Freshwater spray device and spray control method for underwater robot maintenance

Abstract

This application relates to the field of automated cleaning technology, specifically to a freshwater spray device and spray control method for underwater robot maintenance. The method comprises: capturing images of the underwater robot from various angles and converting them into point cloud data, dividing all the point cloud data converted from each image into small blocks; obtaining the difference coefficient between any two adjacent small blocks for each image; obtaining the adjacent small block similarity value and distance complexity of each small block; and then obtaining the neighborhood feature coefficient of each small block; obtaining the minimum number of search neighbors when filtering the point cloud data within each small block, and filtering the point cloud data within each small block; obtaining a three-dimensional model of the underwater robot, and controlling the pressure at the spray head outlet based on the distance between the spray head and the three-dimensional model. This application aims to improve the cleaning effect of the underwater robot by improving the filtering effect of the point cloud data, thereby precisely controlling the pressure at the spray head outlet.

Description

Technical Field

[0001] The present application relates to the field of automated cleaning technology, and in particular to a freshwater spraying device and a spraying control method for underwater robot maintenance. Background Art

[0002] An underwater maintenance robot is an automated device used for underwater operations. After underwater operations, impurities such as salt, silt, and algae from the seawater adhere to the robot's surface. To prevent corrosion and damage to the robot's exterior and components, the robot must be cleaned. During the cleaning process, the spray pressure must be controlled according to the spray distance.

[0003] Existing technologies typically collect point cloud data from the underwater robot and use the resulting 3D model to calculate the distance from the sprinkler head to various parts of the underwater robot. However, this 3D reconstruction requires filtering of the point cloud data. Traditional filtering algorithms use the same minimum number of neighboring points for all point cloud data, which can lead to poor filtering effectiveness for simple structures and less detailed features for complex structures, affecting modeling accuracy and, in turn, compromising the cleaning performance of the underwater robot. Summary of the Invention

[0004] In view of the above, it is necessary to provide a freshwater spray device and spray control method for underwater robot maintenance. Compared with the traditional freshwater spray device and spray control method for underwater robot maintenance, it can improve the filtering effect of point cloud data and accurately control the pressure at the spray head outlet, thereby improving the cleaning effect of the underwater robot:

[0005] In a first aspect, an embodiment of the present application provides a spray control method for a freshwater spray device for underwater robot maintenance, the method comprising the following steps:

[0006] Images of the underwater robot are taken from various angles and converted into point cloud data, respectively, to obtain the grayscale value of each point cloud data; all the point cloud data converted from each image are divided into small blocks;

[0007] For all point cloud data converted from any image, obtain the difference coefficient between any two adjacent small blocks based on the difference in grayscale values ​​of the point cloud data in the two adjacent small blocks;

[0008] Obtain the similarity value of each small block to its adjacent small blocks through the difference coefficient between each small block and all its adjacent small blocks;

[0009] Get the fitting plane of all point cloud data in each small block, and obtain the distance complexity of each small block through the distribution of the distance from all point cloud data in each small block to its fitting plane;

[0010] Obtaining a neighborhood feature coefficient of each small block through the similarity value of the adjacent small blocks and the distance complexity;

[0011] The minimum number of neighboring points to be searched when filtering the point cloud data in each small block is obtained by using the neighborhood characteristic coefficient, and the point cloud data in each small block is filtered;

[0012] The three-dimensional model of the underwater robot is obtained by filtering all the point cloud data converted from all the images, and the pressure at the outlet of the sprinkler head is controlled by the distance between the sprinkler head and the three-dimensional model.

[0013] In one embodiment, the process of obtaining the difference coefficient is as follows:

[0014] Cluster the grayscale values ​​of all point cloud data in each small block, and arrange the grayscale values ​​of all point cloud data in each cluster in ascending order to obtain each grayscale value subsequence;

[0015] The expression of the difference factor between the a-th grayscale value subsequence of the i-th small block and the b-th grayscale value subsequence of its j-th adjacent small block is:

[0016] Where, represents the difference factor between the a-th grayscale value subsequence of the i-th small block and the b-th grayscale value subsequence of its j-th adjacent small block; SI( ) represents the difference measurement function; represents the ath gray value subsequence of the i-th small block; represents the bth gray value subsequence of the jth adjacent block of the i-th block;

[0017] The difference coefficient between the i-th small block and its j-th adjacent small block is obtained by distributing the difference factors between each gray value subsequence of the i-th small block and each gray value subsequence of its j-th adjacent small block.

[0018] In one embodiment, the difference coefficient is the minimum value of all the difference factors calculated between the i-th small block and its j-th adjacent small block.

[0019] In one embodiment, the process of obtaining the similarity values ​​of adjacent small blocks is as follows:

[0020] Cluster the difference coefficients between each small block and all its adjacent small blocks, and take the maximum value in the cluster with the smallest element mean as the difference threshold of each small block;

[0021] The number of difference coefficients between each small block and all its adjacent small blocks that are smaller than the difference threshold is counted as the adjacent small block similarity value of each small block.

[0022] In one embodiment, the distance complexity is obtained by:

[0023] Calculate the average distance between all point cloud data in each small block and its fitting plane;

[0024] Count the number of different values ​​of the distance from all point cloud data in each small block to its fitting plane;

[0025] The distance complexity is determined by the average value and the number.

[0026] In one embodiment, the distance complexity is the product of the average value and the number.

[0027] In one embodiment, the process of obtaining the neighborhood feature coefficient is: mapping the distance complexity into a positive number, and the neighborhood feature coefficient is the ratio of the adjacent small block similarity value to the positive number.

[0028] In one embodiment, obtaining the minimum number of neighboring points to search when filtering the point cloud data in each small block includes:

[0029] Calculating the difference between a preset maximum value and a preset minimum value of the search quantity;

[0030] Calculating the product of the normalized value of the neighborhood characteristic coefficient and the difference;

[0031] The search quantity is positively correlated with the product and the preset minimum value respectively.

[0032] In one embodiment, the search quantity is the sum of a rounded-off result of the product and the preset minimum value.

[0033] Secondly, an embodiment of the present application provides a freshwater spray device for underwater robot maintenance, the device comprising: a binocular stereo camera, a spray head, an air compressor, and a pressure controller; wherein the binocular stereo camera is used to capture images of the underwater robot; the spray head is used to spray fresh water; the air compressor is used to provide pressure; and the pressure controller is used to control the pressure at the spray head outlet.

[0034] This application has at least the following beneficial effects:

[0035] By capturing images of the underwater robot from different angles, this application can cover all parts of the underwater robot and ensure the acquisition of complete point cloud data. The point cloud data can accurately represent the surface features of the underwater robot, providing basic data support for subsequent analysis and processing. By dividing the point cloud data into multiple small blocks and analyzing each small block separately, the characteristics of each local area can be studied in more detail, facilitating subsequent targeted filtering processing of each local area. By comparing the similarity of the grayscale values ​​of adjacent small blocks, it is possible to determine whether adjacent small blocks belong to the same surface of the object or to areas with relatively consistent structures, so that in the subsequent filtering process, greater filtering strength is applied to areas with more consistent surfaces.

[0036] Furthermore, considering that the structural complexity of different parts of the underwater robot varies, the distance between the point cloud data in each small block and its fitting plane is used to analyze the structural complexity of each small block. Subsequently, a stronger filtering is performed on the area with simpler structure.

[0037] Furthermore, by combining the similarity between each small block and the adjacent small blocks, as well as the structural complexity of the location of each small block, the intensity of filtering the point cloud data in each small block is obtained. By conducting a comprehensive analysis from two dimensions, the one-sidedness and defects of single-dimensional analysis can be effectively avoided, and the characteristics of each small block can be more comprehensively evaluated, so as to formulate a more accurate filtering strategy, while ensuring the filtering effect, retaining more detailed features; thereby improving the accuracy of the three-dimensional model, being able to accurately control the pressure at the outlet of the sprinkler head, improving the cleaning effect of the underwater robot, and avoiding damage to the surface coating of the underwater robot due to excessive pressure. BRIEF DESCRIPTION OF THE DRAWINGS

[0038] In order to more clearly illustrate the technical solutions and advantages of the embodiments of the present application or the prior art, the following is a brief introduction to the drawings required for use in the embodiments or the description of the prior art. Obviously, the drawings described below are only some embodiments of the present application. For ordinary technicians in this field, other drawings can be obtained based on these drawings without any creative work.

[0039] Figure 1 A flowchart of a spray control method for a freshwater spray device for underwater robot maintenance provided in one embodiment of the present application;

[0040] Figure 2 Schematic diagram of the process of obtaining distance complexity;

[0041] Figure 3 The figure is a flowchart for obtaining the search quantity. DETAILED DESCRIPTION

[0042] In the description of the embodiments of this application, words such as "exemplary," "or," and "for example" are used to indicate examples, illustrations, or descriptions. Any embodiment or design described as "exemplary" or "for example" in the embodiments of this application should not be construed as being preferred or advantageous over other embodiments or designs. Rather, the use of words such as "exemplary," "or," and "for example" is intended to present the relevant concepts in a concrete manner.

[0043] Unless otherwise defined, all technical and scientific terms used herein have the same meaning as commonly understood by those skilled in the art to which this application relates. The terms used in this application are for the purpose of describing specific embodiments only and are not intended to limit this application. It should be understood that, unless otherwise indicated, " / " represents or.

[0044] It should also be noted that the terms "first" and "second" in this application are used to distinguish similar objects, rather than to describe a specific order or sequence.

[0045] The freshwater spray device for underwater robot maintenance provided in this application includes a binocular stereo camera, a spray head, an air compressor, and a pressure controller; wherein the binocular stereo camera is used to capture images of the underwater robot; the spray head is used to spray fresh water; the air compressor is used to provide pressure; and the pressure controller is used to control the pressure at the spray head outlet.

[0046] The specific scheme of the spray control method of the freshwater spray device for underwater robot maintenance provided by this application is described in detail below with reference to the accompanying drawings.

[0047] like Figure 1 , which shows a flowchart of a spray control method for a freshwater spray device for underwater robot maintenance provided by an embodiment of the present application, the method comprising the following steps:

[0048] Step S1: capture images of the underwater robot from various angles and convert them into point cloud data, obtaining the grayscale value of each point cloud data; and dividing all the point cloud data converted from each image into small blocks.

[0049] A binocular stereo camera is used to capture images of the underwater robot from various angles. The images captured by the binocular stereo camera are first stereo-rectified using the Bouguet algorithm. Then, the disparity map of the image is calculated using the SGBM algorithm. The disparity map is then used to calculate the depth image of the underwater robot using the geometric relationship of binocular vision. Finally, the depth image is converted into point cloud data using the create_from_depth_image function of the PointCloud class in Open3D. The Bouguet algorithm, SGBM algorithm, the geometric relationship of binocular vision, and the create_from_depth_image function of the PointCloud class in Open3D are all well-known technologies and will not be described in detail in this application.

[0050] In this embodiment, an image of the underwater robot is captured every 45 degrees in the horizontal direction.

[0051] All point cloud data converted from each image are evenly divided into small blocks. Before division, the maximum and minimum values ​​of the point cloud data in each direction are obtained to determine the division boundaries. Since the point cloud data is converted from a two-dimensional image, each point cloud data corresponds to a grayscale value in the original image taken by a binocular stereo camera. The grayscale value corresponding to each point cloud data is obtained and recorded as the grayscale value of each point cloud data.

[0052] In this embodiment, the point cloud data is evenly divided into small blocks with a length, width, and height of 0.5 cm. The size of the small blocks can be set by the implementer according to actual conditions, and this application does not impose any special restrictions.

[0053] Step S2: for all point cloud data converted from any image, obtain the difference coefficient between any two adjacent small blocks according to the difference in grayscale values ​​of the point cloud data in the two adjacent small blocks.

[0054] Typically, for ease of manufacturing, the same part of an underwater robot is typically only one color. When operating deep underwater, the robot's surface carries impurities such as salt, silt, and algae. These impurities can cause changes in the robot's point cloud data and its grayscale values. However, the impact of these impurities is limited, and the color of the impurities differs significantly from that of the robot. Different substances can be distinguished by their grayscale values. The grayscale values ​​of all point cloud data within each small block are divided into two parts to characterize the different substances. Specifically, the grayscale values ​​of all point cloud data within each small block are clustered to obtain two clusters. The grayscale values ​​of all point cloud data within each cluster are sorted in ascending order to obtain a subsequence of grayscale values.

[0055] In this embodiment, the grayscale values ​​of all point cloud data in each small block are clustered by the K-means algorithm. On the basis of being able to divide the grayscale values ​​of all point cloud data in each small block into two parts, the implementer may adopt other existing technologies, such as Otsu threshold segmentation algorithm, iterative threshold segmentation, etc., and this application does not impose any special restrictions.

[0056] Under normal circumstances, adjacent locations on the surface of the underwater robot are similar. However, impurities will reduce the similarity. Based on the grayscale values ​​of the normal robot and impurities in each small block, the difference coefficient between any two adjacent small blocks is obtained. The specific process is as follows:

[0057] Where, represents the difference factor between the a-th grayscale value subsequence of the i-th small block and the b-th grayscale value subsequence of its j-th adjacent small block; SI( ) represents the difference measurement function; represents the ath gray value subsequence of the i-th small block; represents the bth gray value subsequence of the jth adjacent block of the i-th block;

[0058] The difference factors of all two gray value subsequences between the i-th small block and its j-th adjacent small block are formed into a set, and the minimum value in the set is taken as the difference coefficient of the i-th small block.

[0059] In this embodiment, the difference measurement function used is the DTW (Dynamic Time Warping) distance. The DTW distance is a well-known technology and will not be described in detail in this application. As other implementation methods, on the basis of being able to measure the differences between grayscale value subsequences, the implementer may adopt other existing technologies, such as Mahalanobis distance, the inverse of cosine similarity, etc., and this application does not impose any special restrictions.

[0060] The coefficient of difference of each small block is calculated using the same calculation method as that of the i-th small block.

[0061] During the manufacturing process of underwater robots, the same part usually has only one color. Although the impurities carried on the surface of the underwater robot after completing its work are different from the underwater robot parts, for small blocks within the same part area, there are parts with a high degree of similarity between adjacent small blocks. As a result, the value of the difference coefficient between small blocks is small, which can reduce the impact of impurities on the actual similarity between adjacent small blocks.

[0062] Step S3: Obtain the adjacent block similarity value of each small block through the difference coefficient between each small block and all its adjacent small blocks.

[0063] When measuring the similarity of colors between different small blocks, although there may be impurities that affect the color of the small blocks, the similarity between the small blocks is still very high by clustering and segmenting the colors. However, the colors of different parts are different, and the similarity between small blocks is low. Therefore, taking the i-th small block as an example, the difference coefficient between the i-th small block and all its adjacent small blocks is obtained and used as the input of the clustering algorithm. Each cluster is output, and the maximum value in the cluster with the smallest element mean is used as the difference threshold of the i-th small block; the cluster with the smallest mean is formed by similar small blocks. Therefore, the difference threshold can effectively determine whether the small blocks are similar, and obtain the adjacent small block similarity value of each small block, specifically:

[0064] The number of difference coefficients between each small block and all its adjacent small blocks that are smaller than the difference threshold is counted as the adjacent small block similarity value of each small block.

[0065] When a small block is located within the underwater robot's shell, the similarity between the different parts of the shell is high due to the large surface area of ​​the underwater robot. Consequently, the number of similarities between the small blocks and their neighbors is high. The high similarity between each small block and its neighbors indicates that the color and texture of each block are relatively uniform, allowing for a larger search number during filtering to reduce the impact of noise and impurities on the point cloud data. This enhanced filtering operation can more effectively remove noise and impurities, thereby improving the consistency between the constructed 3D model and the actual underwater robot. This significantly enhances the cleaning effectiveness of the underwater robot shell, ensuring cleanliness and maintenance efficiency.

[0066] Step S4: Obtain the fitting plane of the point cloud data in each small block, and obtain the distance complexity of each small block through the distribution of the distances from all the point cloud data in each small block to its fitting plane.

[0067] Since different parts of the outer surface of the underwater robot are not a straight surface structure, the shapes of different parts are different. If the shape is more complex, a lower filtering intensity should be used to ensure the details of the complex structure of the underwater robot, such as the robot gripper. If the shape is simpler and closer to a plane, a stronger filtering intensity should be used to reduce the impact of noise and underwater attachments on the underwater robot modeling.

[0068] Based on the above analysis, the fitting planes of all point cloud data in each small block are obtained, and the distances from each point cloud data in each small block to its fitting plane are calculated. Calculation of the distance from a point to a plane is a well-known technique and will not be described in detail in this application.

[0069] In this embodiment, the least squares method is used to obtain the fitting plane. The least squares method is a well-known technology and will not be described in detail in this application. As other implementation methods, on the basis of being able to obtain the fitting plane, the implementer may adopt other existing technologies, such as random sampling consistency algorithm, singular value decomposition, principal component analysis, etc., and this application does not impose any special restrictions.

[0070] Based on the above analysis, the distance complexity of each small block is obtained by distributing the distances from all point cloud data in each small block to its fitting plane. The expression is:

[0071] Where, represents the distance complexity of the i-th small block; Represents the number of different values ​​of the distance from all point cloud data in the i-th block to its fitting plane. For example, when the distance from the point cloud data to its fitting plane is 1, 1, 2, 3, 4, 4 and 4, there are 4 different values. The value of is 4; Indicates the number of point cloud data in the i-th small block; It represents the distance from the xth point cloud data in the ith block to its fitting plane. The process diagram of obtaining distance complexity is as follows: Figure 2 shown.

[0072] When the i-th small block is in a complex structural part of the underwater robot, due to the irregular and varied shapes of the complex structural parts, the greater the distance between the point cloud data in the i-th small block and the fitting plane, and the greater the number of distance types, the lower the filtering intensity needs to be used for the point cloud data of the i-th small block to retain the detailed features, reduce the impact of filtering on the complex structure, and enable more detailed features to be retained in the three-dimensional model.

[0073] Step S5: Obtaining the neighborhood feature coefficient of each small block through the adjacent small block similarity value and the distance complexity.

[0074] Through the above two-dimensional analysis, the greater the number of similarities between a small block and its neighboring small blocks, the lower the structural complexity of the small block, and the higher the filtering strength required to make the filtering result smoother, which can effectively reduce the influence of noise and outliers; conversely, the smaller the filtering strength, the more local details can be retained.

[0075] The neighborhood feature coefficient of each small block is obtained by the similarity value of the adjacent small blocks and the distance complexity, specifically:

[0076] The distance complexity of each small block is mapped to a positive number, and the ratio of the similarity value of each small block to the adjacent small blocks and the positive number is used as the neighborhood feature coefficient of each small block.

[0077] In this embodiment, the method for mapping the distance complexity to a positive number is: calculating the sum of the distance complexity and a preset positive number, wherein the preset positive number is used to avoid the denominator being 0. The value of the preset positive number is preset manually and can be set by the implementer. In this embodiment, the value of the preset positive number is 0.01. There are many methods for mapping data to positive numbers, and the implementer can choose other feasible methods on his own. This application does not impose any special restrictions.

[0078] In practical applications, analyzing across two distinct dimensions can effectively avoid the incompleteness and drawbacks that can arise from single-dimensional analysis. For example, when evaluating small blocks in an underwater robot's shell, relying solely on the number of similarities between a block and its neighbors can overlook the impact of structural complexity on the selection of filter strength, resulting in poor filtering results. Conversely, considering only structural complexity while ignoring similarity can also lead to inaccurate filtering strategies. Combining a comprehensive analysis across these two dimensions allows for a more comprehensive assessment of the block's characteristics and requirements, leading to the development of a more rational and precise filtering scheme that effectively reduces noise while preserving important details.

[0079] Step S6: Obtain the minimum number of neighboring points to be searched when filtering the point cloud data in each small block using the neighborhood feature coefficient, and perform filtering on the point cloud data in each small block.

[0080] The minimum number of neighboring points to be searched when filtering the point cloud data in each small block is obtained through the neighborhood feature coefficient. The expression is:

[0081] Where, Indicates the minimum number of neighboring points to search when filtering the point cloud data in the i-th small block; round() indicates the rounding function; norm() indicates the normalization function; Represents the neighborhood feature coefficient of the i-th small block; Indicates the preset minimum value of the search quantity; Indicates the difference between the preset maximum and minimum values ​​of the search quantity. The flowchart for obtaining the search quantity is as follows: Figure 3 shown.

[0082] In this embodiment, in order to avoid the lack of reference points in the filtering process due to too few neighboring points, resulting in poor filtering effect, and to avoid the excessive filtering due to too many neighboring points, thereby losing important detail information, the preset minimum value and the preset maximum value are set to 10 and 50 respectively. The implementer can set the preset minimum value and the preset maximum value according to the actual situation.

[0083] The point cloud data of the underwater robot and the search number of the minimum neighboring points of each point cloud data are used as input of the statistical filtering algorithm, and the filtered point cloud data is output. The statistical filtering algorithm is a well-known technology and will not be described in detail in this application.

[0084] Step S7: Obtain a three-dimensional model of the underwater robot through all point cloud data after filtering and conversion of all images, and control the pressure at the outlet of the sprinkler head by the distance between the sprinkler head and the three-dimensional model.

[0085] The filtered point cloud data obtained at different angles are matched, and the matched point cloud data are reconstructed in three dimensions to obtain a three-dimensional model of the underwater robot. The holes in the three-dimensional model are then filled using a hole filling algorithm, and texture mapping is used to optimize the realism of the three-dimensional model.

[0086] In this embodiment, an iterative closest point algorithm is used to match point cloud data. The iterative closest point algorithm is a well-known technology and will not be described in detail in this application. As other implementation methods, on the basis of being able to match point cloud data, the implementer may adopt other existing technologies, such as feature point matching algorithms, normal distribution transformation algorithms, etc., and this application does not impose any special restrictions.

[0087] In this embodiment, the Delaunay triangulation method is used for three-dimensional reconstruction. The Delaunay triangulation method is a well-known technology and will not be described in detail in this application. As other implementation methods, on the basis of being able to realize three-dimensional reconstruction of point cloud data, the implementer can adopt other existing technologies, such as Poisson reconstruction, etc., and this application does not impose any special restrictions.

[0088] When spraying an underwater robot, the pressure at the sprinkler outlet needs to be adjusted based on the distance between the sprinkler head and the 3D model. The specific method is: the distance between the sprinkler head and the underwater robot is recorded as the spray distance. The pressure at the sprinkler outlet is recorded at each historical spray distance when the pressure on the underwater robot surface is the same. Using the spray distance as the independent variable and the pressure at the sprinkler outlet as the dependent variable, a pressure-distance function is established to calculate the pressure at the sprinkler outlet and the spray distance. During the actual spraying process, the distance between the underwater robot and the 3D model is used as the independent variable, and the pressure at the sprinkler outlet is calculated using this pressure-distance function, thus achieving precise spraying operation.

[0089] In summary, the present application can capture images of the underwater robot from different angles, thereby covering all parts of the underwater robot and ensuring the acquisition of complete point cloud data. The point cloud data can accurately represent the surface features of the underwater robot, providing basic data support for subsequent analysis and processing. By dividing the point cloud data into multiple small blocks and analyzing each small block separately, the characteristics of each local area can be studied in more detail, facilitating subsequent targeted filtering processing of each local area. By comparing the similarity of the grayscale values ​​of adjacent small blocks, it is possible to determine whether adjacent small blocks belong to the same surface of the object or to areas with relatively consistent structures, facilitating the use of greater filtering strength for areas with more consistent surfaces in the subsequent filtering process.

[0090] Furthermore, considering that the structural complexity of different parts of the underwater robot varies, the distance between the point cloud data in each small block and its fitting plane is used to analyze the structural complexity of each small block. Subsequently, a stronger filtering is performed on the area with simpler structure.

[0091] Furthermore, by combining the similarity between each small block and the adjacent small blocks, as well as the structural complexity of the location of each small block, the intensity of filtering the point cloud data in each small block is obtained. By conducting a comprehensive analysis from two dimensions, the one-sidedness and defects of single-dimensional analysis can be effectively avoided, and the characteristics of each small block can be more comprehensively evaluated, so as to formulate a more accurate filtering strategy, while ensuring the filtering effect, retaining more detailed features; thereby improving the accuracy of the three-dimensional model, being able to accurately control the pressure at the outlet of the sprinkler head, improving the cleaning effect of the underwater robot, and avoiding damage to the surface coating of the underwater robot due to excessive pressure.

[0092] The flowcharts and block diagrams in the accompanying drawings show the possible implementation architectures, functions and operations of the systems, methods and computer program products according to the embodiments of the present disclosure. In this regard, each box in the flowchart or block diagram can represent a module, program segment or part of the code, and the module, program segment or part of the code contains one or more executable instructions for implementing the specified logical functions. In some alternative implementations, the functions marked in the box can also occur in an order different from that marked in the accompanying drawings. For example, two consecutive boxes can actually be executed substantially in parallel, or they can sometimes be executed in the opposite order, which can depend on the functions involved. In the descriptions corresponding to the flowcharts and block diagrams in the accompanying drawings, the operations or steps corresponding to different boxes can also occur in an order different from that disclosed in the description, and sometimes there is no specific order between different operations or steps. For example, two consecutive operations or steps can actually be executed substantially in parallel, or they can sometimes be executed in the opposite order, which can depend on the functions involved. Each block in the block diagrams and / or flowcharts, and combinations of blocks in the block diagrams and / or flowcharts, may be implemented by a dedicated hardware-based system that performs the specified function or action, or may be implemented by a combination of dedicated hardware and computer instructions.

[0093] It is obvious to those skilled in the art that the present application is not limited to the details of the above exemplary embodiments, and that the present application can be implemented in other specific forms without departing from the basic characteristics of the present application. Therefore, from all perspectives, the above embodiments of the present application should be regarded as exemplary and non-restrictive.

Claims

1. A spray control method for a freshwater spray device for underwater robot maintenance, characterized in that: The method comprises the following steps: Images of the underwater robot are taken from various angles and converted into point cloud data, respectively, to obtain the grayscale value of each point cloud data; all the point cloud data converted from each image are divided into small blocks; For all point cloud data converted from any image, obtain the difference coefficient between any two adjacent small blocks based on the difference in grayscale values ​​of the point cloud data in the two adjacent small blocks; Obtain the similarity value of each small block to its adjacent small blocks through the difference coefficient between each small block and all its adjacent small blocks; Get the fitting plane of all point cloud data in each small block, and obtain the distance complexity of each small block through the distribution of the distance from all point cloud data in each small block to its fitting plane; Obtaining a neighborhood feature coefficient of each small block through the similarity value of the adjacent small blocks and the distance complexity; The minimum number of neighboring points to be searched when filtering the point cloud data in each small block is obtained by using the neighborhood characteristic coefficient, and the point cloud data in each small block is filtered; Obtain a three-dimensional model of the underwater robot through all point cloud data after filtering and conversion of all images, and control the pressure at the outlet of the sprinkler head by the distance between the sprinkler head and the three-dimensional model; The process of obtaining the coefficient of difference is as follows: Cluster the grayscale values ​​of all point cloud data in each small block, and arrange the grayscale values ​​of all point cloud data in each cluster in ascending order to obtain each grayscale value subsequence; The expression of the difference factor between the a-th grayscale value subsequence of the i-th small block and the b-th grayscale value subsequence of its j-th adjacent small block is: Where, represents the difference factor between the a-th grayscale value subsequence of the i-th small block and the b-th grayscale value subsequence of its j-th adjacent small block; SI( ) represents the difference measurement function; represents the ath gray value subsequence of the i-th small block; represents the bth gray value subsequence of the jth adjacent block of the i-th block; The difference coefficient between the i-th small block and its j-th adjacent small block is obtained by distributing the difference factors between each gray value subsequence of the i-th small block and each gray value subsequence of its j-th adjacent small block.

2. The spray control method of the fresh water spray device for underwater robot maintenance according to claim 1, characterized in that: The difference coefficient is the minimum value of all the difference factors calculated between the i-th small block and its j-th adjacent small block.

3. The spray control method of the fresh water spray device for underwater robot maintenance according to claim 1, characterized in that: The process of obtaining the similarity values ​​of adjacent small blocks is as follows: Cluster the difference coefficients between each small block and all its adjacent small blocks, and take the maximum value in the cluster with the smallest element mean as the difference threshold of each small block; The number of difference coefficients between each small block and all its adjacent small blocks that are smaller than the difference threshold is counted as the adjacent small block similarity value of each small block.

4. The spray control method of the fresh water spray device for underwater robot maintenance according to claim 1, characterized in that: The method for obtaining the distance complexity is: Calculate the average distance between all point cloud data in each small block and its fitting plane; Count the number of different values ​​of the distance from all point cloud data in each small block to its fitting plane; The distance complexity is determined by the average value and the number.

5. The spray control method of the fresh water spray device for underwater robot maintenance according to claim 4, characterized in that: The distance complexity is the product of the average value and the number.

6. The spray control method of the fresh water spray device for underwater robot maintenance according to claim 1, characterized in that: The process of obtaining the neighborhood feature coefficient is as follows: mapping the distance complexity into a positive number, and the neighborhood feature coefficient is the ratio of the adjacent small block similarity value to the positive number.

7. The spray control method of the fresh water spray device for underwater robot maintenance according to claim 1, characterized in that: The step of obtaining the minimum number of neighboring points to search when filtering the point cloud data in each small block includes: Calculating the difference between a preset maximum value and a preset minimum value of the search quantity; Calculating the product of the normalized value of the neighborhood characteristic coefficient and the difference; The search quantity is positively correlated with the product and the preset minimum value respectively.

8. The spray control method of the fresh water spray device for underwater robot maintenance according to claim 7, characterized in that: The search quantity is the sum of the rounded-off calculation result of the product and the preset minimum value.

9. A freshwater spray device for underwater robot maintenance, using the spray control method of the freshwater spray device for underwater robot maintenance as claimed in claim 1, characterized in that: The device includes: a binocular stereo camera, a spray head, an air compressor, and a pressure controller; wherein the binocular stereo camera is used to capture images of the underwater robot; the spray head is used to spray fresh water; the air compressor is used to provide pressure; and the pressure controller is used to control the pressure at the spray head outlet.

Images