CN111241948B - Method and system for all-weather ship identification - Google Patents

Abstract

The disclosure provides a method and a system for identifying ships in all weather, and belongs to the technical field of image processing. The method comprises the following steps: acquiring a plurality of images around a first ship, wherein the shooting time of the plurality of images is different; identifying the ship in each image by adopting a deep learning algorithm to obtain the position of the second ship in each image; determining a risk level of the second ship according to the position of the second ship in each image; and issuing an alarm according to the danger level of the second ship. The position of the second ship around the first ship in each image can be obtained by acquiring a plurality of images shot at different times around the first ship and identifying the ship in each image by adopting a deep learning algorithm. According to the position of the second ship in each image, the danger level of the second ship can be obtained, and an alarm is sent out according to the danger level of the second ship, so that the first ship can avoid collision with the second ship, and the safety of the ship is guaranteed.

Description

Method and system for identifying ships around the clock

Technical field

The present disclosure relates to the field of image processing technology, and in particular to a method and system for identifying ships around the clock.

Background technique

The transport speed of sea transport is slower than that of air transport, but the single transport volume is much larger than that of air transport, it is better adaptable to cargo, and the transport cost is lower than that of air transport. Therefore, sea transport has become the main mode of transportation in international trade. As the scale of maritime transport continues to expand, a large number of ships enter and leave the ports of large port cities and maritime hub cities every day, leading to congestion in the port area and the risk of collisions between ships. Moreover, as more and more ultra-large container ships are put into use, safety issues in maritime transportation have gradually become prominent.

During the day when visibility is good, port pilots can direct ships to enter the port in an orderly manner, and patrol officers on the ship's bridge will observe surrounding ships, so the safety of ships can be ensured. However, at night when visibility is low, the above methods are no longer used, and the safety of the ship cannot be guaranteed. Although radar can scan surrounding ships, radar is very expensive and cannot be widely used on civilian cargo ships.

Contents of the invention

Embodiments of the present disclosure provide a method and system for all-weather identification of ships, which can use low-cost image acquisition equipment and cooperate with image processing technology to effectively identify surrounding ships and ensure ship safety. It is especially suitable for civilian cargo ships without widespread radar. . The technical solutions are as follows:

On the one hand, embodiments of the present disclosure provide a method for identifying ships around the clock. The method includes:

Obtaining multiple images around the first ship, the multiple images being taken at different times;

Use a deep learning algorithm to identify the ship in each of the images, and obtain the position of the second ship in each of the images;

Determine the danger level of the second ship according to the position of the second ship in each of the images;

Based on the danger level of the second ship, an alarm is issued.

Optionally, the acquisition of multiple images around the first ship includes:

Obtain environmental information around the first ship;

determining a visibility level around the first ship based on the environmental information;

When the visibility level reaches the set standard, control the camera to continuously capture multiple images around the first ship;

When the visibility level does not reach the set standard, the laser pan/tilt is controlled to continuously capture multiple images around the first ship.

Optionally, using a deep learning algorithm to identify the second ship in each of the images includes:

extract image features from the image;

Generate candidate regions based on the image features;

extract regional features according to the image features and the candidate regions;

According to the regional characteristics, the category of the candidate region is determined, and the position of the second ship in the image is obtained.

Optionally, determining the risk level of the second ship based on the position of the second ship in each of the images includes:

Determine the distance between the second ship and the first ship at the shooting time of the image based on the position of the second ship in the image;

Determine the course and speed of the second ship based on the distance between the second ship and the first ship at the shooting time of each of the images;

The danger level of the second ship is determined based on the course, speed, and distance of the second ship from the first ship.

Optionally, issuing an alarm according to the danger level of the second ship includes:

In order from high to low risk levels, the headings, speeds, and distances to the first ship of multiple second ships are sequentially output to generate alarms.

On the other hand, embodiments of the present disclosure provide an all-weather ship identification system, which system includes:

An acquisition module, configured to acquire multiple images around the first ship, where the multiple images were taken at different times;

An identification module used to use a deep learning algorithm to identify the ship in each of the images and obtain the position of the second ship in each of the images;

A rating module, configured to determine the risk level of the second ship based on the position of the second ship in each of the images;

An alarm module is used to issue an alarm according to the danger level of the second ship.

Optionally, the acquisition module includes:

An information acquisition submodule, used to acquire environmental information around the first ship;

A level determination submodule, configured to determine the visibility level around the first ship based on the environmental information;

The shooting control submodule is used to control the camera to continuously shoot multiple images around the first ship when the visibility level reaches the set standard; when the visibility level does not reach the set standard, control the laser pan/tilt to continuously shoot all images. Multiple images of the surroundings of the first ship.

Optionally, the identification module includes:

a convolutional layer for extracting image features from said image;

A region proposal network, used to generate candidate regions based on the image features;

A pooling layer, used to extract regional features based on the image features and the candidate regions;

A classifier, configured to determine the category of the candidate area according to the area characteristics and obtain the position of the second ship in the image.

Optionally, the rating module includes:

A first determination sub-module configured to determine the distance between the second ship and the first ship at the shooting time of the image based on the position of the second ship in the image;

a second determination sub-module, configured to determine the course and speed of the second ship based on the distance between the second ship and the first ship at the shooting time of each of the images;

The third determination sub-module is used to determine the risk level of the second ship based on the course, speed, and distance of the second ship from the first ship.

Optionally, the alarm module is used to:

In order from high to low risk levels, the headings, speeds, and distances to the first ship of multiple second ships are sequentially output to generate alarms.

The beneficial effects brought by the technical solutions provided by the embodiments of the present disclosure are:

By acquiring multiple images taken at different times around the first ship, and using a deep learning algorithm to identify the ship in each image, the position of the second ship around the first ship in each image can be obtained. According to the position of the second ship in each image, the danger level of the second ship can be obtained. An alarm is issued according to the danger level of the second ship, which helps the first ship avoid collision with the second ship and ensures the safety of the ship. Moreover, the price of image acquisition equipment and image processing equipment is much lower than that of radar, which can greatly reduce the implementation cost of ship identification system, and is especially suitable for civilian cargo ships without widespread radar.

Description of the drawings

In order to more clearly illustrate the technical solutions in the embodiments of the present disclosure, the drawings needed to be used in the description of the embodiments will be briefly introduced below. Obviously, the drawings in the following description are only some embodiments of the present disclosure. For those of ordinary skill in the art, other drawings can also be obtained based on these drawings without exerting creative efforts.

Figure 1 is an application scenario diagram of an all-weather ship identification method provided by an embodiment of the present disclosure;

Figure 2 is a flow chart of an all-weather ship identification method provided by an embodiment of the present disclosure;

Figure 3 is a flow chart of another method for identifying ships all day long provided by an embodiment of the present disclosure;

Figure 4 is a schematic structural diagram of the Faster RCNN algorithm model provided by the embodiment of the present disclosure;

Figure 5 is a schematic structural diagram of a convolutional layer provided by an embodiment of the present disclosure;

Figure 6 is a schematic structural diagram of a region recommendation network provided by an embodiment of the present disclosure;

Figure 7 is a schematic structural diagram of a classifier provided by an embodiment of the present disclosure;

Figure 8 is a schematic structural diagram of an all-weather ship identification system provided by an embodiment of the present disclosure.

Detailed ways

In order to make the purpose, technical solutions and advantages of the present disclosure clearer, the embodiments of the present disclosure will be described in further detail below in conjunction with the accompanying drawings.

Figure 1 is an application scenario diagram of an all-weather ship identification method provided by an embodiment of the present disclosure. Referring to FIG. 1 , there are multiple second ships 20 surrounding the first ship 10 . Each of the multiple second ships 20 has the potential to collide with the first ship 10 , which needs to be recognized by the first ship 10 and avoided. Since the courses, speeds, and distances between the plurality of second ships 20 and the first ship 10 are different, the probability of collision between the plurality of second ships 20 and the first ship 10 is different.

Embodiments of the present disclosure provide a method for identifying ships around the clock. Figure 2 is a flow chart of a method for identifying ships all day long provided by an embodiment of the present disclosure. Referring to Figure 2, the method includes:

Step 101: Acquire multiple images around the first ship, and the multiple images are taken at different times.

In this embodiment, the first ship is a ship that needs to identify surrounding ships to avoid collision. The area around the first ship refers to the area with the first ship as the center and a set distance (such as 15km) as the radius.

Step 102: Use a deep learning algorithm to identify the ship in each image and obtain the position of the second ship in each image.

In this embodiment, the second ship is a ship located around the first ship.

Step 103: Determine the danger level of the second ship based on the position of the second ship in each image.

Step 104: Issue an alarm according to the danger level of the second ship.

By acquiring multiple images taken at different times around the first ship and using a deep learning algorithm to identify the ship in each image, the embodiment of the present disclosure can obtain the position of the second ship around the first ship in each image. According to the position of the second ship in each image, the danger level of the second ship can be obtained. An alarm is issued according to the danger level of the second ship, which helps the first ship avoid collision with the second ship and ensures the safety of the ship. Moreover, the price of image acquisition equipment and image processing equipment is much lower than that of radar, which can greatly reduce the implementation cost of ship identification system, and is especially suitable for civilian cargo ships without widespread radar.

The embodiment of the present disclosure provides another all-weather method for identifying ships, which is an optional implementation of the method for all-weather identification of ships shown in Figure 2 . Figure 3 is a flow chart of another method for identifying ships all day long provided by an embodiment of the present disclosure. Referring to Figure 3, the method includes:

Step 201: Obtain environmental information around the first ship.

In this embodiment, the environmental information may include light intensity and humidity. The intensity of light will directly affect the level of visibility. At the same time, the level of humidity will affect the formation of fog and the diffusion of impurities such as dust, which indirectly affects the height of visibility. Therefore, obtaining light intensity and humidity as environmental information is beneficial to accurately determine visibility levels.

Optionally, this step 201 may include:

Use a light sensor to measure the light intensity around the first ship;

The humidity around the first ship is measured using one of a humidity sensor, a moisture meter, and a weather radar.

Step 202: Determine the visibility level around the first ship based on the environmental information.

In this embodiment, visibility levels may include non-harsh environment and harsh environment. In practical applications, visibility levels can be divided directly by whether the set standards are met. When the visibility level reaches the set standard, the visibility is high and the environment is non-harsh. If the visibility level does not reach the set standard, the visibility is low and the environment is harsh.

Optionally, this step 202 may include:

When the light intensity around the first ship is above the set intensity and the humidity around the first ship is below the set humidity, it is determined that the visibility level around the first ship reaches the set standard;

When the light intensity around the first ship is below the set intensity, or the humidity around the first ship is above the set humidity, it is determined that the visibility level around the first ship has not reached the set standard.

When the light intensity and humidity meet the requirements at the same time, the visibility level is judged to have reached the set standard. The higher requirements for the visibility level are conducive to ensuring the clarity of the images captured by the camera.

For example, the set intensity can be 100 lux and the set humidity can be 50%; the set standard can be that the light intensity is above 100 lux and the humidity is below 50%.

Step 203: When the visibility level reaches the set standard, control the camera to continuously shoot multiple images around the first ship, and the shooting times of the multiple images are different.

In practical applications, the camera can be a wide-angle camera to meet the shooting needs of different areas.

Step 204: When the visibility level does not reach the set standard, control the laser pan/tilt to continuously shoot multiple images around the first ship, and the shooting times of the multiple images are different.

In this embodiment, by executing step 201, step 202 and step 203 in sequence, or executing step 201, step 202 and step 204 in sequence, multiple images around the first ship can be acquired.

In practical applications, the above process can be automatically controlled by the device, or the image acquisition device can be manually switched to capture images around the first ship. The image acquisition device may be disposed on the mast of the first ship, or around the hull of the first ship, for example, at least one image acquisition device may be disposed on each side. For the first ship with a longer hull, multiple image acquisition devices can be arranged at certain intervals on the side.

The embodiment of the present disclosure determines the visibility level around the first ship by obtaining the environmental information around the first ship, and controls the laser suitable for harsh environments when the visibility around the first ship is low based on whether the visibility level reaches the set standard. The PTZ captures images around the first ship to ensure that the captured images are clear enough to identify the ship, and when the visibility around the first ship is high, the camera is controlled to capture images around the first ship to ensure that the captured images are clear enough to be identified When leaving the ship, avoid long-term working of the laser gimbal to extend the service life of the laser gimbal.

Step 205: Use a deep learning algorithm to identify the ship in each image and obtain the position of the second ship in each image. This step 205 is performed after step 203 or step 204.

In this embodiment, a fast region convolutional neural network (English: faster region convolutional neural networks, abbreviation: Faster RCNN) algorithm can be used to identify ships in images, effectively improving the detection speed, detecting collision threats in a timely manner, and providing alarms to ensure the safety of ships. safety.

Optionally, this step 205 may include:

Extract image features from images;

Generate candidate regions based on image features;

Extract regional features based on image features and candidate areas;

According to the regional characteristics, the category of the candidate area is determined, and the position of the second ship in the image is obtained.

The above steps can be implemented on the basis of the existing Faster RCNN algorithm, which is more convenient to implement.

Figure 4 is a schematic structural diagram of the Faster RCNN algorithm model provided by the embodiment of the present disclosure. Referring to Figure 4, the model of the FasterRCNN algorithm includes convolutional layers (English: conv layers) 31, region proposal networks (English: regionproposal networks, abbreviation: PRN) 32, pooling layers (English: roi pooling) 33 and classifiers (English: roi pooling) 33 :classification)34. The convolution layer 31 extracts image feature information (English: image feature maps) from the image (English: image) 30 and outputs them to the region proposal network 32 and the pooling layer 33 respectively. The region proposal network 32 generates candidate regions (English: region proposals) based on image features and outputs them to the pooling layer 33 . The pooling layer 33 extracts feature information (English: proposal feature maps) of the candidate area based on the feature information of the image and the candidate area and outputs it to the classifier 34. The classifier 34 determines the category of the candidate area and thus the position of the target in the image based on the feature information of the candidate area.

Figure 5 is a schematic structural diagram of a convolution layer provided by an embodiment of the present disclosure. Referring to FIG. 5 , the convolution layer 31 may include multiple convolution kernels 41 , multiple activation functions (English: rectified linear unit, relu for short) 42 and multiple pooling layers 43 . The convolution kernel 41 can perceive local features; the activation function 42 can increase the nonlinearity of the neural network model; and the pooling layer 43 can aggregate statistics on features. For example, as shown in Figure 5, the convolution layer 31 may include 13 convolution kernels 41, 13 activation functions 42 and 4 pooling layers 43. According to the order of image processing, the convolution kernel 41, activation function Function 42, convolution kernel 41, activation function 42, pooling layer 43, convolution kernel 41, activation function 42, convolution kernel 41, activation function 42, pooling layer 43, convolution kernel 41, activation function 42, convolution Kernel 41, activation function 42, convolution kernel 41, activation function 42, pooling layer 43, convolution kernel 41, activation function 42, convolution kernel 41, activation function 42, convolution kernel 41, activation function 42, pooling layer 43, convolution kernel 41, activation function 42, convolution kernel 41, activation function 42, convolution kernel 41, activation function 42.

Figure 6 is a schematic structural diagram of a region recommendation network provided by an embodiment of the present disclosure. Referring to Figure 6, the region proposal network 32 first uses the convolution kernel 41 and the activation function 42 for processing, and generates a large number of candidate frames (English: anchors), and then is divided into two paths: after passing through the convolution kernel 41, the softmax logistic regression function is used 44 is classified to determine whether the candidate frame belongs to the foreground (English: positive) or the background (English: negative), and a reshape layer (English: reshape layer) 45 is set before and after the softmax logistic regression function to facilitate the classification of the softmax logistic regression function; another After passing through the convolution kernel 41, the offset of the candidate box is calculated through bounding box regression. The proposal layer (English: proposal layer) 46 obtains the candidate area based on the classification result and offset of the candidate frame.

Figure 7 is a schematic structural diagram of a classifier provided by an embodiment of the present disclosure. Referring to Figure 7, the classifier 34 may include a fully connected layer 47, which may establish connections between each neuron of the upper layer and all the neurons of the next layer. Illustratively, as shown in Figure 7, the classifier 34 first processes the fully connected layer 47, the activation function 42, the fully connected layer 47, and the activation function 42 and then divides it into two paths: one path passes through the fully connected layer 47 and is output as the result. ; After the other path passes through the fully connected layer 47, the softmax logistic regression function 44 is used for classification, and the offset of the candidate frame is calculated through border regression to improve the accuracy of the candidate area.

Optionally, before step 205, the method may also include:

Obtain multiple images with the location of the marked ship;

The model of the deep learning algorithm is trained using multiple images with the location of the ship marked.

In practical applications, the back propagation algorithm can be used to iteratively update the parameters in the deep learning algorithm model until the number of iterations reaches the set number or the loss function is within the set range.

Optionally, after step 205, the method may also include:

Store multiple images around the first ship;

The model of the deep learning algorithm is updated based on multiple images around the first ship.

Use the images obtained during the actual application to train the deep learning algorithm again, so that the updated deep learning algorithm is more consistent with the actual situation and improves the accuracy of recognition.

In practical applications, multiple images can be stored in industrial hard drives to work in harsh environments such as high salinity, high humidity, high temperature or low temperature, and can store massive amounts of data.

In practical applications, the process of training the model of the deep learning algorithm using multiple images around the first ship is similar to the process of training the model of the deep learning algorithm using multiple images with the location of the ship marked. Here, No more details.

Optionally, the method may also include:

A deep learning algorithm is used to determine the type of the second ship.

While using deep learning algorithms to identify ships in images, it can also determine the type of ship, without the need for human judgment based on the shape of obstacles scanned by radar.

In practical applications, the risk level of the second ship can also be determined according to the type of the second ship. For example, when other parameters are the same, the second ships are sorted in descending order of tonnage.

Step 206: Determine the danger level of the second ship based on the position of the second ship in each image.

Optionally, this step 206 may include:

The first step is to determine the distance between the second ship and the first ship at the time when the image was taken based on the position of the second ship in the image;

The second step is to determine the course and speed of the second ship based on the distance between the second ship and the first ship at the shooting time of each image;

The third step is to determine the danger level of the second ship based on its course, speed, and distance from the first ship.

According to the position of the second ship in each image, the course, speed, and distance between the second ship and the first ship can be obtained, and then the probability of collision between the second ship and the first ship can be analyzed, and the second ship can be obtained The hazard level of the ship is consistent with the actual situation and the accuracy is high.

In practical applications, the first step may include:

According to the position of the second ship in the image, the distance between the second ship and the first ship in the image is obtained;

According to the distance in the image and the zoom ratio of the image, the actual distance between the second ship and the first ship is obtained.

For example, if the distance between the second ship and the first ship in the image is 10cm and the scale of the image is 1:10000, then the actual distance between the second ship and the first ship is 1km.

For example, the second step may include:

When the distance between the second ship and the first ship decreases, determining the course of the second ship toward the first ship;

When the distance between the second ship and the first ship increases, it is determined that the course of the second ship is away from the first ship.

For example, if the distance between the second ship and the first ship decreases from 10km to 5km, the course of the second ship will be towards the first ship; another example, if the distance between the second ship and the first ship increases from 5km to 10km, the course of the second ship deviates from the first ship.

In practical applications, the second step can include:

The speed of the second ship is calculated based on the change in distance between the second ship and the first ship and the shooting interval of the images.

For example, if the distance between the second ship and the first ship is reduced by 0.1 km and the image shooting interval is 5 seconds, the speed of the second ship is 72 km/h.

Optionally, the third step can include:

Determine the gear to which the second ship belongs based on the distance between the second ship and the first ship and the distance range of the multiple gears;

Determine the course and speed of the second ship, and sort the second ships in the same gear.

For example, in the high-level threat level, the distance between the second ship and the first ship is less than 5km; in the medium-level threat level, the distance between the second ship and the first ship is between 5km and 10km. between; in the low-level threat level, the distance between the second ship and the first ship is between 10km and 15km; in the non-threat level, the distance between the second ship and the first ship is 15km above.

Among the second ships in the same gear, the second ship facing toward the first ship is arranged in front of the second ship facing away from the first ship.

Among the second ships in the same gear and facing the same direction, they are arranged in descending order of speed.

Step 207: Issue an alarm according to the danger level of the second ship.

Optionally, this step 207 may include:

In order from high to low risk levels, the courses, speeds, and distances to the first ship of multiple second ships are sequentially output to generate alarms.

Alarms are issued in sequence from high to low risk levels, which will help the first ship avoid the second ship with the highest probability of collision as soon as possible, effectively ensuring the safety of the ship.

In practical applications, the device that sends the alarm can be separately installed in the bridge of the first ship, or can also be installed in the alarm system of the first ship.

The embodiment of the present disclosure provides an all-weather ship identification system, which is suitable for implementing the all-weather ship identification method shown in Figure 2 or Figure 3 . Figure 8 is a schematic structural diagram of an all-weather ship identification system provided by an embodiment of the present disclosure. Referring to Figure 8, the system includes:

The acquisition module 301 is used to acquire multiple images around the first ship, and the multiple images are taken at different times;

The identification module 302 is used to identify the ship in each image using a deep learning algorithm and obtain the position of the second ship in each image;

The rating module 303 is used to determine the risk level of the second ship based on the position of the second ship in each image;

The alarm module 304 is used to issue an alarm according to the danger level of the second ship.

Embodiments of the present disclosure

Optionally, the acquisition module 301 may include:

Information acquisition submodule, used to acquire environmental information around the first ship;

The level determination submodule is used to determine the visibility level around the first ship based on environmental information;

The shooting control submodule is used to control the camera to continuously shoot multiple images around the first ship when the visibility level reaches the set standard; when the visibility level does not reach the set standard, control the laser pan/tilt to continuously shoot multiple images around the first ship. image.

Optionally, the identification module 302 may include:

Convolutional layers, used to extract image features from images;

Region proposal network, used to generate candidate regions based on image features;

The pooling layer is used to extract regional features based on image features and candidate regions;

The classifier is used to determine the category of the candidate area based on the area characteristics and obtain the position of the second ship in the image.

Optionally, the rating module 303 may include:

The first determination sub-module is used to determine the distance between the second ship and the first ship at the shooting time of the image based on the position of the second ship in the image;

The second determination sub-module is used to determine the course and speed of the second ship based on the distance between the second ship and the first ship at the shooting time of each image;

The third determination sub-module is used to determine the risk level of the second ship based on the course, speed, and distance of the second ship from the first ship.

Optionally, the alarm module 304 can be used to,

In order from high to low risk levels, the courses, speeds, and distances to the first ship of multiple second ships are sequentially output to generate alarms.

Optionally, the acquisition module 301 can also be used to:

Obtain multiple images with the vessel's location annotated.

Accordingly, the system may also include:

The training module is used to train the model of the deep learning algorithm using multiple images with the location of the ship marked.

Optionally, the system can also include:

A storage module used to store multiple images around the first ship;

The update module is used to update the model of the deep learning algorithm based on multiple images around the first ship.

It should be noted that when the system for identifying ships all-weather in the above embodiments is used to identify ships all-weather, the division of the above-mentioned functional modules is only used as an example. In practical applications, the above functions can be allocated to different functional modules as needed. Completion means dividing the internal structure of the system into different functional modules to complete all or part of the functions described above. In addition, the all-weather ship identification system provided by the above embodiments and the all-weather ship identification method embodiments belong to the same concept. The specific implementation process can be found in the method embodiments and will not be described again here.

The above serial numbers of the embodiments of the present disclosure are only for description and do not represent the advantages and disadvantages of the embodiments.

Those of ordinary skill in the art can understand that all or part of the steps to implement the above embodiments can be completed by hardware, or can be completed by instructing relevant hardware through a program. The program can be stored in a computer-readable storage medium. The above-mentioned The storage media mentioned can be read-only memory, magnetic disks or optical disks, etc.

The above are only optional embodiments of the present disclosure and are not intended to limit the present disclosure. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principles of the present disclosure shall be included in the protection of the present disclosure. within the range.

Claims (8)

1. A method of all-weather identification of a vessel, the method comprising:

acquiring a plurality of images around a first ship, wherein the shooting time of the plurality of images is different;

identifying the ship in each image by adopting a deep learning algorithm to obtain the position of a second ship in each image;

determining a distance between the shooting time of the second ship in the image and the first ship according to the position of the second ship in the image;

determining a heading of the second vessel toward the first vessel when a distance between the second vessel and the first vessel decreases; determining that the heading of the second vessel deviates from the first vessel when the distance between the second vessel and the first vessel increases;

calculating the navigational speed of the second ship according to the distance change between the second ship and the first ship and the shooting interval of the image;

determining a gear of a dangerous grade to which the second ship belongs according to the distance between the second ship and the first ship and the distance range of a plurality of gears; in a high threat gear, the distance between the second vessel and the first vessel is below 5 km; in a gear of a medium-grade threat, the distance between the second ship and the first ship is between 5km and 10 km; in a low-level threat gear, the distance between the second ship and the first ship is between 10km and 15 km; in a threat-free gear, the distance between the second vessel and the first vessel is above 15 km;

determining the course and the navigational speed of the second ship, and sequencing the second ship with the same gear; in the second vessel of the same gear, the second vessel facing the first vessel is arranged in front of the second vessel facing away from the first vessel; in the second ship with the same gear and the same orientation, arranging the second ship according to the sequence from the high speed to the low speed;

and sending out an alarm according to the danger level of the second ship.

2. The method of claim 1, wherein the acquiring a plurality of images of the surroundings of the first vessel comprises:

acquiring environmental information around the first ship;

determining a visibility level around the first vessel according to the environmental information;

when the visibility level reaches a set standard, controlling a camera to continuously shoot a plurality of images around the first ship;

and when the visibility level does not reach the set standard, controlling the laser cradle head to continuously shoot a plurality of images around the first ship.

3. The method of claim 1 or 2, wherein said identifying a second vessel in each of said images using a deep learning algorithm comprises:

extracting image features from the image;

generating a candidate region according to the image characteristics;

extracting region features according to the image features and the candidate regions;

and determining the category of the candidate region according to the region characteristics to obtain the position of the second ship in the image.

4. A method according to claim 1 or 2, wherein said alerting according to the risk level of the second vessel comprises:

and sequentially outputting the heading, the navigational speed and the distance between the second ships and the first ships for alarming according to the sequence of the dangerous level from high to low.

5. A system for all-weather identification of a vessel, the system comprising:

the acquisition module is used for acquiring a plurality of images around the first ship, and the shooting time of the plurality of images is different;

the identification module is used for identifying the ship in each image by adopting a deep learning algorithm to obtain the position of the second ship in each image;

the grading module is used for determining the distance between the shooting time of the second ship in the image and the first ship according to the position of the second ship in the image;

determining a heading of the second vessel toward the first vessel when a distance between the second vessel and the first vessel decreases; determining that the heading of the second vessel deviates from the first vessel when the distance between the second vessel and the first vessel increases;

calculating the navigational speed of the second ship according to the distance change between the second ship and the first ship and the shooting interval of the image;

determining a gear of a dangerous grade to which the second ship belongs according to the distance between the second ship and the first ship and the distance range of a plurality of gears; in a high threat gear, the distance between the second vessel and the first vessel is below 5 km; in a gear of a medium-grade threat, the distance between the second ship and the first ship is between 5km and 10 km; in a low-level threat gear, the distance between the second ship and the first ship is between 10km and 15 km; in a threat-free gear, the distance between the second vessel and the first vessel is above 15 km;

determining the course and the navigational speed of the second ship, and sequencing the second ship with the same gear; in the second vessel of the same gear, the second vessel facing the first vessel is arranged in front of the second vessel facing away from the first vessel; in the second ship with the same gear and the same orientation, arranging the second ship according to the sequence from the high speed to the low speed;

and the alarm module is used for sending out an alarm according to the danger level of the second ship.

6. The system of claim 5, wherein the acquisition module comprises:

the information acquisition sub-module is used for acquiring the environmental information around the first ship;

the grade determining submodule is used for determining the visibility grade around the first ship according to the environmental information;

the shooting control sub-module is used for controlling a camera to continuously shoot a plurality of images around the first ship when the visibility level reaches a set standard; and when the visibility level does not reach the set standard, controlling the laser cradle head to continuously shoot a plurality of images around the first ship.

7. The system of claim 5 or 6, wherein the identification module comprises:

a convolution layer for extracting image features from the image;

the regional suggestion network is used for generating candidate regions according to the image characteristics;

a pooling layer for extracting region features according to the image features and the candidate regions;

and the classifier is used for determining the category of the candidate region according to the region characteristics to obtain the position of the second ship in the image.

8. The system of claim 5 or 6, wherein the alarm module is configured to,

and sequentially outputting the heading, the navigational speed and the distance between the second ships and the first ships for alarming according to the sequence of the dangerous level from high to low.