Navigation mark fault intelligent identification system based on AI artificial intelligence

Abstract

The invention discloses a navigation mark fault intelligent identification system based on AI artificial intelligence in the technical field of image identification. The navigation mark fault intelligent identification system comprises a camera, a platform layer, an algorithm layer and a hardware layer used for configuring the system. The platform layer comprises the following modules: an AI middle station which is used as a basic platform of the system and is responsible for data input, processing, task management, computing power scheduling, alarm processing and data processing and storage; the data input processing service module is responsible for receiving data from different cameras and providing support for video stream access of the different cameras and camera control access; the task management module is used for automatically scheduling inspection tasks according to a preset inspection plan and monitoring task execution conditions; the method is scientific and reasonable, and realizes real-time monitoring, automatic identification and intelligent analysis of the state of the navigation mark through the AI technology, thereby improving the efficiency and accuracy of navigation mark patrol inspection, reducing the cost of manpower and material resources, and improving the navigation safety guarantee level.

Description

Technical Field

[0001] The present invention belongs to the technical field of image recognition, and in particular is an intelligent recognition system for navigation mark faults based on AI artificial intelligence. Background Art

[0002] As the marine economy is increasingly prosperous, the importance of maritime safety as the cornerstone of the development of the marine industry is self-evident. As an important navigation facility in navigation, navigation marks provide accurate navigation information for ships and are a strong guarantee for the safe navigation of ships.

[0003] With the increase in navigation activities and the rapid growth in the number of navigation marks, navigation mark patrol and inspection work faces huge challenges. Traditional manual inspection methods are inefficient, costly, and difficult to achieve real-time monitoring of navigation mark status, which to some extent affects the level of navigation safety assurance. With the rapid development of artificial intelligence technology, new ideas and methods have been provided to solve the difficulties of navigation mark patrol and inspection. Breakthroughs in AI technology, especially in computer vision, machine learning, and deep learning, have made it possible to automate and intelligently patrol and inspect navigation marks.

[0004] In summary, the traditional manual inspection method is inefficient, costly and difficult to achieve real-time monitoring, which has become a problem that technicians in this field urgently need to solve. Therefore, it is necessary to propose an intelligent identification system for navigation mark faults based on AI artificial intelligence. Summary of the invention

[0005] In order to solve the above problems, the purpose of the present invention is to provide an intelligent identification system for navigation mark faults based on AI artificial intelligence. Through AI technology, real-time monitoring, automatic identification and intelligent analysis of the navigation mark status can be realized, thereby greatly improving the efficiency and accuracy of navigation mark patrol and inspection, reducing manpower and material costs, and improving the level of navigation safety.

[0006] In order to achieve the above-mentioned purpose, the technical solution of the present invention is as follows: an intelligent identification system for navigation mark faults based on AI artificial intelligence, including several cameras for real-time shooting of navigation mark image data, a platform layer, an algorithm layer and a hardware layer for system counterweight; the platform layer includes the following modules:

[0007] AI middle platform: As the basic platform of the system, it is responsible for data input, processing, task management, computing power scheduling, alarm processing, and data processing and storage.

[0008] Data input processing service module: responsible for receiving image data from different cameras, providing support for different camera video stream access and camera control access.

[0009] Task management module: automatically schedules inspection tasks according to the preset inspection plan and monitors the execution of tasks.

[0010] Computing power scheduling module: allocates computing resources according to task requirements.

[0011] AI alarm processing service module: When the AI ​​algorithm detects an abnormal state of the navigation mark, the system automatically generates an alarm message and sends it to the on-duty personnel.

[0012] The algorithm layer contains the following modules:

[0013] Navigation mark recognition algorithm module: Use the trained convolutional neural network to extract and classify the features of navigation mark images to achieve automatic recognition of navigation marks.

[0014] Appearance status recognition algorithm module: Through the deep learning algorithm, the features of the beacon appearance image are extracted and classified to realize the automatic recognition of the appearance status of the beacon. When the beacon is identified to be damaged, tilted, or dimly lit beyond the preset range, it is marked as abnormal.

[0015] Light quality status recognition algorithm module: Use image processing algorithm to process and analyze the image of navigation light quality, compare and analyze the recognized status with the known correct status of the navigation light, and realize abnormal status judgment of the navigation light status.

[0016] Furthermore, the hardware layer includes the following modules:

[0017] Server: Deploy AI middleware and algorithm layers to provide computing and data processing capabilities, and flexibly increase GPU computing power according to task requirements.

[0018] Storage device: used to store data collected by the camera and data processed by the AI ​​platform.

[0019] Network equipment: used to build a network environment and realize data transmission and sharing.

[0020] Auxiliary equipment: It is used to collect the frequency and intensity of water fluctuations in the preset area to assist the platform layer in real-time monitoring of navigation marks.

[0021] Furthermore, the convolutional neural network in the algorithm layer is trained based on the navigation mark image data model and the navigation mark feature data model.

[0022] Furthermore, the image data in the algorithm layer is cleaned and labeled before use.

[0023] Furthermore, after collecting the navigation mark images in the algorithm layer, negative sample data is collected, and the convolutional neural network is used to train based on the negative sample data model to improve the generalization ability in actual scenarios.

[0024] Furthermore, the trained algorithm model is evaluated and tested in the algorithm layer, and the performance of the model is measured using accuracy and recall indicators.

[0025] Furthermore, after the image data in the algorithm layer is collected, cleaned and annotated, the target detection algorithm model YOLOv8 is used to perform posture detection and damage detection on the navigation mark data.

[0026] Furthermore, grid search and random search methods are used in the algorithm layer to search for parameter combinations by adjusting the number of training samples, learning rate, and batch size parameters.

[0027] Furthermore, the appearance state recognition algorithm module is also used to obtain water fluctuation frequency and intensity data collected by auxiliary equipment. When the appearance state recognition algorithm module identifies the tilt condition of the beacon, it reduces the identified tilt amplitude by 0% to 30% based on the water fluctuation frequency and intensity data, and then judges the tilt condition based on the preset range.

[0028] Furthermore, the network equipment is also used to obtain public weather forecast information through the network. When the navigation mark is in severe weather with low visibility, the severe weather period is divided into several detection segments based on the time nodes when the visibility is above a preset value during the severe weather period according to the weather forecast information. In each detection segment, the water fluctuation frequency and intensity data collected by each auxiliary device are collected to generate a fluctuation function with time as the X-axis and fluctuation intensity as the Y-axis. The camera is used to collect images of the corresponding areas of the auxiliary devices at the time node at the end of each detection segment in the order of the integral calculation value of the fluctuation function of each auxiliary device.

[0029] The following principles and beneficial effects are achieved by adopting the above scheme:

[0030] 1. The system of the present invention uses the video resources of the North Sea Beacon Office and high-definition cameras to collect marine videos in real time, establishes a deep learning algorithm based on AI image analysis technology, analyzes and identifies the video images of the camera, intelligently identifies the beacons at sea, and provides services for identifying the appearance status, light quality, light frequency and color of the beacons through AI algorithms. In combination with the supporting AI middle station, the abnormal appearance events of the beacons found can be recorded, alarmed, reviewed and feedback can be given to realize the closed-loop management of the beacon status. The camera hardware equipment collects image data of the beacon and its surrounding environment in real time. The collected data includes images of the beacon in normal and abnormal states, as well as negative sample data of non-beacon objects. The collected image data is cleaned and annotated to ensure data quality and provide a reliable data set for subsequent algorithm training. Real-time monitoring, automatic identification and intelligent analysis of beacons are realized through AI technology, which greatly improves the efficiency and accuracy of beacon inspections.

[0031] 2. Algorithm processing in the present invention: Navigation mark recognition, using convolutional neural network to extract features and classify navigation mark images to realize automatic recognition of navigation marks. Appearance state recognition, using deep learning algorithm to extract features and classify navigation mark appearance images to determine whether the appearance of the navigation mark is intact. Light quality state recognition, using image processing algorithm to analyze the navigation mark light image, compare with the standard light state, and determine whether the light is abnormal. Use target detection algorithm YOLOv8 for posture detection and damage detection to improve the accuracy and efficiency of recognition. Finally, the algorithm parameters are optimized through grid search and random search methods to ensure the optimal algorithm performance. By collecting and analyzing a large amount of data, data support is provided for navigation mark management and maintenance, which helps to make more scientific and reasonable decisions. The system promptly detects and handles navigation mark failures, improves the level of navigation safety, and reduces navigation accidents caused by navigation mark failures.

[0032] 3. System scheduling and management in the present invention: The AI ​​middle station is the core of the system and is responsible for data processing, task management, computing power scheduling and alarm processing. The system automatically schedules inspection tasks according to the preset inspection plan and monitors the execution of tasks. The system dynamically allocates computing resources according to task requirements to improve system operation efficiency. When the system detects that the navigation mark is abnormal, it automatically generates an alarm message and sends it to the on-duty personnel. Automated monitoring reduces the need for manual inspections and reduces manpower and material costs.

[0033] 4. The appearance state recognition algorithm module in the present invention not only relies on traditional images to judge the inclination of the beacon, but also innovatively introduces water fluctuation frequency and intensity data. These data are collected through auxiliary equipment and can reflect the impact of water flow on the beacon. When identifying the tilt condition, the algorithm first reduces the initially identified tilt amplitude by 0% to 30% based on the water fluctuation data. This step is intended to reduce misjudgment caused by water flow fluctuations and improve the accuracy of tilt recognition. Subsequently, the algorithm compares the adjusted tilt amplitude with the preset range to determine whether the beacon is in an abnormal tilt state.

[0034] 5. The network equipment in the present invention obtains public weather forecast information in real time, especially forecasts for low-visibility severe weather. When it is predicted that the navigation mark may be in a low-visibility environment, the algorithm subdivides the severe weather period into multiple detection segments according to the time nodes in the weather forecast when the visibility is above the preset value. In each detection segment, the system collects the water fluctuation data of each auxiliary device and generates a fluctuation function, with time as the X-axis and the fluctuation intensity as the Y-axis, to intuitively display the water fluctuation situation. At the end of each detection segment, the camera does not directly scan the entire area, but prioritizes the collection of images of areas with larger fluctuations according to the order of the integral calculation values ​​of the fluctuation functions of each auxiliary device. This strategy can make more effective use of camera resources, give priority to areas that may need more attention due to the influence of water flow or wind and waves, and improve monitoring efficiency and accuracy. BRIEF DESCRIPTION OF THE DRAWINGS

[0035] Figure 1 Schematic diagram of the platform layer of an embodiment of the present invention.

[0036] Figure 2 Schematic diagram of algorithm layers of an embodiment of the present invention.

[0037] Figure 3 Schematic diagram of the hardware layer of an embodiment of the present invention. DETAILED DESCRIPTION

[0038] The following is further described in detail through specific implementation methods:

[0039] Example 1

[0040] Basically as attached Figure 1-Figure 3 As shown: A navigation mark fault intelligent identification system based on AI artificial intelligence includes several cameras for real-time shooting of navigation mark image data, a platform layer, an algorithm layer and a hardware layer for system configuration; the platform layer includes the following modules:

[0041] AI middle platform: As the basic platform of the system, it is responsible for data input, processing, task management, computing power scheduling, alarm processing, and data processing and storage. Its main function is to provide users with a friendly interface for viewing the damage identification of navigation marks, monitoring the operating status of the system, viewing the historical identification status information of navigation marks, and performing system settings and configurations.

[0042] Data input processing service module: responsible for receiving image data from different cameras, providing support for different camera video stream access and camera control access.

[0043] Task management module: automatically schedules inspection tasks according to the preset inspection plan and monitors the execution of tasks.

[0044] Computing power scheduling module: allocates computing resources according to task requirements to improve system operation efficiency.

[0045] AI alarm processing service module: When the AI ​​algorithm detects an abnormal state of the navigation mark, the system automatically generates an alarm message and sends it to the on-duty personnel.

[0046] The algorithm layer contains the following modules:

[0047] Navigation mark recognition algorithm module: Use the trained convolutional neural network to extract and classify the features of navigation mark images to achieve automatic recognition of navigation marks.

[0048] Appearance status recognition algorithm module: Through the deep learning algorithm, the features of the beacon appearance image are extracted and classified to realize the automatic recognition of the appearance status of the beacon. When the beacon is identified to be damaged, tilted, or dimly lit beyond the preset range, it is marked as abnormal.

[0049] Light quality status recognition algorithm module: Use image processing algorithm to process and analyze the image of navigation light quality, compare and analyze the recognized status with the known correct status of the navigation light, and realize abnormal status judgment of the navigation light status.

[0050] In order to improve the performance of the algorithm model in actual recognition, in addition to collecting a large amount of image data related to navigation marks, a large number of navigation mark images obtained in actual scenes are also required. In the algorithm layer, the convolutional neural network is trained based on the navigation mark image data model and the navigation mark feature data model. The image needs to contain navigation mark images in normal and abnormal states for algorithm training. In order to ensure that the collected image data is available, the image data needs to be cleaned and labeled. In addition to collecting navigation mark images, some negative sample data, such as non-navigation mark objects, can also be added, and then the convolutional neural network is trained based on the negative sample data model to improve the generalization ability in actual scenes. . In the algorithm layer, the trained algorithm model is evaluated and tested, and the accuracy and recall rate indicators are used to measure the performance of the model, and the model is further optimized based on the test results.

[0051] After the preliminary data collection, cleaning and annotation, the target detection algorithm model YOLOv8 is used to perform posture detection and damage detection on the navigation mark data, and the model is trained and optimized. Grid search, random search and other methods are used to adjust parameters such as the number of training samples, learning rate, batch size, and other hyperparameter optimization to systematically search for the optimal parameter combination to improve the performance and accuracy of the algorithm.

[0052] The hardware layer includes the following modules:

[0053] Server: Deploy the AI ​​middle platform and algorithm layer to provide powerful computing and data processing capabilities. The GPU computing power can be flexibly increased according to task requirements to speed up the efficiency of AI recognition.

[0054] Storage device: used to store data collected by the camera and data processed by the AI ​​platform.

[0055] Network equipment: used to build a stable network environment and achieve high-speed data transmission and sharing.

[0056] Auxiliary equipment: It is used to collect the frequency and intensity of water fluctuations in the preset area to assist the platform layer in real-time monitoring of navigation marks.

[0057] Specific implementation steps:

[0058] First, we collect image data of the beacon and its surrounding environment from multiple cameras, including images of the beacon in normal and abnormal states. We also add negative sample data, such as non-beacon objects, changes in the natural environment, etc., and then use a convolutional neural network to train the model based on the negative sample data to improve the generalization ability in actual scenarios.

[0059] The collected image data is cleaned to remove blurry, repeated or irrelevant images. The cleaned images are annotated to clarify the location, normal status and abnormal status of the navigation mark in each image.

[0060] The algorithm layer selects the algorithm model YOLOv8 for attitude detection and damage detection of navigation marks. The model is trained using labeled image data, and feature extraction and classification are performed using a convolutional neural network. A similar training process is also performed for the appearance state recognition algorithm and the light quality state recognition algorithm.

[0061] The algorithm layer uses grid search, random search and other methods to adjust parameters such as the number of training samples, learning rate, batch size, and other hyperparameters to optimize model performance. Evaluate the model performance under different parameter combinations and select the optimal parameter combination.

[0062] Deploy the AI ​​platform and algorithm layer on the server to ensure sufficient computing power and data processing capabilities. Install and configure storage devices to store data collected by the camera and data processed by the AI ​​platform. Build network equipment to ensure data transmission and sharing. Deploy auxiliary equipment in the designated area, including sensors for monitoring the frequency and intensity of water fluctuations.

[0063] Start the AI ​​middle platform and load the trained algorithm model. Configure the camera to ensure that the video stream can be smoothly connected to the system. According to the preset inspection plan, the task management module automatically schedules the inspection tasks. The computing power scheduling module allocates computing resources according to task requirements.

[0064] The camera continuously collects image data of the beacon and its surroundings. The data input processing service receives and processes this data and sends it to the algorithm layer for identification. The beacon identification algorithm, appearance status identification algorithm, and light quality status identification algorithm process the image data to determine the status of the beacon. When the AI ​​algorithm detects that the beacon status is abnormal, the AI ​​alarm processing service automatically generates an alarm message. The alarm message is sent to the on-duty personnel through the preset communication channel so that timely measures can be taken.

[0065] The AI ​​platform provides system operation status monitoring function to ensure stable system operation. Regularly check system logs to promptly identify and handle potential problems. Continuously collect new image data for continuous optimization and updating of models. Regularly retrain the model to adapt to new environments and requirements.

[0066] Example 2

[0067] Basically as attached Figure 1-Figure 3 As shown: Different from the above-mentioned embodiment, the appearance state recognition algorithm module is also used to obtain water fluctuation frequency and intensity data collected by auxiliary equipment. When the appearance state recognition algorithm module identifies the tilt condition of the beacon, the identified tilt amplitude is reduced by 0% to 30% based on the water fluctuation frequency and intensity data, and then the tilt condition is judged based on the preset range.

[0068] The network equipment is also used to obtain public weather forecast information through the network. When the navigation mark is in severe weather with low visibility, the severe weather period is divided into several detection segments based on the time nodes when the visibility is above the preset value during the severe weather period according to the weather forecast information. In each detection segment, the water fluctuation frequency and intensity data collected by each auxiliary equipment are collected to generate a fluctuation function with time as the X-axis and fluctuation intensity as the Y-axis. The camera is used to collect images of the corresponding areas of the auxiliary equipment at the time node at the end of each detection segment in the order of the integral calculation value of the fluctuation function of each auxiliary equipment.

[0069] Specific implementation steps:

[0070] The auxiliary equipment continuously collects water fluctuation frequency and intensity data and transmits the data to the system. The camera collects image data of the navigation mark area regularly or on demand. The water fluctuation data and image data are pre-processed by cleaning, denoising and other operations to improve the accuracy of subsequent processing.

[0071] The appearance state recognition algorithm module receives water fluctuation data and adjusts the initially identified tilt amplitude by 0% to 30% based on the frequency and intensity of water fluctuations. The adjusted tilt amplitude is compared with the preset tilt range to determine whether the navigation mark is in a tilted state. The network equipment regularly obtains public weather forecast information through the network, analyzes the weather forecast information, and identifies the bad weather period with low visibility of the navigation mark. Based on the time nodes when the visibility is above the preset value during the bad weather period, the period is subdivided into several detection segments.

[0072] In each detection section, the system collects the water fluctuation frequency and intensity data collected by each auxiliary device. With time as the X-axis and the fluctuation intensity as the Y-axis, a fluctuation function of each auxiliary device is generated. The end time node of each detection section is monitored. At the end of each detection section, the integral value of the fluctuation function of each auxiliary device is calculated and sorted according to the size of the integral. According to the order of the integral calculation value, the system gives priority to collecting images of the corresponding areas of the auxiliary equipment with higher integral values. The system further analyzes the collected images to detect whether there are any abnormalities with the navigation mark, including severe tilt, damage, etc. If an abnormality is detected, the system generates an alarm message and sends it to the relevant personnel. Based on Example 1, Example 2 adds the ability to predict and respond to severe weather, and improves the practicality and efficiency of the intelligent identification system for navigation mark failures.

[0073] The above is only an embodiment of the present invention. The common sense such as the known specific structure and characteristics in the scheme is not described in detail here. The ordinary technicians in the relevant field know all the common technical knowledge in the technical field of the invention before the application date or priority date, can obtain all the existing technologies in the field, and have the ability to apply the conventional experimental means before that date. The ordinary technicians in the relevant field can improve and implement this scheme in combination with their own abilities under the enlightenment given by this application. Some typical known structures or known methods should not become obstacles for ordinary technicians in the relevant field to implement this application. It should be pointed out that for those skilled in the art, without departing from the structure of the present invention, several deformations and improvements can be made, which should also be regarded as the protection scope of the present invention, which will not affect the effect of the implementation of the present invention and the practicality of the patent. The protection scope required by this application shall be based on the content of its claims, and the specific implementation methods and other records in the specification can be used to interpret the content of the claims.

Claims

1. An intelligent identification system for navigation mark faults based on AI artificial intelligence, characterized by: It includes several cameras for real-time shooting of navigation image data, platform layer, algorithm layer and hardware layer for system configuration; the platform layer includes the following modules: AI middle platform: As the basic platform of the system, it is responsible for data input, processing, task management, computing power scheduling, alarm processing, and data processing and storage; Data input processing service module: responsible for receiving image data from different cameras, providing support for different camera video stream access and camera control access; Task management module: automatically schedules inspection tasks according to the preset inspection plan and monitors the execution of tasks; Computing power scheduling module: allocates computing resources according to task requirements; AI alarm processing service module: When the AI ​​algorithm detects an abnormal state of the navigation mark, the system automatically generates an alarm message and sends it to the on-duty personnel; The algorithm layer contains the following modules: Navigation mark recognition algorithm module: Use the trained convolutional neural network to extract features and classify navigation mark images to achieve automatic recognition of navigation marks; Appearance status recognition algorithm module: The deep learning algorithm is used to extract and classify the appearance image of the navigation mark to realize automatic recognition of the appearance status of the navigation mark. When the navigation mark is found to be damaged, tilted, or dimly lit beyond the preset range, it is marked as abnormal. Light quality status recognition algorithm module: Use image processing algorithm to process and analyze the image of navigation light quality, compare and analyze the recognized status with the known correct status of the navigation light, and realize abnormal status judgment of the navigation light status.

2. The AI-based navigation mark fault intelligent identification system according to claim 1 is characterized by: The hardware layer includes the following modules: Server: Deploy AI middleware and algorithm layers to provide computing and data processing capabilities, and flexibly increase GPU computing power according to task requirements; Storage device: used to store data collected by the camera and data processed by the AI ​​platform; Network equipment: used to build a network environment to achieve data transmission and sharing; Auxiliary equipment: It is used to collect the frequency and intensity of water fluctuations in the preset area to assist the platform layer in real-time monitoring of navigation marks.

3. The AI-based intelligent navigation mark fault identification system according to claim 2 is characterized by: The convolutional neural network in the algorithm layer is trained based on the navigation mark image data model and the navigation mark feature data model.

4. The AI-based intelligent navigation mark fault identification system according to claim 3 is characterized by: Image data in the algorithm layer is cleaned and labeled before use.

5. The AI-based intelligent navigation mark fault identification system according to claim 4 is characterized by: After collecting the navigation mark images in the algorithm layer, negative sample data is collected. The convolutional neural network is used to train based on the negative sample data model to improve the generalization ability in actual scenarios.

6. The AI-based intelligent navigation mark fault identification system according to claim 5 is characterized by: In the algorithm layer, the trained algorithm model is evaluated and tested, and the performance of the model is measured using accuracy and recall indicators.

7. The AI-based intelligent navigation mark fault identification system according to claim 6 is characterized by: After the image data in the algorithm layer is collected, cleaned and annotated, the target detection algorithm model YOLOv8 is used to perform posture detection and damage detection on the navigation mark data.

8. The AI-based intelligent navigation mark fault identification system according to claim 7 is characterized by: Grid search and random search methods are used in the algorithm layer to search for parameter combinations by adjusting the number of training samples, learning rate, and batch size parameters.

9. The AI-based intelligent navigation mark fault identification system according to claim 8 is characterized by: The appearance state recognition algorithm module is also used to obtain water fluctuation frequency and intensity data collected by auxiliary equipment. When the appearance state recognition algorithm module identifies the tilt condition of the beacon, it reduces the identified tilt amplitude by 0% to 30% based on the water fluctuation frequency and intensity data, and then judges the tilt condition based on the preset range.

10. The AI-based intelligent navigation mark fault identification system according to claim 9 is characterized in that: The network equipment is also used to obtain public weather forecast information through the network. When the navigation mark is in bad weather with low visibility, the bad weather period is divided into several detection sections based on the time nodes when the visibility is above the preset value in the bad weather period of the weather forecast information. In each detection section, the water fluctuation frequency and intensity data collected by each auxiliary equipment are collected to generate a fluctuation function with time as the X-axis and fluctuation intensity as the Y-axis. The camera is used to collect images of the corresponding areas of the auxiliary devices at the end time of each detection segment in the order of the integral calculation value of the wave function of each auxiliary device.

Images