128]

[0129] The 5×4 point cloud feature matrix is ​​standardized as follows:

[0130]

[0131] The standardized 5×4 matrix is ​​fitted with a 5×1 (n×1) label vector using partial least squares to obtain a 4×1 (m×1) regression coefficient vector as follows:

[0132]

[0133] After obtaining the regression coefficient vector, the values ​​in the regression coefficients are sorted in descending order of absolute value. According to step S3241, these values ​​are retained to obtain the feature dimensions corresponding to -0.37, -0.16, and -0.12, which are the minimum point, variance, and maximum point, respectively. In other words, the minimum point, variance, and maximum point are the three target feature dimensions with the highest correlation to whether the pixel grid in the reference point cloud data contains ship hull point clouds.

[0134] Continue to refer to Figure 1 In S4, the ship hull in the target area is identified based on the target feature dimension.

[0135] It is important to understand that the most relevant feature dimensions obtained above are most correlated with whether there are ship hull point clouds in the pixel grids of the reference point cloud data. Therefore, the ship hull in the target area is identified according to the aforementioned non-human-selected feature dimensions, and the identification accuracy is naturally the highest.

[0136] In some embodiments of this application, the hull in the target region is identified based on the target feature dimension, which can be done according to, as follows: Figure 8 Perform the steps shown.

[0137] Reference Figure 8 The diagram illustrates a flowchart of identifying the ship's hull in a target region based on target feature dimensions, as shown in an embodiment of this application. The flowchart specifically includes steps S41 to S42:

[0138] S41. Obtain the point cloud data to be identified in the target area.

[0139] S42. Identify the ship's hull based on the point cloud data to be identified and the target feature dimensions.

[0140] It should be noted that the point cloud data to be identified here is point cloud data within the target area collected based on the actual needs of ship hull identification.

[0141] It is important to understand that, when identifying point cloud data within a corresponding target region, the ship's hull can be identified by using a set number of target feature dimensions corresponding to that target region. This process can eliminate the influence of environmental factors such as geographical or climatic factors, thus making the ship identification process more accurate.

[0142] In some embodiments of this application, the target feature dimension has three dimensions. Based on the point cloud data to be identified and the target feature dimension, the ship hull is identified, which can be done according to the following... Figure 9 Perform the steps shown.

[0143] Reference Figure 9 The diagram illustrates a flowchart of ship hull identification based on point cloud data to be identified and target feature dimensions in an embodiment of this application, specifically including steps S421 to S422:

[0144] S421. Obtain the point cloud feature values ​​of each pixel grid corresponding to the point cloud data to be identified in the three target feature dimensions.

[0145] S422, convert the point cloud feature values ​​in the three target feature dimensions into RGB three-channel color information.

[0146] S423. Based on RGB three-channel color information, generate an image to be identified, and identify the ship hull of the target area through the image to be identified.

[0147] It should be noted that the target feature dimension has three dimensions here to facilitate conversion into RGB three-channel color information.

[0148] Specifically, the highest value of the target feature dimension point cloud feature value with a set number of dimensions in each pixel cell can be associated with 255, the lowest value of the target feature dimension point cloud feature value with a set number of dimensions in each pixel cell can be associated with 0, and the remaining values ​​of the target feature dimension point cloud feature value with a set number of dimensions in each pixel cell can be converted to values ​​between 0 and 255 proportionally.

[0149] It is important to understand that target detection algorithms can be used for ship hull feature recognition. Specifically, convolutional neural network feature extraction algorithms can be used for target feature recognition. Using convolutional neural network extraction algorithms can improve the accuracy of ship hull recognition.

[0150] The beneficial effects of this invention are as follows: In the process of identifying a ship hull using point cloud data within a target area, the lack of normal point cloud data due to weather conditions or geographical factors can lead to low accuracy in identifying the ship hull based on predetermined feature dimensions (such as maximum height, intensity value corresponding to the maximum height, and density). However, by using the point cloud feature values ​​of each pixel grid across multiple feature dimensions, and whether the reference point cloud data in each pixel grid contains ship hull point clouds, the inherent feature dimensions of the pixel grid corresponding to the point cloud data can be filtered out. This allows for the selection of target feature dimensions that are relevant to the accuracy of ship hull identification. As a result, the influence of subjective human judgment on ship hull identification can be reduced during the identification process within the target area, thereby improving the accuracy of ship hull identification to a certain extent.

[0151] Secondly, based on the same inventive concept, this application also discloses a method for identifying ship hulls.

[0152] Reference Figure 10 The diagram shows a flowchart of a ship identification method according to an embodiment of this application.

[0153] Reference Figure 10 As shown, the hull identification method disclosed in the embodiments of this application includes at least steps S100 to S300, which are described in detail below:

[0154] S100. Determine multiple sets of target feature dimensions according to any of the above methods, wherein each set of target feature dimensions includes at least one target feature dimension.

[0155] S200. Verify the target feature dimensions of each group based on the verification data, and calculate the root mean square error corresponding to each target feature dimension.

[0156] S300 identifies the ship hull in the target region based on a set of target feature dimensions that minimize the root mean square error.

[0157] It should be understood that the root mean square error mentioned in steps S200 and S300 can also be calculated using other error calculation methods.

[0158] It is important to understand that by setting different or the same number of iterations for the same iterative experiment, there may be differences in a certain target feature dimension between the retained target feature dimensions of different groups. If they are different, regression analysis is performed on the last retained target feature dimension to obtain the target feature dimension with the smallest error, thereby further improving the accuracy of ship identification.

[0159] Specifically, different iteration numbers are set. If the target feature dimension with the same number of retained dimensions is retained in multiple iterations, then the target feature dimension with the same number of retained dimensions is taken as the final result. If the target feature dimension with the same number of retained dimensions is different in multiple iterations, then 70% of the total number of pixel grids (n) of point cloud data in multiple image detection regions is taken as the modeling set, and 30% of the total number of pixel grids (n) of point cloud data in multiple image detection regions is taken as the validation set. Models are built for the modeling set according to the target feature dimension with the different number of retained dimensions and the label vectors. The target feature dimension with the different number of retained dimensions is taken into the corresponding model for prediction according to the different number of retained dimensions, and the predicted vector is obtained. The root mean square error is calculated on the predicted vector and the label vector of the validation set, and the target feature dimension with the smallest root mean square error is taken as the final result.

[0160] To enable those skilled in the art to better understand the above embodiments, the steps in the above embodiments regarding verifying each group of target feature dimensions based on verification data and calculating the root mean square error corresponding to each group of target feature dimensions will be explained below with reference to a specific example:

[0161] Assuming 100 scenes were captured, and each scene contains 2500 pixel grids representing point clouds, then n = 100 × 2500 = 250000

[0162] Assuming the number of feature dimensions m is 100, the three target feature dimensions selected after 10, 20, and 30 iterations are as follows:

[0163]

[0164] The three selected feature dimensions were used for regression analysis at different times. The 250,000 pixels of the dot cloud were randomly divided into a 70% modeling set and a 30% validation set, as follows:

[0165]

[0166] The modeling set uses the first set of target feature dimensions (lowest point, highest point, density) and the third set of target feature dimensions (lowest point, highest point, height value variance) combined with the label vector to build models respectively.

[0167] The validation set is used to predict the corresponding model using the first set of target feature dimensions (lowest point, highest point, density) and the third set of target feature dimensions (lowest point, highest point, height value variance). The result is a 75000×1 prediction vector. The root mean square error is calculated by comparing the prediction vector with the label vector of the validation set.

[0168] Assuming the root mean square error (RMSE) of the first set of target feature dimensions (lowest point, highest point, density) is 1.3, and the RMS error of the third set of target feature dimensions (lowest point, highest point, height value variance) is 1.1, then the third set of target feature dimensions (lowest point, highest point, height value variance) will be converted into color information of the three channels of the image.

[0169] Thirdly, based on the same inventive concept, this application discloses a hull identification device 1100, which can be used to execute the hull identification method of the first aspect of the above embodiments of this application. For details not disclosed in the device embodiments of this application, please refer to the embodiments of the hull identification method of the first aspect of this application.

[0170] Reference Figure 11 The diagram shown is a block diagram of the ship identification device 1100 in an embodiment of this application.

[0171] Reference Figure 11 As shown, the ship identification device 1100 according to an embodiment of this application includes: an acquisition unit 1101, a data processing unit 1102, an iteration unit 1103, and an identification unit 1104.

[0172] The acquisition unit 1101 is used to acquire reference point cloud data and target point cloud data of the target area; the data processing unit 1102 is used to statistically analyze the point cloud feature values ​​of each pixel grid in multiple feature dimensions, and the point cloud feature values ​​are used to characterize the features of the point cloud within the pixel grid; the iteration unit 1103 is used to filter the target feature dimension from multiple feature dimensions based on the point cloud feature values ​​of each pixel grid in multiple feature dimensions and whether the reference point cloud data in each pixel grid contains the hull point cloud; the recognition unit 1104 is used to recognize the hull in the target area based on the target feature dimension.

[0173] In some embodiments of this application, based on the foregoing scheme, the acquisition unit 1101 is configured to: acquire reference point cloud data of the target area, and determine multiple pixel grids in a top view based on the reference point cloud data.

[0174] In some embodiments of this application, based on the foregoing scheme, the data processing unit 1102 is configured as follows:

[0175] The point cloud feature values ​​of each pixel grid are statistically analyzed across multiple feature dimensions. These point cloud feature values ​​are used to characterize the features of the point cloud within each pixel grid.

[0176] In some embodiments of this application, based on the foregoing scheme, the iteration unit 1103 is configured as follows:

[0177] Multiple feature dimensions are defined as candidate feature dimensions. Based on the point cloud feature values ​​of each pixel grid on the candidate feature dimensions, and whether the reference point cloud data in each pixel grid contains the hull point cloud, the target feature dimension is determined from the candidate feature dimensions. If the number of target feature dimensions exceeds the set number of dimensions, the target feature dimension is used as a new candidate feature dimension, and the process of iteratively determining the target feature dimension based on the point cloud feature values ​​of each pixel grid on the candidate feature dimensions, and whether the reference point cloud data in each pixel grid contains the hull point cloud, is repeated until the number of target feature dimensions meets the set number of dimensions.

[0178] In some embodiments of this application, based on the foregoing scheme, the iteration unit 1103 is further configured as follows:

[0179] Based on the point cloud feature values ​​of each pixel in the candidate feature dimensions, the point cloud feature matrix is ​​determined; each pixel is labeled according to whether the reference point cloud data in each pixel contains the hull point cloud, thus determining the label vector; the point cloud feature matrix and the label vector are fitted to obtain the regression coefficients of each feature dimension, and the absolute value of the regression coefficient is used to characterize the contribution of the corresponding feature dimension to determining whether the reference point cloud data in the pixel contains the hull point cloud; based on the regression coefficients of each feature dimension, the target feature dimension is determined from the candidate feature dimensions.

[0180] In some embodiments of this application, based on the foregoing scheme, the iteration unit 1103 is further configured as follows:

[0181] Based on the number of dimensions of multiple feature dimensions, determine the number of target feature dimensions to retain; according to the number of target feature dimensions to retain, determine the target feature dimensions from the candidate feature dimensions based on the regression coefficients of each feature dimension.

[0182] In some embodiments of this application, based on the foregoing scheme, the iteration unit 1103 is further configured as follows:

[0183] If the number of dimensions of multiple feature dimensions is less than or equal to the number of dimensions threshold, the number of target feature dimensions to be retained is determined according to the set number of feature dimensions to be removed; if the number of dimensions of multiple feature dimensions is greater than the number of dimensions threshold, the number of feature dimensions to be removed is calculated using an exponential decay method, and the number of target feature dimensions to be retained is determined according to the calculated number of feature dimensions to be removed.

[0184] In some embodiments of this application, based on the foregoing scheme, the iteration unit 1103 is further configured as follows:

[0185] Sort the absolute values ​​of the regression coefficients of each feature dimension in descending order; determine the candidate feature dimensions with the highest number of retained coefficients as the target feature dimensions.

[0186] In some embodiments of this application, based on the foregoing scheme, the identification unit 1104 is configured as follows:

[0187] Acquire point cloud data of the target area to be identified; identify the ship hull based on the point cloud data to be identified and the target feature dimensions with a set number of dimensions.

[0188] In some embodiments of this application, based on the foregoing scheme, the identification unit 1104 is further configured as follows:

[0189] Obtain the target feature dimension point cloud feature values ​​of the point cloud data to be identified in each pixel grid, with a set number of dimensions (3). Assign the highest value of the target feature dimension point cloud feature values ​​of the set number of dimensions in each pixel grid to 255, and the lowest value to 0. Convert the remaining value of the target feature dimension point cloud feature values ​​of the set number of dimensions in each pixel grid to integers between 0 and 255 proportionally, and then convert them into RGB three-channel color information for each pixel grid. Generate the image to be identified from the RGB three-channel color information of each pixel grid. Perform ship hull feature recognition on the image to be identified.

[0190] Fourthly, based on the same inventive concept, this application discloses a hull identification device 2100, which can be used to execute the hull identification method of the second aspect of the above-described embodiments of this application. For details not disclosed in the device embodiments of this application, please refer to the embodiments of the hull identification method of the second aspect of this application.

[0191] See Figure 12 The diagram shown illustrates a block diagram of another ship identification device 2100 according to an embodiment of this application.

[0192] Reference Figure 12 As shown, the ship identification device 2100 disclosed in the embodiments of this application includes a determining unit 2101, an analysis unit 2102, and an identification unit 2103:

[0193] The determining unit 2101 is used to determine multiple sets of target feature dimensions according to any of the above methods; the analyzing unit 2102 verifies each set of target feature dimensions based on the verification data and calculates the root mean square error corresponding to each set of target feature dimensions; the identifying unit 2103 is used to identify the ship hull in the target area based on the set of target feature dimensions with the smallest root mean square error.

[0194] Fifthly, based on the same inventive concept, embodiments of this application also provide a ship identification electronic device, see reference. Figure 13The diagram shows a schematic of the structure of a ship identification electronic device according to an embodiment of this application. The ship identification electronic device includes one or more memories 1204, one or more processors 1202, and at least one computer program (program code) stored in the memory 1204 and executable on the processor 1202. When the processor 1202 executes the computer program, it implements the ship identification method of the first aspect mentioned above.

[0195] Among them, Figure 13 In this document, a bus architecture (represented by bus 1200) is used. Bus 1200 may include any number of interconnected buses and bridges, linking various circuits including one or more processors represented by processor 1202 and memory represented by memory 1204. Bus 1200 may also link various other circuits such as peripheral devices, voltage regulators, and power management circuits, which are well known in the art and therefore will not be described further herein. Bus interface 1205 provides an interface between bus 1200 and receiver 1201 and transmitter 1203. Receiver 1201 and transmitter 1203 may be the same element, i.e., a transceiver, providing a unit for communicating with various other devices over a transmission medium. Processor 1202 is responsible for managing bus 1200 and general processing, while memory 1204 may be used to store data used by processor 1202 during operation.

[0196] The beneficial effects of the embodiments of the second to third aspects described above can be referred to the beneficial effects of the first aspect and the embodiments of the first aspect described above, and will not be repeated here.

[0197] The functions described herein may be implemented in hardware, software executed by a processor, firmware, or any combination thereof. If implemented in software executed by a processor, the functions may be stored as one or more instructions or codes on or transmitted via a computer-readable medium. Other examples and embodiments are within the scope and spirit of this application and the appended claims. For example, due to the nature of software, the functions described above may be implemented using software executed by a processor, hardware, firmware, hardwired, or any combination thereof. Furthermore, the functional units may be integrated into a single processing unit, or each unit may exist physically separately, or two or more units may be integrated into a single unit.

[0198] In the several embodiments provided in this application, it should be understood that the disclosed technical content can be implemented in other ways. The device embodiments described above are merely illustrative; for example, the division of units can be a logical functional division, and in actual implementation, there may be other division methods. For instance, multiple units or components may be combined or integrated into another system, or some features may be ignored or not executed. Furthermore, the displayed or discussed mutual coupling, direct coupling, or communication connection may be through some interfaces; the indirect coupling or communication connection between units or modules may be electrical or other forms.

[0199] The units described as separate components may or may not be physically separate. Similarly, the components of the control device may or may not be physical units; they may be located in one place or distributed across multiple units. Some or all of the units can be selected to achieve the purpose of this embodiment, depending on actual needs.

[0200] If the integrated unit is implemented as a software functional unit and sold or used as an independent product, it can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, or all or part of the technical solution, can be embodied in the form of a software product. This computer software product is stored in a storage medium and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute all or part of the steps of the methods described in the various embodiments of this application. The aforementioned storage medium includes various media capable of storing program code, such as a USB flash drive, read-only memory (ROM), random access memory (RAM), portable hard drive, magnetic disk, or optical disk.

[0201] The above description is merely an embodiment of this application and is not intended to limit this application. Various modifications and variations can be made to this application by those skilled in the art. Any modifications, equivalent substitutions, improvements, etc., made within the spirit and principles of this application should be included within the scope of the claims of this application.

Claims

1. A method for identifying ship hulls, characterized in that, The method includes: Acquire reference point cloud data for the target area, and determine multiple pixel grids in a top view based on the reference point cloud data; The point cloud feature values ​​of each pixel grid are statistically analyzed in multiple feature dimensions. These point cloud feature values ​​are used to characterize the features of the point cloud within the pixel grid. Based on the point cloud feature values ​​of each pixel grid in the multiple feature dimensions, and whether the reference point cloud data in each pixel grid contains the hull point cloud, the target feature dimension is selected from the multiple feature dimensions. Based on the target feature dimensions, the ship hull in the target region is identified; The step of filtering a target feature dimension from the multiple feature dimensions based on the point cloud feature values ​​of each pixel grid on the multiple feature dimensions and whether the reference point cloud data in each pixel grid contains a hull point cloud includes: defining the multiple feature dimensions as candidate feature dimensions; determining a target feature dimension from the candidate feature dimensions based on the point cloud feature values ​​of each pixel grid on the candidate feature dimensions and whether the reference point cloud data in each pixel grid contains a hull point cloud; if the number of dimensions of the target feature dimension exceeds a set number of dimensions, then the target feature dimension is used as a new candidate feature dimension, and the step of iteratively determining the target feature dimension from the candidate feature dimensions based on the point cloud feature values ​​of each pixel grid on the candidate feature dimensions and whether the reference point cloud data in each pixel grid contains a hull point cloud is returned, until the number of dimensions of the target feature dimension meets the set number of dimensions; The step of determining the target feature dimension from the candidate feature dimensions based on the point cloud feature values ​​of each pixel grid on the candidate feature dimension and whether the reference point cloud data in each pixel grid contains hull point clouds includes: determining a point cloud feature matrix based on the point cloud feature values ​​of each pixel grid on the candidate feature dimension; labeling each pixel grid according to whether the reference point cloud data in each pixel grid contains hull point clouds to determine a label vector; fitting the point cloud feature matrix and the label vector to obtain regression coefficients for each feature dimension, the absolute value of the regression coefficients being used to characterize the contribution of the corresponding feature dimension to determining whether the reference point cloud data in the pixel grid contains hull point clouds; and determining the target feature dimension from the candidate feature dimensions based on the regression coefficients of each feature dimension.

2. The method according to claim 1, characterized in that, The determination of the target feature dimension from the candidate feature dimensions based on the regression coefficients of each feature dimension includes: Based on the number of dimensions of the multiple feature dimensions, determine the number of target feature dimensions to retain; Based on the number of target feature dimensions retained, and using the regression coefficients of each feature dimension, the target feature dimension is determined from the candidate feature dimensions.

3. The method according to claim 2, characterized in that, Determining the number of target feature dimensions to retain based on the number of dimensions of the plurality of feature dimensions includes: If the number of dimensions of the multiple feature dimensions is less than or equal to the number of dimensions threshold, then the number of target feature dimensions to be retained is determined according to the set number of feature dimensions to be removed. If the total number of feature dimensions exceeds the threshold, the number of feature dimensions to be removed is calculated using an exponential decay method. Based on the calculated number of feature dimensions to be removed, the number of target feature dimensions to be retained is determined.

4. The method according to claim 2, characterized in that, The step of determining the target feature dimension from the candidate feature dimensions based on the regression coefficients of each feature dimension, according to the number of retained target feature dimensions, includes: Sort the regression coefficients of each feature dimension in descending order of absolute value; The candidate feature dimensions with the highest number of retained items are determined as the target feature dimensions.

5. The method according to claim 1, characterized in that, The process of identifying the ship hull in the target region based on the target feature dimension includes: Acquire the point cloud data to be identified in the target area; The ship hull is identified based on the point cloud data to be identified and the target feature dimensions.

6. The method according to claim 5, characterized in that, The target feature dimension has three dimensions. The process of identifying the ship hull based on the point cloud data to be identified and the target feature dimensions includes: Obtain the point cloud feature values ​​of each pixel grid corresponding to the point cloud data to be identified in the three target feature dimensions; The point cloud feature values ​​in the three target feature dimensions are converted into RGB three-channel color information. Based on the RGB three-channel color information, an image to be identified is generated, and the ship hull of the target area is identified using the image to be identified.

7. A method for identifying a ship's hull, characterized in that, The method includes: The method according to any one of claims 1 to 4 determines multiple sets of target feature dimensions, wherein each set of target feature dimensions includes at least one target feature dimension; The target feature dimensions of each group are validated based on the validation data, and the root mean square error corresponding to each target feature dimension is calculated. The ship hull in the target region is identified based on the set of target feature dimensions that minimize the root mean square error.

8. A ship hull identification electronic device, characterized in that, The method includes one or more processors and one or more memories, wherein at least one piece of program code is stored in the one or more memories, and the at least one piece of program code is loaded and executed by the one or more processors to implement the method as claimed in any one of claims 1-7.

Images