Ship early warning image recognition system

Abstract

The invention discloses a ship early warning image recognition system, belongs to the technical field of image processing, and solves the problems that under the condition that fog is thin, an observer on a ship can observe the condition of a water area with naked eyes without defogging processing, and under the condition that the fog is thick and heavy, the water area cannot be observed with the naked eyes. The ship defogging system can only play a role in properly assisting sailors, and cannot provide early warning for the sailors in the process of fog from appearance to thickening. Comprising an image acquisition module, an image defogging module, a data processing module and an early warning module, the image acquisition module is used for acquiring images in front of a ship and feeding back the images to the image defogging module, and the image defogging module is used for defogging the images in front of the ship, providing fogged images and feeding back the fogged images to the data processing module. According to the invention, the image acquisition module acquires images in front of a ship in fog, filters parts shielded by the fog, retains clear contents for identification, calculates visibility, warns sailors, avoids misjudgment, and ensures sailing safety.

Description

Technical Field

[0001] The present invention relates to the technical field of image processing, and in particular to a ship warning image recognition system. Background Art

[0002] Ships are a kind of water transportation with a long history. They are supported by the buoyancy of water and travel on the water with the help of various power sources such as sails, steam engines, diesel engines or nuclear power. From the wooden ships of ancient Egypt to the modern giant cargo ships, ships have made great developments in design, size and function. They are not only used to transport goods and passengers, but also undertake important tasks such as military, scientific research and exploration. With the advancement of technology, modern ships pay more and more attention to environmental protection and efficiency. The application of innovative technologies such as liquefied natural gas (LNG) powered ships and sail-assisted propulsion systems has enabled the shipping industry to continue to move forward on the path of sustainable development.

[0003] At present, with the increasing prosperity of water transportation, its order has also improved, which has led to increasingly heavy tasks for staff who maintain water transportation order. Ship monitoring is one of the important responsibilities. However, in autumn and winter, whether in the early morning, evening or daytime, thick water fog often covers the water surface, obscuring the ships and making it difficult to monitor effectively. This poses a challenge to the clear detection of ships on the river.

[0004] To address this problem, it is basically necessary to use the system. At present, a ship defogging and restoration method and system with application number CN202410029086.4 is actually used. The technical points are as follows: S1. Obtain clear airspace ship pictures and airspace ship pictures with different levels of water fog, and perform Fourier transform on each airspace ship picture to obtain the corresponding frequency domain ship picture; add all corresponding airspace ship pictures and frequency domain ship pictures and input them into the original water fog concentration classification model for multiple rounds of model training to obtain the target water fog concentration classification model; S2. Obtain the original training set, which is divided into multiple groups, each group including clear ship pictures and corresponding random levels of water fog ship pictures; add a water fog concentration level guidance task to the original Unet model to obtain a category Unet model; S3. Randomly add a kind of Gaussian noise to all clear ship pictures in the original training set and all water fog ship pictures. Input them into the target water fog concentration classification model to judge the water fog level as the first round of training sets, and input them into the category Unet model for the first round of model training to obtain the first round of noise value and the first round of more A new category Unet model is generated; a first noise loss value is calculated based on the noise values ​​of all clear ship images, the first round noise value, and the noise values ​​of all water fog ship images; S4, all clear ship images in the original training set are randomly added with a Gaussian noise and the noise value of the previous round is added, and all water fog ship images minus the noise value of the previous round are used as the current round training set, and are input into the last round updated category Unet model for current round model training to obtain the current round noise value and the current round updated category Unet model; the current noise loss value is calculated based on the noise values ​​of all clear ship images, the current round noise value, and the noise values ​​of all water fog ship images; if the current noise loss value is within the preset noise loss value range, the model training is stopped, the current round updated category Unet model is used as the target category Unet model, and the number of training rounds and the Gaussian noise added in each round are saved; S5, after the ship image to be defogged is input into the target water fog concentration classification model to judge the water fog level, the reverse reasoning defogging is performed according to the number of training rounds and the Gaussian noise added in each round to obtain the target clear ship image.

[0005] The original water mist concentration classification model is modified by the above technical solution, and the obtained target water mist concentration classification model is more accurate in water mist classification. The ship water mist concentration level is used as a guide in the original Unet model to achieve demisting according to different levels; the demisting processing is combined in the air domain and frequency domain, and the target clear ship picture obtained is more accurate.

[0006] However, the ship picture finally obtained by the above method is ultimately a ship picture synthesized from all clear ship pictures obtained after multiple rounds of model training. Its main purpose is to provide a certain reference for the crew. However, at present, this type of defogging method and system mainly cooperates with the crew to explore the surrounding waters. In the end, the crew still plays a decisive role. Therefore, the current situation of this type of ship defogging system is rather embarrassing. When the fog is thin, the observers on the ship can observe the water conditions with the naked eye without defogging. When the fog is heavy, the ship defogging system can only play an appropriate role in assisting the crew, and cannot provide early warning for the crew from the appearance to the thickening of the fog.

[0007] Therefore, a ship early warning image recognition system is proposed to solve or alleviate the above problems. Summary of the invention

[0008] The purpose of the present invention is to solve the shortcomings of the prior art and to propose a ship warning image recognition system.

[0009] In order to achieve the above object, the present invention adopts the following technical solutions:

[0010] A ship warning image recognition system comprises an image acquisition module, an image defogging module, a data processing module, and a warning module, wherein the image acquisition module is used to acquire a picture in front of a ship and feed it back to the image defogging module, the image defogging module is used to provide a post-fog image after defogging the picture in front of the ship and feed it back to the data processing module, and the data processing module is used to match the visibility of the ship according to the post-fog image and control the warning module to issue a warning and feedback according to the visibility of the ship.

[0011] Preferably, it also includes a storage module, which is coupled to the data processing module and is used to store information of the data processing module.

[0012] Preferably, the image acquisition module comprises a standard high-definition camera, and the standard high-definition camera is used to capture the image in front of the ship and feed it back to the image defogging module.

[0013] Preferably, the image defogging module is used to provide a defogged image after defogging the image in front of the ship and feed it back to the data processing module, comprising the following steps:

[0014] Acquire the image in front of the ship from the image acquisition module;

[0015] Preprocess the image in front of the ship to obtain a denoised image;

[0016] Use the denoised image to calculate the minimum value of the three color channels in the neighborhood of each pixel to obtain the dark channel image Among them, Ic (x, y) is the pixel value on color channel C, J dark (x,y) is the dark channel image;

[0017] Take the maximum brightness value in the denoised image to estimate the atmospheric light Among them, I σ (x,y) is the denoised image;

[0018] According to the dark channel image J dark (x, y) and atmospheric light A estimate transmittance Among them, ω is a constant, usually with a value between 0.95-0.99;

[0019] Set the transmittance threshold t min , and distinguish and mark the fog-covered area in the image ahead of the ship based on the transmittance t(x,y) <t min ,then mask(x,y)=1;

[0020] Filter the marked area in the image in front of the ship and output the image after fog.

[0021] Preferably, the data processing module is used to match the visibility of the ship according to the image after fog and control the early warning module to give early warning and feedback according to the visibility of the ship, including the following steps:

[0022] The bounding boxes of multiple objects are extracted from the fogged image through the object detection algorithm. Each bounding box gives the position of the object. The bounding box of each object is set to [x1, x1, y2, y2], where (x1, y1) is the coordinate of the upper left corner and (x2, y2) is the coordinate of the lower right corner.

[0023] Analyze all detected objects to get the center coordinates of the objects

[0024] Set the reference position of the shooting point to (x ship ,y ship ), calculate the Euclidean distance between the object and the reference position of the shooting point

[0025] Calculate and sort the Euclidean distances of all objects, and select the three objects with the largest distances TOP3 = Sort (D obj1 ,D obj2 ,D obj3 ,…,D objn );

[0026] In the foggy image, the three objects with the largest distance are extracted and normalized;

[0027] The three normalized objects are subjected to feature extraction through a convolutional neural network, and the object category is predicted using a softmax layer to obtain multiple similar categories;

[0028] Compare each object with the largest distance to its respective similar category and output the similarity;

[0029] Set the similarity threshold and take the object with the highest average similarity among the three objects with the largest distance as the object with the largest distance.

[0030] Preferably, the three normalized objects are subjected to feature extraction through a convolutional neural network, comprising the following steps:

[0031] Use a convolutional neural network to extract the features of the images of the three objects with the largest distance: F(x,y)=(1*K)(x,y), where K is the convolution kernel, * is the convolution operation, and x and y are the spatial coordinates of the images of the three objects with the largest distance;

[0032] Use ReLU activation function for nonlinear transformation ReLU(x)=max(0,x);

[0033] Downsampling using pooling layers Among them, pooling window is the pooling window, i and j are the offsets within the window;

[0034] The outputs of the convolution and pooling layers are flattened and connected to the fully connected layer to obtain the feature vector V = W·F+b of the object, where W is the weight matrix, b is the bias, and F is the feature of the images of the three objects with the largest distance.

[0035] Preferably, the method of using a softmax layer to predict the object category and obtain multiple similar categories includes the following steps:

[0036] The feature vector V output by the fully connected layer is used to predict the object category using the softmax layer. For multi-classification tasks, the category prediction formula is: Among them, V c is the score of the feature vector V corresponding to category c, and p(y=c|V) is the predicted probability of belonging to category c.

[0037] Preferably, the step of comparing the objects with the largest distances with their respective similar categories and outputting the similarity comprises the following steps:

[0038] Use cosine similarity to calculate the similarity between the object with the largest distance and multiple similar categories. Among them, ∥V1∥ and ∥V2∥ are the L2 norms of the feature vectors V1 and V2 respectively, and · represents the product point of the vectors.

[0039] Preferably, the data processing module is used to match the visibility of the ship according to the image after fog and control the early warning module to give early warning and feedback according to the visibility of the ship, and further comprises the following steps:

[0040] Determine the object with the largest distance and take the center coordinates (x center ,y center ), determine the focal length of the image acquisition module, determine the presence of the bow in the image in front of the ship and the actual width of the bow;

[0041] Determine the pixel width of the bow through the image in front of the ship and calculate the proportional relationship = actual width of the bow / pixel width of the bow;

[0042] Calculate the actual width W of the center coordinate of the object with the largest distance p =w p × proportional relationship, w p is the pixel width of the center coordinates of the object with the largest distance;

[0043] Apply the collinear equation to calculate the distance between the center coordinates of the object with the largest distance and the image acquisition unit Among them, x is the pixel coordinate of the coordinate point in the photo.

[0044] Preferably, the data processing module is used to match the visibility of the ship according to the image after fog and control the early warning module to issue an early warning according to the visibility of the ship, and further includes the following steps:

[0045] The distance between the center coordinates of the object with the largest distance and the image acquisition unit is determined as visibility. If the visibility is ≥ 200 meters, no warning is required. If the visibility is < 200 meters, the warning module is controlled to warn the crew.

[0046] The present invention has the following beneficial effects:

[0047] The present invention can collect the front picture of the ship through the image acquisition module during the ship's sailing process, and when there is fog in the front picture, filter the part of the front picture that is completely blocked by the fog, and only retain the content that has not been completely blocked by the fog in the front picture, and at the same time, identify and judge according to the remaining filtered content, and obtain the distance between the position of the point with the highest clarity and recognition and the highest similarity after final verification and the ship as visibility. Under this method, a relatively actual ship visibility is obtained, so as to warn the crew and avoid the crew from making wrong judgments on the ship visibility due to the picture obtained by the conventional defogging system. At the same time, after the visibility is lower than a certain value, the crew is clearly warned, so that the crew can turn on the radar and other equipment in advance, and arrange more experienced crew members to drive the ship or issue more accurate instructions to avoid collisions and the like during the sailing process. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the drawings required for use in the embodiments are briefly introduced below. It should be understood that the following drawings only show certain embodiments of the present invention and therefore should not be regarded as limiting the scope. For ordinary technicians in this field, other related drawings can be obtained based on these drawings without creative work.

[0049] Figure 1 It is a structural block diagram of the present invention.

[0050] 1. Image acquisition module; 2. Image defogging module; 3. Data processing module; 4. Early warning module; 5. Storage module. DETAILED DESCRIPTION

[0051] In order to make the purpose, technical solutions and advantages of the embodiments of the present invention clearer, the technical solutions in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Generally, the components of the embodiments of the present invention described and shown in the drawings here can be arranged and designed in various different configurations.

[0052] Therefore, the following detailed description of the embodiments of the present invention provided in the accompanying drawings is not intended to limit the scope of the invention claimed for protection, but merely represents selected embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0053] It should be noted that similar reference numerals and letters denote similar items in the following drawings, and therefore, once an item is defined in one drawing, further definition and explanation thereof is not required in subsequent drawings.

[0054] In the description of the present invention, it should be understood that the terms "center", "up", "down", "left", "right", "vertical", "horizontal", "inside", "outside" and the like indicate orientations or positional relationships based on the orientations or positional relationships shown in the accompanying drawings, or are the orientations or positional relationships in which the product of the invention is conventionally placed when in use, or are the orientations or positional relationships conventionally understood by those skilled in the art. They are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, be constructed and operated in a specific orientation, and therefore should not be understood as a limitation on the present invention.

[0055] Furthermore, the terms “first”, “second”, “third”, etc. are merely used for distinguishing descriptions and are not to be understood as indicating or implying relative importance.

[0056] In the description of the present invention, it is also necessary to explain that, unless otherwise clearly specified and limited, the terms "set", "install", "connect", and "connect" should be understood in a broad sense, for example, it can be a fixed connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection, or it can be indirectly connected through an intermediate medium, or it can be the internal communication of two elements. For ordinary technicians in this field, the specific meanings of the above terms in the present invention can be understood according to specific circumstances.

[0057] A ship warning image recognition system, such as Figure 1 As shown, it includes an image acquisition module 1, an image defogging module 2, a data processing module 3, an early warning module 4, and a storage module 5. The image acquisition module 1 is used to acquire the image in front of the ship and feed it back to the image defogging module 2. The image defogging module 2 is used to provide a post-fog image after defogging the image in front of the ship and feed it back to the data processing module 3. The data processing module 3 is used to match the visibility of the ship according to the post-fog image and control the early warning module 4 to give early warning and feedback according to the visibility of the ship. The storage module 5 is coupled to the data processing module 3, and the storage module 5 is used to store the information of the data processing module 3.

[0058] The image acquisition module 1 includes a standard high-definition camera, which is used to collect the image in front of the ship and feed it back to the image defogging module 2. The storage module 5 includes a memory. The image defogging module 2 and the data processing module 3 include a processor and a computer program stored in the memory and running on the processor. The early warning module 4 includes an alarm.

[0059] The system that determines visibility and issues early warnings during ship navigation can capture images in front of the bow through the image acquisition unit carried on the ship.

[0060] When affected by fog, the system will automatically identify and filter out areas completely blocked by fog, retaining only those partially blocked or unblocked images.

[0061] Then, image processing technology is used to analyze the remaining visible content, identify the target objects with the highest clarity and recognition, and calculate the actual distance between these target objects and the ship. This is used as a measure of visibility. This can effectively provide more realistic visibility data and help crew members avoid misjudgments due to images from traditional defog systems. When the system detects that visibility is reduced to a preset threshold, it will issue an early warning, prompting crew members to start auxiliary equipment such as radar, and recommend that more experienced crew members take over driving or issue navigation instructions to prevent possible collisions at sea.

[0062] Preferably, the image defogging module 2 is used to provide a defogging image after defogging the image in front of the ship and feed it back to the data processing module 3, including the following steps:

[0063] Acquire the image in front of the ship from the image acquisition module 1;

[0064] Preprocess the image in front of the ship to obtain a denoised image;

[0065] Use the denoised image to calculate the minimum value of the three color channels in the neighborhood of each pixel to obtain the dark channel image Among them, I c (x, y) is the pixel value on color channel C, J dark (x,y) is the dark channel image;

[0066] Take the maximum brightness value in the denoised image to estimate the atmospheric light Among them, I σ (x,y) is the denoised image;

[0067] According to the dark channel image J dark (x, y) and atmospheric light A estimate transmittance Among them, ω is a constant, usually with a value between 0.95-0.99;

[0068] Set the transmittance threshold t min , and distinguish and mark the fog-covered area in the image ahead of the ship based on the transmittance t(x,y) <t min ,then mask(x,y)=1;

[0069] Filter the marked area in the image in front of the ship and output the image after fog.

[0070] Through the above method steps, the system can process the picture in front of the ship, so that when the picture in front of the ship is blocked by fog, the part of the picture in front of the ship that is completely blocked by fog can be filtered, and only the part of the picture in front of the ship that can still be seen clearly is retained. In subsequent steps, these parts that can still be seen clearly can be used to calculate the visibility of the ship.

[0071] Preferably, the data processing module 3 is used to match the visibility of the ship according to the image after fog and control the early warning module 4 to give early warning and feedback according to the visibility of the ship, including the following steps:

[0072] The bounding boxes of multiple objects are extracted from the fogged image through the object detection algorithm. Each bounding box gives the position of the object. The bounding box of each object is set to [x1, x1, y2, y2], where (x1, y1) is the coordinate of the upper left corner and (x2, y2) is the coordinate of the lower right corner.

[0073] Analyze all detected objects to get the center coordinates of the objects

[0074] Set the reference position of the shooting point to (x ship ,y ship ), calculate the Euclidean distance between the object and the reference position of the shooting point

[0075] Calculate and sort the Euclidean distances of all objects, and select the three objects with the largest distances TOP3 = Sort (D obj1 ,D obj2 ,D obj3 ,…,D objn );

[0076] In the foggy image, the three objects with the largest distance are extracted and normalized;

[0077] The three normalized objects are subjected to feature extraction through a convolutional neural network, and the object category is predicted using a softmax layer to obtain multiple similar categories;

[0078] Compare each object with the largest distance to its respective similar category and output the similarity;

[0079] Set the similarity threshold and take the object with the highest average similarity among the three objects with the largest distance as the object with the largest distance.

[0080] Through the above method steps, the system can identify the target of the still visible part of the fogged image obtained after defogging, and then identify the target objects with the farthest Euclidean distance. From these farthest objects, a convolutional neural network is used to match them. It is required to select an object with the highest clarity and the farthest distance as a benchmark, so that the system can calculate the distance between the object and the ship as visibility feedback.

[0081] Preferably, the three normalized objects are subjected to feature extraction through a convolutional neural network, comprising the following steps:

[0082] Use a convolutional neural network to extract the features of the images of the three objects with the largest distance: F(x,y)=(1*K)(x,y), where K is the convolution kernel, * is the convolution operation, and x and y are the spatial coordinates of the images of the three objects with the largest distance;

[0083] Use ReLU activation function for nonlinear transformation ReLU(x)=max(0,x);

[0084] Downsampling using pooling layers Among them, pooling window is the pooling window, i and j are the offsets within the window;

[0085] The outputs of the convolution and pooling layers are flattened and connected to the fully connected layer to obtain the feature vector V = W·F+b of the object, where W is the weight matrix, b is the bias, and F is the feature of the images of the three objects with the largest distance.

[0086] Preferably, the softmax layer is used to predict the object category to obtain multiple similar categories, including the following steps:

[0087] The feature vector V output by the fully connected layer is used to predict the object category using the softmax layer. For multi-classification tasks, the category prediction formula is: Among them, V c is the score of the feature vector V corresponding to category c, and p(y=c|V) is the predicted probability of belonging to category c.

[0088] Preferably, comparing each object with the largest distance with its respective similar category and outputting the similarity comprises the following steps:

[0089] Use cosine similarity to calculate the similarity between the object with the largest distance and multiple similar categories. Among them, ∥V1∥ and ∥V2∥ are the L2 norms of the feature vectors V1 and V2 respectively, and · represents the product point of the vectors.

[0090] Preferably, the data processing module 3 is used to match the visibility of the ship according to the image after fog and control the early warning module 4 to give early warning and feedback according to the visibility of the ship, and also includes the following steps:

[0091] Determine the object with the largest distance and take the center coordinates (x center ,y center ), determine the focal length of the image acquisition module 1, determine the presence of a bow in the image in front of the ship and the actual width of the bow;

[0092] Determine the pixel width of the bow through the image in front of the ship and calculate the proportional relationship = actual width of the bow / pixel width of the bow;

[0093] Calculate the actual width W of the center coordinate of the object with the largest distance p =w p × proportional relationship, w p is the pixel width of the center coordinates of the object with the largest distance;

[0094] Apply the collinear equation to calculate the distance between the center coordinates of the object with the largest distance and the image acquisition unit Among them, x is the pixel coordinate of the coordinate point in the photo.

[0095] Preferably, the data processing module 3 is used to match the visibility of the ship according to the image after fog and control the early warning module 4 to issue an early warning according to the visibility of the ship, and further includes the following steps:

[0096] The distance between the center coordinates of the object with the largest distance and the image acquisition unit is determined as visibility. If the visibility is ≥ 200 meters, no warning is required. If the visibility is < 200 meters, the warning module 4 is controlled to warn the crew.

[0097] Through the above steps and methods, the system can calculate the approximate distance between the target object and the ship based on the obtained target object, and then obtain the corresponding visibility. The visibility can be used as a reference for the crew, and can also be compared with the threshold value set later, thereby realizing the action of automatic early warning.

[0098] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. For those skilled in the art, the present invention may have various modifications and variations. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention shall be included in the protection scope of the present invention.

Claims

1. A ship warning image recognition system, characterized in that: The system comprises an image acquisition module (1), an image defogging module (2), a data processing module (3), and an early warning module (4); the image acquisition module (1) is used to acquire a picture in front of a ship and feed it back to the image defogging module (2); the image defogging module (2) is used to provide a post-fog image after defogging the picture in front of the ship and feed it back to the data processing module (3); the data processing module (3) is used to match the visibility of the ship according to the post-fog image and control the early warning module (4) to issue an early warning and feedback according to the visibility of the ship.

2. A ship warning image recognition system according to claim 1, characterized in that: It also comprises a storage module (5), which is coupled to the data processing module (3) and is used to store information of the data processing module (3).

3. A ship warning image recognition system according to claim 1, characterized in that: The image acquisition module (1) comprises a standard high-definition camera, which is used to acquire images in front of the ship and feed them back to the image defogging module (2).

4. A ship warning image recognition system according to claim 1, characterized in that: The image defogging module (2) is used to provide a defogging image after defogging the image in front of the ship and feed it back to the data processing module (3), and comprises the following steps: Acquire the image in front of the ship from the image acquisition module (1); Preprocess the image in front of the ship to obtain a denoised image; Use the denoised image to calculate the minimum value of the three color channels in the neighborhood of each pixel to obtain the dark channel image Among them, I c (x, y) is the pixel value on color channel C, J dark (x,y) is the dark channel image; Take the maximum brightness value in the denoised image to estimate the atmospheric light Among them, I σ (x,y) is the denoised image; According to the dark channel image J dark (x, y) and atmospheric light A estimate transmittance Among them, ω is a constant, usually with a value between 0.95-0.99; Set the transmittance threshold t min , and distinguish and mark the fog-covered area in the image in front of the ship based on the transmittance t(x,y) Ift(x,y) <t min ,thenmask(x,y)=1; Filter the marked area in the image in front of the ship and output the image after fog.

5. A ship warning image recognition system according to claim 1, characterized in that: The data processing module (3) is used to match the visibility of the ship according to the post-fog image and control the early warning module (4) to perform early warning and feedback according to the visibility of the ship, and comprises the following steps: The bounding boxes of multiple objects are extracted from the fogged image through the object detection algorithm. Each bounding box gives the position of the object. The bounding box of each object is set to [x1, x1, y2, y2], where (x1, y1) is the coordinate of the upper left corner and (x2, y2) is the coordinate of the lower right corner. Analyze all detected objects to get the center coordinates of the objects Set the reference position of the shooting point to (x ship ,y ship ), calculate the Euclidean distance between the object and the reference position of the shooting point Calculate and sort the Euclidean distances of all objects, and select the three objects with the largest distances TOP3 = Sort (D obj1 ,D obj2 ,D obj3 ,…,D objn ); In the foggy image, the three objects with the largest distance are extracted and normalized; The three normalized objects are subjected to feature extraction through a convolutional neural network, and the object category is predicted using a softmax layer to obtain multiple similar categories; Compare each object with the largest distance to its respective similar category and output the similarity; Set the similarity threshold and take the object with the highest average similarity among the three objects with the largest distance as the object with the largest distance.

6. A ship warning image recognition system according to claim 5, characterized in that: The three normalized objects are subjected to feature extraction through a convolutional neural network, comprising the following steps: Use convolutional neural network to extract the features of the three objects with the largest distance F(x,y)= ( 1*K)(x,y), where K is the convolution kernel, * is the convolution operation, and x and y are the spatial coordinates of the images of the three objects with the largest distance; Use ReLU activation function for nonlinear transformation ReLU(x)=max(0,x); Downsampling using pooling layers Among them, poolingwindow is the pooling window, i and j are the offsets within the window; The outputs of the convolution and pooling layers are flattened and connected to the fully connected layer to obtain the feature vector V = W·F+b of the object, where W is the weight matrix, b is the bias, and F is the feature of the images of the three objects with the largest distance.

7. A ship warning image recognition system according to claim 5, characterized in that: The softmax layer is used to predict the object category and obtain multiple similar categories, including the following steps: The feature vector V output by the fully connected layer is used to predict the object category using the softmax layer. For multi-classification tasks, the category prediction formula is: Among them, V c is the score of the feature vector V corresponding to category c, and p(y=c|V) is the predicted probability of belonging to category c.

8. A ship warning image recognition system according to claim 5, characterized in that: The method of comparing each object with the largest distance with its respective similar category and outputting the similarity comprises the following steps: Use cosine similarity to calculate the similarity between the object with the largest distance and multiple similar categories. Among them, ||V1|| and ||V2|| are the L2 norms of the eigenvectors V1 and V2 respectively, and · represents the product point of the vectors.

9. A ship warning image recognition system according to claim 5, characterized in that: The data processing module (3) is used to match the visibility of the ship according to the image after fog and control the early warning module (4) to perform early warning and feedback according to the visibility of the ship, and also includes the following steps: Determine the object with the largest distance and take the center coordinates (x center ,y center ), determine the focal length of the image acquisition module (1), determine the presence of a bow in the image in front of the ship and the actual width of the bow; Determine the pixel width of the bow through the image in front of the ship and calculate the proportional relationship = actual width of the bow / pixel width of the bow; Calculate the actual width W of the center coordinate of the object with the largest distance p =w p × proportional relationship, w p is the pixel width of the center coordinates of the object with the largest distance; Apply the collinear equation to calculate the distance between the center coordinates of the object with the largest distance and the image acquisition unit Among them, x is the pixel coordinate of the coordinate point in the photo.

10. A ship warning image recognition system according to claim 9, characterized in that: The data processing module (3) is used to match the visibility of the ship according to the image after fog and control the early warning module (4) to issue an early warning according to the visibility of the ship, and further comprises the following steps: The distance between the center coordinates of the object with the largest distance and the image acquisition unit is determined as visibility. If the visibility is ≥ 200 meters, no warning is required. If the visibility is < 200 meters, the warning module (4) is controlled to warn the crew.

Images