Unmanned ship deck damage area detection system and method
By using drones to autonomously collect data and deep learning algorithms to identify damage to ship decks, and combining the mapping relationship between pixels and actual dimensions, the problem of low efficiency, poor accuracy, and high safety risks in traditional inspection has been solved, achieving efficient and accurate damage detection and data management.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- SANDIANSHUI NEW ENERGY TECH (ANHUI) CO LTD
- Filing Date
- 2026-04-22
- Publication Date
- 2026-07-31
AI Technical Summary
Traditional ship deck damage detection methods are inefficient, inaccurate, pose high safety risks, and have poor data traceability. They cannot achieve full coverage and high-precision area measurement, and data management is chaotic.
High-definition images are collected by drones during autonomous flight. Deep learning algorithms are used to identify damaged areas and calculate their areas, establishing a mapping relationship between pixels and actual physical dimensions to achieve standardized data management.
It achieves efficient and accurate damage detection, with a coverage rate of 99%, an area measurement error of ≤3%, high safety, standardized data management, and supports damage trend analysis and early warning throughout the entire life cycle.
Smart Images
Figure CN122492792A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the fields of ship structure inspection and UAV application technology, and more specifically, to a UAV ship deck damage area detection system and method. Background Technology
[0002] As the core load-bearing structure of a ship, the deck is constantly exposed to the marine environment, facing multiple effects such as wave impact, cargo abrasion, corrosion, and collisions. This makes it prone to damage issues such as coating deterioration, steel plate cracking, pitting, and localized peeling. If deck damage is not detected and repaired in a timely manner, the damaged area will continue to expand, affecting the deck's load-bearing capacity and even leading to structural failure, threatening the ship's navigational safety.
[0003] Traditional ship deck damage inspection mainly relies on manual boarding inspections, which has many technical bottlenecks and industry pain points: (1) Low detection efficiency and incomplete coverage: The ship deck area is large (the deck area of a large container ship can reach thousands of square meters). Manual inspection requires checking each area one by one. The inspection of a single ship takes up to several hours. Furthermore, the deck edge, high areas, and cargo hold perimeter are easily missed, resulting in blind spots in the inspection. (2) Low accuracy and strong subjectivity in measuring damaged area: Manual measurement of damaged area often uses tools such as tape measure and ruler. Due to the irregular shape of the damaged area and the limitation of the measurement angle, the measurement error can reach 20%-30%. Moreover, the measurement results depend on the experience of the operator and lack standardized quantitative basis. (3) High operational safety risks: Deck inspection requires personnel to climb and walk, and the swaying of the ship and the influence of sea wind during the voyage can easily cause personnel to fall or fall; at the same time, factors such as salt spray and humidity in the marine environment further increase the operational risks. (4) Data management is chaotic and traceability is poor: manual inspection data is mainly based on paper records and photo archives. Information such as the location, area and development trend of damage lacks systematic management, making it impossible to achieve historical data comparison and analysis and full life cycle tracking, which makes it difficult to support preventive maintenance decisions. Summary of the Invention
[0004] This invention aims to solve the problems of low efficiency, poor accuracy, high safety risks, and poor data traceability in traditional ship deck damage detection. It provides a UAV ship deck damage area detection system and method, which realizes automatic identification of deck damage areas, accurate boundary extraction, area quantification calculation, and full-process data management through UAV autonomous data collection and visual intelligent analysis, providing efficient, accurate, and safe technical support for ship deck repair and safety assessment.
[0005] To achieve the above objectives, in a first aspect, the present invention proposes an unmanned aerial vehicle (UAV) ship deck damage area detection system, comprising: The UAV visual acquisition module is used to plan flight routes based on target ship information and acquire high-definition images of the ship's deck via the UAV. The deck damage identification module is used to preprocess the high-definition image and identify the damaged area and extract the boundary contour of the damaged area through a deep learning algorithm. The damaged area calculation module is used to calculate the actual area of the damaged area by integrating UAV flight parameters, camera parameters and reference calibration information to establish a mapping relationship between pixels and actual physical size. The data management module is used to store the detection data containing the actual area, perform damage trend analysis, and generate detection reports.
[0006] Secondly, this invention proposes a method for detecting the damaged area of a ship's deck using an unmanned aerial vehicle (UAV), applied to the system described in the first aspect, comprising the following steps: S1: Import target vessel information, plan the detection area, set flight parameters and enter reference object size information; S2: The drone flies autonomously along the planned route and avoids obstacles in real time, collecting high-definition images of the deck and transmitting them back in real time. S3: Preprocess the acquired high-definition image, use deep learning algorithm to identify the damaged area and extract the boundary contour of the damaged area; S4: Based on the boundary of the damaged area, the pixel scale is calculated by integrating the UAV flight parameters and camera parameters, and after calibration with the reference object size information entered in the detection planning step, the actual area of the damaged area is calculated. S5: Summarize the detection data containing the actual area, generate a detection report, and complete data storage and archiving.
[0007] The beneficial effects of this invention are as follows: 1. Significantly improved detection efficiency and coverage: By using drones for autonomous data collection, the detection time for a single ship deck is reduced from several hours to tens of minutes, and comprehensive coverage of areas that are difficult for humans to reach is achieved, eliminating detection blind spots.
[0008] 2. Significantly improved area measurement accuracy: By integrating UAV flight parameters and camera parameters, and using reference objects on the deck for pixel scale calibration, accurate calculation of irregular damage areas was achieved. The relative measurement error was significantly lower than that of traditional manual measurement, providing reliable data for maintenance decisions.
[0009] 3. High safety of the inspection process: The use of drones for contactless remote inspection avoids the need for inspection personnel to board the ship for high-risk operations, effectively avoiding potential safety hazards.
[0010] 4. Standardized and intelligent data management: By establishing a database to systematically store and manage test data, it is possible to compare and analyze historical data and provide early warnings of damage development, providing technical support for implementing preventive maintenance strategies and helping to reduce maintenance costs and safety risks.
[0011] The present invention has other features and advantages, which will be apparent from or will be set forth in detail in the accompanying drawings and the following detailed description, which together serve to explain the particular principles of the invention. Attached Figure Description
[0012] The above and other objects, features and advantages of the present invention will become more apparent from the accompanying drawings, in which like reference numerals generally denote like parts.
[0013] Figure 1 A schematic diagram of a drone-based ship deck damage area detection system according to an embodiment of the present invention is shown.
[0014] Figure 2 A step diagram of a method for detecting the damaged area of a ship deck using an unmanned aerial vehicle (UAV) according to an embodiment of the present invention is shown. Detailed Implementation
[0015] Existing UAV visual inspection technology has been applied in damage detection in fields such as buildings and roads, but its adaptability to ship deck inspection is insufficient: it lacks the ability to identify damage in complex backgrounds on ship decks (such as containers, equipment, and railing obstructions), does not consider the impact of ship swaying and changes in ocean lighting on image acquisition, and the calculation of damage area is not accurately calibrated in conjunction with the ship deck coordinate system, thus failing to meet the professional and high-precision requirements of ship deck damage inspection.
[0016] This invention provides a system and method for detecting the damaged area of ship decks using unmanned aerial vehicles (UAVs). Through autonomous visual acquisition, intelligent damage identification, and accurate area calculation by UAVs, it achieves non-contact, high-efficiency, and high-precision detection of ship deck damage, realizes standardized management of damage information and full life cycle tracking, reduces detection safety risks, and provides a scientific basis for ship deck repair and safety assessment.
[0017] The invention will now be described in more detail with reference to the accompanying drawings. While preferred embodiments of the invention are shown in the drawings, it should be understood that the invention can be implemented in various forms and should not be limited to the embodiments set forth herein. Rather, these embodiments are provided so that the invention will be thorough and complete, and will fully convey the scope of the invention to those skilled in the art.
[0018] Example 1
[0019] This embodiment provides an unmanned aerial vehicle (UAV) system for detecting the damaged area of a ship's deck. This system adopts an integrated architecture of "autonomous UAV data acquisition + visual intelligent analysis + precise data processing," such as... Figure 1 As shown, the system in this embodiment includes: a UAV visual acquisition module, a deck damage identification module, a damage area calculation module, and a data management module. These modules work collaboratively to achieve intelligent detection of ship deck damage area throughout the entire process. The core process is: detection area planning → autonomous UAV flight and image acquisition → intelligent deck damage identification → precise extraction of damage boundaries → quantitative calculation of damage area → data archiving and report generation.
[0020] The core functional modules of the system in this embodiment will be described in detail below.
[0021] The UAV visual acquisition module is used to plan flight routes based on target ship information and acquire high-definition images of the ship's deck via the UAV. Specifically, the UAV visual acquisition module, as the core of on-site data acquisition, refers to a module responsible for acquiring high-precision image data on-site. Using a UAV as a carrier, it can be equipped with a high-definition visible light camera, a stabilization gimbal, a Global Positioning System (GPS), and a BeiDou Navigation Satellite System, and integrates an intelligent obstacle avoidance system to ensure the quality of image acquisition and flight safety in complex environments such as ship decks, adapting to the complex environment and high-precision acquisition requirements of ship decks. In a preferred embodiment, the specific functional configuration of this module includes: Visual equipment configuration: Equipped with a 4K high-definition visible light camera (resolution ≥3840×2160) and a three-axis stabilized gimbal, it has autofocus and exposure compensation functions and can adapt to complex lighting conditions such as strong ocean light, backlight, and cloudy days. Flight control capabilities: Supports GPS + Beidou dual-mode positioning, integrates millimeter-wave radar and visual obstacle avoidance system, can automatically avoid obstacles such as deck equipment, railings, containers, etc., and has a minimum safe distance of ≥1 meter from deck obstacles; Autonomous flight planning: Built-in digital model library of ship decks, supports custom detection area, flight altitude (default 5-10 meters, can be dynamically adjusted), and shooting overlap rate (heading overlap rate ≥80%, lateral overlap rate ≥60%). It can realize one-click take-off, autonomous cruise, and automatic return. The detection time of a single ship deck is ≤30 minutes (for decks under 1000㎡). Image acquisition and transmission: It adopts a combination of "fixed-point shooting + flight path cruise shooting" mode to ensure that no damaged areas are missed; the acquired images are transmitted back to the ground terminal in real time with a transmission delay of ≤300ms, supports wireless network (4G / 5G network or Wi-Fi) transmission, and has the function of resuming interrupted transmission to avoid data loss.
[0022] This module is designed to replace manual boarding inspections, aiming to solve the problems of low efficiency and high safety risks associated with traditional inspection methods. Leveraging the high mobility of drones, they can quickly cover a vast deck area and acquire image data from angles inaccessible to the human eye via onboard high-definition cameras, thereby achieving a comprehensive understanding of the deck's condition.
[0023] The deck damage identification module is used to preprocess the high-definition image and identify the damaged area and extract the boundary contour of the damaged area through a deep learning algorithm. Specifically, the deck damage identification module refers to a module that processes acquired images based on deep learning algorithms (such as Mask R-CNN), capable of semantic segmentation of the damaged area, identifying and distinguishing various typical damage types from the deck background at the pixel level, and accurately outlining the boundaries of the damaged area using contour extraction algorithms. In a preferred embodiment, the specific functional configuration of this module includes: Image preprocessing: To address image blurring and noise interference caused by changes in marine ambient lighting and ship swaying, adaptive illumination equalization, noise reduction filtering, and image sharpening algorithms are employed to improve image clarity and the ability to identify damage features. Damage type identification: A database of ship deck damage features was established (including 6 typical types of damage such as coating damage, steel plate cracking, pitting, and localized peeling). An improved Mask R-CNN algorithm was used to achieve semantic segmentation of the damaged area, accurately distinguishing the damage from deck equipment, shadows, normal textures, and other backgrounds. The damage identification accuracy rate was ≥96%. Precise boundary extraction: Through the contour extraction algorithm (Canny edge detection + morphological processing), the contour of the damaged area is automatically extracted, and the contour distortion caused by minor noise is eliminated. The boundary extraction error is ≤1 pixel.
[0024] The purpose of this module is to use computer vision technology to replace human eyes in making judgments, overcoming the subjectivity and fatigue of manual recognition. Through training with a deep learning model, the system can accurately separate real structural damage from complex deck backgrounds (such as rust, water stains, shadows, and equipment obstructions), solving the problem of low accuracy in ship deck scenarios for general detection schemes in existing technologies, and providing reliable input for subsequent precise quantification.
[0025] The damaged area calculation module is used to calculate the actual area of the damaged area by integrating UAV flight parameters, camera parameters and reference calibration information to establish a mapping relationship between pixels and actual physical size. Specifically, the damage area calculation module is a functional module that combines spatial calibration and coordinate transformation to quantify the pixel area in an image into the actual physical area with high precision. It establishes a mapping relationship between pixels and physical dimensions (pixel scale) by fusing real-time positioning data, attitude data, and camera intrinsic and extrinsic parameters from the UAV, and performs real-time calibration using known-sized reference objects on the deck to accurately calculate the actual area of the damaged region. In a preferred embodiment, the specific functional configuration of this module includes: Image spatial calibration: Based on UAV positioning data (latitude, longitude, altitude), camera intrinsic parameters (focal length, pixel size), and extrinsic parameters (attitude angle), a mapping relationship between image pixels and the actual physical dimensions of the deck is established using the formula S=(H The pixel scale is calculated as dx / f, where S is the pixel scale (unit: mm / pixel), H is the drone's flight altitude (unit: mm), dx is the camera pixel size (unit: mm / pixel), and f is the camera focal length (unit: mm). Area calculation methods: For regular-shaped damage (rectangles, circles), geometric formulas are used to calculate the area; for irregular-shaped damage, pixel counting (combining boundary extraction results to count the total number of pixels in the damaged area) is used, calculated using the formula A=N. S 2 Calculate the actual damaged area, where A is the actual damaged area (unit: m²), N is the total number of pixels in the damaged area, and S is the pixel scale (unit: m / pixel). Accuracy calibration mechanism: Real-time calibration is performed using reference objects of known dimensions on the deck (such as railings and container corner pieces) to correct scale errors caused by changes in flight altitude and attitude, with relative area measurement error ≤3%.
[0026] A key technical feature of this module is that it not only integrates the UAV's flight parameters (such as flight altitude) and camera parameters to calculate a basic pixel scale, but also further calibrates this pixel scale using a reference object of known size placed on the ship's deck. This design aims to address the issue of minute changes in flight altitude and attitude angles caused by factors such as airflow and ship sway during UAV flight. These changes directly affect the accuracy of the pixel scale and are a major source of area calculation errors. By introducing a reference object for real-time calibration, the system can dynamically compensate for these errors, thereby achieving high-precision measurement of the damaged area and solving the technical pain point of large measurement errors in the background technology.
[0027] The data management module is used to store the detection data containing the actual area, perform damage trend analysis, and generate detection reports.
[0028] Specifically, the data management module is used to achieve standardized management and traceability of damage inspection data. It is a platform for managing the entire data lifecycle, responsible for establishing a damage inspection database, structurally storing all inspection information, supporting multi-dimensional data retrieval, and possessing damage trend analysis capabilities. It can track damage changes and provide early warnings, ultimately automatically generating standardized inspection reports. In a preferred embodiment, the specific functional configuration of this module includes: Data storage function: Establish a ship deck damage detection database to store original images, damage identification results, boundary contour data, area calculation results, detection time, UAV flight parameters and other information. It supports multi-dimensional retrieval by ship name, deck area, detection time and damage type, and the data retention period is ≥5 years. Damage trend analysis: Automatically records the detection data of the same damaged area in previous tests, generates a damage area change curve, predicts the damage development rate (damage area change curve), and triggers an early warning when the damage area growth rate exceeds a preset threshold (e.g., triggering an early warning when the monthly growth rate of the damage area is ≥5%). Automatic report generation: Built-in standardized inspection report templates automatically generate inspection reports containing basic ship information, inspection tasks, damage distribution diagrams, detailed information on individual damage (location, type, area, boundary map), area statistics summary, maintenance recommendations, etc., and supports export in PDF / Word format.
[0029] This module is designed to address the problems of chaotic data management and difficulty in traceability in traditional inspection methods. By establishing a structured local or cloud database, it systematically links and stores information such as original images, damage location, type, area, and inspection time for each inspection. This not only provides ship managers with a clear and complete deck health record, but also lays the foundation for damage trend analysis and preventative maintenance through data accumulation, enabling full lifecycle management of deck structure health.
[0030] Example 2
[0031] This embodiment also provides a method for detecting the damaged area of a ship's deck using an unmanned aerial vehicle (UAV), applied to the system described in Embodiment 1, such as... Figure 2 As shown, the method includes the following steps: S1 (Detection Preparation and Planning): Import target vessel information, plan the detection area, set flight parameters and enter reference object size information; Specifically, users can import the target ship's deck structure drawings or select the corresponding ship type template through a ground terminal, customize the inspection area (such as the main deck and cargo hold deck edge), and set parameters such as flight altitude, shooting overlap rate, and safety distance. The system will automatically generate the optimal inspection route with full coverage and no blind spots. At the same time, the system will input the deck reference object size information for subsequent area calibration.
[0032] S2 (Autonomous Image Acquisition by Drone): The drone flies autonomously along a planned route and avoids obstacles in real time, acquiring high-definition images of the deck and transmitting them back in real time. Specifically, the drone takes off from a designated take-off and landing point on the ship's deck and autonomously performs inspection tasks according to the planned route. It uses a stabilized gimbal to keep the camera vertically downward and adopts a combination of fixed-point shooting and route-following shooting to collect high-definition images of the deck (heading overlap rate of no less than 80% and lateral overlap rate of no less than 60%). The high-definition images of the deck collected in real time are transmitted back to the ground terminal via wireless network (4G / 5G network or Wi-Fi). If the network is interrupted, the drone will automatically resume transmission. During the flight, the drone will automatically avoid obstacles using millimeter-wave radar and a visual obstacle avoidance system. For suspected damaged areas, the drone will automatically reduce its altitude (as low as 3 meters) to take close-up pictures to improve the identification of damage details.
[0033] This step is the starting point of the entire data chain, and its purpose is to acquire high-quality raw data. By replacing manpower with drones, the risks of working at heights and the limitations of visibility faced by traditional inspection methods are solved, enabling safe and wide-angle data collection.
[0034] S3 (Intelligent Deck Damage Recognition): The acquired high-definition images are preprocessed, and deep learning algorithms are used to identify the damaged areas and extract the boundary contours of the damaged areas. Specifically, the ground terminal uses adaptive illumination equalization, noise reduction filtering, and image sharpening algorithms to preprocess the acquired high-definition images. It identifies damaged areas using an improved Mask R-CNN algorithm and uses a contour extraction algorithm that combines Canny edge detection and morphological processing to extract pixel-level boundaries of damaged areas, complete semantic segmentation and boundary extraction, automatically label damage type and location coordinates (based on the ship deck coordinate system), and generate a damage distribution heat map.
[0035] This step is the core of intelligent analysis. Its principle is to use an algorithmic model to replace the human eye in making professional judgments, deepening the granularity of recognition from the object level to the pixel level. This allows for the precise depiction of irregular, damaged outlines, providing high-precision input for area calculation and directly improving the accuracy of the final result. Through this step, damage can be automatically and objectively located and depicted from complex image backgrounds, solving the problems of subjective bias and inefficiency inherent in manual interpretation.
[0036] S4 (Accurate Calculation of Damaged Area): Based on the boundary of the damaged area, the pixel scale is calculated by integrating the UAV flight parameters and camera parameters, and after calibration with the reference object size information entered in the detection planning step, the actual area of the damaged area is calculated. Specifically, the system calculates the pixel scale S=(H) based on the UAV flight parameters and camera parameters. dx) / f, and perform real-time accuracy calibration of the scale using a reference object of known size on the deck; for each damaged area, calculate the total number of pixels N based on the boundary extraction results, and calculate the area using the formula A=N. S² yields the actual damaged area A, and automatically calculates the damaged area of a single region, the total damaged area, and the percentage of each type of damaged area.
[0037] The purpose of this step is to eliminate measurement errors caused by flight disturbances, which is crucial to ensuring the accuracy of the final area data. By introducing a reference object for closed-loop calibration, the fundamental problem of large area calculation errors due to inaccurate scales in existing technologies is solved.
[0038] S5 (Data Archiving and Report Generation): Summarize the detection data containing the actual area, generate a detection report, and complete data storage and archiving.
[0039] Specifically, the system synchronously uploads the original images, damage identification results, boundary contour data, area calculation results, detection time, and flight parameters to the ship deck damage detection database; it supports historical data comparison and trend analysis, can automatically retrieve previous detection data for the same damaged area, generate a damage area change curve, and trigger an early warning when the damage area growth rate exceeds a preset threshold; the system automatically generates standardized detection reports (including a damage distribution diagram and area statistics results), and users can view, edit, and export reports as needed to support maintenance decisions and safety assessments.
[0040] This step, by storing structured data in a database, solves the problems of traditional paper records being easily lost and difficult to analyze, providing a data foundation for the full lifecycle management of ships. Simultaneously, this step elevates data management from simple storage to intelligent analysis and early warning, enabling this method not only to diagnose current problems but also to predict future risks, achieving a shift from passive response to proactive prevention.
[0041] Compared with the prior art, the innovation of this invention is mainly reflected in the following aspects: (1) Adaptation design for complex ship deck environments: In response to special scenarios such as dense equipment on ship decks, variable lighting, and ship swaying, the UAV obstacle avoidance mechanism and image preprocessing algorithm are optimized to solve the pain points of traditional visual detection being easily interfered with and having low recognition accuracy, so as to achieve full coverage and blind spot detection.
[0042] (2) Integrated technology for damage identification and boundary extraction: The improved Mask R-CNN algorithm is adopted to simultaneously complete the identification of damage type and semantic segmentation. Combined with the precise contour extraction algorithm, the accuracy of damage boundary extraction is ensured, providing a reliable basis for area calculation.
[0043] (3) High-precision area calculation method with multi-parameter fusion: integrate UAV positioning, camera parameters and reference calibration to establish a precise mapping relationship between pixels and actual size, solve the problem of difficult measurement and large error of irregular damaged area, and the relative measurement error is ≤3%.
[0044] (4) Full life cycle data management system: Establish a damage detection database to realize standardized storage, trend analysis and traceability of damage information, support preventive maintenance of ship decks and break through the limitations of chaotic traditional manual inspection data management.
[0045] Based on the above innovations, and from the perspective of practical application effects, the technical effects of this invention are reflected in the following aspects: (1) Significantly improved detection efficiency: The single-ship deck inspection time is reduced from several hours to less than 30 minutes, with an efficiency increase of 6-8 times. At the same time, it covers blind spots that are difficult for humans to reach, with a detection coverage rate of ≥99%.
[0046] (2) Significantly improved area measurement accuracy: The relative error of damaged area measurement is ≤3%, which is far superior to the accuracy of manual measurement, providing accurate data support for maintenance plan formulation and safety assessment.
[0047] (3) Operational safety risks are completely reduced: The drone remote inspection avoids personnel climbing and walking on the deck, eliminating safety risks such as falls and ensures the personal safety of inspection personnel.
[0048] (4) Data management standardization upgrade: realize the systematic storage, traceability and trend analysis of data throughout the damage detection process, provide technical support for the health management of the ship deck throughout its entire life cycle, and reduce maintenance costs and safety risks caused by the expansion of damage.
[0049] This invention is technologically mature and highly adaptable, and can be widely applied to the following ship deck damage detection scenarios: (1) Routine inspection of shipping companies: Provide shipping companies with efficient and accurate deck damage detection tools to achieve regular inspection and damage tracking, and ensure the safety of ship navigation.
[0050] (2) Ship repair shop maintenance and inspection: support the shipyard's damage assessment before deck repair and acceptance inspection after repair, improve maintenance quality and efficiency.
[0051] (3) Statutory inspection by ship surveying agencies: Provide standardized deck damage detection methods for ship surveying agencies (such as CCS, BV, LR, etc.) to meet the inspection requirements of statutory ship inspection and classification inspection for deck structure.
[0052] (4) Safety spot checks by maritime regulatory authorities: Provide non-invasive testing tools to maritime authorities to improve the coverage and efficiency of ship deck safety supervision and promptly identify safety hazards.
[0053] The various embodiments of the present invention have been described above. These descriptions are exemplary and not exhaustive, nor are they limited to the disclosed embodiments. Many modifications and variations will be apparent to those skilled in the art without departing from the scope and spirit of the described embodiments.
Claims
1. A system for detecting the damaged area of a ship's deck using an unmanned aerial vehicle (UAV), characterized in that, include: The UAV visual acquisition module is used to plan flight routes based on target ship information and acquire high-definition images of the ship's deck via the UAV. The deck damage identification module is used to preprocess the high-definition image and identify the damaged area and extract the boundary contour of the damaged area through a deep learning algorithm. The damaged area calculation module is used to calculate the actual area of the damaged area by integrating UAV flight parameters, camera parameters and reference calibration information to establish a mapping relationship between pixels and actual physical size. The data management module is used to store the detection data containing the actual area, perform damage trend analysis, and generate detection reports.
2. The system according to claim 1, characterized in that, The UAV visual acquisition module is equipped with a 4K high-definition camera and a three-axis stabilization gimbal to keep the camera shooting vertically downwards. It supports GPS and Beidou dual-mode positioning to obtain UAV location information in real time. It integrates millimeter-wave radar and visual obstacle avoidance system to automatically avoid deck obstacles. It also supports wireless network image transmission and has a breakpoint resume function to transmit the acquired images back in real time.
3. The system according to claim 1, characterized in that, The deck damage identification module is configured as follows: Adaptive illumination equalization, noise reduction filtering, and image sharpening algorithms are employed to improve image clarity and the ability to identify damage features. An improved Mask R-CNN algorithm is used to identify the damage type and perform semantic segmentation on the preprocessed image; A contour extraction algorithm combining Canny edge detection and morphological processing is used to extract pixel-level boundaries of damaged areas.
4. The system according to claim 1, characterized in that, The damaged area calculation module is configured as follows: According to the formula S=(H dx) / f calculates the pixel scale S, where H is the drone's flight altitude, dx is the camera pixel size, and f is the camera focal length; Real-time calibration is performed using a reference object of known size on the deck to correct for errors in the pixel scale bar S. Based on the pixel scale S and the total number of pixels N in the damaged area, the formula A=N is used. S² is used to calculate the actual damaged area A.
5. The system according to claim 1, characterized in that, The data management module is configured as follows: Establish a database for ship deck damage detection, storing original images, damage identification results, boundary contour data, area calculation results, detection time, and flight parameters, and supporting multi-dimensional retrieval by ship name, deck area, detection time, and damage type; Automatically records all detection data for the same damaged area, generates a curve showing the change in the damaged area, and triggers an alert when the growth rate of the damaged area exceeds a preset threshold. Automatically generate standardized inspection reports that include a diagram of the damage distribution and area statistics.
6. A method for detecting the damaged area of a ship's deck using an unmanned aerial vehicle (UAV), applied to the system described in any one of claims 1 to 5, characterized in that, Includes the following steps: S1: Import target vessel information, plan the detection area, set flight parameters and enter reference object size information; S2: The drone flies autonomously along the planned route and avoids obstacles in real time, collecting high-definition images of the deck and transmitting them back in real time. S3: Preprocess the acquired high-definition image, use deep learning algorithm to identify the damaged area and extract the boundary contour of the damaged area; S4: Based on the boundary of the damaged area, the pixel scale is calculated by integrating the UAV flight parameters and camera parameters, and after calibration with the reference object size information entered in the detection planning step, the actual area of the damaged area is calculated. S5: Summarize the detection data containing the actual area, generate a detection report, and complete data storage and archiving.
7. The method according to claim 6, characterized in that, Step S2 includes: The drone flies autonomously along a planned route, automatically avoiding deck obstacles using millimeter-wave radar and a visual obstacle avoidance system. The camera is kept vertically downward using a three-axis stabilized gimbal, and high-definition images of the deck are acquired using a combination of fixed-point shooting and cruise shooting, with a directional overlap rate of no less than 80% and a lateral overlap rate of no less than 60%. The high-definition images collected are transmitted back to the ground terminal in real time via wireless network, and the interruption resume function is automatically enabled when the network is interrupted.
8. The method according to claim 6, characterized in that, Step S3 includes: Adaptive illumination equalization, noise reduction filtering, and image sharpening algorithms are used to preprocess the acquired high-definition images; An improved Mask R-CNN algorithm is used to identify the damage type and perform semantic segmentation on the preprocessed image; A contour extraction algorithm combining Canny edge detection and morphological processing is used to extract pixel-level boundaries of damaged areas.
9. The method according to claim 6, characterized in that, Step S4 includes: According to the formula S=(H dx) / f calculates the pixel scale S, where H is the flight altitude, dx is the pixel size, and f is the focal length; The pixel scale S is calibrated in real time using a reference object of known size on the deck; Based on the calibrated pixel scale S and the total number of pixels N in the damaged area, the formula A=N is used. S² is used to calculate the actual damaged area A.
10. The method according to claim 6, characterized in that, Step S5 includes: The original image, damage identification results, boundary contour data, actual damage area calculation results, detection time and flight parameters are stored in the ship deck damage detection database. Retrieve historical detection data for the same damaged area, generate a curve showing the change in the damaged area, and trigger an alert when the growth rate of the damaged area exceeds a preset threshold. Generate a standardized inspection report that includes a diagram of the damage distribution and area statistics.