Marine navigation mark identification method, system, device and storage medium

Abstract

The present invention discloses a method, system, device and storage medium for identifying a maritime beacon, which can be applied to the field of maritime traffic technology. The method of the present invention comprises: constructing a training set for maritime beacon detection; constructing a beacon recognition model according to a preset target recognition algorithm including a YOLOv2 algorithm; training the beacon recognition model using a training set; acquiring image data to be identified; inputting the image data to be identified into the trained beacon recognition model to obtain a beacon recognition result for the sea area corresponding to the image data to be identified. The present invention constructs a beacon recognition model according to a preset target recognition algorithm including a YOLOv2 algorithm, and trains the beacon recognition model through a training set for maritime beacon detection, and then identifies the beacon in the sea area corresponding to the image data to be identified through the trained beacon recognition model, so as to improve the accuracy and real-time performance of the beacon recognition result, thereby transmitting the correct beacon location information to the ship.

Description

Technical Field

[0001] The present invention relates to the field of marine traffic technology, and in particular to a marine navigation mark identification method, system, device and storage medium. Background Art

[0002] A navigation mark is a navigation aid that guides ships to navigate. Its main function is to mark the scope of the waterway, indicate obstacles or provide danger signals, and it plays an important role in ensuring the safety of ships' navigation. Due to the influence of strong winds, waves and tides, the design position of the navigation mark, the location of the placement and the location observed by the navigator are inconsistent. In previous navigation activities, the specific location of the navigation mark was mostly identified by the naked eye, which posed a safety hazard. In related technologies, the existing recognition method for navigation marks is generally based on the use of deep learning networks for target recognition based on marine targets such as ships, islands and icebergs, while the automatic recognition of navigation marks currently uses traditional image processing methods for recognition, which not only has a slow recognition speed, but also has low robustness and cannot be applied to all sea areas. At the same time, there are also studies on the spatiotemporal prediction of the location of navigation marks in specific sea areas. These predictions are not necessarily completely accurate and can easily convey incorrect navigation mark location information to ships. Summary of the invention

[0003] The present invention aims to solve at least one of the technical problems existing in the prior art. To this end, the present invention provides a method, system, device and storage medium for identifying a marine beacon, which can effectively improve the accuracy of the beacon identification result.

[0004] On the one hand, an embodiment of the present invention provides a method for identifying a marine beacon, comprising the following steps:

[0005] Construct a training set for marine navigation mark detection;

[0006] Constructing a navigation mark recognition model according to a preset target recognition algorithm, wherein the preset target recognition algorithm includes a YOLOv2 algorithm;

[0007] Using the training set to train the navigation mark recognition model;

[0008] Acquire image data to be identified;

[0009] The image data to be identified is input into the trained navigation mark identification model to obtain the navigation mark identification result of the sea area corresponding to the image data to be identified.

[0010] In some embodiments, the step of constructing a training set for marine navigation mark detection includes:

[0011] Get several navigation mark images;

[0012] According to the plurality of navigation mark images, training samples are expanded through a deep convolutional generative adversarial network and a feature pyramid network;

[0013] marking the navigation mark position information and navigation mark category of the plurality of navigation mark images and the navigation mark images obtained by expansion;

[0014] All the labeled navigation mark images are saved in the training set.

[0015] In some embodiments, the method of expanding training samples based on the plurality of navigation mark images by using a deep convolutional generative adversarial network and a feature pyramid network includes:

[0016] Inputting the plurality of navigation mark images into a generator of a deep convolutional adversarial network to generate a navigation mark image feature map that meets a first preset requirement, and performing multiple deconvolutions on the navigation mark image feature map to obtain a first image of the same size as the navigation mark image;

[0017] Inputting the first image into a discriminator of a deep convolutional adversarial generative network, reconvolving the first image, and retaining the first image that meets the second preset requirement;

[0018] The plurality of navigation mark images and the first image meeting the second preset requirement are input into a feature pyramid network, and an enlargement operation and a reduction operation are respectively performed on each input image.

[0019] In some embodiments, the marking of the navigation mark position information and the navigation mark category of the plurality of navigation mark images and the expanded navigation mark images includes:

[0020] The navigation mark position information and navigation mark category of the plurality of navigation mark images and the expanded navigation mark images are marked by using a preset marking method.

[0021] In some embodiments, the step of training the navigation mark identification model using the training set includes:

[0022] Inputting the training set into the navigation mark recognition model, and using a preset clustering algorithm to analyze the size differences of target boxes on each image in the training set;

[0023] The distance between the target frames on each image in the training set is determined according to the target frame size difference.

[0024] In some embodiments, the using a preset clustering algorithm to analyze the target box size differences on each image in the training set includes:

[0025] Determine a merging ratio of target boxes on any two images in the training set;

[0026] The target frame size difference between the two images is determined according to the merging ratio.

[0027] In some embodiments, inputting the image data to be identified into the trained navigation mark identification model to obtain the navigation mark identification result of the sea area corresponding to the image data to be identified includes:

[0028] Inputting the image data to be identified into the trained navigation mark identification model;

[0029] Determining that the image data to be identified is video data, and determining a magnitude relationship between the number of frames of the video data and a first preset number of frames;

[0030] When the frame number of the video data is less than the first preset frame number, the beacon recognition model detects the beacon on each frame image in the video data in real time; when the frame number of the video data is greater than or equal to the first preset frame number, the beacon recognition model detects the beacon on the image corresponding to the second preset frame number in the video data per second in real time.

[0031] On the other hand, an embodiment of the present invention provides a marine navigation mark identification system, including:

[0032] The first building module is used to build a training set for marine navigation mark detection;

[0033] A second building module is used to build a navigation mark recognition model according to a preset target recognition algorithm, wherein the preset target recognition algorithm includes a YOLOv2 algorithm;

[0034] A training module, used for training the navigation mark recognition model using the training set;

[0035] An acquisition module, used for acquiring image data to be identified;

[0036] The recognition module is used to input the image data to be recognized into the trained navigation mark recognition model to obtain the navigation mark recognition result of the sea area corresponding to the image data to be recognized.

[0037] On the other hand, an embodiment of the present invention provides a marine navigation mark identification device, including:

[0038] at least one memory for storing a program;

[0039] At least one processor is used to load the program to execute the marine aid identification method.

[0040] On the other hand, an embodiment of the present invention provides a storage medium, in which a computer-executable program is stored. When the computer-executable program is executed by a processor, it is used to implement the method for identifying maritime beacons.

[0041] The embodiment of the present invention provides a method for identifying a marine beacon, which has the following beneficial effects:

[0042] This embodiment first constructs a training set for marine beacon detection, and constructs a beacon recognition model according to preset target recognition algorithms including the YOLOv2 algorithm. After the beacon recognition model is trained with the training set, image data to be recognized is obtained, and then the beacon in the sea area corresponding to the image data to be recognized is recognized with the trained beacon recognition model, so as to improve the accuracy and real-time performance of the beacon recognition result, thereby delivering the correct beacon location information to the ship.

[0043] Additional aspects and advantages of the present invention will be given in part in the following description and in part will be obvious from the following description, or will be learned through practice of the present invention. BRIEF DESCRIPTION OF THE DRAWINGS

[0044] The present invention will be further described below with reference to the accompanying drawings and embodiments, wherein:

[0045] Figure 1 A schematic diagram of an implementation environment of a method for identifying a marine beacon according to an embodiment of the present invention;

[0046] Figure 2 A flow chart of a method for identifying a marine beacon according to an embodiment of the present invention;

[0047] Figure 3 A schematic diagram of the structure of a deep convolutional adversarial generation network according to an embodiment of the present invention;

[0048] Figure 4 Schematic diagram of the YOLOv2 algorithm framework of an embodiment of the present invention;

[0049] Figure 5 A schematic diagram of training loss in an embodiment of the present invention;

[0050] Figure 6 This is a diagram of the k-means beacon aiming frame clustering result according to an embodiment of the present invention. DETAILED DESCRIPTION

[0051] Embodiments of the present invention are described in detail below, examples of which are shown in the accompanying drawings, wherein the same or similar reference numerals throughout represent the same or similar elements or elements having the same or similar functions. The embodiments described below with reference to the accompanying drawings are exemplary and are only used to explain the present invention, and cannot be understood as limiting the present invention.

[0052] In the description of the present invention, it should be understood that descriptions involving orientations, such as up, down, front, back, left, right, etc., and orientations or positional relationships indicated are based on the orientations or positional relationships shown in the accompanying drawings, and 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.

[0053] In the description of the present invention, "several" means more than one, "many" means more than two, "greater than", "less than", "exceed", etc. are understood to exclude the number itself, and "above", "below", "within", etc. are understood to include the number itself. If there is a description of "first" or "second", it is only used for the purpose of distinguishing the technical features, and cannot be understood as indicating or implying the relative importance or implicitly indicating the number of the indicated technical features or implicitly indicating the order of the indicated technical features.

[0054] In the description of the present invention, unless otherwise clearly defined, terms such as setting, installing, connecting, etc. should be understood in a broad sense, and technicians in the relevant technical field can reasonably determine the specific meanings of the above terms in the present invention in combination with the specific content of the technical solution.

[0055] In the description of the present invention, the description with reference to the terms "one embodiment", "some embodiments", "illustrative embodiments", "examples", "specific examples", or "some examples" means that the specific features, structures, materials or characteristics described in conjunction with the embodiment or example are included in at least one embodiment or example of the present invention. In this specification, the schematic representation of the above terms does not necessarily refer to the same embodiment or example. Moreover, the specific features, structures, materials or characteristics described may be combined in any one or more embodiments or examples in a suitable manner.

[0056] A navigation mark is a navigation aid that guides ships. Its main function is to mark the scope of the waterway, indicate obstacles or provide danger signals. It plays an important role in ensuring the safety of ships. Accurate navigation mark information can effectively improve the safety of ships during navigation. In related technologies, the navigation mark identification process has problems such as poor timeliness and low accuracy.

[0057] In view of this, an embodiment of the present invention provides a method for identifying marine beacons. The method in this embodiment constructs a beacon identification model through preset target recognition algorithms including the YOLOv2 algorithm to improve the effectiveness of the beacon identification model in the actual beacon identification process, and after training the beacon identification model with the constructed training set, the beacon identification model is used to identify the beacons in the actual image data, thereby improving the accuracy of the beacon identification results.

[0058] First, refer to Figure 1, an embodiment of the present invention provides a schematic diagram of an implementation environment of a method for identifying a maritime beacon. The hardware and software components of the environment of this embodiment mainly include an operation terminal 101 and a server 102, and the operation terminal 101 is in communication connection with the server 102. The method for identifying a maritime beacon can be configured to be executed on the operation terminal 101 alone, or can be configured to be executed on the server 102 alone, or can be executed based on the interaction between the operation terminal 101 and the server 102. The specific selection can be made according to the actual application situation, and this embodiment does not specifically limit this.

[0059] Specifically, the operation terminal 101 in this embodiment may include but is not limited to any one or more of a smart watch, a smart phone, a computer, a personal digital assistant (PDA), an intelligent voice interaction device, a smart home appliance or a ship terminal. The server 102 may be an independent physical server, or a server cluster or distributed system composed of multiple physical servers, or a cloud server that provides basic cloud computing services such as cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network services, cloud communications, middleware services, domain name services, security services, CDN (Content Delivery Network), and big data and artificial intelligence platforms. The operation terminal 101 and the server 102 may establish a communication connection through a wireless network or a wired network, which uses standard communication technology and / or protocols. The network may be set to the Internet or any other network, such as but not limited to any combination of a local area network (LAN), a metropolitan area network (MAN), a wide area network (WAN), a mobile, wired or wireless network, a private network or a virtual private network.

[0060] Since the shapes and colors of special navigation marks vary from region to region, the embodiment of the present invention takes a general navigation mark as an example to illustrate the specific process of the embodiment of the present application. The general navigation mark includes a left mark, a right mark, a north mark, a south mark, an east mark, a west mark and an isolated dangerous object mark.

[0061] Reference Figure 2 The embodiment of the present invention provides a method for identifying a marine beacon, which can be configured in at least one of an operation terminal or a server. The method of this embodiment includes but is not limited to:

[0062] Step 210: construct a training set for marine navigation mark detection.

[0063] In this embodiment, the navigation mark images can be collected from the web page, or obtained from the preset collection terminal, and the preset collection terminal can control the preset collection terminal to focus on shooting the navigation mark images when collecting the navigation mark images. Then, the training samples are expanded based on the collected several navigation mark images through the deep convolutional generative adversarial network and the feature pyramid network. Then, the navigation mark position information and navigation mark category of several navigation mark images and the expanded navigation mark images are annotated in a preset standard manner. For example, the navigation mark images can be annotated by the labelling tool to obtain the location information and category of the navigation mark on each navigation mark image for the navigation mark recognition model to learn and use. When the labelling tool is used to annotate the navigation mark image, it is only necessary to pass in the picture, manually select the target box containing the navigation mark target, and set the navigation mark category, and the labelling tool can automatically complete the annotation.

[0064] In this embodiment, the deep convolutional adversarial network is as follows: Figure 3 As shown. Among them, the deep convolutional generative adversarial network is based on game theory and the generative adversarial network as the prototype. The main components are the generator and the discriminator. During the processing, several images of navigation marks are input into the generator of the deep convolutional generative adversarial network to generate a navigation mark image feature map that meets the first preset requirement through the generator, and the navigation mark image feature map is deconvolved multiple times to obtain a first image of the same size as the navigation mark image. Then, the first image is input into the discriminator of the deep convolutional generative adversarial network to re-convolute the first image through the discriminator and retain the first image that meets the second preset requirement. For example, when a navigation mark image is input into the generator of a deep convolutional generative adversarial network, the generator will first generate a 100*1 (first preset requirement) navigation mark image feature map, enlarge the feature map through multiple deconvolutions, and finally obtain an image (first image) of the same size as the original data set image. The discriminator reconvolves the first image and determines whether the image meets the navigation mark image requirements (second preset requirements). If so, the first image will be added to the data set. If not, the first image will be removed.

[0065] After the navigation mark images are screened, the navigation mark images retained in the training set are input into the feature pyramid, so that each input navigation mark image can be enlarged and reduced through the feature pyramid network. In order to make the image conform to the network input size, when enlarging the image, the area containing the navigation mark target is manually selected; when reducing the image, the image is filled with 0 elements around it, thereby enriching the data in the data set to better train the navigation mark recognition model.

[0066] Step 220: construct a navigation mark recognition model according to a preset target recognition algorithm, wherein the preset target recognition algorithm includes a YOLOv2 algorithm with less computational effort, so as to increase the speed of data processing by the navigation mark recognition model during application, thereby improving timeliness. For example, when the navigation mark recognition model is constructed using the YOLOv2 algorithm, the YOLOv2 algorithm framework is as follows: Figure 4 As shown in the figure. YOLOv2 has a small amount of calculation and can perform real-time calculations. After experiments, Under the RTX2070SUPER graphics card hardware environment, the number of navigation mark images detected per second is 30.

[0067] Step 230: Use the training set to train the navigation mark recognition model.

[0068] In this embodiment, training is performed using mini-batch stochastic gradient descent, and the size of the mini-batch is 128. All weights are initialized to a normal distribution with a mean of 0 and a variance of 0.02. In LeakyReLU, the slope of the leak is set to 0.2. The hyperparameters adjusted by the optimizer use Adam, with a learning rate of 0.001, a momentum of 0.5, a decay coefficient of 0.0005, a stochastic gradient descent of 0.9, and a confidence threshold of 0.3, which helps stabilize the training of the beacon recognition model. During the training process, after continuous training and debugging, such as Figure 5 As shown in the figure, when the number of training iterations is 500, the algorithm can reach the optimal loss value, and the algorithm loss value does not decrease after another 30 rounds of training. Therefore, the number of iterations is set to 500.

[0069] Specifically, after obtaining the navigation mark position information of each navigation mark image in the training set, the target frame size of each navigation mark can be obtained. Since the size of the navigation mark target frame is inconsistent with that of other target frames, during the training process, after the training set is input into the navigation mark recognition model, a preset clustering algorithm is used to analyze the target frame size differences on each image in the training set; then the distance between the target frames on each image in the training set is determined based on the target frame size differences. Among them, the merging ratio of the target frames on any two images in the training set can be determined first, and then the target frame size difference on any two images can be determined based on the merging ratio. For example, the k-means algorithm is used to perform cluster analysis on the navigation mark frame. Among them, the K-means algorithm is a classic algorithm for clustering based on distance, which can classify all coordinate points on the plane according to the minimum distance principle.

[0070] When the K-means algorithm is used to analyze the distance of the target box, a larger true box in the iteration is more likely to produce a larger error than a smaller true box. Therefore, the distance judgment formula of the K-means algorithm can be changed to formula (1):

[0071] d(box,centroid)=1-IOU(box,centroid) Formula (1)

[0072] In formula (1), IOU (box, centroid) is the intersection-over-union ratio of the two target boxes, box and centroid, that is, the intersection area of ​​the two target boxes overlaps with their centroid area. When the two target boxes are similar, they are judged to be close to each other. When the size difference of the two target boxes is too large, they are judged to be close to each other.

[0073] like Figure 6 As shown in the figure, the dot coordinates represent the size information of all the beacon target frames participating in the clustering, and the five-pointed star coordinates represent the five anchor frames selected according to the size of these target frames. The position of the anchor frame is obtained by the improved k-means algorithm.

[0074] Step 240: Acquire the image data to be identified, and input the image data to be identified into the trained navigation mark identification model to obtain the navigation mark identification result of the sea area corresponding to the image data to be identified.

[0075] In this embodiment, after the training of the navigation mark recognition model is completed, the navigation mark recognition model can be applied to the actual navigation mark recognition process. Specifically, when the acquired image data to be recognized is a picture, the navigation mark recognition model can process more than 30 pictures per second. When the acquired image data to be recognized is video data, the relationship between the number of frames of the video data and the first preset number of frames is determined; when the number of frames of the video data is less than the first preset number of frames, the navigation mark recognition model detects the navigation mark on each frame of the video data in real time; when the number of frames of the video data is greater than or equal to the first preset number of frames, the navigation mark recognition model detects the navigation mark on the image corresponding to the second preset number of frames in the video data per second in real time. For example, when the number of frames of the video data is less than 30 (the first preset number of frames), each frame of the video can be detected; when the number of frames of the video data is greater than 30 (the first preset number of frames), 30 frames are selected from the video images per second for detection.

[0076] In this embodiment, the acquired image data to be recognized can be synchronously input into other algorithms to compare the effect with the method of this embodiment. Specifically, as shown in Table 1:

[0077] Table 1

[0078] Methods for participating in the comparison Small target accuracy / % Average accuracy / % Map FPS YOLOv2 (anchor boxes not improved) 90.0 94.0 0.93 30.2 Faster R-CNN 96.6 97.0 0.95 10.9 The navigation mark recognition model of this embodiment 96.6 96.0 0.94 30.2

[0079] It can be seen from Table 1 that the method of this embodiment can take into account both detection accuracy and detection speed. It should be noted that when the frame rate per second (FPS) is less than 30, the algorithm is not suitable for real-time detection.

[0080] At the same time, when the training set provided by this embodiment is used, the navigation mark recognition model provided by this embodiment has a high accuracy rate, as shown in Table 2:

[0081] Table 2

[0082]

[0083] It can be seen from Table 2 that the training set provided by this embodiment greatly improves the accuracy of marine landmark identification and can simultaneously ensure the detection speed.

[0084] The embodiment of the present invention provides a marine navigation mark identification system, including:

[0085] The first building module is used to build a training set for marine navigation mark detection;

[0086] A second building module is used to build a navigation mark recognition model according to a preset target recognition algorithm, wherein the preset target recognition algorithm includes a YOLOv2 algorithm;

[0087] A training module, used for training the navigation mark recognition model using the training set;

[0088] An acquisition module, used for acquiring image data to be identified;

[0089] The recognition module is used to input the image data to be recognized into the trained navigation mark recognition model to obtain the navigation mark recognition result of the sea area corresponding to the image data to be recognized.

[0090] The contents of the method embodiments of the present invention are all applicable to the system embodiments. The functions specifically implemented by the system embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method.

[0091] An embodiment of the present invention provides a marine navigation mark identification device, comprising:

[0092] at least one memory for storing a program;

[0093] At least one processor for loading the program to execute Figure 2 The method of identifying maritime aids to navigation is shown.

[0094] The contents of the method embodiments of the present invention are all applicable to the device embodiments. The functions specifically implemented by the device embodiments are the same as those of the above method embodiments, and the beneficial effects achieved are also the same as those achieved by the above method embodiments.

[0095] An embodiment of the present invention provides a storage medium, wherein a computer executable program is stored, and when the computer executable program is executed by a processor, it is used to implement Figure 2 The method of identifying maritime aids to navigation is shown.

[0096] The embodiment of the present invention further provides a computer program product or a computer program, wherein the computer program product or the computer program includes computer instructions, and the computer instructions are stored in a computer-readable storage medium. A processor of a computer device can read the computer instructions from the computer-readable storage medium, and the processor executes the computer instructions, so that the computer device executes Figure 2 The method of identifying maritime aids to navigation is shown.

[0097] The embodiments of the present invention are described in detail above in conjunction with the accompanying drawings, but the present invention is not limited to the above embodiments. Various changes can be made within the knowledge of ordinary technicians in the relevant technical field without departing from the purpose of the present invention. In addition, the embodiments of the present invention and the features in the embodiments can be combined with each other without conflict.

Claims

1. A method for identifying a marine beacon, characterized in that: The following steps are involved: Construct a training set for marine navigation mark detection; Constructing a navigation mark recognition model according to a preset target recognition algorithm, wherein the preset target recognition algorithm includes a YOLOv2 algorithm; Using the training set to train the navigation mark recognition model; Acquire image data to be identified; Inputting the image data to be identified into the trained navigation mark identification model to obtain the navigation mark identification result of the sea area corresponding to the image data to be identified; The construction of a training set for marine navigation mark detection includes: Get several navigation mark images; According to the plurality of navigation mark images, training samples are expanded through a deep convolutional generative adversarial network and a feature pyramid network; marking the navigation mark position information and navigation mark category of the plurality of navigation mark images and the navigation mark images obtained by expansion; Saving all the labeled navigation mark images in a training set; The method of expanding training samples based on the plurality of navigation mark images by using a deep convolutional generative adversarial network and a feature pyramid network includes: Input the plurality of navigation mark images into a generator of a deep convolutional adversarial network to generate a navigation mark image feature map that meets a first preset requirement, and perform multiple deconvolutions on the navigation mark image feature map to obtain a first image of the same size as the navigation mark image, where the first preset requirement is 100*1; Inputting the first image into a discriminator of a deep convolutional adversarial network, reconvolving the first image, and retaining the first image that meets a second preset requirement, where the second preset requirement is used to characterize a navigation mark image requirement; The plurality of navigation mark images and the first image meeting the second preset requirement are input into a feature pyramid network, and each input image is enlarged and reduced respectively, wherein the enlargement operation is used to represent that when the image is enlarged, an area containing the navigation mark target is manually selected, and the reduction operation is used to represent that when the image is reduced, the area around the image is filled with 0 elements, thereby enriching the data in the data set.

2. A method for identifying a marine beacon according to claim 1, characterized in that: The marking of the navigation mark position information and navigation mark category of the plurality of navigation mark images and the expanded navigation mark images includes: The navigation mark position information and navigation mark category of the plurality of navigation mark images and the expanded navigation mark images are marked by using a preset marking method.

3. A method for identifying a marine beacon according to claim 1, characterized in that: The adopting the training set to train the navigation mark recognition model comprises: Inputting the training set into the navigation mark recognition model, and using a preset clustering algorithm to analyze the size differences of target boxes on each image in the training set; The distance between the target frames on each image in the training set is determined according to the target frame size difference.

4. A method for identifying a marine beacon according to claim 3, characterized in that: The using a preset clustering algorithm to analyze the target frame size differences on each image in the training set includes: Determine a merging ratio of target boxes on any two images in the training set; The target frame size difference between the two images is determined according to the merging ratio.

5. A method for identifying marine navigation marks according to claim 1, characterized in that: The step of inputting the image data to be identified into the trained navigation mark identification model to obtain the navigation mark identification result of the sea area corresponding to the image data to be identified includes: Inputting the image data to be identified into the trained navigation mark identification model; Determining that the image data to be identified is video data, and determining a magnitude relationship between the number of frames of the video data and a first preset number of frames; When the frame number of the video data is less than the first preset frame number, the beacon recognition model detects the beacon on each frame image in the video data in real time; when the frame number of the video data is greater than or equal to the first preset frame number, the beacon recognition model detects the beacon on the image corresponding to the second preset frame number in the video data per second in real time.

6. A marine navigation mark identification system, characterized in that: include: The first building module is used to build a training set for marine navigation mark detection; A second building module is used to build a navigation mark recognition model according to a preset target recognition algorithm, wherein the preset target recognition algorithm includes a YOLOv2 algorithm; A training module, used for training the navigation mark recognition model using the training set; An acquisition module, used for acquiring image data to be identified; A recognition module, used for inputting the image data to be recognized into the trained navigation mark recognition model to obtain the navigation mark recognition result of the sea area corresponding to the image data to be recognized; The construction of a training set for marine navigation mark detection includes: Get several navigation mark images; According to the plurality of navigation mark images, training samples are expanded through a deep convolutional generative adversarial network and a feature pyramid network; marking the navigation mark position information and navigation mark category of the plurality of navigation mark images and the navigation mark images obtained by expansion; Saving all the labeled navigation mark images in a training set; The method of expanding training samples based on the plurality of navigation mark images by using a deep convolutional generative adversarial network and a feature pyramid network includes: Input the plurality of navigation mark images into a generator of a deep convolutional adversarial network to generate a navigation mark image feature map that meets a first preset requirement, and perform multiple deconvolutions on the navigation mark image feature map to obtain a first image of the same size as the navigation mark image, where the first preset requirement is 100*1; Inputting the first image into a discriminator of a deep convolutional adversarial network, reconvolving the first image, and retaining the first image that meets a second preset requirement, where the second preset requirement is used to characterize a navigation mark image requirement; The plurality of navigation mark images and the first image meeting the second preset requirement are input into a feature pyramid network, and each input image is enlarged and reduced respectively, wherein the enlargement operation is used to represent that when the image is enlarged, an area containing the navigation mark target is manually selected, and the reduction operation is used to represent that when the image is reduced, the area around the image is filled with 0 elements, thereby enriching the data in the data set.

7. A marine navigation mark identification device, characterized in that: include: at least one memory for storing a program; At least one processor is used to load the program to execute the marine navigation mark identification method as described in any one of claims 1-5.

8. A storage medium, characterized in that: A computer-executable program is stored therein, and when the computer-executable program is executed by a processor, it is used to implement the marine navigation mark identification method as described in any one of claims 1-5.

Images