A method for detecting and analyzing infrared features of a ship wake

Abstract

This invention belongs to the field of ship motion analysis technology, and particularly relates to a method for detecting and analyzing the infrared features of ship wakes. The method includes the following steps: acquiring infrared images of ship wakes and converting them to grayscale; converting the original pseudo-color infrared image and the visible light fused infrared image into grayscale images; filtering and noise reduction; fusing ship wake infrared images at different spatiotemporal scales; and edge extraction, considering the addition of gradient amplitude detection in the diagonal direction to the original coordinate axis direction to enhance the extraction of effective edge features from the ship wake grayscale image. The ship wake infrared feature detection and analysis method proposed in this application can effectively solve the problems of high difficulty in feature extraction from original ship infrared trajectory images and unclear edge features. While achieving data enhancement, it reduces the existence of unnecessary data, lowers the burden of system analysis and computation, and provides a technical foundation for further improving the efficiency of ship infrared trajectory analysis.

Description

[0001] Technology Neighborhood

[0002] This invention belongs to the field of ship motion analysis technology, and in particular relates to a method for detecting and analyzing the infrared features of ship wakes. Background Technology

[0003] During ship navigation, the interaction between the ship's wake and the water creates a distinct water surface trace. Depending on the diffusion time and water surface conditions, this trace can remain or propagate for tens of minutes, with a length ranging from hundreds to thousands of meters. Due to the interaction between the ship and the water, as well as the temperature differences at different depths, this water surface trace generates corresponding trajectory information on infrared images. By analyzing and extracting this trajectory information, various characteristic information about the ship and its motion can be obtained, such as ship type, tonnage, speed, and power type. This method is known as infrared feature analysis of propagating wakes. Currently, wake feature analysis mainly relies on the ripples and turbulent trajectories exhibited by the surface water along the ship's path. Although it has obvious V-shaped oblique waves and traveling wave characteristics, it is still necessary to extract the subtle turbulence signals within it in order to obtain more detailed motion state information. The current problem is that, except in very rare cases, traditional infrared feature images of ship wakes often retain information about water changes caused by surface winds and ocean currents, which obscures or interferes with the detailed information, making it impossible to directly analyze and extract using the original infrared image, thus limiting the application of ship infrared wake feature analysis. Summary of the Invention

[0004] The purpose of this invention is to provide a method for detecting and analyzing the infrared features of ship wakes by optimizing the representation effect of infrared images of ship wakes, thereby reducing the difficulty of image processing while revealing the infrared trajectory features, based on practical needs.

[0005] To achieve the above objectives, the present invention adopts the following technical solution.

[0006] A method for detecting and analyzing the infrared features of ship wakes includes the following steps:

[0007] Step 1: Acquire infrared images of ship wakes and convert them to grayscale; convert the original pseudo-color infrared images and visible light fused infrared images to grayscale.

[0008] Step 2, Filtering and Noise Reduction Processing, includes the following steps:

[0009] 2a. To eliminate systematic interference and improve image smoothness to reveal the trajectory, a filter is constructed to perform noise reduction and edge trimming on the infrared image of the ship's wake; including the filter used for noise reduction and edge trimming, which can be represented as: ;

[0010] in Infrared images of ship wakes At coordinates pixels grayscale value, This refers to the coordinates after filtering. The pixel value of the pixel; Infrared images of ship wakes China-Israel point The set of neighborhoods of the center pixel. For pixels within the neighborhood set, The coordinates within the neighborhood set are pixels grayscale value; It refers to the neighborhood Standard deviation of the spatial location of an intrapixel It refers to the neighborhood Standard deviation of grayscale values ​​of internal pixels; It refers to the neighborhood The average grayscale value of each pixel;

[0011] 2b. Adjust the neighborhood size to perform noise reduction and edge trimming on the infrared image of the ship's wake, and analyze the filtering effect of the filter corresponding to different neighborhood sizes through peak signal-to-noise ratio to determine the appropriate filtering parameters;

[0012] 2c. Filter the image using spatial orientation filters of different window sizes to obtain filtered images at different scales as the original data; that is, the spatial orientation filter used for spatial orientation information optimization can be represented as:

[0013]

[0014] in This is an infrared image of a ship's wake after filtering. At coordinates pixels grayscale value, Indicates the aspect ratio of the filter window. This refers to the direction of filtering. The filter wavelength; It is the standard deviation of the Gaussian function;

[0015] Step 3: Fusion processing of infrared images of ship wakes at different spatiotemporal scales; specifically including:

[0016] 3a. Based on the infrared images of ship wakes acquired in step two at different spatiotemporal scales, perform fusion processing to obtain the ship wake infrared images. Enhanced image after fusion The pixel grayscale representation is ;

[0017] in, This refers to the weights for smooth edge feature fusion. This refers to the weighting of spatial and orientational features. ;

[0018] 3b. Extract the wake feature data from the ship wake infrared image and remove background pixels. Specifically:

[0019] For pixel size grayscale level is grayscale image of ship wake Obtain the number of pixels at different gray levels. , Then its total number of pixels The normalized histogram distribution probability of the image can then be expressed as: , ;

[0020] Define background pixel grayscale segmentation threshold , The grayscale value is located in The grayscale value range of the background pixels is located in The grayscale value range of the feature pixels;

[0021] Then the grayscale image of the ship's stern The probability that a medium pixel belongs to a background pixel. The probability of belonging to a feature pixel Average grayscale value of background pixels The average gray value of the feature pixels The average gray level of pixels at different gray levels is The overall gray level mean of the image. Then there is ;

[0022] Obtain inter-class variance ;

[0023] grayscale image of ship stern In terms of background pixel grayscale segmentation threshold The value that maximizes the inter-class variance, i.e. ;

[0024] Based on the aforementioned steps, determine the optimal background pixel grayscale segmentation threshold for each image. Then, the grayscale value range is... By setting the grayscale value of each pixel to zero, background elements in the grayscale image of the ship's wake can be removed, thus obtaining the grayscale image of the ship's wake. Background-removed grayscale image ;

[0025] Step 4: Edge extraction. This involves adding gradient magnitude detection along the diagonal direction to the existing coordinate axes to enhance the extraction of effective edge features from the ship's wake grayscale image. Specifically, this includes:

[0026] 4a. The grayscale image of the ship's wake is smoothed using the derivative of a first-order two-dimensional Gaussian function to improve the effectiveness of edge extraction; the pixel grayscale representation of the smoothed image is as follows: ,in It refers to a two-dimensional Gaussian function and ,in This refers to the standard deviation of the Gaussian function;

[0027] 4b. Calculate the gradient magnitude and direction along the coordinate axes and diagonal;

[0028] gradient in the horizontal direction of the X-axis

[0029]

[0030] gradient along the Y-axis object direction

[0031]

[0032] Gradient along the first diagonal direction

[0033]

[0034] Gradient along the second diagonal direction

[0035]

[0036] Projecting the gradients along the first and second diagonals onto the X and Y axes and superimposing them, we obtain the corrected X-axis gradient and Y-axis gradient as follows:

[0037] ;

[0038] ;

[0039] Then gradient magnitude ;

[0040] Direction angle ;

[0041] 4c. Gradient interpolation, using The sliding window compares the center pixel with those adjacent to it along any gradient direction within the sliding window. The average gradient value of each pixel is used to calculate the grayscale image of the ship's wake. If the gradient value of the center pixel is less than the average gradient value in a certain direction, the center pixel is considered not to be an edge pixel in that direction; otherwise, it is considered to be an edge pixel in that direction. This process yields the grayscale image of the ship's wake after gradient interpolation optimization. ;

[0042] 4d. Extracting grayscale images of ship sterns histogram Take histogram Valley points are used as the threshold for edge detection to obtain the edge image of the ship's wake. Using histograms Using the valley point as a threshold, the edge image of the ship's wake is analyzed. Nonmaximum suppression is performed to obtain a binary edge image. Searching for images of the edge of a ship's wake All non-edge connected regions For each non-edge connected region Perform region searches separately until the boundary is a single pixel that is isotropically connected;

[0043] The rules and steps for area search are as follows:

[0044] S1, Define the point set Images containing the edge of a ship's wake All edge points in the set; define the point set. Images containing the edge of a ship's wake All edge points in the set; define the point set. Define a point set This indicates that it belongs to a non-edge connected region. The edge and belonging to ;

[0045] S2. Based on the grayscale image of the ship's stern Pair of grayscale values ​​of middle pixels Sort the pixels in the data, starting from the point with the lowest gray value and moving towards the nearest point in the set. An edge search is performed to ensure that the edges of the final result contain non-maximum suppression points. Each point is confirmed to have only one adjacent non-edge connected region; if more than one is found, the search stops, continuing until the point set is reached. All pixels in the search have been completed;

[0046] S3. Based on the grayscale image of the ship's stern Pair of grayscale values ​​of middle pixels Sort the pixels in the data, starting from the point with the lowest gray value and moving towards the nearest point in the set. An edge search is performed to ensure that the edges of the final result contain non-maximum suppression points. Each point is confirmed to have only one adjacent non-edge connected region; if more than one is found, the search stops, continuing until the point set is reached. All pixels in the search have been completed;

[0047] S4. Determine non-edge connected regions based on search results. For the new edge, repeat steps S2-S3 until the non-edge connected region is completed. Search;

[0048] 4e. Extract edges based on the updated edge pixels to obtain complete ship wake information.

[0049] In a further improvement or specific implementation of the aforementioned method for detecting and analyzing ship wake infrared features, step three involves using a local information entropy weighting method to enhance local features for complex sea surface conditions, resulting in ship wake infrared images. At the coordinates of the pixel The method for calculating information entropy is as follows:

[0050] ;

[0051] in Infrared images of ship wakes The grayscale value of the middle pixel The total number of pixels; Infrared images of ship wakes gray levels, Infrared images of ship wakes The total number of gray levels; based on this, the feature fusion expression of the ship wake infrared image, which incorporates information entropy, is corrected to... ; It is the information entropy feature weight. ;

[0052] In a further improvement or specific implementation of the aforementioned method for detecting and analyzing the infrared features of ship wakes, step 3b further includes removing residual pixel information based on the V-shaped distribution area and the outer edge of the ripple area of ​​the ship wake grayscale image.

[0053] 1. In a further improvement or specific implementation of the aforementioned method for detecting and analyzing ship wake infrared features, step two further includes selecting a suitable filter window size and filter direction, and selectively filtering image features with specific spatial dimensions and directions to obtain infrared images of international ship wakes with different spatial dimensions and directions.

[0054] In a further improvement or specific implementation of the aforementioned method for detecting and analyzing the infrared features of ship wakes, in step 2c, the filter window size is... ,in It is a positive integer. .

[0055] In a further improvement or specific implementation of the aforementioned method for detecting and analyzing the infrared features of ship wakes, the fusion processing in step 3a may also include pixel normalization processing of the fused grayscale image. Attached Figure Description

[0056] Figure 1 This is the image of a ship's wake infrared image before noise reduction and edge trimming;

[0057] Figure 2 This is a denoised and edge-trimmed image of an infrared image of a ship's wake.

[0058] Figure 3 It is a filtered image based on different spatial dimensions and orientations;

[0059] Figure 4 It is a filtered image based on different spatial dimensions and orientations. Detailed Implementation

[0060] The present invention will be described in detail below with reference to specific embodiments.

[0061] To meet the needs of feature analysis of ship infrared wake images, this application provides a method for optimizing original ship wake infrared image data, enhancing its wake feature representation capability, and eliminating invalid information for ship wake infrared feature detection and analysis.

[0062] This method mainly consists of the following basic steps:

[0063] Step 1: Acquire infrared images of the ship's wake and convert them to grayscale.

[0064] Based on the differences in the acquisition principles and methods of different wake acquisition systems, ship wake features may be in the form of pseudo-color infrared images, visible light fused infrared images, and grayscale infrared images. Based on the needs of ship wake feature analysis in this application, the original pseudo-color infrared images and visible light fused infrared images need to be converted into grayscale images in order to more intuitively and simply represent the corresponding trajectory features.

[0065] Step 2: Filtering and noise reduction processing;

[0066] Since the acquisition system used to acquire raw infrared images usually relies on multiple different infrared imaging units, the differences in the system properties and working conditions of each infrared imaging unit will cause infrared system non-uniformity interference in the acquired infrared wake images. In order to eliminate this systematic interference, improve the smoothness of the image to reveal the trajectory, and preserve the edge information of the ship wake infrared image for easy extraction, a filter is constructed to perform noise reduction and edge trimming processing on the ship wake infrared image.

[0067] Including the first filter used for noise reduction and trimming, it can be represented as: ;

[0068] in Infrared images of ship wakes At coordinates pixels grayscale value, This refers to the coordinates after filtering. The pixel value of the pixel; Infrared images of ship wakes China-Israel point The set of neighborhoods of the center pixel. For pixels within the neighborhood set, The coordinates within the neighborhood set are pixels grayscale value; It refers to the neighborhood Standard deviation of the spatial location of an intrapixel It refers to the neighborhood Standard deviation of grayscale values ​​of internal pixels; It refers to the neighborhood The average grayscale value of each pixel;

[0069] Noise reduction and edge trimming of ship wake infrared images are performed by selecting an appropriate neighborhood size. The filtering effect of filters corresponding to different neighborhood sizes is analyzed based on the peak signal-to-noise ratio to determine suitable filtering parameters. Figure 1 , Figure 2 As shown, after noise reduction and edge trimming, noise signals can be effectively reduced while effectively preserving local core edge data, and the edge information is made more complete and clear, which is conducive to further obtaining boundary information.

[0070] Local spatial distribution and orientation data in ship wake infrared images are important parameters for studying ship motion characteristics. At different spatial scales, the different sizes of spatial and orientation information exhibited by ship wake infrared images help to extract detailed elements of ship motion in different times and spaces. In order to enhance spatial and orientation information and better extract the required information from complex ship wake infrared images, spatial orientation filters with different window sizes are used to filter the images and obtain filtered images at different scales as raw data.

[0071] That is, a spatial orientation filter used for optimizing spatial orientation information. The spatial orientation filter constructed in this invention can be expressed as:

[0072]

[0073] in This is an infrared image of a ship's wake after filtering. At coordinates pixels grayscale value, Indicates the aspect ratio of the filter window. This refers to the filtering direction (i.e., the direction of the texture to be extracted). The filter wavelength; It is the standard deviation of the Gaussian function; as a preferred embodiment, in step 2c, the filter window can be a rectangular window with consistent length, width, and height, and the window size is... ,in It is a positive integer. .

[0074] Since the wakes in infrared images of international ship wakes have obvious spatial and azimuth features, based on the aforementioned spatial azimuth filter, such as Figure 3 , Figure 4 As shown, by selecting appropriate filter window size and filter direction, image features with specific spatial size and direction can be filtered in a targeted manner to obtain infrared images of international ship wakes with different spatial size and direction features. This can effectively enhance the efficiency and accuracy of feature extraction from international ship wake infrared images, reduce unnecessary data processing, and improve the efficiency of the method.

[0075] Step 3: Fusion processing of infrared images of ship wakes at different spatiotemporal scales; specifically including:

[0076] Based on the above filtering and necessary preprocessing, infrared images containing different spatiotemporal features can be extracted from the original ship wake infrared image. To effectively represent the feature data in the original ship wake infrared image, further image fusion is performed using the images obtained after the aforementioned filtering process. Enhanced image after fusion The pixel grayscale representation is ;

[0077] in, This refers to the weights for smooth edge feature fusion. This refers to the weighting of spatial and orientational features. ;

[0078] To reduce the difficulty of data processing in subsequent steps, in addition to the above, pixel normalization processing is also performed on the grayscale image after fusion processing.

[0079] In particular, for complex sea surface conditions, the presence of eddies and waves leads to a large amount of complex local feature information in ship wake infrared images. This local information often clusters around a specific center point or region. In this case, considering the similarity of these clustered local features at the pixel level and their inherent information distribution, a local information entropy weighting method is used to enhance these local features. At the coordinates of the pixel The method for calculating information entropy is as follows:

[0080] ;

[0081] in Infrared images of ship wakes The grayscale value of the middle pixel The total number of pixels; Infrared images of ship wakes gray levels, Infrared images of ship wakes The total number of gray levels; based on this, the feature fusion expression of the ship wake infrared image, which incorporates information entropy, is corrected to... ; It is the information entropy feature weight. ;

[0082] The aforementioned feature enhancement processing effectively improves the expressive power of wake features in ship wake infrared images. However, in the specific data analysis process, it is still necessary to extract these wake feature data from the ship wake infrared images and remove background pixels other than non-feature information. Specifically:

[0083] For pixel size grayscale level is grayscale image of ship wake Obtain the number of pixels at different gray levels. , Then its total number of pixels The normalized histogram distribution probability of the image can then be expressed as: , ;

[0084] Define background pixel grayscale segmentation threshold , The grayscale value is located in The grayscale value range of the background pixels is located in The grayscale value range of the feature pixels;

[0085] Then the grayscale image of the ship's stern The probability that a medium pixel belongs to a background pixel. The probability of belonging to a feature pixel Average grayscale value of background pixels The average gray value of the feature pixels The average gray level of pixels at different gray levels is The overall gray level mean of the image. Then there is ;

[0086] Obtain inter-class variance ;

[0087] grayscale image of ship stern In terms of background pixel grayscale segmentation threshold The value that maximizes the inter-class variance, i.e. ;

[0088] Based on the aforementioned steps, determine the optimal background pixel grayscale segmentation threshold for each image. Then, the grayscale value range is... By setting the grayscale value of each pixel to zero, background elements in the grayscale image of the ship's wake can be removed, thus obtaining the grayscale image of the ship's wake. Background-removed grayscale image ;

[0089] In particular, by utilizing the distinct distribution characteristics of grayscale images of ship wakes, after removing background elements, further removal can be performed based on the V-shaped distribution area and the pixel information remaining outside the ripple area of ​​the grayscale image of ship wakes.

[0090] Based on the feature pixel images of the grayscale images of ship wakes obtained from the aforementioned steps, key feature pixel information such as the shape, distribution, and direction of the wake is preserved. On this basis, by analyzing its contour boundary features and performing edge extraction, clearer and more complete ship wake information can be obtained.

[0091] Edge extraction is an important part of various image processing tasks. Currently, commonly used edge extraction algorithms mainly utilize the changes in gradient magnitude and direction along two coordinate axes to extract image contours. While this approach has universality, the contour extraction results vary greatly depending on the specific image. Considering the unique V-shaped wavy edge features of the ship wake grayscale image in this application, we propose adding gradient magnitude detection along the diagonal direction to the existing coordinate axis directions to enhance the extraction of effective edge features from the ship wake grayscale image. Specifically, this can be divided into the following steps:

[0092] 4a. The grayscale image of the ship's wake is smoothed using the derivative of a first-order two-dimensional Gaussian function to improve the effectiveness of edge extraction; the pixel grayscale representation of the smoothed image is as follows: ,in It refers to a two-dimensional Gaussian function and ,in This refers to the standard deviation of the Gaussian function;

[0093] 4b. Calculate the gradient magnitude and direction along the coordinate axes and diagonal;

[0094] gradient in the horizontal direction of the X-axis

[0095]

[0096] gradient along the Y-axis object direction

[0097]

[0098] Gradient along the first diagonal direction

[0099]

[0100] Gradient along the second diagonal direction

[0101]

[0102] Projecting the gradients along the first and second diagonals onto the X and Y axes and superimposing them, we obtain the corrected X-axis gradient and Y-axis gradient as follows:

[0103] ;

[0104] ;

[0105] Then gradient magnitude ;

[0106] Direction angle ;

[0107] 4c. Gradient interpolation: During the calculation of gradient magnitude and direction, false edge information generated by factors such as noise interference will be retained. To ensure the effectiveness of the edges and improve recognition accuracy, local gradient interpolation analysis is used to optimize the image magnitude distribution. Specifically:

[0108] use The sliding window compares the center pixel with those adjacent to it along any gradient direction within the sliding window. The average gradient value of each pixel is used to calculate the grayscale image of the ship's wake. If the gradient value of the center pixel is less than the average gradient value in a certain direction, the center pixel is considered not to be an edge pixel in that direction; otherwise, it is considered to be an edge pixel in that direction. This process yields the grayscale image of the ship's wake after gradient interpolation optimization. ;

[0109] 4d. Extracting grayscale images of ship sterns histogram Take histogram Valley points are used as the threshold for edge detection to obtain the edge image of the ship's wake. Using histograms Using the valley point as a threshold, the edge image of the ship's wake is analyzed. Nonmaximum suppression is performed to obtain a binary edge image. Searching for images of the edge of a ship's wake All non-edge connected regions For each non-edge connected region Perform region searches separately until the boundary is a single pixel that is isotropically connected;

[0110] The rules and steps for area search are as follows:

[0111] S1, Define the point set Images containing the edge of a ship's wake All edge points in the set; define the point set. Images containing the edge of a ship's wake All edge points in the set; define the point set. Define a point set This indicates that it belongs to a non-edge connected region. The edge and belonging to ;

[0112] S2. Based on the grayscale image of the ship's stern Pair of grayscale values ​​of middle pixels Sort the pixels in the data, starting from the point with the lowest gray value and moving towards the nearest point in the set. An edge search is performed to ensure that the edges of the final result contain non-maximum suppression points. Each point is confirmed to have only one adjacent non-edge connected region; if more than one is found, the search stops, continuing until the point set is reached. All pixels in the search have been completed;

[0113] S3. Based on the grayscale image of the ship's stern Pair of grayscale values ​​of middle pixels Sort the pixels in the data, starting from the point with the lowest gray value and moving towards the nearest point in the set. An edge search is performed to ensure that the edges of the final result contain non-maximum suppression points. Each point is confirmed to have only one adjacent non-edge connected region; if more than one is found, the search stops, continuing until the point set is reached. All pixels in the search have been completed;

[0114] S4. Determine non-edge connected regions based on search results. For the new edge, repeat steps S2-S3 until the non-edge connected region is completed. Search;

[0115] 4e. Extract edges based on the updated edge pixels to obtain complete ship wake information.

[0116] The infrared feature detection and analysis method for ship wakes proposed in this application can effectively solve the problems of high difficulty in feature extraction and unclear edge features in the original ship infrared trajectory images. While achieving data enhancement, it reduces the existence of unnecessary data and lowers the burden of system analysis and calculation, providing a technical foundation for further improving the efficiency of ship infrared trajectory analysis.

[0117] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, and are not intended to limit the scope of protection of the present invention. Although the present invention has been described in detail with reference to preferred embodiments, those skilled in the art should understand that modifications or equivalent substitutions can be made to the technical solutions of the present invention without departing from the essence and scope of the technical solutions of the present invention.

Claims

1. A method for detecting and analyzing the infrared features of ship wakes, characterized in that, Includes the following steps: Step 1: Acquire infrared images of ship wakes and convert them to grayscale; convert the original pseudo-color infrared images and visible light fused infrared images to grayscale. Step 2, Filtering and Noise Reduction Processing, includes the following steps: 2a. To eliminate systematic interference and improve the smoothness of the image to reveal the trajectory, a filter is constructed to perform noise reduction and edge trimming on the infrared image of the ship's wake. Including filters used for noise reduction and trimming, it can be represented as: ; in Infrared images of ship wakes At coordinates pixels grayscale value, This refers to the coordinates after filtering. The pixel value of the pixel; Infrared images of ship wakes China-Israel point The set of neighborhoods of the center pixel. For pixels within the neighborhood set, The coordinates within the neighborhood set are pixels grayscale value; It refers to the neighborhood Standard deviation of the spatial location of an intrapixel It refers to the neighborhood Standard deviation of grayscale values ​​of internal pixels; It refers to the neighborhood The average grayscale value of each pixel; 2b. Adjust the neighborhood size to perform noise reduction and edge trimming on the infrared image of the ship's wake, and analyze the filtering effect of the filter corresponding to different neighborhood sizes through peak signal-to-noise ratio to determine the appropriate filtering parameters; 2c. Filter the image using spatial orientation filters of different window sizes to obtain filtered images at different scales as the original data; that is, the spatial orientation filter used for spatial orientation information optimization can be represented as: in This is an infrared image of a ship's wake after filtering. At coordinates pixels grayscale value, Indicates the aspect ratio of the filter window. This refers to the direction of filtering. The filter wavelength; It is the standard deviation of the Gaussian function; Step 3: Fusion processing of infrared images of ship wakes at different spatiotemporal scales; specifically including: 3a. Based on the infrared images of ship wakes acquired in step two at different spatiotemporal scales, perform fusion processing to obtain the ship wake infrared images. Enhanced image after fusion The pixel grayscale representation is ; in, This refers to the weights for smooth edge feature fusion. This refers to the weighting of spatial and orientational features. ; 3b. Extract the wake feature data from the ship wake infrared image and remove background pixels. Specifically: For pixel size grayscale level is grayscale image of ship wake Obtain the number of pixels at different gray levels. , Then its total number of pixels The normalized histogram distribution probability of the image can then be expressed as: , ; Define background pixel grayscale segmentation threshold , The grayscale value is located in The grayscale value range of the background pixels is located in The grayscale value range of the feature pixels; Then the grayscale image of the ship's stern The probability that a medium pixel belongs to a background pixel. The probability of belonging to a feature pixel Average grayscale value of background pixels The average gray value of the feature pixels The average gray level of pixels at different gray levels is The overall gray level mean of the image. Then there is ; Obtain inter-class variance ; grayscale image of ship stern In terms of background pixel grayscale segmentation threshold The value that maximizes the inter-class variance, i.e. ; Based on the aforementioned steps, determine the optimal background pixel grayscale segmentation threshold for each image. Then, the grayscale value range is... By setting the grayscale value of each pixel to zero, background elements in the grayscale image of the ship's wake can be removed, thus obtaining the grayscale image of the ship's wake. Background-removed grayscale image ; Step 4: Edge extraction. This involves adding gradient magnitude detection along the diagonal direction to the existing coordinate axes to enhance the extraction of effective edge features from the ship's wake grayscale image. Specifically, this includes: 4a. The grayscale image of the ship's wake is smoothed using the derivative of a first-order two-dimensional Gaussian function to improve the effectiveness of edge extraction; the pixel grayscale representation of the smoothed image is as follows: ,in It refers to a two-dimensional Gaussian function and ,in This refers to the standard deviation of the Gaussian function; 4b. Calculate the gradient magnitude and direction along the coordinate axes and diagonal; gradient in the horizontal direction of the X-axis gradient along the Y-axis object direction Gradient along the first diagonal direction Gradient along the second diagonal direction Projecting the gradients along the first and second diagonals onto the X and Y axes and superimposing them, we obtain the corrected X-axis gradient and Y-axis gradient as follows: ； ； Then gradient magnitude ; Direction angle ; 4c. Gradient interpolation, using The sliding window compares the center pixel with those adjacent to it along any gradient direction within the sliding window. The average gradient value of each pixel is used to calculate the grayscale image of the ship's wake. If the gradient value of the center pixel is less than the average gradient value in a certain direction, the center pixel is considered not to be an edge pixel in that direction; otherwise, it is considered to be an edge pixel in that direction. This process yields the grayscale image of the ship's wake after gradient interpolation optimization. ; 4d. Extracting grayscale images of ship sterns histogram Take histogram Valley points are used as the threshold for edge detection to obtain the edge image of the ship's wake. Using histograms Using the valley point as a threshold, the edge image of the ship's wake is analyzed. Nonmaximum suppression is performed to obtain a binary edge image. Searching for images of the edge of a ship's wake All non-edge connected regions For each non-edge connected region Perform region searches separately until the boundary is a single pixel that is isotropically connected; The rules and steps for area search are as follows: S1, Define the point set Images containing the edge of a ship's wake All edge points in the set; define the point set. Images containing the edge of a ship's wake All edge points in the set; define the point set. Define a point set This indicates that it belongs to a non-edge connected region. The edge and belonging to ; S2. Based on the grayscale image of the ship's stern Pair of grayscale values ​​of middle pixels Sort the pixels in the data, starting from the point with the lowest gray value and moving towards the nearest point in the set. An edge search is performed to ensure that the edges of the final result contain non-maximum suppression points. Each point is confirmed to have only one adjacent non-edge connected region; if more than one is found, the search stops, continuing until the point set is reached. All pixels in the search have been completed; S3. Based on the grayscale image of the ship's stern Pair of grayscale values ​​of middle pixels Sort the pixels in the data, starting from the point with the lowest gray value and moving towards the nearest point in the set. An edge search is performed to ensure that the edges of the final result contain non-maximum suppression points. Each point is confirmed to have only one adjacent non-edge connected region; if more than one is found, the search stops, continuing until the point set is reached. All pixels in the search have been completed; S4. Determine non-edge connected regions based on search results. For the new edge, repeat steps S2-S3 until the non-edge connected region is completed. Search; 4e. Extract edges based on the updated edge pixels to obtain complete ship wake information.

2. The method for detecting and analyzing the infrared features of ship wakes according to claim 1, characterized in that, In step three, for complex sea surface conditions, a local information entropy weighting method is used to enhance local features, and the ship wake infrared image is used. At the coordinates of the pixel The method for calculating information entropy is as follows: ； in Infrared images of ship wakes The grayscale value of the middle pixel The total number of pixels; Infrared images of ship wakes gray levels, Infrared images of ship wakes The total number of gray levels; based on this, the feature fusion expression of the ship wake infrared image, which incorporates information entropy, is corrected to... ; It is the information entropy feature weight. .

3. The method for detecting and analyzing the infrared features of ship wakes according to claim 1, characterized in that, Step 3b further includes further removing pixels based on the V-shaped distribution area of ​​the ship's wake grayscale image and the residual pixel information outside the ripple area.

4. The method for detecting and analyzing the infrared features of ship wakes according to claim 1, characterized in that, Step two further includes selecting an appropriate filter window size and filter direction, and selectively filtering image features with specific spatial dimensions and directions to obtain infrared images of international ship wakes with different spatial dimensions and directions.

5. The method for detecting and analyzing the infrared features of ship wakes according to claim 1, characterized in that, In step 2c, the filter window size is ,in It is a positive integer. .

6. The method for detecting and analyzing the infrared features of ship wakes according to claim 1, characterized in that, The fusion process in step 3a further includes performing pixel normalization on the fused grayscale image.

Images