Methods and systems for laser cleaning and intelligent defect identification of ship surfaces

Abstract

The invention discloses a method and system for laser cleaning and intelligent defect recognition of a ship surface, and relates to the technical field of image recognition. The method comprises the following steps: acquiring images of a ship surface by a camera carried by an unmanned aerial vehicle to obtain a color image and a spectral image of the ship surface; separating the concave defects on the ship surface in the color image of the ship surface by setting a first rust separation strategy to obtain a first rust label; setting a second rust separation strategy by combining a constant light compensation coefficient of the hull color to separate biological mucus in the first rust label to obtain a second rust label; calculating the defect rust corrosion value by the second rust label; and establishing a laser parameter-corrosion value association model, taking the defect rust corrosion value as an input parameter, and outputting the corresponding laser wavelength and power. The invention realizes the accurate recognition of ship surface defects and the intelligent adjustment of laser cleaning parameters, and solves the influence of light on the image recognition of the ship surface.

Description

Technical Field

[0001] The present invention relates to the field of image recognition technology, and specifically to a method and system for laser cleaning and intelligent defect recognition of ship surfaces. Background Art

[0002] With the continuous development of the global marine economy, the safety and efficiency of ships as the main means of marine transportation have received widespread attention. However, during the long-term operation of ships at sea, their surfaces will inevitably be corroded by the marine environment (such as seawater, salt, oil pollution, etc.), resulting in defects such as rust and coating aging on the hull surface. These defects not only affect the appearance of the ship, but more importantly, reduce the structural strength and service life of the ship, and increase the operating risk. However, since ships are easily attached to marine biological mucus in the water environment, including algae, seaweed and marine animal debris, it will have a certain impact on the identification of hull surface defects, increase the error rate of defect identification, and thus increase the cost of automated cleaning.

[0003] Traditional ship surface cleaning methods, such as sandblasting and chemical cleaning, can remove surface rust and coating to a certain extent, but they have disadvantages such as high labor intensity, serious environmental pollution, and easy damage to the ship itself. Therefore, finding an efficient, environmentally friendly, and non-damaging cleaning method for the ship surface has become an urgent problem to be solved in the field of ship maintenance and upkeep.

[0004] In recent years, laser cleaning technology, with its unique advantages, has gradually shown broad application prospects in the field of ship surface cleaning. Laser cleaning technology uses high-energy-density laser beams to irradiate the surface of the ship. Through the interaction between the laser and the material, the rust, coating and other pollutants on the surface are quickly gasified, melted or peeled off, thereby achieving the purpose of cleaning. This technology has the advantages of non-contact, environmental protection, non-destructive, easy automation and intelligence, and can meet the high requirements of ship surface cleaning.

[0005] However, the application of laser cleaning technology in the field of ship surface cleaning still faces some challenges. Among them, how to overcome the influence of light on the ship surface, accurately identify the defects on the ship surface, and adjust the laser cleaning parameters accordingly is the key to achieving efficient and accurate cleaning.

[0006] For example, the existing Chinese patent with publication number CN113920407A discloses a ship target recognition method and system based on multi-band remote sensing image fusion, including the following steps: obtaining a multi-band remote sensing image that has undergone multi-band registration, extracting remote sensing image features of multiple bands through a convolutional neural network, and based on a generative model, continuously back-propagating to continuously update the pixels of the fused image, and after achieving the best effect, outputting the fused image to the ship target detection step; receiving the fused image output by the multi-band image fusion step, completing the detection of the ship target in the fused image based on an end-to-end target detection network, and outputting a slice image of the ship target and a rough classification result of the ship category to the ship target recognition step; obtaining the slice image of the ship target output by the ship target detection step, and after passing through a refined discrimination network, outputting the final recognition result of the ship category. However, this design may be affected by multiple factors such as image quality, lighting conditions, and ship posture, resulting in insufficient classification accuracy. If the distribution of ship categories in the training data is unbalanced, the model may be biased towards categories with a larger number, resulting in poor recognition of categories with a smaller number. For this reason, the present invention provides a method and system for laser cleaning and intelligent defect recognition of ship surfaces. Summary of the invention

[0007] The purpose of the present invention is to provide a method and system for laser cleaning and intelligent defect recognition of ship surface to solve the existing problems raised in the above background technology.

[0008] To achieve the above object, the present invention provides the following technical solution: a method for laser cleaning and intelligent defect recognition of a ship surface, comprising the following steps:

[0009] S1. Capture images of the ship surface by using a camera mounted on a drone to obtain a color image and a spectral image of the ship surface;

[0010] S2, separating the ship surface concave defects in the ship surface color image by setting a first rust separation strategy to obtain a first rust label;

[0011] S3, setting a second rust separation strategy by combining the constant color of the hull and the light compensation coefficient, separating the biological mucus in the first rust label to obtain a second rust label;

[0012] S4, calculating the defect rust corrosion value through the second rust label;

[0013] S5. By establishing a laser parameter-corrosion value association model, the defect rust corrosion value is used as an input parameter, and the corresponding laser wavelength and power are output.

[0014] The present invention is further improved in that the first rust separation strategy specifically comprises the following steps:

[0015] S21, extracting defective parts in the ship surface color image by using a Sobel operator to obtain a first defect set;

[0016] S22, calculating the edge clarity of defects in the first defect set to obtain a first defect edge clarity data set ;

[0017] S23. Set edge defect clarity threshold , extract the first defect edge clarity data set The edge definition is greater than the edge defect definition threshold The corresponding defect image is extracted as the first rust label, the first defect edge clarity dataset The edge definition is less than or equal to the edge defect definition threshold The defect image corresponding to the data is taken as a concave defect.

[0018] A further improvement of the present invention is that the edge clarity is obtained by extracting the vector edges in the first defect set and using the gradient amplitude of the Sobel operator to replace the standard deviation and covariance of the pixels in the structural similarity index. The specific steps include:

[0019] S221, extracting pixels on both sides of the vector edge of any defect edge to obtain a pixel data set within the defect edge and edge pixel dataset , and record the corresponding position, where Indicates that the defect The pixels inside the edge of the vector edge, Indicates that the defect The pixels outside the edge of the vector edge; Indicates the number of pixels passed by any defect edge;

[0020] S222. Calculate the edge clarity of the defect ,in, Indicates that the defect The edge pixels of the vector edge and the defect The difference in the horizontal gradient of the pixels outside the edge of the vector edge, Indicates that the defect The edge pixels of the vector edge and the defect The difference in the vertical gradient of the pixels outside the edge of the vector edge; ;

[0021] S223, repeating step S221 and step S222 until the edge clarity of all defect data in the first defect set is calculated and included in the first defect edge clarity data set middle.

[0022] The present invention is further improved in that the second rust separation strategy specifically comprises the following steps:

[0023] S31, extracting the area corresponding to the first rust label in the ship surface color image, and including it in a first defect surface color image dataset; extracting the area corresponding to the first rust label in the ship surface spectral image, and including it in a first defect spectral image dataset;

[0024] S32, calculating the constant-light compensation coefficient of the hull color through the ship surface color image, and calculating the defect color representation value through the data in the first defect surface color image data set;

[0025] S33, calculating the defect spectrum feature influence value through the data in the first defect spectrum image data set;

[0026] S34, combining the hull color constant-light compensation coefficient, the defect color representation total value and the defect spectral characteristic influence value to obtain a defect rust approach value;

[0027] S35, setting a rust approach threshold, listing the data in the first rust tag whose defect rust approach value is greater than the rust approach threshold as a second rust tag, and listing the data in the first rust tag whose defect rust approach value is less than or equal to the rust approach threshold as biological slime.

[0028] A further improvement of the present invention is that the specific calculation steps of the hull color constant-light compensation coefficient include:

[0029] S321, looping through all pixel RGB values ​​of the ship surface color image to obtain all pixel brightness values , and arranged in order from low to high to obtain a sequence of pixel brightness values ​​of the ship surface color image; where R is the red channel color value, G is the green channel color value, and B is the blue channel color value;

[0030] S322, extracting the sum of the first 5% of the pixel brightness values ​​of the ship surface color image sequence as a reference compensation value , and extract the corresponding number of pixels ;

[0031] S323, calculate the constant-light compensation coefficient of the hull according to the reference compensation value .

[0032] The present invention is further improved in that the specific calculation step of the total value of defect color representation includes: extracting the gray value sequence of all pixels within the contour of any defect in the first defect surface color image data set ,in Indicates that the defect The gray value of a pixel, , m represents the number of pixels in the defect; extract the RGB component values ​​of all pixels in the defect, The red component of a pixel is represented by , No. The green component of a pixel is expressed as , No. The blue component of a pixel is expressed as ; Calculate the The color representation value of a pixel is expressed as ,in, represents the gray value weight, Indicates color component weight; sets color representation segmentation value , and , dividing the color representation value into four pixel color representation intervals according to the color representation segmentation value, including , , , , count the distribution rate of the color representation values ​​of the pixels in the defect area in the above color representation interval, and extract the average color representation value of the pixels in the color representation interval with the highest distribution rate as the total color representation value of the area .

[0033] A further improvement of the present invention is that the specific steps of the defect spectrum characteristic influence value include:

[0034] S331, extracting a near-infrared band reflectivity dataset of any defect in the first defect spectral image dataset , red light band reflectance dataset , UV band reflectance dataset and green band reflectance dataset , where a represents the number of near-infrared bands, b represents the number of red light bands, c represents the number of ultraviolet bands, and d represents the number of green light bands; Indicates The spectral reflectance of the near-infrared band is Indicates The spectral reflectance of the red light band is Indicates The spectral reflectance of the ultraviolet band is Indicates The spectral reflectance of the green light band is , , , ;

[0035] S332, set the defect sensitive band combination value ,in It represents the red light band sensitivity value of the defect, and the calculation formula is: ;

[0036] S333, calculate the grayscale mean value within the contour of the defect , set the rust sensitive band combination extraction formula , where the function Indicates the calculation of correlation coefficient, function Indicates maximum value extraction, function Representation Extraction The defect sensitive band combination value of the corresponding near-infrared band and ultraviolet band is taken as the defect spectral characteristic influence value of the defect .

[0037] The present invention is further improved in that the specific calculation process of the defect rust approximation value includes: extracting the total defect color representation value of any defect of the first rust label and defect spectral characteristics influence value , then the defect rust approximation value formula of the defect is: ,in Indicates the total value weight of the defect color representation of the defect, Indicates the defect spectrum feature influence value weight of the defect.

[0038] The present invention is further improved in that S4 includes calculating the defect area of ​​any defect in the second rust tag, and the defect area is calculated by finding all connected domains of the defect through the findContours function and then using the contourArea function to count the number of pixels contained in each marked connected domain, which is recorded as the area of ​​the defect. , get the defect rust corrosion value of the defect .

[0039] A further improvement of the present invention is that the laser parameter-corrosion value association model is based on a support vector machine model, calculates the defect rust corrosion values ​​of historical defects and the corresponding laser wavelength and power as training data, uses the defect rust corrosion values ​​of all defects in the second rust label as input data of the laser parameter-corrosion value association model, and outputs the corresponding laser wavelength and power.

[0040] On the other hand, the present invention provides a system for laser cleaning and intelligent defect recognition of a ship surface, comprising:

[0041] The data acquisition module is used to acquire images of the ship surface through a camera mounted on a drone to obtain a color image and a spectral image of the ship surface;

[0042] A first rust separation module is used to separate the ship surface concave defects in the ship surface color image to obtain a first rust label;

[0043] A second rust separation module, used for separating the biological mucus in the first rust tag by combining the constant color of the hull-light compensation coefficient to obtain a second rust tag;

[0044] The laser parameter output module is used to calculate the defect rust corrosion value through the second rust label, establish a laser parameter-corrosion value association model, use the defect rust corrosion value as an input parameter, and output the corresponding laser wavelength and power.

[0045] Compared with the prior art, the present invention has the following beneficial effects:

[0046] 1. The present invention firstly uses the first rust separation strategy to extract the dent defects using the Sobel operator and other technologies, and further distinguishes the rust from the biological slime through the constant color-light compensation coefficient of the hull;

[0047] 2. Secondly, by calculating the total value of the color representation of the defect, the spectral characteristic influence value and the defect area, the rust approach value and corrosion value of the defect are obtained. Based on these key parameters, the system establishes a laser parameter-corrosion value association model, automatically determines the laser wavelength and power suitable for cleaning, and realizes the accurate identification of ship surface defects and the intelligent adjustment of laser cleaning parameters. It not only improves the efficiency and accuracy of ship cleaning, but also provides an intelligent and automated solution for ship maintenance. BRIEF DESCRIPTION OF THE DRAWINGS

[0048] Figure 1 This is a flow chart of the method for laser cleaning and intelligent defect recognition of ship surface according to the present invention;

[0049] Figure 2 This is a flow chart of the first rust separation strategy in the method for laser cleaning and intelligent defect identification of ship surface of the present invention;

[0050] Figure 3 This is a flow chart of the second rust separation strategy in the method for laser cleaning and intelligent defect identification of ship surface of the present invention;

[0051] Figure 4 This is a framework diagram of the laser cleaning and intelligent defect recognition system for ship surfaces according to the present invention. DETAILED DESCRIPTION

[0052] The technical solution of the present invention is described in detail below through the accompanying drawings and specific embodiments. It should be understood that the embodiments of the present invention and the specific features in the embodiments are detailed descriptions of the technical solution of the present invention, rather than limitations on the technical solution of the present invention. The embodiments of the present invention and the technical features in the embodiments may be combined with each other unless there is a conflict.

[0053] The term "and / or" is only a description of the association relationship between related objects, indicating that there can be three relationships. For example, A and / or B can mean: A exists alone, A and B exist at the same time, and B exists alone. In addition, the character " / " generally indicates that the related objects are in an "or" relationship.

[0054] Example 1

[0055] Figure 1 The flowchart of the method for laser cleaning and intelligent defect recognition of ship surface disclosed in this embodiment is shown, and the steps are as follows:

[0056] S1. Capture images of the ship surface by using a camera mounted on a drone to obtain a color image and a spectral image of the ship surface;

[0057] S2, separating the ship surface concave defects in the ship surface color image by setting a first rust separation strategy to obtain a first rust label; Figure 2 This is a flow chart of the first rust separation strategy in the method for laser cleaning and intelligent defect recognition of ship surface of the present invention; the specific steps of the first rust separation strategy include:

[0058] S21, extracting defective parts in the ship surface color image by using a Sobel operator to obtain a first defect set;

[0059] S22, calculating the edge clarity of defects in the first defect set to obtain a first defect edge clarity data set The edge clarity is obtained by extracting the vector edges in the first defect set and using the gradient amplitude of the Sobel operator to replace the standard deviation and covariance of the pixels in the structural similarity index. The specific steps include:

[0060] S221, extracting pixels on both sides of the vector edge of any defect edge to obtain a pixel data set within the defect edge and edge pixel dataset , and record the corresponding position, where Indicates that the defect The pixels inside the edge of the vector edge, Indicates that the defect The pixels outside the edge of the vector edge; Indicates the number of pixels passed by any defect edge;

[0061] S222. Calculate the edge clarity of the defect ,in, Indicates that the defect The edge pixels within the edge of the vector edge and the defect The difference in the horizontal gradient of the pixels outside the edge of the vector edge, Indicates that the defect The edge pixels within the edge of the vector edge and the defect The difference in the vertical gradient of the pixels outside the edge of the vector edge; ;

[0062] S223, repeating step S221 and step S222 until the edge clarity of all defect data in the first defect set is calculated and included in the first defect edge clarity data set middle.

[0063] S23. Set edge defect clarity threshold , extract the first defect edge clarity data set The edge definition is greater than the edge defect definition threshold The corresponding defect image is extracted as the first rust label, the first defect edge clarity dataset The edge definition is less than or equal to the edge defect definition threshold The defect image corresponding to the data is taken as a concave defect.

[0064] S3, setting a second rust separation strategy by combining the constant color of the hull and the light compensation coefficient, separating the biological mucus in the first rust label to obtain a second rust label; Figure 3 This is a flow chart of the second rust separation strategy in the method for laser cleaning and intelligent defect recognition of ship surface of the present invention. The specific steps of the second rust separation strategy include:

[0065] S31, extracting the area corresponding to the first rust label in the ship surface color image, and including it in a first defect surface color image dataset; extracting the area corresponding to the first rust label in the ship surface spectral image, and including it in a first defect spectral image dataset;

[0066] S32, calculating the constant color-light compensation coefficient of the hull through the surface color image of the ship, and calculating the defect color representation value through the data in the first defect surface color image data set; the specific calculation steps of the constant color-light compensation coefficient of the hull include:

[0067] S321, looping through all pixel RGB values ​​of the ship surface color image to obtain all pixel brightness values , and arranged in order from low to high to obtain a sequence of pixel brightness values ​​of the ship surface color image; where R is the red channel color value, G is the green channel color value, and B is the blue channel color value;

[0068] S322, extracting the sum of the first 5% of the pixel brightness values ​​of the ship surface color image sequence as a reference compensation value , and extract the corresponding number of pixels ;

[0069] S323, calculate the constant-light compensation coefficient of the hull according to the reference compensation value .

[0070] S33, calculating the defect spectrum feature influence value through the data in the first defect spectrum image data set; the specific steps of calculating the defect spectrum feature influence value include:

[0071] S331. The structure of the ship surface is relatively sensitive to infrared band, red light band, ultraviolet band and green light band, so extract the near-infrared band reflectivity data set of any defect in the first defect spectral image data set. , red light band reflectance dataset , UV band reflectance dataset and green band reflectance dataset , where a represents the number of near-infrared bands, b represents the number of red light bands, c represents the number of ultraviolet bands, and d represents the number of green light bands; Indicates The spectral reflectance of the near-infrared band is Indicates The spectral reflectance of the red light band is Indicates The spectral reflectance of the ultraviolet band is Indicates The spectral reflectance of the green light band is , , , ;

[0072] S332, set the defect sensitive band combination value ,in Indicates the red light band sensitivity value of the defect. Defects on the hull surface are brown, while algae and other substances are green. In order to better distinguish between ship surface corrosion and biological slime, set The calculation formula is ;

[0073] S333, calculate the grayscale mean value within the contour of the defect Since the gray value can be regarded as a representation of the severity of the defect, the rust sensitive band combination extraction formula is set by exploring the correlation between the gray value mean value within the defect contour and the sensitive band , where the function Indicates the calculation of correlation coefficient, function Indicates maximum value extraction, function Representation Extraction The defect sensitive band combination value of the corresponding near-infrared band and ultraviolet band is taken as the defect spectral characteristic influence value of the defect .

[0074] The specific calculation process of the defect rust approach value includes: extracting the total defect color representation value of any defect of the first rust label and defect spectral characteristics influence value , then the defect rust approximation value formula of the defect is: ,in Indicates the total value weight of the defect color representation of the defect, Indicates the defect spectrum feature influence value weight of the defect.

[0075] S34, combining the constant-light compensation coefficient of the hull color, the total value of the defect color representation and the influence value of the defect spectrum characteristic, to obtain the defect rust approach value; the specific calculation steps of the total value of the defect color representation include: extracting the gray value sequence of all pixels within the contour of any defect in the first defect surface color image data set ,in Indicates that the defect The gray value of a pixel, , m represents the number of pixels in the defect; extract the RGB component values ​​of all pixels in the defect, The red component of a pixel is represented by , No. The green component of a pixel is expressed as , No. The blue component of a pixel is expressed as ; Calculate the The color representation value of a pixel is expressed as ,in, represents the gray value weight, Indicates color component weight; sets color representation segmentation value , and , dividing the color representation value into four pixel color representation intervals according to the color representation segmentation value, including , , , , count the distribution rate of the color representation values ​​of the pixels in the defect area in the above color representation interval, and extract the average color representation value of the pixels in the color representation interval with the highest distribution rate as the total color representation value of the area .

[0076] S35, setting a rust approach threshold, listing the data in the first rust tag whose defect rust approach value is greater than the rust approach threshold as a second rust tag, and listing the data in the first rust tag whose defect rust approach value is less than or equal to the rust approach threshold as biological slime.

[0077] S4, calculating the rust corrosion value of the defect through the second rust label; including calculating the defect area of ​​any defect in the second rust label, the defect area is calculated by finding all connected domains of the defect through the findContours function and then using the contourArea function to count the number of pixels contained in each marked connected domain, and record it as the area of ​​the defect , get the defect rust corrosion value of the defect .

[0078] S5. By establishing a laser parameter-corrosion value association model based on a support vector machine model, the defect rust corrosion value of the historical defect and the corresponding laser wavelength and power are calculated as training data, the defect rust corrosion value of all defects in the second rust label is used as the input data of the laser parameter-corrosion value association model, and the corresponding laser wavelength and power are output.

[0079] The threshold and weight may be set by default according to the present invention, or may be set by an operator.

[0080] Example 2

[0081] Figure 4 The framework diagram of the laser cleaning and intelligent defect recognition system for ship surface of the present invention is shown. Based on the same inventive concept as that of Example 1, the present invention provides a laser cleaning and intelligent defect recognition system for ship surface, including:

[0082] The data acquisition module is used to acquire images of the ship surface through a camera mounted on a drone to obtain a color image and a spectral image of the ship surface;

[0083] A first rust separation module is used to separate the ship surface concave defects in the ship surface color image to obtain a first rust label;

[0084] A second rust separation module, used for separating the biological mucus in the first rust tag by combining the constant color of the hull-light compensation coefficient to obtain a second rust tag;

[0085] The laser parameter output module is used to calculate the defect rust corrosion value through the second rust label, establish a laser parameter-corrosion value association model, use the defect rust corrosion value as an input parameter, and output the corresponding laser wavelength and power.

[0086] It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as methods, systems, or computer program products. Therefore, the present invention may take the form of a complete hardware embodiment, a complete software embodiment, or an embodiment combining software and hardware. Furthermore, the present invention may take the form of a computer program product implemented on one or more computer-usable storage media (including but not limited to disk storage, CD-ROM, optical storage, etc.) containing computer-usable program code.

[0087] The present invention is described with reference to flowcharts and / or block diagrams of methods, devices (systems), and computer program products according to embodiments of the present invention. It should be understood that each process and / or block in the flowchart and / or block diagram, as well as the combination of processes and / or blocks in the flowchart and / or block diagram, can be implemented by computer program instructions. These computer program instructions can be provided to a processor of a general-purpose computer, a special-purpose computer, an embedded processor, or other programmable data processing device to produce a machine, so that the instructions executed by the processor of the computer or other programmable data processing device generate instructions for implementing the processes in the flowchart and / or block diagram. Figure 1 A process or multiple processes and / or boxes Figure 1 A device that provides the functions specified in a block or multiple blocks.

[0088] These computer program instructions may also be stored in a computer-readable memory capable of directing a computer or other programmable data processing device to operate in a specific manner, so that the instructions stored in the computer-readable memory produce an article of manufacture comprising an instruction device, which implements the process Figure 1 A process or multiple processes and / or boxes Figure 1 A function specified in one or more boxes.

[0089] These computer program instructions can also be loaded onto a computer or other programmable data processing device so that a series of operating steps are executed on the computer or other programmable device to produce a computer-implemented process, thereby providing instructions for implementing the process. Figure 1 A process or multiple processes and / or boxes Figure 1 The steps for the functions specified in one or more boxes.

[0090] The embodiments of the present invention are described above in conjunction with the accompanying drawings, but the present invention is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the enlightenment of the present invention, ordinary technicians in this field can also make many forms without departing from the scope of protection of the purpose of the present invention and the claims, which all fall within the protection of the present invention.

Claims

1. Laser cleaning and intelligent defect recognition method for ship surface, It is characterized in that The following steps are involved: S1. Capture images of the ship surface by using a camera mounted on a drone to obtain a color image and a spectral image of the ship surface; S2, separating the ship surface concave defects in the ship surface color image by setting a first rust separation strategy to obtain a first rust label; S3, setting a second rust separation strategy by combining the constant color of the hull and the light compensation coefficient, separating the biological mucus in the first rust label to obtain a second rust label; S4, calculating the defect rust corrosion value through the second rust label; S5, by establishing a laser parameter-corrosion value association model, taking the defect rust corrosion value as an input parameter, and outputting the corresponding laser wavelength and power; The specific steps of the first rust separation strategy include: S21, extracting defective parts in the ship surface color image by using a Sobel operator to obtain a first defect set; S22, calculating the edge clarity of defects in the first defect set to obtain a first defect edge clarity data set ; S23. Set edge defect clarity threshold , extract the first defect edge clarity data set The edge definition is greater than the edge defect definition threshold The corresponding defect image is extracted as the first rust label, the first defect edge clarity dataset The edge definition is less than or equal to the edge defect definition threshold The defect image corresponding to the data is taken as a concave defect; The edge clarity is obtained by extracting the vector edges in the first defect set and using the gradient amplitude of the Sobel operator to replace the standard deviation and covariance of the pixels in the structural similarity index. The specific steps include: S221, extracting pixels on both sides of the vector edge of any defect edge to obtain a pixel data set within the defect edge and edge pixel dataset , and record the corresponding position, where Indicates that the defect The pixels inside the edge of the vector edge, Indicates that the defect The pixels outside the edge of the vector edge; Indicates the number of pixels passed by any defect edge; S222. Calculate the edge clarity of the defect ,in, Indicates that the defect The edge pixels of the vector edge and the defect The difference in the horizontal gradient of the pixels outside the edge of the vector edge, Indicates that the defect The edge pixels within the edge of the vector edge and the defect The difference in the vertical gradient of the pixels outside the edge of the vector edge; ; S223, repeating step S221 and step S222 until the edge clarity calculation of all defect data in the first defect set is completed, and obtaining a first defect edge clarity data set ; The specific steps of the second rust separation strategy include: S31, extracting the area corresponding to the first rust label in the ship surface color image, and including it in a first defect surface color image dataset; extracting the area corresponding to the first rust label in the ship surface spectral image, and including it in a first defect spectral image dataset; S32, calculating the constant-light compensation coefficient of the hull color through the ship surface color image, and calculating the defect color representation value through the data in the first defect surface color image data set; S33, calculating the defect spectrum feature influence value through the data in the first defect spectrum image data set; S34, combining the hull color constant-light compensation coefficient, the defect color representation total value and the defect spectral characteristic influence value to obtain a defect rust approach value; S35, setting a rust approach threshold, listing the data in the first rust tag whose defect rust approach value is greater than the rust approach threshold as a second rust tag, and listing the data in the first rust tag whose defect rust approach value is less than or equal to the rust approach threshold as biological slime; The specific calculation steps of the hull color constant-light compensation coefficient include: S321, looping through all pixel RGB values ​​of the ship surface color image to obtain all pixel brightness values , and arranged in order from low to high to obtain a sequence of pixel brightness values ​​of the ship surface color image; where R is the red channel color value, G is the green channel color value, and B is the blue channel color value; S322, extracting the sum of the first 5% of the pixel brightness value sequence of the ship surface color image as a reference compensation value , and extract the corresponding number of pixels ; S323, calculate the constant-light compensation coefficient of the hull color according to the reference compensation value ; The specific calculation steps of the defect color representation total value include: extracting the gray value sequence of all pixels within the contour of any defect in the first defect surface color image data set , the gray value sequence includes , Indicates that the defect The gray value of a pixel, , m represents the number of pixels in the defect; extract the RGB component values ​​of all pixels in the defect, The red component of a pixel is represented by , No. The green component of a pixel is expressed as , No. The blue component of a pixel is expressed as ; Calculate the The color representation value of a pixel is expressed as ,in, represents the gray value weight, Indicates color component weight; sets color representation segmentation value , and , dividing the color representation value into four pixel color representation intervals according to the color representation segmentation value, including , , , , count the distribution rate of the color representation values ​​of the pixels in the defect area in the above color representation interval, and extract the average color representation value of the pixels in the color representation interval with the highest distribution rate as the total color representation value of the area ; The specific steps of the defect spectrum characteristic influencing value include: S331, extracting a near-infrared band reflectivity dataset of any defect in the first defect spectral image dataset , red light band reflectance dataset , UV band reflectance dataset and green band reflectance dataset , where a represents the number of near-infrared bands, b represents the number of red light bands, c represents the number of ultraviolet bands, and d represents the number of green light bands; Indicates The spectral reflectance of the near-infrared band is Indicates The spectral reflectance of the red light band is Indicates The spectral reflectance of the ultraviolet band is Indicates The spectral reflectance of the green light band is , , , ; S332, set the defect sensitive band combination value ,in It represents the red light band sensitivity value of the defect, and the calculation formula is: ; S333, calculate the grayscale mean value within the contour of the defect , set the rust sensitive band combination extraction formula , where the function Indicates the calculation of correlation coefficient, function Represents maximum value extraction, function Representation Extraction The defect sensitive band combination value of the corresponding near-infrared band and ultraviolet band is taken as the defect spectral characteristic influence value of the defect ; The specific calculation process of the defect rust approach value includes: extracting the total defect color representation value of any defect of the first rust label and defect spectral characteristics influence value , then the defect rust approximation value formula of the defect is: ,in Indicates the total value weight of the defect color representation of the defect, Indicates the defect spectrum characteristic influence value weight of the defect; S4 includes: calculating the defect area of ​​any defect in the second rust tag, the defect area is calculated by finding all connected domains of the defect through the findContours function, and then using the contourArea function to count the number of pixels contained in each marked connected domain, which is recorded as the area of ​​the defect , get the defect rust corrosion value of the defect .

2. The method for laser cleaning and intelligent defect recognition on the surface of a ship according to claim 1, Features: The laser parameter-corrosion value association model is based on a support vector machine model. The defect rust corrosion values ​​of historical defects and the corresponding laser wavelength and power are used as training data. The defect rust corrosion values ​​of all defects in the second rust label are used as input data of the laser parameter-corrosion value association model, and the corresponding laser wavelength and power are output.

3. A system for laser cleaning and intelligent defect recognition on a ship surface, for implementing a method for laser cleaning and intelligent defect recognition on a ship surface as claimed in any one of claims 1 to 2, It is characterized in that include: The data acquisition module is used to acquire images of the ship surface through a camera mounted on a drone to obtain a color image and a spectral image of the ship surface; A first rust separation module is used to separate the ship surface concave defects in the ship surface color image to obtain a first rust label; A second rust separation module, used for separating the biological mucus in the first rust tag by combining the constant color of the hull-light compensation coefficient to obtain a second rust tag; The laser parameter output module is used to calculate the defect rust corrosion value through the second rust label, establish a laser parameter-corrosion value association model, use the defect rust corrosion value as an input parameter, and output the corresponding laser wavelength and power.

Images