A method and system for identifying water vessels based on infrared thermal imaging

Abstract

The present invention relates to the technical field of water vessel identification, and provides a method for identifying water vessels based on infrared thermal imaging, including: obtaining a plurality of original images by using an infrared thermal imaging camera; inputting each original image into a preset water vessel identification model to identify water and vessels to be identified in each original image, and determining a first coordinate range of the water and a second coordinate range of the vessels to be identified in each original image; determining alternative vessels in each original image according to the first coordinate range and the second coordinate range; obtaining the running trajectories of the alternative vessels according to the alternative vessels in each original image; and determining vessel targets according to the running trajectories of the alternative vessels. The infrared thermal imaging camera obtains the original images, uses the first coordinate range of the water and the second coordinate range of the vessels to be identified to exclude the interference of the coast, and the vessel movement trajectory conforms to the linear law to exclude the interference in the water area, thereby improving the high-precision and accurate identification of vessels in day and night water area monitoring.

Description

Technical Field

[0001] The present invention relates to the technical field of water vessel identification, and particularly to a method and system for water vessel identification based on infrared thermal imaging. Background Art

[0002] With the development of technology, marine resources have gradually attracted the attention of various countries. Maritime transportation has become increasingly prosperous, and maritime conflicts have also occurred from time to time, seriously affecting ship navigation and the water ecological environment. In order to better protect water resources, it is necessary to conduct safety monitoring and management of water areas.

[0003] In water area safety monitoring and management, the vessels traveling on the water area are the main management targets. When using image-based monitoring systems, monitoring personnel need to constantly view single or multiple monitoring screens. Not only is the screen single, but the water surface reflection and extremely poor image effects at night seriously affect the monitoring personnel's viewing of the monitoring screens, thereby affecting the monitoring effect of vessels on the water area.

[0004] In the prior art, visible light images are usually used for vessel identification. Although AI recognition and judgment are applied through visible light images, similar or approximate houses, pavilions and other targets on the shore can be clearly filtered out according to the color characteristics of the imaging. However, at night, there is not enough light on the water surface. In addition, there are foggy and rainy weather conditions on the water area, which will seriously affect the imaging effect of visible light images, resulting in a very low accuracy of the visible light image recognition method and unable to achieve 24-hour monitoring of the water area. In the prior art, thermal imaging images are used to identify vessels. Although 24-hour monitoring can be achieved, due to the imaging characteristics, similar or approximate houses, pavilions and other targets on the shore cannot be filtered out through the characteristics of the imaging color, resulting in a low recognition accuracy and a high false alarm rate. Summary of the Invention

[0005] Aiming at the deficiencies of the prior art, the present invention proposes a method for water vessel identification based on infrared thermal imaging, which improves the accuracy of vessel identification.

[0006] In a first aspect, the present invention provides a method for water vessel identification based on infrared thermal imaging.

[0007] In a first implementable manner, a method for water vessel identification based on infrared thermal imaging includes:

[0008] Obtaining a plurality of original images by using an infrared thermal imaging camera;

[0009] Inputting each original image into a preset water vessel identification model to identify the water and the vessels to be identified in each original image, and determining the first coordinate range of the water and the second coordinate range of the vessels to be identified in each original image;

[0010] Determine the alternative vessels in each original image according to the first coordinate range and the second coordinate range;

[0011] Obtain the alternative vessel movement trajectories based on the alternative vessels in each original image;

[0012] Determine the vessel targets according to the alternative vessel movement trajectories.

[0013] Combined with the first implementation manner, in the second implementation manner, the preset water area vessel recognition model is constructed in the following way:

[0014] Use an infrared thermal imaging camera to capture water area images and vessel images, and form a training set with the water area images and vessel images;

[0015] Construct an end-to-end initial neural network model, and train the end-to-end initial neural network model with the training set to obtain the water area vessel recognition model.

[0016] Combined with the first implementation manner, in the third implementation manner, determining the alternative vessels in each original image according to the first coordinate range and the second coordinate range includes:

[0017] Compare the first coordinate range and the second coordinate range in the same original image respectively;

[0018] When the second coordinate range is within the first coordinate range, determine the vessel to be recognized corresponding to the second coordinate range as the alternative vessel.

[0019] Combined with the first implementation manner, in the fourth implementation manner, obtaining the alternative vessel movement trajectories based on the alternative vessels in each original image includes:

[0020] Sort the original images in the order of shooting;

[0021] Determine the center point coordinates of the same alternative vessel in each original image respectively;

[0022] Sort the center point coordinates in the arrangement order of the original images to which they belong;

[0023] Plot the sorted center point coordinates as the alternative vessel movement trajectories.

[0024] Combined with the second implementation manner, in the fifth implementation manner, determining the vessel targets according to the alternative vessel movement trajectories includes:

[0025] Judge whether the alternative vessel movement trajectories of each alternative vessel conform to the linear movement law; if they conform, determine the alternative vessels whose alternative vessel movement trajectories conform to the linear movement law as the vessel targets.

[0026] In a second aspect, the present invention provides a water area vessel recognition system based on infrared thermal imaging.

[0027] In a sixth implementation manner, a water area vessel recognition system based on infrared thermal imaging includes:

[0028] An AI model analysis module, configured to obtain multiple original images of an infrared thermal imaging camera; and use a preset water area vessel recognition model to recognize the original images, so as to obtain water and vessels to be recognized in each original image;

[0029] A calculation module, configured to determine a first coordinate range of water and a second coordinate range of vessels to be recognized in each original image, and determine alternative vessels in each original image according to the first coordinate range and the second coordinate range; determine vessel targets according to the alternative vessels in each original image.

[0030] Combined with the sixth implementation manner, in a seventh implementation manner, determining vessel targets according to the alternative vessels in each original image includes:

[0031] Determine the center points of the alternative vessels;

[0032] Sort each frame of the original images in the shooting order;

[0033] Determine the order of each frame of the original images as the moving order of the corresponding center points of the alternative vessels;

[0034] Judge whether the alternative vessels meet a first rule according to the moving order of the center points of the alternative vessels, where the first rule is that the alternative vessels move in one direction;

[0035] Judge whether the alternative vessels meet a second rule according to the moving order of the center points of the alternative vessels, where the second rule is that the movement of the alternative vessels conforms to the linear motion law;

[0036] Determine the alternative vessels that meet both the first rule and the second rule as target vessels.

[0037] Combined with the sixth implementation manner, in an eighth implementation manner, it further includes:

[0038] An identification overlay module, configured to obtain the recognition frame of the vessel target and label recognition text on the recognition frame;

[0039] An image output module, configured to output the current frame image in which the vessel target is recognized to a preset client, and the recognition frame and recognition text of the vessel target are marked on the current frame image.

[0040] Combined with the eighth implementation manner, in a ninth implementation manner, obtaining the recognition frame of the vessel target includes:

[0041] Obtain the maximum and minimum horizontal and vertical coordinates of the vessel target;

[0042] Determine the rectangular frame of the vessel based on the maximum and minimum horizontal and vertical coordinates, set the attributes and color of the rectangular frame, and determine the rectangular frame as the recognition frame of the vessel target.

[0043] Combined with the sixth implementation method, in the tenth implementation method, it further includes:

[0044] An early warning module, which is used to continue taking tracking images of the vessel target after the vessel target is recognized;

[0045] Determine the running track of the vessel target based on the tracking images;

[0046] In the case where the vessel target changes its running direction or stops moving, give an early warning and output the current frame tracking image of the vessel target at the same time.

[0047] As can be seen from the above technical solutions, the beneficial technical effects of the present invention are as follows:

[0048] 1. Using an infrared thermal imaging camera as the front-end image sensor, compared with a visible light camera, it can effectively filter the influence of water surface waves and ripples on the image. At the same time, the characteristics of infrared thermal imaging ensure that the camera can image day and night, avoiding the problem that the visible light camera cannot be used on the water surface at night.

[0049] 2. The safety risk of vessels traveling at night on water is much higher than that during the day. However, the existing technologies cannot be effectively used at night or have high false alarms and low accuracy for day and night recognition. Through the method provided in this embodiment, an infrared thermal imaging camera is used to obtain the original image, and then the interference of the coast is initially excluded by using the first coordinate range of the water and the second coordinate range of the vessel to be recognized. The interference within the water area is further excluded by using the fact that the vessel's movement track conforms to a linear law, improving the high-precision and accurate recognition of vessels in day and night water area monitoring, thereby ensuring the day and night driving safety of vessels on the water. BRIEF DESCRIPTION OF THE DRAWINGS

[0050] In order to more clearly illustrate the specific implementation manners of the present invention or the technical solutions in the prior art, the following will briefly introduce the drawings required for use in the description of the specific implementation manners or the prior art. In all the drawings, similar elements or parts are generally identified by similar reference numerals. In the drawings, the elements or parts are not necessarily drawn to scale.

[0051] Figure 1 It is a schematic structural diagram of a method for identifying water area vessels based on infrared thermal imaging provided in this embodiment;

[0052] Figure 2 It is a schematic diagram of a system for identifying water area vessels based on infrared thermal imaging provided in this embodiment. Specific Embodiments

[0053] The embodiments of the technical solution of the present invention will be described in detail below with reference to the accompanying drawings. The following embodiments are only used to illustrate the technical solution of the present invention more clearly, so they are only examples and cannot be used to limit the protection scope of the present invention.

[0054] It should be noted that unless otherwise specified, the technical terms or scientific terms used in this application should be of the ordinary meaning understood by those skilled in the art to which the present invention belongs. The terms "first", "second", etc. in the description and claims of the embodiments of this disclosure and the above-mentioned drawings are used to distinguish similar objects and do not have to be used to describe a specific order or sequence. It should be understood that such data can be interchanged under appropriate circumstances for the implementation of the embodiments of this disclosure described here. In addition, the terms "include" and "have" and any variations thereof are intended to cover non-exclusive inclusion. Unless otherwise specified, the term "plurality" means two or more. In the embodiments of this disclosure, the character " / " means that the objects before and after are in an "or" relationship. For example, A / B means: A or B. The term "and / or" is a description of the association relationship of objects, indicating that three relationships can exist. For example, A and / or B means: A or B, or, A and B these three relationships. The term "corresponding" can refer to an association relationship or a binding relationship. A corresponding to B means that there is an association relationship or a binding relationship between A and B.

[0055] Combined with Figure 1 As shown, this embodiment provides a method for identifying water area vessels based on infrared thermal imaging, including:

[0056] Step S01: Use an infrared thermal imaging camera to obtain multiple original images;

[0057] Step S02: Input each original image into a preset water area vessel identification model to identify the water and the vessels to be identified in each original image, and determine the first coordinate range of the water and the second coordinate range of the vessels to be identified in each original image;

[0058] Step S03: Determine the alternative vessels in each original image according to the first coordinate range and the second coordinate range;

[0059] Step S04: Obtain the running trajectories of the alternative vessels according to the alternative vessels in each original image;

[0060] Step S05: Determine the vessel target according to the running trajectories of the alternative vessels.

[0061] In some embodiments, an infrared thermal imaging camera is used to shoot a water area monitoring video, and each frame image in the water area monitoring video is the original image.

[0062] The safety risks of vessels navigating at night on water are far higher than during the day. However, existing technologies cannot be effectively used at night, or have high false alarms and low accuracy for day-night identification. Through the method provided in this embodiment, an original image is obtained using an infrared thermal imaging camera, and then the interference of the coast is initially excluded using the first coordinate range of water, and the interference within the water area is further excluded using the vessel movement trajectory, improving the high-precision and accurate identification of vessels in day-night water area monitoring, thereby ensuring the day-night navigation safety of vessels on water.

[0063] Optionally, the preset water area vessel identification model is constructed as follows: Use an infrared thermal imaging camera to capture water area images and vessel images, and form a training set from the water area images and vessel images; construct an end-to-end initial neural network model, and train the end-to-end initial neural network model using the training set to obtain the water area vessel identification model.

[0064] In some embodiments, use an infrared thermal imaging camera to capture a first image containing only water area, a second image containing only vessels, and a third image containing both water area and vessels. There are no less than 1000 images for each of the first image, the second image, and the third image. Manually screen out these three types of images, and mark the water or vessels in each type of image. Then form a training set from the marked images of each type.

[0065] In some embodiments, the end-to-end initial neural network model (i.e., the end-to-end network) adopts the artificial intelligence framework of YOLOV3. Select the images in the training set as the sample input of the model, and use the annotations of the images as the labels of the samples to train the end-to-end initial neural network model to obtain the water area vessel identification model. The end-to-end initial neural network model divides the input image into an m*m grid, and each cell in the grid detects the targets within the cell, determines whether the target category is water or a vessel. If the target category is water or a vessel, then output the target category and the position of the target category, thus completing the output from the input of the original image to the target category and the position of the target category.

[0066] Optionally, determining the first coordinate range of water in each original image includes: Using a clustering method to cluster the coordinate points of the category of water, determining the largest clustering result as the water area, determining the boundary coordinates of the water area, and using the boundary coordinates of the water area as the first coordinate range of water.

[0067] Optionally, determining the second coordinate range of the vessels to be identified in each original image includes: After the preset water area vessel identification model outputs all the coordinates of the category of vessels, using a clustering method to cluster the coordinate points of the category of vessels, determining each clustering result as each vessel to be identified, and determining the boundary coordinates of each vessel to be identified as the second coordinate range of each vessel to be identified.

[0068] In some embodiments, the clustering method includes one of hierarchical clustering, K-means clustering, and second-order clustering.

[0069] Optionally, after inputting the original infrared thermal imaging image into the water area vessel recognition model, the water area vessel recognition model identifies the water and vessels in the original image and directly outputs the boundary coordinates of the water and the boundary coordinates of the vessels.

[0070] In some embodiments, the end-to-end initial neural network model (i.e., the end-to-end network) adopts the artificial intelligence framework of YOLOV5 and performs preprocessing such as "normalization" and "graying" on a large number of collected infrared images. These preprocessed images are manually labeled with class labels of vessels or water. The higher the labeling quality, the better the detection effect. After the image labeling is completed, they are placed in a specific directory according to the specification, and then a certain proportion of the training set and test set are generated. Then, the YOLOV5 is trained using the labeled images to obtain the water area vessel recognition model. Since the NPU supports the pytorch framework, the trained model is converted into a model under the pytorch framework, and then a ".rknn" file is generated and finally deployed to the front end.

[0071] The working principle at the front end is that the front-end infrared thermal imager acquires an infrared image and transmits this frame of infrared image to the NPU (neural-network process units) for analysis. The NPU performs a convolution operation on this frame of image and outputs its feature map, that is, determines the image of the vessel to be recognized, and then transmits it to the calculation module of the CPU for further calculation. The CPU (Central Processing Unit) compares the feature map with the recognition frame, generates the classification to which the target object belongs and its specific coordinates, and then superimposes the category and coordinate information on the image and outputs it to the monitoring display, thus completing the front-end target detection of the infrared image.

[0072] Optionally, determining the alternative vessels in each original image according to the first coordinate range and the second coordinate range includes: respectively comparing the first coordinate range and the second coordinate range in the same original image; in the case where the second coordinate range is within the first coordinate range, determining the vessel to be recognized corresponding to the second coordinate range as the alternative vessel.

[0073] In the same original image, compare the first coordinate range of the water and the second coordinate range of the vessel, exclude the vessels to be recognized outside the first coordinate range, and determine the vessels to be recognized within the first coordinate range as the alternative vessels. In this way, by identifying the water and the vessels and then using the water and the vessels to exclude the objects similar to vessels on the shore in the original infrared thermal imaging image, such as houses, pavilions, etc., various interference targets on the shore are eliminated, which is beneficial to improving the accuracy of vessel recognition.

[0074] Optionally, obtaining the running trajectories of alternative vessels according to the alternative vessels in each original image includes: sorting the original images in the order of shooting; respectively determining the central point coordinates of the same alternative vessel in each original image; sorting the central point coordinates in the arrangement order of the original images to which they belong; and plotting the sorted central point coordinates into the running trajectories of the alternative vessels.

[0075] Optionally, determining the vessel target according to the running trajectories of the alternative vessels includes: judging whether the running trajectories of the alternative vessels conform to the linear running law; if so, determining the alternative vessels whose running trajectories of the alternative vessels conform to the linear running law as the vessel targets. In this way, by analyzing and judging whether the running trajectories of the alternative vessels move linearly, the alternative vessels that do not conform to the linear movement law are excluded, so that static interference targets such as islands and docks on the water surface can be filtered out, and the accuracy of vessel recognition can be improved.

[0076] Combined with Figure 2 As shown in the figure, this embodiment provides a water area vessel recognition system based on infrared thermal imaging, including: an AI model analysis module 211, configured to obtain multiple original images of an infrared thermal imaging camera; and use a preset water area vessel recognition model to recognize the original images, and obtain water and vessels to be recognized in each original image; a calculation module 212, configured to determine a first coordinate range of the water and a second coordinate range of the vessels to be recognized in each original image, and determine alternative vessels in each original image according to the first coordinate range and the second coordinate range; and determine vessel targets according to the alternative vessels in each original image.

[0077] In some embodiments, the AI chip can support the artificial intelligence framework of YOLOV3, and the preset water area vessel recognition model is burned into the AI chip. After the thermal imaging camera captures the original image of the water area, the original image is sent to the AI chip, and the AI chip analyzes the original image, recognizes the vessels to be recognized and the water in the original image, determines alternative vessels in each original image according to the first coordinate range and the second coordinate range, and outputs the coordinates of the alternative vessels.

[0078] Optionally, determining the vessel target according to the alternative vessels in each original image includes: determining the central points of the alternative vessels; sorting the original images of each frame in the shooting order; determining the order of the original images of each frame as the moving order of the central points of the corresponding alternative vessels; judging whether the alternative vessels meet the first rule according to the moving order of the central points of the alternative vessels, and the first rule is that the alternative vessels run in one direction; judging whether the alternative vessels meet the second rule according to the moving order of the central points of the alternative vessels, and the second rule is that the running of the alternative vessels conforms to the linear motion law; and determining the alternative vessels that meet both the first rule and the second rule as the target vessels.

[0079] Optionally, determine the maximum and minimum horizontal and vertical coordinates of the alternative vessels, and determine the middle value of the maximum and minimum abscissas as the abscissa of the center point, and determine the middle value of the maximum and minimum ordinates as the ordinate of the center point.

[0080] Optionally, it further includes: an identification overlay module 213 for obtaining the recognition frame of the vessel target and annotating the recognition text on the recognition frame; an image output module 214 for outputting the current frame image of the recognized vessel target to a preset client, and the recognition frame and recognition text of the vessel target are marked on the current frame image.

[0081] Optionally, it further includes: an original image acquisition module for taking pictures of the water area with an infrared thermal imaging camera to obtain the original image and sending the original image to the AI model analysis module; an image display module for receiving the current frame image sent by the image output module and displaying the current frame image of the recognized vessel target through a display screen.

[0082] In some embodiments, in combination with Figure 2 As shown, a water area vessel recognition system based on infrared thermal imaging includes step S200: taking pictures of the water area with an infrared thermal imaging camera to obtain the original image; step S201: recognizing the original vessel target and outputting it to a preset user terminal; S202: performing image display on the vessel target. Among them, in step S200, the obtained original image is sent to the AI model analysis module, the AI model analysis module recognizes the water and the vessel to be recognized in the original image and sends it to the calculation module, the calculation module recognizes the vessel target, performs identification overlay on the vessel target through the identification overlay module, and then outputs the current frame image of the recognized vessel target to a preset client through the image output module, and the client displays the current frame image.

[0083] Optionally, obtaining the recognition frame of the vessel target includes: obtaining the maximum and minimum horizontal and vertical coordinates of the vessel target; determining the rectangular frame of the vessel according to the maximum and minimum horizontal and vertical coordinates, setting the attributes and colors of the rectangular frame, and determining the rectangular frame as the recognition frame of the vessel target.

[0084] Optionally, the attributes of the rectangular frame include the thickness and line shape of the line, etc.

[0085] Optionally, obtaining the recognition frame of the vessel target includes: after the calculation model determines the alternative vessels, determining the upper left and lower right coordinates of each alternative vessel, drawing a rectangular frame in the image according to the upper left and lower right coordinates of the alternative vessel, setting the thickness and color of the rectangular frame, and writing the word 'boat' at the upper left corner of each recognized rectangular frame.

[0086] In one embodiment, after the AI model analysis module identifies the vessel to be identified and water through a preset water area vessel identification model, it determines the coordinate ranges of the vessel to be identified and water, deletes the vessels to be identified outside the coordinate range of water, and retains the vessels to be identified within the coordinate range of water as alternative vessels. Then, it records the coordinates of the alternative vessels in each frame of the original image, determines the recognition frames of the alternative vessels according to the coordinates, takes the center points of the recognition frames as the center points of the alternative vessels, determines the shooting order of the original image as the arrangement order of the center points of the alternative vessels in the original image, draws the running trajectory in the two-dimensional coordinate system according to the arrangement order of the center points, and judges whether the center points of the alternative vessels move in the same direction and conform to the linear motion law. If both are met, it outputs as a vessel target and marks the recognition frame and recognition text. The image output module encodes and outputs the current frame image marked with the recognition frame and recognition text according to the required protocol, and sends the current frame image to a preset client or a preset display.

[0087] In some embodiments, in this solution, the calculation module compiles the analysis results of the AI model analysis module into a fixed algorithm according to the objective laws of things. By using the objective conditions of the known vessels moving on the water, that is, the running trajectory showing a linear law as the screening condition, the accuracy of vessel identification is improved, the problem of inability to monitor at night and false alarms under the condition of lack of color features of pixel points is avoided, and the reliability and accuracy of the day and night efficient application of the vessel identification technology are improved.

[0088] Optionally, it further includes: an early warning module, which is used to continue shooting the tracking images of the vessel target after identifying the vessel target; determining the running trajectory of the vessel target according to the tracking images; and giving an early warning and outputting the current frame tracking image of the vessel target at the same time when the vessel target changes its running direction or stops moving. By using the early warning module to track and give early warnings to the vessels on the water area, the safety of vessel operation is improved, and it is also beneficial to the supervision of the vessels on the water area.

[0089] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit them; although the present invention has been described in detail with reference to the foregoing embodiments, those of ordinary skill in the art should understand that they can still modify the technical solutions described in the foregoing embodiments, or perform equivalent replacements on some 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 invention, and they should all be covered by the scope of the claims and the description of the present invention.

Claims

1. A method for identifying water vessels based on infrared thermal imaging, characterized in that, Including: Obtaining multiple original images by using an infrared thermal imaging camera, taking a water area image and a ship image with the infrared thermal imaging camera, and forming a training set with the water area image and the ship image; Constructing an end-to-end initial neural network model, and training the end-to-end initial neural network model with the training set to obtain a water area ship recognition model; Inputting each original image into a preset water area ship recognition model, identifying the water and the ship to be recognized in each original image, and determining the first coordinate range of the water and the second coordinate range of the ship to be recognized in each original image; Determining alternative ships in each original image according to the first coordinate range and the second coordinate range; Obtaining the running trajectories of the alternative ships according to the alternative ships in each original image; Determining ship targets according to the running trajectories of the alternative ships; Among them, determining alternative ships in each original image according to the first coordinate range and the second coordinate range includes: Comparing the first coordinate range and the second coordinate range in the same original image respectively; When the second coordinate range is within the first coordinate range, determining the ship to be recognized corresponding to the second coordinate range as an alternative ship; Among them, obtaining the running trajectories of the alternative ships according to the alternative ships in each original image includes: Sorting each original image in the order of shooting; Respectively determining the center point coordinates of the same alternative ship in each original image; Sorting the center point coordinates in the arrangement order of the original images to which they belong; Plotting the sorted center point coordinates into the running trajectories of the alternative ships; Judging whether the running trajectories of the alternative ships conform to the linear running law; if they conform, determining the alternative ships whose running trajectories of the alternative ships conform to the linear running law as ship targets; Among them, determining ship targets according to the running trajectories of the alternative ships includes: Judging whether the running trajectories of the alternative ships conform to the linear running law and only run in one direction; if they conform, determining the alternative ships whose running trajectories of the alternative ships conform to the linear running law and only run in one direction as ship targets.

2. An infrared thermal imaging-based water vessel recognition system, characterized in that, This system uses a water area ship recognition method based on infrared thermal imaging as described in claim 1, including: An AI model analysis module, configured to obtain multiple original images of an infrared thermal imaging camera; and use a preset water area ship recognition model to recognize the original images, and obtain the water and the ships to be recognized in each original image; A calculation module, configured to determine the first coordinate range of the water and the second coordinate range of the ship to be recognized in each original image, and determine alternative ships in each original image according to the first coordinate range and the second coordinate range; determining ship targets according to the alternative ships in each original image.

3. The system according to claim 2, wherein Determining ship targets according to the alternative ships in each original image includes: Determining the center points of the alternative ships; Sorting each frame of the original images in the shooting order; Determining the order of each frame of the original images as the moving order of the corresponding alternative ship center points; Judging whether the alternative ships meet the first rule according to the moving order of the alternative ship center points, and the first rule is that the alternative ships run in one direction; Judge whether the alternative vessel meets the second rule according to the moving order of the center points of the alternative vessels. The second rule is that the movement of the alternative vessel conforms to the linear motion law; Determine the alternative vessels that meet both the first rule and the second rule as the target vessels.

4. The system according to claim 2, wherein It also includes: An identification overlay module for obtaining the recognition frame of the vessel target and annotating the recognition text on the recognition frame; An image output module for outputting the current frame image of the recognized vessel target to a preset client, where the recognition frame and recognition text of the vessel target are annotated on the current frame image.

5. The system according to claim 4, characterized in that, Obtaining the recognition frame of the vessel target includes: Obtaining the maximum and minimum horizontal and vertical coordinates of the vessel target; Determine the rectangular frame of the vessel according to the maximum and minimum horizontal and vertical coordinates, set the attributes and colors of the rectangular frame, and determine the rectangular frame as the recognition frame of the vessel target.

6. The system according to claim 2, wherein It also includes: An early warning module for continuously capturing the tracking image of the vessel target after the vessel target is recognized; Determine the running track of the vessel target according to the tracking image; In the case where the vessel target changes its running direction or stops moving, give an early warning and output the current frame tracking image of the vessel target at the same time.

Images