Active azimuth history graph suspected target prompting method based on image segmentation
By employing image segmentation and connected component labeling techniques, the problem of target identification difficulties in low-frequency sonar systems has been solved, enabling efficient localization and identification of suspected targets.
Patent Information
- Application Number
- CN202511524811.6
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-10-24
- Publication Date
- 2026-01-27
AI Technical Summary
In low-frequency sonar systems, targets appear as isolated bright spots in active azimuth history maps, making it difficult to manually screen underwater targets and severely affecting situation assessment efficiency.
Image segmentation is performed using the Otsu's method, combined with morphological dilation and connected component labeling. The centroid and Euclidean distance are calculated to remove large-area interference, identify small targets, and mark the locations of suspected targets.
It improves the accuracy and efficiency of small target identification, assisting sonar operators in quickly identifying suspected targets.
Smart Images
Figure CN121415232A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of low-frequency sonar system technology, specifically to a method for identifying suspected targets using an active orientation history map based on image segmentation. Background Technology
[0002] Underwater target image processing technology plays a crucial role in underwater detection, marine exploration, and resource exploration. However, due to the strong reverberation and numerous interferences during operation, active sonar often appears as a single isolated bright spot in the active azimuth history map of low-frequency sonar systems, while the detected image is filled with numerous stars, making manual target selection difficult and severely impacting the efficiency of sonar operators in assessing the situation. Summary of the Invention
[0003] To address the shortcomings of existing technologies, the present invention aims to provide an active orientation history map-based method for identifying potential targets based on image segmentation.
[0004] To achieve the above objectives, the present invention provides the following technical solution: a method for active orientation history mapping-based suspected target hinting based on image segmentation, comprising the following steps:
[0005] (1) Input the active orientation history map, use the maximum inter-class variance method to calculate the gray-level segmentation threshold of the image pixels, and generate a binary image;
[0006] (2) Perform a dilation operation on the binary image using cross-shaped structural elements;
[0007] (3) Perform 8-connected region labeling on the dilated binary image to obtain a connected region labeled image;
[0008] (4) Calculate the area and centroid of each connected region in the connected region labeled image;
[0009] (5) Remove connected regions with an area greater than a preset threshold to obtain a discrete target region image;
[0010] (6) Traverse the non-zero centroids in the discrete target region image and calculate the sum of Euclidean distances from each centroid to its 5 nearest neighbor connected region centroids using breadth-first search (BFS);
[0011] (7) Sort the Euclidean distances in descending order, select the five largest Euclidean distances and their corresponding discrete target region centroids, and mark the centroid positions on the active orientation history map.
[0012] In some embodiments, according to step (1), an active azimuth history map is input. The image was obtained using the maximum inter-class variance method. Pixel grayscale segmentation threshold Using segmentation threshold Obtain a binary image The specific method is as follows:
[0013] (1.1) Calculation Pixel grayscale average ;
[0014] (1.2) Traversing the segmentation threshold The traversal range is pixel grayscale range ;
[0015] (1.3) Based on the threshold of the current traversal , to divide the grayscale range of pixels Divided into target grayscale range and background grayscale range ;
[0016] (1.4) Calculate separately and The proportion of pixels to the total number of pixels , and grayscale mean , ;
[0017] (1.5) Calculation and Between-class variance , is represented as:
[0018]
[0019] (1.6) Repeat steps (1.2)-(1.5) to... Maximum The target grayscale range is determined as the final segmentation threshold. and background grayscale range Obtain a binary image ,
[0020] .
[0021] In some embodiments, the dilation operation of the cross-shaped structural element in step (2) is used to connect fragmented targets in the binary image, eliminating target breakage caused by noise.
[0022] Specifically, a cross-shaped structural element is adopted. right The expansion operation is represented as:
[0023]
[0024] In some embodiments, the binarized image is processed according to step (3). By performing 8-connected component labeling, a connected component labeled image is obtained. Specifically, it includes the following processes:
[0025] (3.1) Initialize the connected component labeling image , is represented as:
[0026] ;
[0027] (3.2) Traversing the binarized image and connected component labeled image ,like and (Unlabeled foreground pixels), labeled ,Will Add to queue ;
[0028] (3.3) When If not empty, retrieve the position of the first pixel in the queue. For all pixel positions within its 8-neighborhood (in and If it is within the image range, ,and Then mark and add it to the queue. ;
[0029] (3.4) Clear the queue. (Update the marker value for the next region).
[0030] In some embodiments, the connected component labeling image is calculated according to step (4). Area of each connected region and center of mass Specifically, it includes the following processes:
[0031] (4.1) No. Area of each connected region Defined as the number of all pixels in the region, expressed as:
[0032]
[0033] in , Image size, i.e. ;
[0034] (4.2) No. The centroid of each connected region (Pixel coordinates, rounded down) is defined as the coordinates of all pixels within this region. Coordinate mean and The mean of the coordinates is expressed as:
[0035]
[0036] .
[0037] In some embodiments, according to step (5), the connected component marker image is removed. Large connected regions are used to obtain discrete target region images. Discrete target region area and center of mass , is represented as:
[0038]
[0039]
[0040] .
[0041] In some embodiments, according to step (6), the non-zero centroids of the discrete target region are traversed. Calculate the sum of its Euclidean distances to its 5 nearest connected regions. ,right Sort and label the original images. Centroids of the top 5 discrete target regions Specifically, it includes the following processes:
[0042] (6.1) Traverse the non-zero centroids of the discrete target region ,Will Add to queue , ;
[0043] (6.2) When If not empty, retrieve the position of the first pixel in the queue. For all pixel positions within its 8-neighborhood (in , and If it falls within the image range, add it to the queue. If it is within the image range, and (in , For queue (The current number of non-zero coordinates) is added to the queue. Calculate distance , ;
[0044] (6.3) When When, output the sum of distances. The queue is cleared, and the sum of Euclidean distances from the non-zero centroid of the next discrete target region to its five nearest connected regions is calculated.
[0045] (6.4) Sort in descending order (from largest to smallest), and the sorted result satisfies ,at this time ;
[0046] (6.5) Mark on the original image The first 5 centroid positions.
[0047] In some embodiments, the marking method in step (7) includes drawing crosshairs, circles or numbers on the active orientation history map corresponding to the centroid position to visualize the suspected target position.
[0048] In some embodiments, after marking the centroid in step (7), the location information of the suspected target is output, including the azimuth angle and the corresponding distance value in the active azimuth history map.
[0049] Compared with existing technologies, the beneficial effects of this invention are: image segmentation using the maximum inter-class variance method effectively separates the target from the background; morphological dilation and connected region labeling are used to calculate the area and centroid of connected regions and remove large interference areas, thus identifying the connected regions where small targets are located, improving the accuracy of small target identification; and the combination of neighborhood search to calculate distance further optimizes the localization effect of small targets, assisting sonar operators in situation assessment and improving situation assessment efficiency.
[0050] Details of one or more embodiments of this application are set forth in the following drawings and description to make other features, objects and advantages of this application more readily apparent. The embodiments of this application will provide a detailed description and understanding of the application. Attached Figure Description
[0051] Figure 1 This is a flowchart of the process of the present invention;
[0052] Figure 2 This is the active orientation history diagram of the present invention;
[0053] Figure 3 This is a binarized diagram of the active orientation history of the present invention;
[0054] Figure 4 This is the active orientation history expansion diagram of the present invention;
[0055] Figure 5 This is a diagram showing the connected regions of the present invention;
[0056] Figure 6 This is a discrete target region diagram of the present invention;
[0057] Figure 7 This is a diagram illustrating the intended meaning of the present invention. Detailed Implementation
[0058] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0059] Please see Figure 1 This invention provides a technical solution: a method for active orientation history mapping-based suspected target hinting based on image segmentation, comprising the following steps:
[0060] (1) Input the active orientation history map, use the maximum inter-class variance method to calculate the gray-level segmentation threshold of the image pixels, and generate a binary image;
[0061] (2) Perform a dilation operation on the binary image using cross-shaped structural elements;
[0062] (3) Perform 8-connected region labeling on the dilated binary image to obtain a connected region labeled image;
[0063] (4) Calculate the area and centroid of each connected region in the connected region labeled image;
[0064] (5) Remove connected regions with an area greater than a preset threshold to obtain a discrete target region image;
[0065] (6) Traverse the non-zero centroids in the discrete target region image and calculate the sum of Euclidean distances from each centroid to its 5 nearest neighbor connected region centroids using breadth-first search (BFS);
[0066] (7) Sort the Euclidean distances in descending order, select the five largest Euclidean distances and their corresponding discrete target region centroids, and mark the centroid positions on the active orientation history map.
[0067] Specifically, the following steps are included:
[0068] Step 1: Input active orientation time map ,like Figure 2 As shown, the image is obtained using the Otsu's method. Pixel grayscale segmentation threshold Using segmentation threshold Obtain a binary image ,like Figure 3 As shown, the specific process includes the following:
[0069] (1.1) Calculation Pixel grayscale average .
[0070] (1.2) Traversing the segmentation threshold The traversal range is pixel grayscale range .
[0071] (1.3) Based on the threshold of the current traversal , to divide the grayscale range of pixels Divided into target grayscale range and background grayscale range .
[0072] (1.4) Calculate separately and The proportion of pixels to the total number of pixels , and grayscale mean , .
[0073] (1.5) Calculation and Between-class variance , can be represented as:
[0074]
[0075] (1.6) Repeat steps (2.2)-(2.5) to... Maximum The target grayscale range is determined as the final segmentation threshold. and background grayscale range Obtain a binary image .
[0076]
[0077] Step 2: Use a cross-shaped structural element right Perform an expansion operation, such as Figure 4 As shown, it can be represented as:
[0078]
[0079]
[0080] Step 3: Binarize the image By performing 8-connected component labeling, a connected component labeled image is obtained. ,like Figure 5 As shown, the specific process includes the following:
[0081] (3.1) Initialize the connected component labeling image , can be represented as:
[0082]
[0083] (3.2) Traversing the binarized image and connected component labeled image ,like and (Unlabeled foreground pixels), labeled ,Will Add to queue .
[0084] (3.3) When If not empty, retrieve the position of the first pixel in the queue. For all pixel positions within its 8-neighborhood (in and If it is within the image range, ,and Then mark and add it to the queue. .
[0085] (3.4) Clear the queue. (Update the marker value for the next region).
[0086] Step 4: Calculate the connected component label image Area of each connected region and center of mass Specifically, it includes the following processes:
[0087] (4.1) No. Area of each connected region Defined as the number of all pixels in this region, it can be expressed as:
[0088]
[0089] in , Image size, i.e. .
[0090] (4.2) No. The centroid of each connected region (Pixel coordinates, rounded down) is defined as the coordinates of all pixels within this region. Coordinate mean and The mean of the coordinates can be expressed as:
[0091]
[0092]
[0093] Step 5: Remove connected component markers from the image Large connected regions are used to obtain discrete target region images. ,like Figure 6 The area of the discrete target region is shown. and center of mass , can be represented as:
[0094]
[0095]
[0096]
[0097] Step 6: Traverse the non-zero centroids of the discrete target region Calculate the sum of its Euclidean distances to its 5 nearest connected regions. ,right Sort and label the original images. Centroids of the top 5 discrete target regions Specifically, it includes the following processes:
[0098] (6.1) Traverse the non-zero centroids of the discrete target region ,Will Add to queue , .
[0099] (6.2) When If not empty, retrieve the position of the first pixel in the queue. For all pixel positions within its 8-neighborhood (in , and If it falls within the image range, add it to the queue. If it is within the image range, and (in , For queue (The current number of non-zero coordinates) is added to the queue. Calculate distance , .
[0100] (6.3) When When, output the sum of distances. The queue is cleared, and the sum of Euclidean distances from the non-zero centroid of the next discrete target region to the five nearest connected regions is calculated.
[0101] (6.4) Sort in descending order (from largest to smallest), and the sorted result satisfies ,at this time .
[0102] (6.5) Mark on the original image The first 5 centroid positions, such as Figure 7 As shown.
[0103] Step 7: Sort the Euclidean distances in descending order, select the top 5 largest Euclidean distances and their corresponding discrete target region centroids, and mark the centroid positions on the active azimuth history map. The marking method includes drawing crosshairs, circles, or numbers on the active azimuth history map corresponding to the centroid positions to visualize the suspected target positions. After marking the centroids, output the suspected target position information, which includes the azimuth angle and the corresponding distance value in the active azimuth history map.
[0104] Through the technical solution of this application
[0105] (1) Morphological area filtering: This invention performs a cross-dilation operation on the binarized image, connects the fragmented targets, and then performs area threshold filtering to eliminate large-area interference, such as Figures 2 to 5 As shown, it can be observed that by performing dilation operation after segmentation using the maximum inter-class variance, connecting fragment targets, and then performing area threshold filtering, reverberation and large-area interference regions can be effectively removed.
[0106] (2) Spatial isolation calculation: This invention calculates the sum of Euclidean distances between the centroid of the discrete target region and its neighboring connected regions, utilizes the isolation characteristics of small targets, distinguishes targets from reverberation and noise interference, and marks isolated bright spots, such as... Figure 7 As shown, this invention first uses the BFS search method to calculate the sum of Euclidean distances to the centroids of the five nearest connected regions, sorts the calculation results, and marks the centroid positions of the five regions with the largest sum of Euclidean distances, which can be used to identify the location of isolated bright spots.
[0107] The above embodiments merely illustrate several implementation methods of this application, and while the descriptions are relatively specific and detailed, they should not be construed as limiting the scope of the invention patent. It should be noted that those skilled in the art can make various modifications and improvements without departing from the concept of this application, and these all fall within the protection scope of this application. Therefore, the protection scope of this patent application should be determined by the appended claims.
[0108] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.
Claims
1. A method for active orientation history mapping-based suspected target cues based on image segmentation, characterized in that: Includes the following steps: (1) Input the active orientation history map, use the maximum inter-class variance method to calculate the gray-level segmentation threshold of the image pixels, and generate a binary image; (2) Perform a dilation operation on the binary image using cross-shaped structural elements; (3) Perform 8-connected region labeling on the dilated binary image to obtain a connected region labeled image; (4) Calculate the area and centroid of each connected region in the connected region labeled image; (5) Remove connected regions with an area greater than a preset threshold to obtain a discrete target region image; (6) Traverse the non-zero centroids in the discrete target region image and calculate the sum of Euclidean distances from each centroid to its 5 nearest neighbor connected region centroids using breadth-first search (BFS); (7) Sort the Euclidean distances in descending order, select the five largest Euclidean distances and their corresponding discrete target region centroids, and mark the centroid positions on the active orientation history map.
2. The method for active orientation history mapping-based suspected target hinting based on image segmentation according to claim 1, characterized in that: According to step (1), input the active orientation history map. The image was obtained using the maximum inter-class variance method. Pixel grayscale segmentation threshold Using segmentation threshold Obtain a binary image The specific method is as follows: (1.1) Calculation Pixel grayscale average ; (1.2) Traversing the segmentation threshold The traversal range is pixel grayscale range ; (1.3) Based on the threshold of the current traversal , to divide the grayscale range of pixels Divided into target grayscale range and background grayscale range ; (1.4) Calculate separately and The proportion of pixels to the total number of pixels , and grayscale mean , ; (1.5) Calculation and Between-class variance , is represented as: (1.6) Repeat steps (1.2)-(1.5) to... Maximum The target grayscale range is determined as the final segmentation threshold. and background grayscale range Obtain a binary image , 。 3. The method for active orientation history mapping-based suspected target hinting based on image segmentation according to claim 2, characterized in that: The dilation operation of the cross-shaped structural element in step (2) is used to connect fragmented targets in the binary image, eliminating target breakage caused by noise. Specifically, a cross-shaped structural element is adopted. right The expansion operation is represented as: 。 4. The active orientation history map suspected target hinting method based on image segmentation according to claim 3, characterized in that: According to step (3), the binarized image By performing 8-connected component labeling, a connected component labeled image is obtained. Specifically, it includes the following processes: (3.1) Initialize the connected component labeling image , is represented as: ； (3.2) Traversing the binarized image and connected component labeled image ,like and (Unlabeled foreground pixels), labeled ,Will Add to queue ; (3.3) When If not empty, retrieve the position of the first pixel in the queue. For all pixel positions within its 8-neighborhood (in and If it is within the image range, ,and Then mark and add it to the queue. ; (3.4) Clear the queue. (Update the marker value for the next region).
5. The active orientation history map suspected target hinting method based on image segmentation according to claim 4, characterized in that: Based on step (4), calculate the connected component labeling image. Area of each connected region and center of mass Specifically, it includes the following processes: (4.1) No. Area of each connected region Defined as the number of all pixels in the region, expressed as: in , Image size, i.e. ; (4.2) No. The centroid of each connected region (Pixel coordinates, rounded down) is defined as the coordinates of all pixels within this region. Coordinate mean and The mean of the coordinates is expressed as: 。 6. The active orientation history map suspected target hinting method based on image segmentation according to claim 5, characterized in that: According to step (5), remove the connected component markers from the image. Large connected regions are used to obtain discrete target region images. Discrete target region area and center of mass , is represented as: 。 7. The active orientation history map suspected target hinting method based on image segmentation according to claim 6, characterized in that: According to step (6), traverse the non-zero centroids of the discrete target region. Calculate the sum of its Euclidean distances to its 5 nearest connected regions. ,right Sort and label the original images. Centroids of the top 5 discrete target regions Specifically, it includes the following processes: (6.1) Traverse the non-zero centroids of the discrete target region ,Will Add to queue , ; (6.2) When If not empty, retrieve the position of the first pixel in the queue. For all pixel positions within its 8-neighborhood (in , and If it falls within the image range, add it to the queue. If it is within the image range, and (in , For queue (The current number of non-zero coordinates) is added to the queue. Calculate distance , ; (6.3) When When, output the sum of distances. The queue is cleared, and the sum of Euclidean distances from the non-zero centroid of the next discrete target region to its five nearest connected regions is calculated. (6.4) Sort in descending order (from largest to smallest), and the sorted result satisfies ,at this time ; (6.5) Mark on the original image The first 5 centroid positions.
8. The active orientation history map suspected target hinting method based on image segmentation according to claim 7, characterized in that: The marking method in step (7) includes drawing crosshairs, circles or numbers on the corresponding centroid position on the active orientation history map to visualize the suspected target position.
9. The active orientation history map suspected target hinting method based on image segmentation according to claim 8, characterized in that: After marking the centroid in step (7), the location information of the suspected target is output. The location information includes the azimuth angle and the corresponding distance value in the active azimuth history map.