A ship target identification method and system based on an improved trisetum vulgare optimization algorithm

Abstract

The application provides a ship target recognition method and system based on an improved wild oat optimization algorithm, and relates to the technical field of target recognition.The technical points of the application include: acquiring an infrared image containing a ship target at sea;acquiring an optimal segmentation threshold value by using the improved wild oat algorithm, and segmenting the infrared image by using the optimal segmentation threshold value to acquire an optimal segmentation image; and post-processing the optimal segmentation image to acquire a ship target image; wherein the improvements of the improved wild oat algorithm include: improving an initial segmentation threshold value based on a histogram peak value; introducing a dynamic attenuation diffusion coefficient to improve a seed generation update formula; and introducing a connected region penalty term to improve a fitness function.The application enhances the anti-interference ability and stability of target recognition in a complex environment.The application is suitable for application scenarios such as infrared imaging, target recognition and other application scenarios requiring accurate image segmentation.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of target recognition, in particular to a ship target recognition method and system based on an improved wild oat optimization algorithm. BACKGROUND

[0002] With the continuous progress of infrared imaging technology and image feature extraction technology, their application in the field of high-light target recognition is increasingly integrated. In particular, in airborne infrared images, the application of image feature extraction technology makes the identification of high-light targets more accurate and efficient. By analyzing infrared images or radar images, the position, size and moving direction of ships can be automatically identified, which is crucial for optimizing route planning and traffic management. In addition, image processing technology also plays an important role in detecting abnormal behaviors of ships, such as illegal intrusion and collision risk, providing strong protection for maritime safety. Further, by identifying the activity patterns of ships, the development and utilization of marine resources can be evaluated, providing scientific basis for the sustainable management of marine resources. Integrating image feature extraction technology into ship monitoring systems can achieve accurate and efficient identification of marine ship targets, thereby providing strong technical support for maritime traffic monitoring, ship safety detection and marine resource investigation. The application of these technologies not only improves the safety and efficiency of maritime operations, but also lays a solid foundation for the rational development and utilization of marine resources.

[0003] However, the existing target recognition method has the problems of incomplete segmentation of small targets in complex background, excessive dominance of large background area and strong parameter dependence, which further leads to low accuracy of ship target recognition. SUMMARY

[0004] In view of the above problems, the present application proposes a ship target recognition method and system based on an improved wild oat optimization algorithm.

[0005] According to an aspect of the present application, a ship target recognition method based on an improved wild oat optimization algorithm is proposed, which comprises:

[0006] obtaining an infrared image containing a marine ship target;

[0007] using the improved wild oat algorithm to obtain an optimal segmentation threshold, and using the optimal segmentation threshold to segment the infrared image to obtain an optimal segmentation image;

[0008] post-processing the optimal segmentation image to obtain a ship target image.

[0009] Further, the improvements of the improved wild oat algorithm include: improving the initial segmentation threshold based on the histogram peak value; introducing a dynamic decay diffusion coefficient to improve the seed generation update formula; and introducing a connected region penalty term to improve the fitness function.

[0010] Furthermore, the improved formula for the initial segmentation threshold is:

[0011] ;

[0012] Where H(b) represents the count of the image histogram in the b-th pixel interval; B represents the set of pixel intervals; This represents the second derivative of the histogram in the b-th pixel interval.

[0013] Furthermore, the improved seed generation and update formula is as follows:

[0014] ;

[0015] in, This represents the i-th seed in the (t+1)-th iteration; This represents the optimal seed for the t-th iteration; A random number uniformly distributed in the interval [0,1]. This represents the diffusion coefficient at the t-th iteration. , Here, t is the decay factor, and t is the current iteration number. The initial diffusion coefficient; The function is used to restrict the seed value to a reasonable range, and its mathematical definition is as follows:

[0016] .

[0017] Furthermore, the improved fitness function formula is as follows:

[0018] ;

[0019] in, This represents the number of pixels in the connected regions obtained after each iteration of image segmentation. This represents the total number of pixels in the image. To adjust the weight of the penalty term by controlling the parameters; The inter-class variance is between the target and the background.

[0020] Furthermore, the post-processing of the optimal segmented image includes: deleting the largest connected region in the optimal segmented image; removing blemishes and filling holes.

[0021] According to another aspect of the present invention, a ship target recognition system based on an improved wild oat optimization algorithm is proposed to implement the aforementioned ship target recognition method based on an improved wild oat optimization algorithm; the system includes:

[0022] An image acquisition module configured to acquire infrared images containing maritime targets;

[0023] a most optimal segmentation threshold determination module configured to obtain a most optimal segmentation threshold by using the improved wild oat optimization algorithm, and to segment the infrared image by using the most optimal segmentation threshold to obtain a most optimal segmentation image;

[0024] a target recognition module configured to post-process the most optimal segmentation image to obtain a ship target image.

[0025] According to another aspect of the present application, a terminal device is provided, comprising a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, and the processor implements the ship target recognition method based on the improved wild oat optimization algorithm when executing the computer program.

[0026] According to another aspect of the present application, a computer readable storage medium is provided, comprising a stored computer program, wherein the computer readable storage medium controls the device where the computer readable storage medium is located to execute the ship target recognition method based on the improved wild oat optimization algorithm when the computer program runs.

[0027] The present application has the following beneficial technical effects:

[0028] The present application provides a ship target recognition method and system based on the improved wild oat optimization algorithm, which has many significant advantages compared with the prior art. The present application uses the improved wild oat optimization algorithm to obtain the most optimal segmentation threshold, and uses the most optimal segmentation threshold to segment the infrared image to obtain the most optimal segmentation image. The most optimal segmentation image is post-processed to obtain a ship target image. In the improved wild oat optimization algorithm, the initialization method of histogram peak value analysis is introduced to effectively solve the convergence instability problem caused by traditional random initialization. Secondly, the introduction of the dynamic attenuation mechanism in the seed update formula enables the algorithm to automatically reduce the search range in the later iteration, which has a faster convergence speed compared with the traditional fixed step method. Thirdly, the improved fitness function effectively avoids the over-segmentation phenomenon in complex background by adding connectivity constraints. Finally, the post-processing process effectively overcomes the defect of large-area background dominating the segmentation result in the traditional method by using the maximum region suppression, and cooperates with the hole filling operation to greatly enhance the anti-interference ability and stability of target recognition in complex environment. The present application is suitable for application scenarios such as infrared imaging and target recognition that require accurate image segmentation. BRIEF DESCRIPTION OF DRAWINGS

[0029] The above and other objects, features and advantages of the present application will become more apparent from the following detailed description when taken in conjunction with the accompanying drawings in which a number of embodiments of the present application are illustrated by way of example and not limitation. In the drawings:

[0030] Figure 1 is a flow chart of a ship target recognition method based on an improved wild oat optimization algorithm according to an embodiment of the present application;

[0031] Figure 2 is a gray scale image of an airborne infrared image according to an embodiment of the present application;

[0032] Figure 3 is an example diagram of a segmentation result of an existing wild oat algorithm according to an embodiment of the present application;

[0033] Figure 4 is an example diagram of a segmentation result of an improved wild oat algorithm according to an embodiment of the present application;

[0034] Figure 5 is an example diagram after removing the largest region in post-processing according to an embodiment of the present application;

[0035] Figure 6 is an example diagram of a final target recognition result after removing noise and filling holes in post-processing according to an embodiment of the present application. DETAILED DESCRIPTION

[0036] The principles and spirits of the present application will be described below with reference to several exemplary embodiments. It should be understood that these embodiments are given only to enable those skilled in the art to better understand and implement the present application, and do not limit the scope of the present application in any way. On the contrary, these embodiments are provided to make the present disclosure more thorough and complete, and to fully convey the scope of the present disclosure to those skilled in the art.

[0037] Those skilled in the art know that the embodiments of the present application can be implemented as a system, device, apparatus, method or computer program product. Therefore, the present disclosure can be embodied in the form of entire hardware, entire software (including firmware, resident software, microcode, etc.), or a combination of hardware and software. In this context, it should be understood that any number of elements in the drawings are used for illustration only and not limitation, and any naming is only for distinction and does not have any limiting meaning.

[0038] The present application proposes a ship target recognition method and system based on an improved wild oat optimization algorithm to meet the actual needs of accurate extraction of target regions in aerial infrared image processing. The present application can effectively distinguish and segment the real target region and non-target region, improve the accuracy and efficiency of water target recognition, and solve the problems of incomplete segmentation of small targets, excessive dominance of large background regions and strong parameter dependence of traditional segmentation algorithms in complex background.

[0039] An embodiment of the present application proposes a ship target recognition method based on an improved wild oat optimization algorithm, as shown in the following formula: Figure 1As shown, the method comprises:

[0040] S1, acquiring an infrared image containing a marine ship target;

[0041] S2, acquiring an optimal segmentation threshold value by using an improved wild oat algorithm, and segmenting the infrared image by using the optimal segmentation threshold value to acquire an optimal segmentation image; the improvement of the improved wild oat algorithm comprises: improving an initial segmentation threshold value based on a histogram peak value; introducing a diffusion coefficient with dynamic attenuation to improve a seed generation update formula; and introducing a connected region penalty term to improve a fitness function;

[0042] S3, post-processing the optimal segmentation image to acquire a ship target image.

[0043] The method starts from S1, in which an infrared image containing a marine ship target is acquired.

[0044] According to an embodiment of the present application, the infrared image is an image containing a marine ship taken by an unmanned aerial vehicle carrying an infrared camera in the air, Figure 2 A gray-scale image of the airborne infrared image is shown. As can be seen from Figure 2 , the airborne infrared image contains the reflection of multiple ships on the water surface, the overall light of the image is dark, and the high-light ship target is obviously visible; however, although the ship target is recognizable, it is difficult to accurately distinguish the target from the background due to the low contrast between the background and the ship target.

[0045] Then, in S2, an optimal segmentation threshold value is acquired by using an improved wild oat algorithm, and the infrared image is segmented by using the optimal segmentation threshold value to acquire an optimal segmentation image.

[0046] According to an embodiment of the present application, the basic algorithm adopts a wild oat (WOA) algorithm. The present application improves the existing wild oat algorithm as follows:

[0047] 1) The traditional random initialization threshold method is improved by an initialization strategy based on histogram peak value analysis.

[0048] In the existing wild oat algorithm, the initialization is random initialization of the initial segmentation threshold value: ; the present application initializes based on histogram peak value detection, that is:

[0049]

[0050] Wherein, H(b) is the count of the image histogram in the bth pixel interval; B is a pixel interval set; is the second derivative (measuring the convexity of the curve) of the histogram in the bth pixel interval. If the second derivative is less than 0, it indicates that the pixel interval is at a local peak (i.e. the histogram has a maximum value near the point).

[0051] It's important to note that each pixel interval can be called a bin. For grayscale images, pixel values ​​can be divided into several intervals called bins, and then the number of pixels within each interval is counted. For example, if pixel values ​​from 0 to 255 are divided into 256 bins, then each bin represents a grayscale value (0, 1, 2, ..., 255); if 0 to 255 are divided into 16 bins, then each bin represents 16 grayscale value ranges (0-15, 16-31, ..., 240-255). H(b) represents the count (number of pixels) of bin b in the histogram. For example, if bin b represents grayscale values ​​from 100 to 115, then H(b) is the total number of pixels in the image with grayscale values ​​in the range of 100 to 115. B is the set of all possible candidate bins used to calculate the initial threshold. In actual calculations, not all bins may be used; instead, certain specific bins may be selected (e.g., only bins with higher grayscale values ​​are considered). For example, in infrared images, targets are usually bright areas, so B may only contain intervals with higher grayscale values.

[0052] In the improved wild oat algorithm, the initial threshold is not randomly selected. Instead, it is determined by analyzing the image histogram to find the most likely peak point (i.e., the boundary between the target and the background). Specifically, it identifies the pixel interval with the smallest second derivative (i.e., the most significant peak point) among all pixel intervals and uses it as the initial segmentation threshold. By locating the most significant peak point in the histogram as the initial solution, the algorithm can start searching from a region close to the optimal solution each time. This fundamentally solves the convergence instability problem caused by the randomness of the initial value, significantly reduces the number of iterations, and improves computational efficiency.

[0053] 2) Improve the seed generation and update formula by introducing a dynamically decaying diffusion coefficient.

[0054] The existing wild oat algorithm's seed update formula is:

[0055]

[0056] in, This represents the i-th seed in the (t+1)-th iteration; This represents the optimal seed for the t-th iteration; The diffusion coefficient is... (constant); A random number uniformly distributed in the interval [0,1].

[0057] This invention improves the seed update formula as follows:

[0058]

[0059] in, denotes the diffusion coefficient at the t-th iteration, , is the decay factor (in this embodiment = 0.73, i.e. alpha is reduced to 73% of the current value each iteration, t is the current iteration number, is the initial diffusion coefficient; The function of the clip function is to limit the seed value within a reasonable range (such as [0, 1] or the image gray value range) to avoid out-of-bounds. The mathematical definition of the clip function is as follows:

[0060]

[0061] In the traditional wild oat algorithm, the diffusion coefficient a is a fixed value, which means that the search range does not change, and the generation range of the seed is always the same in each iteration. If a is too large, the algorithm may oscillate around the optimal solution and cannot be fine-tuned; if a is too small, the algorithm may converge too slowly, especially when the initial solution is far from the optimal solution. The present application replaces the fixed diffusion coefficient with an exponentially decaying dynamic coefficient and introduces the boundary limiting function clip to achieve intelligent adaptive control of the optimization process. The dynamic decay mechanism allows the algorithm to maintain a large step size in the early iterations, quickly explore the global solution, quickly locate the potential optimal solution area, and effectively avoid falling into a local optimum; in the later iterations, the step size is automatically contracted for fine local development, significantly improving the convergence accuracy and speed; this adaptive search strategy perfectly balances the contradiction between global exploration and local development, and compared with the fixed step size method, it achieves faster convergence speed and higher segmentation accuracy while ensuring that it does not fall into a local optimum. The clip function ensures that the generated solution is always within a reasonable range, enhancing the stability and practicality of the algorithm. The combination of the two solves the contradiction between exploration and utilization in the traditional method, significantly improving the convergence efficiency, global optimization ability, and robustness of the algorithm.

[0062] 3) Improve the fitness function by introducing a connected region penalty term.

[0063] For image segmentation, the corresponding existing fitness function is generally:

[0064]

[0065] wherein, is the inter-class variance, , is the proportion of two classes of pixels, , is the variance of two classes of pixels; the inter-class variance is used to measure the dispersion degree of pixel values within two classes (target and background) after segmentation, and the smaller it is, the more similar the pixels of the same class are. The above fitness function only focuses on the distribution of pixel values and ignores the spatial continuity, resulting in a large number of isolated small regions (noise) in the segmentation result. In a complex background, a fragmented segmentation result may be generated.

[0066] The fitness function introduced by the present application includes a connected region penalty term:

[0067]

[0068] wherein, is the number of connected region pixels obtained by image segmentation after each iteration, is the total number of pixels of the image, is an adjustment parameter for controlling the weight of the penalty term, used to balance the inter-class variance and connectivity. Too small will make the penalty term weak, and it may not effectively suppress the fragmented region; too large may be over-smoothed, losing small targets, and in actual application, the size of the adjustment parameter needs to be adjusted according to the actual situation to ensure that the connectivity constraint is effective and does not dominate the optimization process.

[0069] By introducing the connectivity penalty term, the algorithm can effectively reduce the fragmented region. If the segmentation result contains a large number of small connected regions (i.e. is large), the penalty term increases, and the fitness becomes worse. In an ideal segmentation, the target should be a small number of continuous regions (such as a ship as one connected block), and the background is another connected block. The improved fitness function proposed by the present application combines the pixel gray distribution and the spatial continuity constraint. The algorithm after improving the fitness function is guided to the threshold solution space that can maximize the area of the connected region, minimize the number of regions, and enhance the spatial continuity in the optimization process, preferentially selects the threshold with good connectivity, so that the algorithm is more inclined to select the segmentation result with fewer connected regions, thereby suppressing noise, better preserving the complete target through the connectivity penalty term, and effectively avoiding the over-segmentation phenomenon in a complex background, ensuring that the target entity in the output result is complete and the contour is clear.

[0070] An optimal segmentation threshold is obtained by using the improved wild oat algorithm, and the optimal segmentation threshold is used to segment the infrared image to obtain an optimal segmentation image.

[0071] Then, in S3, the optimal segmentation image is post-processed to obtain a ship target image.

[0072] According to the embodiment of the present application, the post-processing includes: 1) maximum region suppression, that is, deleting the largest connected region (usually the land or sky region) in the optimal segmentation image; 2) removing the speckle and performing hole filling, which reduces the interference and fills the holes in the target region to improve the target integrity. The post-processing step improves the effect of target recognition on the basis of improving the wild oat algorithm, effectively eliminates the noise and improves the target integrity through the maximum region filling and hole filling technology.

[0073] Figure 3 is an example of the segmentation result of the existing wild oat algorithm. It can be seen that after the traditional wild oat algorithm processing, the segmentation result shows the outline of the highlighted target, but there are still many noises and regions that are not completely segmented. These small targets and backgrounds are mixed together, resulting in poor segmentation accuracy and effect, especially in the target recognition of the details. Figure 4 is an example of the improved wild oat algorithm segmentation result of the present application. It can be seen that by using the improved wild oat image segmentation method based on dynamic threshold optimization and region filling, the final effect is improved obviously, more small target features are successfully retained, and the recognition accuracy and segmentation clarity of the target region are significantly improved; the background noise is effectively suppressed, the highlighted target region of the segmentation is more obvious, and the fine features are well retained.

[0074] Figure 5 、 Figure 6 is an example of the post-processing result of the present application. It can be seen that after the maximum region filling processing, the outline of the target region becomes more complete, filling the part of the blank area appearing in the previous segmentation, thereby enhancing the overall display effect of the target. Compared with the traditional method, this processing effectively eliminates the small segmentation defects and provides a more coherent target morphology. The image is subjected to speckle removal, and then the hole filling of the target region is performed, further improving the quality of the image. The details of the original target are well retained in the filling process, and the contrast between the black background and the target region is more obvious, making the image of the ship more prominent. The successful implementation of the hole filling helps to ensure the integrity of the target and eliminates the visual interference caused by noise or incomplete segmentation.

[0075] Another embodiment of the present application provides a ship target recognition system based on an improved wild oat optimization algorithm, which is used to implement the ship target recognition method based on the improved wild oat optimization algorithm described in the above embodiment; the system comprises:

[0076] an image acquisition module configured to acquire an infrared image containing a ship target at sea;

[0077] An optimal segmentation threshold determination module configured to obtain an optimal segmentation threshold by using the improved wild oat optimization algorithm, and segment the infrared image by using the optimal segmentation threshold to obtain an optimal segmentation image;

[0078] A target recognition module configured to post-process the optimal segmentation image to obtain a ship target image.

[0079] The function of the ship target recognition system based on the improved wild oat optimization algorithm according to the embodiment of the application can be described by the ship target recognition method based on the improved wild oat optimization algorithm described above, and thus the unexplained part of the system embodiment can be referred to the method embodiment described above, and will not be described here.

[0080] Another embodiment of the application provides a terminal device comprising a processor, a memory, and a computer program stored in the memory and configured to be executed by the processor, wherein the processor implements the ship target recognition method based on the improved wild oat optimization algorithm according to the above embodiment when executing the computer program.

[0081] Another embodiment of the application provides a computer readable storage medium comprising a stored computer program, wherein the computer readable storage medium controls the device where the computer readable storage medium is located to execute the ship target recognition method based on the improved wild oat optimization algorithm according to the above embodiment when the computer program runs.

[0082] It should be noted that although several units, modules or sub-modules are mentioned in the above detailed description, such division is only exemplary and not mandatory. In fact, according to the embodiments of the application, the features and functions of two or more modules described above can be embodied in one module. Conversely, the features and functions of one module described above can be further divided into several modules embodied.

[0083] In addition, although the operations of the method of the application are described in a specific order in the accompanying drawings, this does not require or imply that the operations must be performed in this specific order, or that all the shown operations must be performed to achieve the desired result. Additionally or alternatively, some steps can be omitted, a plurality of steps can be combined into one step, and / or one step can be divided into a plurality of steps.

[0084] Although the spirit and principles of the application have been described with reference to several specific embodiments, it should be understood that the application is not limited to the disclosed specific embodiments, and the division of aspects does not mean that the features in these aspects cannot be combined to benefit. Such division is only for the convenience of expression. The application is intended to cover various modifications and equivalent arrangements included in the spirit and scope of the appended claims.

Claims

1. A ship target recognition method based on an improved wild oat optimization algorithm, characterized in that, The method comprises: acquiring an infrared image containing a marine ship target; acquiring an optimal segmentation threshold value by using an improved wild oat algorithm, and segmenting the infrared image by using the optimal segmentation threshold value to acquire an optimal segmentation image; post-processing the optimal segmentation image to acquire a ship target image; the improvement of the improved wild oat algorithm comprises: improving an initial segmentation threshold value based on a histogram peak value; introducing a dynamic attenuation diffusion coefficient to improve a seed generation update formula; and introducing a connected region penalty term to improve a fitness function; the formula of the improved initial segmentation threshold value is: ； Wherein, H(b) represents the count of the image histogram in the bth pixel interval; B represents the pixel interval set; represents the second derivative of the histogram in the bth pixel interval; the improved seed generation update formula is: ； wherein, denotes the i-th seed of the t+1-th iteration; denotes the best seed of the t-th iteration; is a random number uniformly distributed in the interval [0, 1]; denotes the diffusion coefficient at the t-th iteration, , is a decay factor, t is the current iteration number, is the initial diffusion coefficient; the function is used to limit the seed value within a reasonable range, and its mathematical definition is as follows: ； the improved fitness function formula is: ； wherein, is the number of connected region pixels obtained after image segmentation for each iteration; is the total number of pixels of the image; is the weight of the penalty term controlled by the adjustment parameter; is the inter-class variance of the target and background.

2. The ship target identification method based on the improved wild oat optimization algorithm according to claim 1, characterized in that, the post-processing of the optimal segmentation image comprises: deleting the largest connected region in the optimal segmentation image; and removing speckles and performing hole filling.

3. A ship target identification system based on improved wild oat optimization algorithm, used to implement the ship target identification method based on improved wild oat optimization algorithm in claim 1 or 2, characterized in that, The method comprises: an image acquisition module configured to acquire an infrared image containing a marine ship target; an optimal segmentation threshold value determination module configured to acquire an optimal segmentation threshold value by using an improved wild oat algorithm, and segment the infrared image by using the optimal segmentation threshold value to acquire an optimal segmentation image; a target recognition module configured to post-process the optimal segmentation image to acquire a ship target image.

4. A terminal device, characterized by comprising: The computer readable storage medium comprises a stored computer program, wherein the computer program controls the device where the computer readable storage medium is located to execute the ship target recognition method based on the improved wild oat optimization algorithm when the computer program is running.

5. A computer readable storage medium, characterized in that, The computer readable storage medium comprises a stored computer program, wherein the computer program controls the device where the computer readable storage medium is located to execute the ship target recognition method based on the improved wild oat optimization algorithm when the computer program is running.

Images