Hull metal surface quality detection system and method based on 3D vision
By using a 3D vision-based inspection system that combines laser structured light and multispectral imaging technology, the problems of low accuracy and low automation in ship hull metal surface inspection have been solved. This system enables high-precision, non-contact surface quality inspection and multi-dimensional evaluation, adapts to complex curved surfaces, and improves inspection efficiency and accuracy.
Patent Information
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-10-23
- Publication Date
- 2026-03-10
AI Technical Summary
Existing technologies for inspecting the surface quality of ship hull metals suffer from problems such as low detection accuracy, easy surface damage, low degree of automation, and difficulty in achieving large-area monitoring, especially after shot blasting/sandblasting, where high-precision non-contact inspection cannot be achieved.
A 3D vision-based detection system is adopted, which combines a laser structured light 3D camera and a multispectral imaging module. The system comprehensively scans the metal surface of the ship hull through multi-line laser structured light projection, and collects depth contour images and spectral data. Combined with image processing techniques such as Gaussian filtering and region growing algorithms, the system can accurately detect surface roughness and micro-pits, and establish a quantitative relationship model between the aspect ratio of micro-pits and adhesion.
It achieves high-precision, non-contact inspection of ship hull metal surface quality, improves inspection accuracy and automation, adapts to complex curved surfaces, provides multi-dimensional quality assessment, and enhances inspection efficiency and accuracy.
Smart Images
Figure CN121632024A_ABST
Abstract
Description
TECHNICAL FIELD
[0001] The present application relates to the technical field of ship manufacturing detection, and in particular to a ship body metal surface quality detection system and method based on 3D vision. BACKGROUND
[0002] Currently, the ship structure health is monitored by artificial periodic inspection, and the detection method is mainly point measurement, which is easy to cause missed detection and low reliability. If the arrangement is encrypted, the workload will be doubled, and the measurement cycle is long. It is also difficult to realize remote and large-area monitoring. Currently, the roughness measurement methods are mainly divided into two types: contact type and non-contact type. The most commonly used contact measurement method is to use a stylus instrument to measure, that is, the part surface is taken as the reference, the stylus always vertically contacts the measured surface to make linear scanning motion, and the longitudinal motion of the stylus is monitored. This kind of measurement method has high measurement accuracy and large measurement range, but it is easy to damage the surface of the measured part and the stylus, and the measurement efficiency is low. The non-contact measurement method mainly includes optical measurement and machine vision measurement. The surface roughness is measured by using a spectral confocal sensor as the core; the surface roughness is measured by using a light scattering method. The optical method does not contact the measured surface during the measurement process, which can reduce the wear and scratch of the part. However, manual adjustment of the instrument and reading is required, the human factor has a great influence, the automation degree is not high, and the light source and measurement environment have strict requirements, which limits its application. SUMMARY
[0003] The purpose of the present application is to overcome the above technical deficiencies, and to provide a ship body metal surface quality detection system and method based on 3D vision, which can solve the problem of high-precision and non-contact detection of surface roughness and micro-pit morphology after shot blasting / sand blasting of the ship body metal surface.
[0004] To achieve the above technical purpose, the following technical scheme is adopted in the present application: In a first aspect, a ship body metal surface quality detection system based on 3D vision is provided, comprising: A central control console comprising an industrial computer, a display and a data storage unit; the industrial computer is used for processing the collected data, the display is used for displaying visual images, spectral data and detection results, and the data storage unit is used for storing data; A laser structured light 3D vision detection module comprising a laser structured light 3D camera, which is used for full-scan of the ship body metal surface by using a multi-line laser structured light projection method to collect the depth profile image of the ship body metal surface; A multispectral imaging module is used to synchronously collect the reflected spectrum of the ship body metal surface to obtain spectral data; The data processing module is used to fit the depth contour image and the spectral data to form a depth image, process the depth image to obtain the surface roughness of the hull metal, divide the depth image into micro-pit regions to obtain segmented micro-pits, extract the morphological feature parameters of the segmented micro-pits, establish a quantitative relationship model between the aspect ratio of the micro-pits and the adhesion to obtain an adhesion score, and integrate the surface roughness of the hull metal, the segmented micro-pits and their morphological feature parameters, and the adhesion score to form a hull metal surface quality report.
[0005] Furthermore, the laser structured light 3D vision inspection module also includes a mobile device; the laser structured light 3D camera is fixed on the mobile device, and the laser structured light 3D camera is used to perform translational motion relative to the ship's metal surface through the mobile device using a multi-line laser structured light projection method, to collect depth contour image data of the ship's metal surface and transmit it to the central control console. The mobile device is used to control the laser structured light 3D camera to move at a constant speed along a preset path to fully scan the ship's metal surface.
[0006] Furthermore, the data processing module includes: a surface roughness calculation unit, a micro-pit segmentation unit, a micro-pit adhesion prediction unit, and a quality report production unit; The surface roughness calculation unit is used to fit the depth contour image and the spectral data according to time alignment to form a depth image, perform grayscale processing on the depth image to eliminate color information interference and highlight the depth information, use histogram equalization to enhance image contrast, use the neighborhood interpolation method to estimate and fill in the missing depth values using the surrounding effective pixel values, and combine the overall geometric model of the hull surface for fitting and filling, and perform roughness calculation on the processed depth image data to obtain the roughness of the hull metal surface. The micro-pit segmentation unit is used to smooth the depth image using Gaussian filtering, perform edge detection on the smoothed depth image to highlight the boundary information between the micro-pits and the surrounding surface, determine the width and position of the micro-pit boundary, determine the initial segmentation threshold based on the grayscale difference between the micro-pit depth and the surrounding surface to divide the micro-pit region, use the boundary pixels of the micro-pit region as seed points, and use a region growing algorithm to gradually expand the range of the micro-pit region to obtain the segmented micro-pits, extract the morphological feature parameters of the segmented micro-pits, and count the number and distribution density of micro-pits based on the morphological feature parameters. The micro-pit adhesion prediction unit is used to establish a quantitative relationship model between the aspect ratio of the micro-pit and the adhesion, and outputs an adhesion score based on the aspect ratio and steepness according to the quantitative relationship model. The quality report production unit is used to integrate the surface roughness of the hull metal, the segmented micro-pits and their morphological characteristics, and the adhesion score to form a hull metal surface quality report.
[0007] Furthermore, the multispectral imaging module includes: Multiple filters are coaxially mounted with the laser structured light 3D camera. Each filter collects spectral information from the metal surface of the ship's hull in a preset wavelength band to obtain spectral data.
[0008] Furthermore, the multispectral imaging module also includes: A motor drive unit is used to drive the filter to switch.
[0009] Secondly, this application provides a method for inspecting the surface quality of ship hull metal based on 3D vision, including the following steps: Multi-line laser structured light projection was used to comprehensively scan and acquire depth contour images of the ship's metal surface. When acquiring depth contour images of the ship's metal surface, spectral data is obtained by simultaneously acquiring the reflectance spectrum of the ship's metal surface. The depth profile image and the spectral data are fitted to form a depth image, and the depth image is processed to obtain the surface roughness of the ship hull metal. The depth image is divided into micro-pit regions to obtain segmented micro-pits, and the morphological feature parameters of the segmented micro-pits are extracted. A quantitative relationship model between the depth-to-width ratio of micropits and adhesion is established to obtain an adhesion score; The hull metal surface roughness, the segmented micro-pits and their morphological characteristics, and the adhesion score are integrated to form a hull metal surface quality report.
[0010] Furthermore, before employing multi-line laser structured light projection to comprehensively scan and acquire depth contour images of the ship's metal surface, the method further includes: The laser structured light 3D camera that acquires depth contour images of the ship's metal surface is calibrated using an experimental lookup table method, and the detection accuracy of the laser structured light 3D camera is improved by using stepped test blocks and B-spline interpolation algorithms.
[0011] Furthermore, the step of fitting the depth contour image and the spectral data to form a depth image, and processing the depth image to obtain the surface roughness of the ship hull metal, includes: The depth contour image and the spectral data are fitted together in time to form a depth image. The depth image is then grayscaled to eliminate color information interference and highlight depth information. Histogram equalization is used to enhance image contrast. Based on the neighborhood interpolation method, the missing depth values are estimated and filled in using the surrounding effective pixel values. The missing depth values are then fitted and filled in using the overall geometric model of the ship's surface. Finally, the roughness of the ship's metal surface is calculated by performing roughness calculation on the processed depth image data. Furthermore, the step of dividing the depth image into micro-pit regions to obtain segmented micro-pits, and extracting the morphological feature parameters of the segmented micro-pits, includes: Gaussian filtering is used to smooth the depth image. Edge detection is then performed on the smoothed depth image to highlight the boundary information between the micro-pits and the surrounding surface, determining the width and position of the micro-pit boundary. An initial segmentation threshold is determined based on the grayscale difference between the micro-pit depth and the surrounding surface to divide the micro-pit region. Using the boundary pixels of the micro-pit region as seed points, a region growing algorithm is used to gradually expand the range of the micro-pit region to obtain the segmented micro-pits. The morphological feature parameters of the segmented micro-pits are extracted, and the number and distribution density of micro-pits are counted based on the morphological feature parameters.
[0012] Furthermore, the process of integrating the surface roughness of the hull metal, the segmented micro-pits and their morphological characteristics, and the adhesion score to form a hull metal surface quality report includes: Different quality levels of the ship's metal surface are identified by various colors, and inspection reports are generated for each quality level area. The inspection reports include statistical charts of surface roughness, micro-pit distribution, and adhesion scores.
[0013] Compared with the prior art, the beneficial technical effects of the technical solution provided in this application include: This application combines high-resolution 3D vision technology with advanced image processing algorithms to precisely capture minute imperfections and details on the hull surface, accurately detecting minute changes in surface roughness, precise dimensions and shapes of micro-pits, etc., greatly improving inspection accuracy and the stringency of quality control. Simultaneously, the laser structured light 3D vision inspection module can flexibly adapt to complex curved surfaces and irregular shapes on the hull, ensuring the completeness and accuracy of inspection and overcoming the limitations of traditional methods in complex surface inspection. Furthermore, the multispectral imaging module simultaneously acquires the surface material's reflectance spectrum, providing data support for quality assessment and further improving inspection efficiency. Combining surface roughness measurement, micro-pit segmentation statistics, and micro-pit adhesion prediction models, a multi-dimensional comprehensive assessment of the hull surface quality is conducted, providing quantitative indicators of surface roughness, accurately counting the number, size, and distribution of micro-pits, and predicting their impact on adhesion, providing a comprehensive and accurate basis for subsequent anti-corrosion and coating processes. Attached Figure Description
[0014] To more clearly illustrate the technical solutions in this application, the accompanying drawings used in the embodiments will be briefly described below: Figure 1 This is a flowchart illustrating the 3D vision-based method for inspecting the surface quality of a ship's hull structure in an embodiment of this application. Figure 2 These are images of the height of the ship's metal surface after shot blasting at different scales in the embodiments of this application; Figure 3 This is the image after missing point repair and noise reduction preprocessing in the embodiments of this application; Figure 4 This is an image after segmenting the micro-pits in an embodiment of this application; Figure 5 This is a smoothed image of the micro-pits in the embodiments of this application. Detailed Implementation
[0015] To make the objectives, technical solutions, and advantages of this application clearer, the following detailed description is provided in conjunction with the accompanying drawings and embodiments. It should be understood that the specific embodiments described herein are merely illustrative and not intended to limit the scope of this application.
[0016] Those skilled in the art will understand that in this specification, the word "comprising" is an open-ended expression, meaning that the described feature exists but other features are not excluded. The singular form can also be used in the plural form. "Multiple" means two or more. For example... Figures 1-5 As shown, the ship hull metal surface quality inspection system of this embodiment includes: a central control console, a laser structured light 3D vision inspection module, a multispectral imaging module, and a data processing module.
[0017] The central control console includes an industrial control computer, a display, and a data storage unit; the industrial control computer is used to process the acquired data, the display is used to display visual images, spectral data, and detection results, and the data storage unit is used to store the data.
[0018] The laser structured light 3D vision inspection module includes a laser structured light 3D camera, which is used to comprehensively scan and acquire depth contour images of the ship's metal surface using a multi-line laser structured light projection method; and a multispectral imaging module, which is used to synchronously acquire the reflection spectrum of the ship's metal surface with the laser structured light 3D camera to obtain spectral data.
[0019] The data processing module is used to fit the depth contour image and the spectral data to form a depth image, process the depth image to obtain the surface roughness of the hull metal, divide the depth image into micro-pit regions to obtain segmented micro-pits, extract the morphological feature parameters of the segmented micro-pits, establish a quantitative relationship model between the aspect ratio of the micro-pits and the adhesion to obtain an adhesion score, and integrate the surface roughness of the hull metal, the segmented micro-pits and their morphological feature parameters, and the adhesion score to form a hull metal surface quality report.
[0020] in, Figure 2 These are images of the height of the ship's metal surface after shot blasting at different scales; Figure 3 It is the image after missing point repair and noise reduction preprocessing; Figure 4 This is the image after segmenting the micro-pits;Figure 5 It is a smoothed image of micro-pits.
[0021] The industrial control computer, equipped with data processing and storage capabilities, can simultaneously process data from multiple sensors, supporting multi-task parallel computing to ensure efficient and stable system operation. The display shows visual images, spectral data, and test results, allowing operators to intuitively view the quality of the ship's metal surface. The data storage unit can store various types of data, including raw image data, preprocessed data, test results, and quality reports, meeting subsequent data query and analysis needs.
[0022] Among them, the laser structured light 3D camera uses multi-line laser structured light projection technology. Through a high-precision optical lens and imaging sensor, it can generate multiple parallel and uniformly spaced laser lines, which are projected onto the metal surface of the ship to form a clear light stripe pattern. It has high-resolution imaging capabilities, which can accurately capture subtle depth changes on the surface. It is also equipped with automatic exposure and dynamic focusing functions to adapt to different ambient lighting and changes in the curvature of the ship's surface.
[0023] The laser structured light 3D vision inspection module further includes a mobile device; the laser structured light 3D camera is fixed on the mobile device, and the laser structured light 3D camera is used to perform translational motion relative to the ship's metal surface through the mobile device using a multi-line laser structured light projection method, to collect depth contour image data of the ship's metal surface and transmit it to the central control console. The mobile device is used to control the laser structured light 3D camera to move at a constant speed along a preset path to fully scan the ship's metal surface.
[0024] The mobile device has a smooth and precise linear motion capability. Under computer control, it can move the 3D camera at a constant speed along a preset path to achieve a comprehensive scan of the ship's metal surface, ensuring that the collected depth contour data is complete and continuous.
[0025] In some embodiments of this application, the multispectral imaging module includes: Multiple filters are coaxially mounted with the laser structured light 3D camera. Each filter collects spectral information from the metal surface of the ship's hull in a preset wavelength band to obtain spectral data.
[0026] The filters are precisely aligned with the detection area on the ship's metal surface, ensuring spatial consistency between spectral information and depth data. The multispectral imaging module is equipped with multiple filters covering several key bands from visible light to near-infrared, enabling precise acquisition of the reflectance spectrum of the ship's metal surface at different bands.
[0027] The multispectral imaging module also includes a motor drive unit for driving the filter switching. The switching of each filter is rapid and stable, and automatic switching is achieved through the motor drive unit, ensuring the continuity of spectral data acquisition.
[0028] In some embodiments of this application, the data processing module includes: a surface roughness calculation unit, a micro-pit segmentation unit, a micro-pit adhesion prediction unit, and a quality report production unit.
[0029] The surface roughness calculation unit is used to fit the depth contour image and the spectral data according to time alignment to form a depth image, perform grayscale processing on the depth image to eliminate color information interference and highlight the depth information, use histogram equalization to enhance image contrast, estimate and complete the missing depth value based on the neighborhood interpolation method using the surrounding effective pixel values, and combine it with the overall geometric model of the hull surface for fitting and completion, and perform roughness calculation on the processed depth image data to obtain the roughness of the hull metal surface.
[0030] The micro-pit segmentation unit is used to smooth the depth image using Gaussian filtering, perform edge detection on the smoothed depth image to highlight the boundary information between the micro-pits and the surrounding surface, determine the width and position of the micro-pit boundary, determine the initial segmentation threshold based on the grayscale difference between the micro-pit depth and the surrounding surface to divide the micro-pit region, use the boundary pixels of the micro-pit region as seed points, and use a region growing algorithm to gradually expand the range of the micro-pit region to obtain the segmented micro-pits, extract the morphological feature parameters of the segmented micro-pits, and count the number and distribution density of micro-pits based on the morphological feature parameters.
[0031] The micro-pit adhesion prediction unit is used to establish a quantitative relationship model between the aspect ratio of the micro-pit and the adhesion, and outputs an adhesion score based on the aspect ratio and steepness according to the quantitative relationship model.
[0032] The quality report production unit is used to integrate the surface roughness of the hull metal, the segmented micro-pits and their morphological characteristics, and the adhesion score to form a hull metal surface quality report.
[0033] The surface roughness calculation unit performs grayscale processing on the acquired depth image to eliminate color interference and highlight depth information. Histogram equalization is used to enhance image contrast, making minute height changes on the surface more clearly visible. Based on neighborhood interpolation, missing depth values are estimated and filled in using surrounding effective pixel values to ensure the integrity of the depth image. For small missing areas, the depth value of the missing point is calculated based on the grayscale change trend of surrounding pixels; for larger missing areas, the overall geometric model of the hull surface is used for fitting and filling. The size of the filtering window is dynamically adjusted according to local image features to effectively remove noise. Simultaneously, the image is denoised under multi-scale decomposition to preserve edge details in the depth image and avoid excessive smoothing that could lead to the loss of true surface features. Based on this, roughness calculation according to the ISO 8503 standard is performed. Following the roughness evaluation parameters specified in the ISO 8503 standard, such as the arithmetic mean deviation of the profile (Ra) and the maximum profile height (Rz), numerical analysis methods are used to accurately calculate the surface roughness of the hull metal based on the processed depth image data, providing a quantitative basis for surface quality assessment.
[0034] The micro-pit segmentation unit uses Gaussian filtering to smooth the depth image, reducing the impact of image noise on micro-pit segmentation while preserving the main features of the micro-pit edges. During the filtering process, the Gaussian kernel size and standard deviation are appropriately selected to balance the smoothing effect with detail preservation. Then, edge detection is performed on the smoothed image to highlight the boundary information between the micro-pit and the surrounding surface, enhancing the clarity of the micro-pit contours. Then, non-maximum suppression and double threshold detection are used to further refine the width and position of the micro-pit boundaries, ensuring the accuracy of subsequent segmentation. Then, the initial segmentation threshold is determined based on the gray-level difference between the micro-pit depth and the surrounding surface to initially divide the micro-pit region. Then, using the micro-pit boundary pixels as seed points, a region growing algorithm is used to gradually expand the micro-pit region range to achieve accurate micro-pit segmentation. After segmentation, the morphological features of the micro-pits are analyzed in detail, including parameters such as area, perimeter, shape factor, and aspect ratio, and information such as the number and distribution density of micro-pits is statistically analyzed.
[0035] The micro-pit adhesion prediction unit establishes a quantitative relationship model between the depth-to-width ratio of micro-pits and adhesion based on the morphological feature parameters extracted by the micro-pit segmentation module.
[0036] like Figure 1 As shown, this application proposes a 3D vision-based method for inspecting the surface quality of ship hull structures, including: Step S1. Use multi-line laser structured light projection to fully scan and acquire depth contour images of the ship's metal surface; Step S2. While acquiring the depth contour image of the ship's metal surface, simultaneously acquire the reflectance spectrum of the ship's metal surface to obtain spectral data; Step S3. Fit the depth contour image and the spectral data to form a depth image, and process the depth image to obtain the surface roughness of the ship hull metal; Step S4. Divide the depth image into micro-pit regions to obtain segmented micro-pits, and extract the morphological feature parameters of the segmented micro-pits; Step S5. Establish a quantitative relationship model between the aspect ratio of the micropit and the adhesion to obtain an adhesion score; Step S6. Integrate the surface roughness of the hull metal, the segmented micro-pits and their morphological characteristics, and the adhesion score to form a hull metal surface quality report.
[0037] Furthermore, before employing multi-line laser structured light projection to comprehensively scan and acquire depth contour images of the ship's metal surface, the method further includes: The laser structured light 3D camera that acquires depth contour images of the ship's metal surface is calibrated using an experimental lookup table method, and the detection accuracy of the laser structured light 3D camera is improved by using stepped test blocks and B-spline interpolation algorithms.
[0038] Among them, the laser structured light 3D vision camera was calibrated using an experimental lookup table method. The measurement accuracy was optimized and adjusted, and the calibration process was iterated repeatedly until the system detection accuracy met the basic requirements, ensuring that it could accurately measure the minute depth changes and complex contour features of the ship's metal surface.
[0039] The central control console, laser structured light 3D vision inspection module, and multispectral imaging module are powered on. Each device undergoes self-testing and initialization, including 3D camera parameter configuration (such as laser power, imaging resolution, frame rate, etc.), zeroing and calibrating motion parameters of the moving device, and initializing and correcting the filters and spectrometers of the multispectral imaging module to ensure optimal operating conditions. Based on the size, shape, and accuracy requirements of the metal surface of the ship hull to be inspected, the motion path and speed parameters of the moving device are preset via the central control console. The acquisition positions and angles of the 3D camera and multispectral imaging module are planned to ensure comprehensive coverage of the inspection area while avoiding data redundancy and duplicate acquisitions.
[0040] Furthermore, the laser structured light 3D camera projects multi-line laser structured light onto the ship's metal surface according to preset parameters. Driven by a mobile device, it moves at a constant speed along a preset path, acquiring surface depth images point by point. The central control console receives the depth data transmitted by the 3D camera in real time and performs data buffering and preliminary processing to ensure data integrity and continuity. The multispectral imaging module is simultaneously activated, acquiring spectral information from the ship's metal surface at each location according to a preset spectral band switching sequence. The data acquisition timing of the multispectral imaging module and the laser structured light 3D camera is precisely controlled to ensure temporal consistency between depth and spectral data.
[0041] After receiving the transmitted depth image data, the central control console executes the surface roughness calculation unit. First, it performs grayscale conversion, transforming the color depth image into a grayscale image to eliminate interference from color information on depth analysis. This also reduces data dimensionality and improves subsequent processing efficiency. Histogram equalization technology is then used to enhance the contrast of the grayscale image. Based on the distribution of the image's grayscale histogram, the distribution range of gray levels is automatically adjusted to make the pixel grayscale values in the image more uniform, highlighting the subtle depth variations on the ship's metal surface, enhancing the image's visual effect, and creating favorable conditions for subsequent missing value recovery and noise reduction processing.
[0042] Preferably, missing depth values in the depth image caused by occlusion, reflection interference, or other reasons are filled in. For small, localized missing areas, the depth value of the missing point is accurately estimated by comprehensively considering the gray values and spatial relationships of multiple surrounding pixels according to a weighted allocation principle. For larger, continuous missing areas, reasonable depth data is generated by combining the overall geometric model of the ship's metal surface (such as a planar model or a curved surface fitting model) with fitting analysis of the surrounding valid data, thereby filling in the missing areas, ensuring the integrity of the depth image, and providing a complete data foundation for subsequent surface quality analysis.
[0043] Preferably, the completed depth image undergoes denoising processing. First, an adaptive median filtering method is used to dynamically adjust the size and shape of the filtering window based on the gray-level distribution characteristics of local image regions. This effectively removes noise while avoiding excessive blurring of image edges and details. Then, wavelet transform is used to decompose the image into spaces of different scales. Thresholding and reconstruction are performed on high-frequency and low-frequency coefficients at different scales to further remove random noise interference such as Gaussian noise from the image, while preserving important feature information such as edges and textures of the ship's metal surface. This improves the quality and clarity of the depth image, providing more accurate data support for subsequent operations such as surface roughness measurement and micro-pit segmentation.
[0044] Micro-pit segmentation and statistical analysis: First, the micro-pit segmentation unit processes the image. Then, image smoothing is performed on the denoised depth image. A Gaussian filter is used to convolve the image, filtering out high-frequency noise components while preserving the main edge and contour information. By appropriately selecting the kernel size and standard deviation parameters of the Gaussian filter, the smoothed image still clearly displays the boundary features between the micro-pits and the surrounding metal surface, laying the foundation for subsequent boundary enhancement processing.
[0045] Preferably, edge detection is performed on the smoothed image. By calculating the gray-level gradient of each pixel in the image in the horizontal and vertical directions, the boundary information between the micro-pits and the surrounding surface is highlighted, making the contours of the micro-pits more clearly visible. During the edge detection process, non-maximum suppression and dual-threshold detection methods are used to further refine the width and position of the micro-pit boundaries, remove false edges and discontinuous boundary fragments, and obtain complete and continuous micro-pit boundary information, providing accurate boundary guidance for subsequent micro-pit segmentation operations.
[0046] Preferably, the principle of consistency of elements within the structure is used to blur the image, eliminating small holes and cracks inside the micro-pits caused by factors such as noise, uneven lighting, or surface oxidation, so that the micro-pit area presents a complete and connected state. In the process, the shape and size of the structural elements are reasonably selected to effectively fill the holes inside the micro-pits without excessively changing the original morphological characteristics and boundary positions of the micro-pits, ensuring the authenticity and identifiability of the micro-pits.
[0047] Preferably, based on the difference in grayscale between the depth of the micro-pit and the surrounding metal surface, histogram analysis and the Otsu algorithm are used to determine the initial segmentation threshold, initially dividing the image into micro-pit regions and non-micro-pit regions. Then, multiple seed points are selected near the boundary pixels of the micro-pits. Guided by the micro-pit boundary information and based on region growth conditions (such as the similarity between the pixel grayscale value and the seed point grayscale value, and the spatial proximity between the pixel and the seed point), the micro-pit region is gradually expanded to the surrounding area, achieving accurate micro-pit segmentation. During the region growth process, by setting reasonable growth thresholds and stopping conditions, over-growth or under-growth problems are avoided, ensuring the accuracy and completeness of micro-pit segmentation.
[0048] Furthermore, morphological and statistical analyses are performed on the segmented micro-pit regions. Morphological analysis includes calculating geometric parameters such as the area, perimeter, shape factor, aspect ratio, and roundness of the micro-pits. These parameters can quantitatively describe the size, shape, and complexity of the micro-pits. Statistical analysis includes statistically analyzing the number of micro-pits, their distribution density, and the proportion of micro-pits in different size ranges. These statistical indicators provide a comprehensive understanding of the overall distribution and quality of micro-pits on the ship's metal surface.
[0049] The surface roughness calculation method is based on the surface roughness calculation method of ISO8503 standard, which is performed on the surface roughness calculation unit. First, the coordinate system is corrected and the unit is converted to ensure the consistency between the depth data and the actual physical size. Then, the corresponding surface profile samples are extracted from the depth image. Next, multiple roughness parameters such as the arithmetic mean deviation of the profile (Ra), the maximum height of the profile (Rz), and the root mean square deviation of the profile (Rq) are calculated using numerical analysis methods. These parameters can quantitatively describe the roughness of the ship hull metal surface from different perspectives, providing a quantitative basis for surface quality assessment.
[0050] Data Integration and Report Generation: The processed ship hull metal surface quality data is integrated, including surface roughness parameters, micro-pit distribution and morphological characteristics, and spectral information analysis results. Following a pre-defined data structure and format, this data is categorized, organized, and stored in a linked manner to construct a complete ship hull metal surface quality database, providing data support for subsequent data queries, analysis, and report generation. Based on the integrated quality data, a detailed inspection report is generated, visually displaying the ship hull metal surface quality status.
[0051] Preferably, different colors are used to identify areas of different quality levels on the ship's metal surface, such as green for good quality areas, yellow for quality defects areas, and red for low quality areas, making the test results more intuitive and easy to understand, and facilitating managers and maintenance personnel to quickly understand the overall quality status of the ship's metal surface.
[0052] Furthermore, the step of fitting the depth contour image and the spectral data to form a depth image, and processing the depth image to obtain the surface roughness of the ship hull metal, includes: The depth contour image and the spectral data are fitted together in time to form a depth image. The depth image is then grayscaled to eliminate color information interference and highlight depth information. Histogram equalization is used to enhance image contrast. Based on the neighborhood interpolation method, the missing depth values are estimated and filled in using the surrounding effective pixel values. The missing depth values are then fitted and filled in using the overall geometric model of the ship's surface. Finally, the roughness of the ship's metal surface is calculated by performing roughness calculation on the processed depth image data. Furthermore, the step of dividing the depth image into micro-pit regions to obtain segmented micro-pits, and extracting the morphological feature parameters of the segmented micro-pits, includes: Gaussian filtering is used to smooth the depth image. Edge detection is then performed on the smoothed depth image to highlight the boundary information between the micro-pits and the surrounding surface, determining the width and position of the micro-pit boundary. An initial segmentation threshold is determined based on the grayscale difference between the micro-pit depth and the surrounding surface to divide the micro-pit region. Using the boundary pixels of the micro-pit region as seed points, a region growing algorithm is used to gradually expand the range of the micro-pit region to obtain the segmented micro-pits. The morphological feature parameters of the segmented micro-pits are extracted, and the number and distribution density of micro-pits are counted based on the morphological feature parameters.
[0053] Furthermore, the process of integrating the surface roughness of the hull metal, the segmented micro-pits and their morphological characteristics, and the adhesion score to form a hull metal surface quality report includes: Different quality levels of the ship's metal surface are identified by various colors, and inspection reports are generated for each quality level area. The inspection reports include statistical charts of surface roughness, micro-pit distribution, and adhesion scores.
[0054] Those skilled in the art will understand that the steps, measures, and schemes in the various operations, methods, processes, and procedures discussed in this application can be alternated, modified, rearranged, decomposed, combined, or deleted. The specific embodiments of this application described above do not constitute a limitation on the scope of protection of this application. Any other corresponding changes and modifications made based on the technical concept of this application should be included within the scope of protection of the claims of this application.
Claims
1. A 3D vision-based hull metal surface quality detection system, characterized in that, The application relates to a ship hull metal surface quality detection system. The system comprises a central console, a laser structured light 3D vision detection module, a multispectral imaging module and a data processing module. The central console comprises an industrial computer, a display and a data storage unit; the industrial computer is used for processing collected data; the display is used for displaying visual images, spectral data and detection results; and the data storage unit is used for storing data. The laser structured light 3D vision detection module comprises a laser structured light 3D camera and is used for comprehensively scanning and collecting a ship hull metal surface depth profile image by adopting a multi-line laser structured light projection mode. The multispectral imaging module is used for synchronously collecting a ship hull metal surface reflection spectrum with the laser structured light 3D camera to obtain spectral data.
2. The system of claim 1, wherein, The data processing module is used for fitting the depth profile image and the spectral data to form a depth image, processing the depth image to obtain a ship hull metal surface roughness, dividing a micro-pit region of the depth image to obtain a segmented micro-pit, extracting morphological feature parameters of the segmented micro-pit, establishing a quantitative relationship model between a micro-pit depth-width ratio and an adhesive force to obtain an adhesive force score, and integrating the ship hull metal surface roughness, the segmented micro-pit and the morphological feature parameters and the adhesive force score to form a ship hull metal surface quality report.
3. The system of claim 1, wherein, The laser structured light 3D vision detection module further comprises a moving device; the laser structured light 3D camera is fixed on the moving device; the laser structured light 3D camera is used for performing a translational motion relative to a ship hull metal surface by the moving device through the multi-line laser structured light projection mode, collecting ship hull metal surface depth profile image data and transmitting the ship hull metal surface depth profile image data to the central console; and the moving device is used for controlling the laser structured light 3D camera to move at a constant speed according to a preset path to comprehensively scan a ship hull metal surface. The multispectral imaging module comprises a plurality of optical filters which are coaxially installed with the laser structured light 3D camera; each optical filter is used for collecting spectral information of a ship hull metal surface under a preset wave band to obtain spectral data.
4. The system of claim 3, wherein, The multispectral imaging module further comprises a motor driving device which is used for driving the optical filters to switch. The data processing module comprises a surface roughness calculation unit, a micro-pit segmentation unit, a micro-pit adhesive force prediction unit and a quality report production unit.
5. The system of claim 1, wherein, The surface roughness calculation unit is used for aligning and fitting the depth profile image and the spectral data according to time to form a depth image, performing a grayscale processing on the depth image to eliminate color information interference and highlight depth information, enhancing image contrast by adopting a histogram equalization mode, estimating and completing missing depth values by using surrounding effective pixel values based on a neighborhood interpolation method, fitting and completing the missing depth values in combination with a ship surface overall geometric model, and calculating a ship hull metal surface roughness by processing the depth image data. The micro-pit segmentation unit is configured to smooth the depth image by using a Gaussian filter, perform edge detection on the smoothed depth image, highlight boundary information of the micro-pit and the surrounding surface, determine the width and position of the micro-pit boundary, determine an initial segmentation threshold according to the gray difference between the micro-pit depth and the surrounding surface to divide the micro-pit region, take the boundary pixels of the micro-pit region as seed points, gradually expand the range of the micro-pit region by using a region growing algorithm, obtain the segmented micro-pit, extract morphological feature parameters of the segmented micro-pit, and count the number and distribution density of the micro-pit according to the morphological feature parameters; The micro-pit adhesion force prediction unit is configured to establish a quantitative relationship model between the micro-pit depth-width ratio and the adhesion force, and output the adhesion force score based on the depth-width ratio and the steepness according to the quantitative relationship model; The quality report production unit is configured to integrate the hull metal surface roughness, the segmented micro-pit and the morphological feature parameters thereof, and the adhesion force score to form a hull metal surface quality report.
6. A 3D vision-based hull metal surface quality detection method, characterized in that, The method comprises the following steps: comprehensively scanning the hull metal surface by using a multi-line laser structured light projection mode to collect a hull metal surface depth profile image; synchronously collecting a hull metal surface reflectance spectrum to obtain spectrum data when the hull metal surface depth profile image is collected; fitting the depth profile image and the spectrum data to form a depth image, and processing the depth image to obtain the hull metal surface roughness; dividing the depth image to obtain the segmented micro-pit, and extracting morphological feature parameters of the segmented micro-pit; establishing a quantitative relationship model between the micro-pit depth-width ratio and the adhesion force to obtain the adhesion force score; integrating the hull metal surface roughness, the segmented micro-pit and the morphological feature parameters thereof, and the adhesion force score to form a hull metal surface quality report.
7. The method of claim 6, wherein, Before the step of comprehensively scanning the hull metal surface by using a multi-line laser structured light projection mode to collect a hull metal surface depth profile image, the method further comprises: calibrating the laser structured light 3D camera used to collect the hull metal surface depth profile image by using an experimental table lookup method, and improving the detection accuracy of the laser structured light 3D camera by using a step block and a B-spline interpolation algorithm.
8. The method of claim 6, wherein, The step of fitting the depth profile image and the spectrum data to form a depth image, and processing the depth image to obtain the hull metal surface roughness comprises: aligning the depth profile image and the spectrum data according to time to form a depth image, performing gray scale processing on the depth image to eliminate color information interference and highlight depth information, enhancing the contrast of the image by using a histogram equalization method, estimating and completing the missing depth values by using the surrounding effective pixel values based on a neighborhood interpolation method, fitting and completing the missing depth values in combination with a hull surface overall geometric model, and performing roughness calculation on the processed depth image data to obtain the hull metal surface roughness.
9. The method of claim 8, wherein, The step of dividing the depth image to obtain the segmented micro-pit, and extracting morphological feature parameters of the segmented micro-pit comprises: The depth image is smoothed by using a Gaussian filter, the smoothed depth image is edge detected to highlight the boundary information between the micro-pit and the surrounding surface, the width and position of the micro-pit boundary are determined, the initial segmentation threshold is determined according to the gray difference between the micro-pit depth and the surrounding surface to divide the micro-pit region, the boundary pixels of the micro-pit region are taken as seed points, the region growing algorithm is used to gradually expand the range of the micro-pit region, the segmented micro-pit is obtained, the morphological characteristic parameters of the segmented micro-pit are extracted, and the number and distribution density of the micro-pit are counted according to the morphological characteristic parameters.
10. The method of claim 6, wherein, The ship hull metal surface roughness, the segmented micro-pit and the morphological characteristic parameters thereof, and the adhesion force score are integrated to form a ship hull metal surface quality report, which includes: Different quality levels of the ship hull metal surface are marked by various colors, and detection reports corresponding to the regions of each quality level are generated, the detection reports including statistical charts of the surface roughness, the micro-pit distribution and the adhesion force score.