Infrared ship positioning method based on line inhibition

Abstract

The application discloses an infrared ship positioning method based on line inhibition, and defines high-frequency components and low-frequency components of an infrared ship image to obtain a frequency domain analysis graph; the high-frequency components include edges of clouds, sea-sky lines, wave edges and ship edges; the low-frequency components include smooth areas surrounded by the sky and cloud edges and smooth areas surrounded by the ship edges; longitudinal high-frequency component inhibition is performed on the frequency domain analysis graph to optimize the infrared ship image; transverse gradient transformation is performed on the optimized infrared ship image to obtain a potential ship target area; and ship edge detection is performed on the potential ship target area to realize positioning of the ship target. The method can inhibit background interference such as clouds, sea-sky lines and waves through longitudinal high-frequency component inhibition, and then calculate the transverse gradient of the image after the background is inhibited, and the potential ship target area is obtained through transverse projection, and finally, the positioning of the ship target is realized through combination of canny edge detection and morphological transformation.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of ship positioning, and particularly relates to an infrared ship positioning method based on line suppression. BACKGROUND

[0002] With the development of human science and technology, the importance of marine resources is gradually highlighted, and marine transportation, marine resource exploitation, etc. gradually become an important part of people's life. However, due to the gradual dependence of modern society on the sea, marine navigation, marine personnel search and rescue, piracy or military conflict, etc. Marine safety issues are increasingly prominent. Therefore, effectively detecting or positioning marine ship targets has important practical significance for searching for ships in a wide sea area, fighting pirates and protecting the safety of territorial waters.

[0003] Among various sensors for detecting ships, infrared detectors have the advantages of long observation distance, strong fog penetration ability, etc. They also have good imaging and strong secrecy, and are therefore widely used in marine ship detection. Current ship detection methods for infrared images mainly fall into three categories: background modeling-based methods, threshold-based methods and feature analysis-based methods.

[0004] In the background modeling-based method, existing research combines fractional Fourier transform with high-order statistical filtering to suppress sea clutter, and then uses double-threshold segmentation and morphological transformation to detect ships. There is also a method of integrating global distribution information into a Gaussian model to suppress high-intensity background, and then introducing local variance and adjacent membership to improve fuzzy C-means clustering to achieve automatic detection of ship targets. These background modeling-based methods have good suppression effect on infrared images of static sea surface background. However, for infrared images with a large number of cloud and sea wave interference, the detection accuracy of the algorithm will decrease significantly.

[0005] In the threshold-based method, existing research has suppressed background intensity through gray histogram stretching and Gaussian filtering, and then used an adaptive binarization method to obtain the best threshold to locate the ship target. There is also a method of enhancing target information through a visual saliency model based on graph theory, determining the target potential area through prior knowledge, and finally locating the target position through clustering. The threshold-based method has the advantages of fast detection speed and easy implementation, but only considers gray information, which makes it difficult to extract ship targets in complex backgrounds and low target-to-background contrast.

[0006] In the feature analysis-based method, the existing research fuses three features of multi-scale mean, infrared multi-stage filtering and local gray maximum, uses adaptive threshold and morphological transformation to realize ship segmentation. A saliency map is constructed by comprehensively constructing multiple features of the ship target, and then an adaptive threshold and shape feature are used to obtain the detection target. Three features of intensity feature, local spatial feature and global spatial feature are constructed, and prior knowledge is used to integrate the three features into a fuzzy inference system to directly segment the ship target. The infrared image is converted into a structure tensor feature map, then the sky and land interference is removed by detecting the horizontal line, and then an adaptive maximum histogram entropy based on the gray distribution of the image is proposed to segment the ship target. The feature analysis-based method has achieved good detection effect in some low-contrast or complex background scenes, but the multi-feature method comprehensively judges multiple features, which may interact or overlap information, causing feature redundancy and affecting detection effect. In addition to the above three methods, the existing research also proposes a convex active contour model based on local entropy, which segments the infrared ship target through level set, but when the sea clutter interference is strong, it is difficult to realize the segmentation of the ship target.

[0007] In summary, although the above methods can detect the ship target in the infrared image, when the infrared image has a large number of clouds, sea wave interference, and has a clear sea-sky line, it cannot effectively realize the positioning of the ship target. SUMMARY

[0008] The present application provides an infrared ship positioning method based on line suppression to overcome the technical problem that the conventional method can detect the ship target in the infrared image, but when the infrared image has a large number of clouds, sea wave interference, and has a clear sea-sky line, it cannot effectively realize the positioning of the ship target.

[0009] In order to achieve the above purpose, the technical scheme of the present application is:

[0010] An infrared ship positioning method based on line suppression, comprising the following steps:

[0011] S1: defining the high-frequency component and the low-frequency component of the infrared ship image, and obtaining a frequency domain analysis graph;

[0012] The high-frequency component includes the edge of the cloud, the sea-sky line, the edge of the sea wave and the edge of the ship; the low-frequency component includes the smooth area surrounded by the cloud edge and the smooth area surrounded by the ship edge;

[0013] S2: suppressing the longitudinal high-frequency component of the frequency domain analysis graph, and optimizing the infrared ship image;

[0014] S3: performing a horizontal gradient transformation on the optimized infrared ship image to obtain a potential area of the ship target.

[0015] S4: ship edge detection is performed on the ship target potential area to realize positioning of the ship target.

[0016] Further, the frequency domain analysis graph in S2 is subjected to longitudinal high-frequency component suppression, specifically

[0017] S2.1: the frequency domain analysis graph is subjected to grid division to obtain a plurality of initial images; and fast Fourier transform is performed on each column of image data x(n) of the initial images to obtain frequency domain component values of each row;

[0018] and the frequency domain component values of each row are arranged in descending order;

[0019] The calculation formula X(k) of the fast Fourier transform is

[0020]

[0021] In the formula, k=0, 1, …, N-1; N represents the number of pixels in a column of the image; j represents an imaginary number; and n represents the position of the data;

[0022] S2.2: based on a preset frequency domain component threshold, low-pass filtering processing is performed on the frequency domain components arranged in descending order; and each row of frequency domain components greater than the preset frequency domain component threshold is removed, and each row of frequency domain components less than or equal to the preset frequency domain component threshold is retained;

[0023] The expression of the low-pass filtering processing is

[0024]

[0025] In the formula, R represents an image area of a low-frequency component; and Y(k) represents a filtering result;

[0026] S2.3: the optimized infrared ship image after the low-pass filtering processing is obtained through inverse Fourier transform:

[0027] The calculation formula of the inverse Fourier transform is

[0028]

[0029] In the formula, k=0, 1, …, N-1; N represents the number of pixels in a column of the image; j represents an imaginary number; and n represents the position of the data.

[0030] Further, the infrared ship image in S3 is subjected to horizontal gradient transformation to obtain a ship target potential area; specifically

[0031] S3.1: performing transverse gradient calculation on the optimized infrared ship image to obtain an infrared ship gradient image, separating a ship class in the infrared ship gradient image from an image background based on a minimum intra-class variance method to obtain a binary infrared ship image;

[0032] The ship class includes a ship chord edge class and a ship island edge class.

[0033] The transverse gradient calculation formula is

[0034] z(n) = (z(n+1)-z(n-1)) / 2

[0035] In the formula, n represents an arbitrary point position of a row in the infrared ship image; and z(n) represents a pixel value of a current pixel.

[0036] S3.2: performing horizontal direction projection on the binary infrared ship image to obtain a gradient sum of each row in the binary infrared ship image, and establishing a rectangular coordinate system with the number of rows of the binary infrared ship image as the horizontal coordinate and the gradient sum corresponding to each row of the binary infrared ship image as the vertical coordinate.

[0037] Based on the rectangular coordinate system, the gradient sum of each row in the binary infrared ship image is curve-shaped to obtain a plurality of transverse gradient curves.

[0038] S3.3: calculating an interval gradient sum of an interval formed by each transverse gradient curve and the horizontal coordinate axis, and selecting an endpoint of an interval gradient sum maximum interval.

[0039] According to the endpoint of the interval gradient sum maximum interval, a potential region of a ship target in the binary infrared ship image is determined.

[0040] Further, the calculation formula of the minimum intra-class variance method in S3.1 is

[0041]

[0042] In the formula, a and b respectively represent the ship chord edge class and the ship island edge class. represents an intra-class variance of the ship chord edge class. represents an intra-class variance of the ship island edge class; Wa represents a ratio of the number of pixels of the ship chord edge class to the total number of pixels; Wb represents a ratio of the number of pixels of the ship island edge class to the total number of pixels; Ra and Rb respectively represent a set of pixels of the ship chord edge class and the ship island edge class; ya i and yb i respectively represent pixel values of the pixels of the ship chord edge class and the ship island edge class; μa and μb respectively represent pixel mean values of the ship chord edge class and the ship island edge class; and T is a threshold value for separating the ship chord edge class and the ship island edge class.

[0043] Further, the ship edge detection is performed on the ship target potential area in S4, so as to realize the positioning of the ship target, specifically

[0044] S4.1: performing ship edge detection on the ship target potential area by using the canny edge detection technology, to obtain a ship edge detection image;

[0045] S4.2: performing horizontal expansion of a preset pixel value on the ship edge detection image based on morphological change;

[0046] S4.3: performing connected domain detection on the horizontally expanded ship edge detection image based on the connected domain detection technology, to obtain a plurality of rectangular connected domain areas;

[0047] S4.4: calculating the area of each rectangular connected domain area, selecting the rectangular connected domain area with the largest area as the circumscribed contour of the ship target, and realizing the positioning of the ship target according to the circumscribed contour of the ship target.

[0048] Beneficial effects: the application provides an infrared ship positioning method based on line suppression, the high-frequency component and the low-frequency component of an infrared ship image are defined to obtain a frequency domain analysis image; the high-frequency component includes the edge of a cloud, a sea-sky line, a sea wave edge and a ship edge; the low-frequency component includes a smooth area surrounded by the sky and the cloud edge, and a smooth area surrounded by the ship edge; the frequency domain analysis image is subjected to longitudinal high-frequency component suppression, and the infrared ship image is optimized; the optimized infrared ship image is subjected to horizontal gradient transformation, to obtain a ship target potential area; the ship edge detection is performed on the ship target potential area, to realize the positioning of the ship target. The cloud, the sea-sky line, the sea wave and other background interference are suppressed by longitudinal high-frequency component suppression, the horizontal gradient of the image after the background suppression is calculated, the ship target potential area is obtained by horizontal projection, and finally the positioning of the ship target is realized by combining the canny edge detection and the morphological transformation. BRIEF DESCRIPTION OF DRAWINGS

[0049] In order to more clearly illustrate the technical solutions in the embodiments of the present application or the prior art, the following will briefly introduce the drawings needed to be used in the embodiments or the prior art description. Obviously, the drawings in the following description are some embodiments of the present application, and other drawings can also be obtained by those skilled in the art without any creative labor.

[0050] Figure 1 The flow chart of the infrared ship positioning method based on line suppression of the present application;

[0051] Figure 2 The result images of each stage of the infrared ship positioning method based on line suppression in the present embodiment;

[0052] Figure 3 The infrared image ship positioning result graph for different ship types in the embodiment;

[0053] Figure 4 The infrared image ship positioning result graph in the presence of multiple interference in the embodiment;

[0054] Figure 5 The infrared image ship positioning result graph in the presence of a certain inclination angle in the embodiment;

[0055] Figure 6 The transverse gradient curve graph in the embodiment;

[0056] Figure 7 The flow chart of the infrared image ship positioning method based on line suppression in the embodiment. DETAILED DESCRIPTION

[0057] To make the objects, technical solutions and advantages of the embodiments of the present application clearer, the technical solutions in the embodiments of the present application will be described below in connection with the drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, rather than all the embodiments. Based on the embodiments in the present application, all other embodiments obtained by those of ordinary skill in the art without creative work fall within the protection scope of the present application.

[0058] The embodiment provides an infrared ship positioning method based on line suppression, which comprises the following steps as shown in Figure 1 and Figure 7 .

[0059] S1: defining the high-frequency component and the low-frequency component of the infrared ship image, and obtaining a frequency domain analysis graph;

[0060] The high-frequency component comprises the edge of a cloud, a sea-sky line, the edge of a sea wave and the edge of a ship; and the low-frequency component comprises a smooth area surrounded by the edge of a cloud and a smooth area surrounded by the edge of a ship.

[0061] S2: performing longitudinal high-frequency component suppression on the frequency domain analysis graph, and optimizing the infrared ship image;

[0062] S3: performing transverse gradient transformation on the optimized infrared ship image, and obtaining a potential ship target area;

[0063] S4: performing ship edge detection on the potential ship target area, and realizing ship target positioning.

[0064] The present application aims at the problem that the ship target is difficult to detect when the infrared image has a large number of clouds, sea waves and obvious sea-sky line interference, and gives an infrared image ship positioning method based on row suppression. In order to suppress the background interference such as clouds, sea-sky line and sea waves, a longitudinal high-frequency component suppression algorithm is given; then the transverse gradient is calculated, the clouds and sea waves are removed by transverse projection, and the potential area of the ship target is obtained; finally, the interference on the sea level is removed by combining the canny edge detection and morphological transformation, and the positioning of the ship target is realized.

[0065] In specific embodiments, the frequency domain analysis graph in S2 is subjected to longitudinal high-frequency component suppression, specifically S2.1: the frequency domain analysis graph is subjected to grid division to obtain a plurality of initial images of rows and columns; and the fast Fourier transform is performed on each column image data x(n) of the initial image to obtain each row frequency domain component value;

[0066] and each row frequency domain component value is arranged in descending order;

[0067] The calculation formula X(k) of the fast Fourier transform is

[0068]

[0069] In the formula, k=0, 1, …, N-1; N represents the number of pixels in a column of the image; j represents an imaginary number; n represents the position of the data;

[0070] S2.2: based on a preset frequency domain component threshold, the low-pass filtering processing is performed on each row frequency domain component arranged in descending order; and each row frequency domain component greater than the preset frequency domain component threshold is removed, and each row frequency domain component less than or equal to the preset frequency domain component threshold is retained;

[0071] The expression of the low-pass filtering processing is

[0072]

[0073] In the formula, R represents the image area of the low-frequency component; Y(k) represents the filtering result;

[0074] S2.3: the optimized infrared ship image after the low-pass filtering processing is obtained by inverse Fourier transform:

[0075] The calculation formula of the inverse Fourier transform is

[0076]

[0077] In the formula, k=0, 1, …, N-1; N represents the number of pixels in a column of the image; j represents an imaginary number; n represents the position of the data.

[0078] In the field of image processing, high frequency components and low frequency components are usually defined by frequency domain analysis, in which Fourier transform is often used. High frequency components refer to parts of an image that change more dramatically and are rich in details, which correspond to higher frequency components in the frequency domain. These parts may contain information such as textures, edges, and details in the image. Low frequency components refer to parts of an image that change relatively slowly and have overall trends, which correspond to lower frequency components in the frequency domain. These parts usually contain smooth areas in the image.

[0079] As shown in Figure 2 , in an infrared image, the edges of clouds, the horizon, waves, and the edges of a ship belong to high frequency components, while the uniform areas inside the clouds, the sky, and the ship belong to low frequency components. However, due to the low contrast of infrared images, clouds, the horizon, and waves have similar linear structures to the ship target, which greatly affects the detection of the ship target. For example, in the selected area in Figure 2 (a), if the infrared image is directly subjected to canny edge detection, clouds, the horizon, and waves will all be incorrectly detected as shown in Figure 2 (c), where the detection results of clouds and waves have similar structures to the ship, and the detection results of the horizon and waves are connected to the ship, making it difficult to distinguish the ship target. As can be seen from the result image, the edges of the ship contain both horizontal and vertical linear structures, while the edges of clouds, the horizon, and waves are mainly horizontal linear structures. Therefore, the interference can be suppressed by removing the horizontal linear structures, i.e., removing the high frequency components in each column of data. Therefore, the present patent eliminates background interference by frequency domain low-pass filtering, and suppresses the influence of the edges of clouds, the horizon, and waves on target positioning by performing frequency domain low-pass filtering on each column of data. According to Figure 2 (b), it can be seen that after the suppression of the vertical high frequency components, the infrared image is compared with the area in the original image, and it can be seen that the structure of the clouds is no longer obvious, and most of the horizon and waves are removed. If the suppressed image is subjected to canny edge detection, it can be seen from Figure 2 (d) that the interference of clouds and waves is almost completely suppressed, the horizon is no longer obvious, and the outline of the ship is completely preserved.

[0080] In specific embodiments, although the suppression of vertical high frequency components has obvious effect on background interference removal, there are still some problems. First, clouds and waves may leave some interference, as shown in Figure 2 (d), which may interfere with the direct positioning of the ship target. If morphological transformation or straight line closure operation is directly used, some cloud and wave interference may be incorrectly included in the ship target, resulting in positioning errors.

[0081] Therefore, in S3, the optimized infrared ship image is first subjected to a lateral gradient transformation to obtain the potential region of the ship target; specifically...

[0082] S3.1: Perform lateral gradient calculation on the optimized infrared ship image to obtain the infrared ship gradient image. Based on the minimum intra-class variance method, separate the ship class from the image background in the infrared ship gradient image to obtain the binarized infrared ship image.

[0083] The types of ships include the hull edge type and the island edge type;

[0084] The formula for calculating the lateral gradient is as follows:

[0085] z(n) = (z(n+1) - z(n-1)) / 2

[0086] In the formula: n represents the position of any point in the row number of the infrared ship image; z(n) represents the pixel value of the current pixel;

[0087] Since longitudinal high-frequency component suppression destroys the ship's lateral edge structure while removing clouds, sea horizons and waves, it is difficult to obtain effective information by longitudinal gradient calculation. Since the left and right edges of the ship's chord and island are significantly different from the background, the calculation results based on the lateral gradient are used to separate the left and right edges of the ship's chord and island from the background by the minimum intra-class variance method, thereby achieving image binarization and obtaining the edges of the ship's chord and island.

[0088] S3.2: Project the binarized infrared ship image horizontally to obtain the gradient sum of each row in the binarized infrared ship image, and establish a rectangular coordinate system with the row number of the binarized infrared ship image as the horizontal axis and the gradient sum corresponding to each row number of the binarized infrared ship image as the vertical axis.

[0089] like Figure 6 As shown, the gradient sum of each row in the binarized infrared ship image is curve-based on the Cartesian coordinate system to obtain several horizontal gradient curves; wherein, the method of curve-based implementation is a known prior art and is not the inventive point of this application, and will not be described in detail here.

[0090] S3.3: Calculate the interval gradient sum of the intervals formed by each of the horizontal gradient curves and the horizontal axis, and select the endpoints of the interval with the maximum interval gradient sum;

[0091] Based on the interval gradient and the endpoints of the maximum interval, the potential regions of ship targets in the binarized infrared ship image are determined.

[0092] Depend on Figure 2(e) The result figure shows that only the left and right edges of the ship string and ship island and a small amount of clouds and sea waves are detected due to the calculation of the horizontal gradient of the infrared image. Then the gradient image is projected in the horizontal direction to calculate the gradient sum of each row to generate a horizontal gradient curve. The intervals formed by all curves and the horizontal coordinate axis are detected, the gradient sum in each interval is calculated, and then the maximum interval is obtained by retaining the maximum gradient sum. The two endpoints of the maximum interval are determined to determine the upper and lower boundaries of the potential ship target region in the binary infrared ship image, so as to remove the interference of clouds and sea waves and obtain the potential ship target region, such as Figure 2 (f).

[0093] In specific embodiments, the calculation formula of the minimum intra-class variance method in S3.1 is

[0094]

[0095] In the formula, a and b represent the ship string edge class and the ship island edge class, respectively; S a represents the intra-class variance of the ship string edge class; S b represents the intra-class variance of the ship island edge class; Wa represents the ratio of the number of pixels of the ship string edge class to the total number of pixels; Wb represents the ratio of the number of pixels of the ship island edge class to the total number of pixels; Ra and Rb represent the pixel set of the ship string edge class and the ship island edge class, respectively; ya i and yb i represent the pixel values of the ship string edge class pixels and the ship island edge class pixels, respectively; μa and μb represent the pixel mean values of the ship string edge class and the ship island edge class, respectively, and T is the threshold for dividing the ship string edge class and the ship island edge class.

[0096] In specific embodiments, the ship edge detection on the ship target potential region in S4 is performed to realize the positioning of the ship target, specifically

[0097] S4.1: performing ship edge detection on the ship target potential region by the canny edge detection technology to obtain a ship edge detection image;

[0098] S4.2: performing horizontal expansion of a preset pixel value on the ship edge detection image based on morphological changes;

[0099] S4.3: performing connected domain detection on the horizontally expanded ship edge detection image based on the connected domain detection technology to obtain a plurality of rectangular connected domain regions;

[0100] S4.4: calculating the areas of the rectangular connected domain regions, selecting the rectangular connected domain region with the largest area as the circumscribed contour of the ship target, and realizing the positioning of the ship target according to the circumscribed contour of the ship target.

[0101] After the potential region of the ship target is obtained, the cloud and sea wave interference have been removed, the potential region after the suppression of the longitudinal high frequency component is subjected to Canny edge detection, then the ship edge detection graph is subjected to transverse inflation with a preset pixel value of 5, then the transverse inflation ship edge detection graph is subjected to connected domain detection, the maximum connected domain region is retained, the left and right boundaries of the ship target are obtained, combined with the interval gradient and the two endpoints of the maximum interval obtained in S3.3, and the horizontal coordinates corresponding to the two endpoints of the maximum interval are the upper and lower boundaries of the potential region of the ship target, so the outer contour of the ship target is obtained, as shown in Figure 2 (g)、 Figure 2 (h) and Figure 2 (i) shown.

[0102] Specifically, in order to illustrate the positioning effect of the algorithm, four representative infrared ship images are selected for experiment, as shown in Figure 3 , Figure 4 and Figure 5 . Among them, Figure 2 and Figure 3 are different in ship type, Figure 4 has more complex cloud interference, Figure 5 the camera has a certain tilt angle when shooting, and the ship is far away from the camera shooting position, which has more serious sea horizon interference.

[0103] Figure 3 (a), Figure 4 (a) and Figure 5 (a) are infrared image original graphs; Figure 3 (b), Figure 4 (b) and Figure 5 (b) are infrared images after the suppression of the longitudinal high frequency component; since the suppressed infrared image cannot directly show the effect of the algorithm, the infrared image original graph and the suppressed image are subjected to Canny edge detection respectively, as shown in Figure 3 (c), Figure 4 (c), Figure 5 (c) and Figure 3 (d), Figure 4 (d), Figure 5 (d) shown; Figure 3 (e), Figure 4 (e), Figure 5 (e) is the transverse gradient binary image obtained by the minimum intra-class variance method, Figure 3 (f), Figure 4 (f), Figure 5 (f) in the red box region is the potential region of the ship target, Figure 3 (g), Figure 4 (g), Figure 5 (g) and Figure 3 (h),Figure 4 (h) Figure 5 (h) shows the results of Canny edge detection and morphological transformation of the potential target area of ​​the ship. Figure 3 (i) Figure 3 (i) and Figure 2 (i) is the positioning result of the infrared ship positioning algorithm based on line suppression.

[0104] Figure 3 In (a), the low cloud content makes it possible to... Figure 2 In image (c), the upper half has less interference, while the water surface has no interference. Figure 3 (a) Bright, the edges of the waves are more distinct, making Figure 3 (d) has no effect on suppressing ocean waves. Figure 3 (d) Obviously, the pixels on the chord are distributed in alternating bright and dark patterns, making the position of the chord easier to obtain, such as... Figure 3 As shown in (e), the potential area of ​​the ship target is finally obtained. Figure 4 (f). Figure 4 (g) shows that a small bright spot at the sea-line position on the right side of the image is causing interference, resulting in incorrect detection of interference after Canny edge detection. However, through... Figure 4 (h) Removed after morphological transformation; localization results are shown below. Figure 4 (i).

[0105] Figure 4 (a) shows that there is more severe cloud interference in the image, with the clouds connected in patches and having obvious edges between them and the sky, as illustrated in the diagram. Figure 4 As shown in (c) Figure 4 (d) While the mid-longitudinal high-frequency suppression algorithm removed most of the wave interference, it failed to effectively suppress significant cloud interference. This resulted in some clouds being incorrectly detected during subsequent target potential area detection and target localization. Figure 4 (e) Top right corner and Figure 5 (g) In the upper left corner region, the algorithm obtains the potential area of ​​the ship target by projecting horizontally and retaining the maximum peak. Figure 5 (f) Combined with morphological transformation Figure 5 (h) The interference was removed, and the results are shown below. Figure 5 (i).

[0106] Figure 5 (a) The ship is at a certain angle of inclination, and because the ship is far from the shooting location, the influence of the horizon on the positioning results is stronger. Figure 5 (c) and Figure 5 The comparison in (d) shows that, due to the reasons mentioned above, the sea-line area after longitudinal high-frequency component suppression was not completely removed. However, because the lateral gradient map was used when detecting the potential target area, the interference did not affect the positioning results, as shown in the figure.Figure 5 (e)、 ​ (f) shows. The subsequent edge detection gets part of the sea-sky line, as shown in ​ (g) and ​ (f) on the left, but the morphological transformation effectively removes the interference, and gets the accurate ship target positioning result ​ (i).

[0107] The patent first analyzes the shortcomings of the existing infrared image ship target extraction algorithm: for the problem that the ship target is difficult to detect when the infrared image has a large number of clouds, sea waves and obvious sea-sky line interference, a kind of infrared image ship positioning method based on line suppression is given. In order to suppress the background interference such as clouds, sea-sky line and sea waves, a kind of longitudinal high-frequency component suppression algorithm is given; then the transverse gradient is calculated, and the transverse projection is used to remove the clouds and sea waves, and the potential area of the ship target is obtained; finally, the canny edge detection and morphological transformation are combined to remove the interference on the sea level, and the positioning of the ship target is realized. And by selecting three infrared ship images with different background interference to verify the positioning performance of the method, the accuracy of the ship target positioning is effectively verified.

[0108] Finally, it should be noted that: the above embodiments are only used to illustrate the technical solutions of the present application, and not to limit them; although the present application has been described in detail with reference to the foregoing embodiments, those skilled in the art should understand that: it can still modify the technical solutions recorded in the foregoing embodiments, or make equivalent replacement for part or all of the technical features; and these modifications or replacements do not make the essence of the corresponding technical solutions deviate from the scope of the technical solutions of the embodiments of the present application.

Claims

1. A method for infrared ship positioning based on line suppression, characterized in that, The method comprises the following steps: S1: defining high-frequency components and low-frequency components of an infrared ship image, and obtaining a frequency domain analysis graph; The high-frequency components include cloud edges, sea-sky lines, sea wave edges, and ship edges; the low-frequency components include smooth areas surrounded by the sky and cloud edges, and smooth areas surrounded by ship edges; S2: performing longitudinal high-frequency component suppression on the frequency domain analysis graph to optimize the infrared ship image, specifically S2.1: performing grid division on the frequency domain analysis graph to obtain a plurality of initial images in rows and columns; And for each column of image data of the initial image Perform a fast Fourier transform to obtain the frequency domain component values ​​of each row; and arranging the frequency domain component values of each row in descending order; The calculation formula of the fast Fourier transform To In the formula: ; represents the number of pixels of an image column; ; represents an imaginary number; n represents the position of data; S2.2: performing low-pass filtering processing on the frequency domain components arranged in descending order based on a preset frequency domain component threshold value; and removing the frequency domain components of each row that are greater than the preset frequency domain component threshold value, and retaining the frequency domain components of each row that are less than or equal to the preset frequency domain component threshold value; The expression of the low-pass filtering processing is In the formula, R represents an image region where a low frequency component is located. represents a filtering result. S2.3: obtaining the optimized infrared ship image after the low-pass filtering processing through inverse Fourier transform: The calculation formula of the inverse Fourier transform is wherein: ; represents the number of pixels of an image column; ; represents an imaginary number; n represents the position of data; S3: performing horizontal gradient transformation on the optimized infrared ship image to obtain a potential area of a ship target, specifically S3.1: performing horizontal gradient calculation on the optimized infrared ship image to obtain an infrared ship gradient image, separating the class of a ship in the infrared ship gradient image from the background of the image based on the minimum intra-class variance method, and obtaining a binary infrared ship image; The class of the ship includes a ship chord edge class and a ship island edge class; The horizontal gradient calculation formula is In the formula: represents the position of an arbitrary point in the row in the infrared ship image; represents the pixel value of the current pixel; S3.2: performing horizontal direction projection on the binary infrared ship image to obtain the gradient sum of each row in the binary infrared ship image, and establishing a rectangular coordinate system with the number of rows of the binary infrared ship image as the horizontal coordinate and the gradient sum corresponding to each row of the binary infrared ship image as the vertical coordinate; Based on the rectangular coordinate system, the gradient sum of each row in the binary infrared ship image is curve-fitted to obtain a plurality of horizontal gradient curves; S3.3: calculating the interval gradient sum of the interval formed by each horizontal gradient curve and the horizontal coordinate axis, and selecting the endpoints of the interval with the maximum interval gradient sum; According to the endpoints of the interval with the maximum interval gradient sum, the potential area of the ship target in the binary infrared ship image is determined; S4: performing ship edge detection on the potential area of the ship target to realize positioning of the ship target.

2. The method according to claim 1, wherein, The calculation formula of the minimum intra-class variance method in S3.1 is In the formula: a , b respectively represent the ship chord edge class and the ship island edge class; represents the intra-class variance of the ship chord edge class; represents the intra-class variance of the ship island edge class; represents the ratio of the number of pixels of the ship chord edge class to the total number of pixels; represents the ratio of the number of pixels of the ship island edge class to the total number of pixels; Ra and Rb respectively represent the pixel set of the ship chord edge class and the ship island edge class; and respectively represent the pixel value of the ship chord edge class pixel and the ship island edge class pixel; and respectively represent the pixel mean of the ship chord edge class and the ship island edge class, is the threshold value for segmenting the ship chord edge class and the ship island edge class.

3. The method of claim 1, wherein the method is a line-of-sight suppression based infrared ship positioning method. In S4, the ship edge detection on the potential area of the ship target to realize positioning of the ship target is specifically S4.1: performing ship edge detection on the potential area of the ship target through a canny edge detection technology to obtain a ship edge detection graph; S4.2: performing horizontal expansion of a preset pixel value on the ship edge detection graph based on morphological changes; S4.3: performing connected domain detection on the horizontally expanded ship edge detection graph based on a connected domain detection technology to obtain a plurality of rectangular connected domain areas; S4.4: Calculate the area of each rectangular connected domain region, select the rectangular connected domain region with the largest area as the circumscribed contour of the ship target, and realize the positioning of the ship target according to the circumscribed contour of the ship target.

Images