Fan fault area positioning and screening method for three-dimensional image recognition

Abstract

The application relates to a fan fault area positioning and screening method for three-dimensional image recognition. The fault area positioning method uses an edge detection algorithm to detect the edges of a fan surface, counts the edge point position quantity change and the edge contour quantity change in a decreasing parameter mode, so that the most suitable parameter of the edge detection algorithm is found, and the area where the edges are located is the fault area. The fault area screening method constructs a screening rule in an HSV color space according to the color characteristics of the fault to discard part of the suspected fault area, so that the fault area screening is realized. The method utilizes the existing information in the three-dimensional image to realize the positioning and screening of the fan fault area and improves the accuracy of the fan surface fault recognition.

Description

Technical Field

[0001] This invention relates to the field of wind farm fault identification, and more specifically to a method for locating and screening wind turbine fault areas for three-dimensional image recognition. Background Technology

[0002] With the rapid development of offshore wind farms in my country, the problems of traditional manual inspections being labor-intensive and inefficient have become increasingly prominent. However, my country's independently developed drone technology is among the world's leading technologies, and drones are widely used in the inspection of various accidents and malfunctions. Using drones in conjunction with cameras, infrared thermal imagers, or lidar to capture images of various parts of wind turbines and construct 3D images for fault inspection can effectively improve the inspection efficiency of deep-sea wind farms and reduce the workload of inspection personnel. However, how to intelligently use 3D images to identify typical faults in wind farms still requires further in-depth research.

[0003] This invention proposes a method for locating and screening fault areas of wind turbines for three-dimensional image recognition. Based on existing three-dimensional images, it aims to locate and screen some suspected fault areas by utilizing information such as point coordinates, visible light color, thermal imaging color, and area. This helps to build a typical fault model of deep-sea wind farms, intelligently identify surface faults of wind turbine units, and promote the development of intelligent inspection of deep-sea wind farms. Summary of the Invention

[0004] This invention proposes a method for locating and filtering wind turbine fault areas using 3D image recognition. It utilizes the point coordinates, visible light color, and thermal imaging color information in existing 3D images, and improves the Canny operator to locate fault areas and construct color filtering rules to filter fault areas. This method can help build typical fault models of deep-sea wind farms and promote the intelligent inspection of deep-sea wind farms.

[0005] To achieve the above effects, this invention discloses a method for locating and screening fault areas in wind turbines using three-dimensional image recognition, specifically including the following:

[0006] S1. Fault Area Location: Edge detection algorithms are used to detect the edges of the fan surface, and the change in the number of detected edges (num) is used as the basis for the location. con and the change in the number of edge points num p Determine the parameters required for the edge detection algorithm; based on the center coordinates of each edge, determine the location of all edges, and consider the area enclosed by these edges and its vicinity as suspected fault areas;

[0007] S2. Fault Area Filtering: Based on the HSV color space, construct fault visible light color filtering rules to filter out some suspected fault areas; and retain all suspected fault areas formed by thermal imaging colors.

[0008] Furthermore, the edge detection algorithm in S1 is the Canny operator, and the method for determining the parameters required for the edge detection algorithm is as follows: Let L canny =L M T canny =2*L canny Compare the results with the previous edge detection, and calculate the num. con and num p If condition 1 is satisfied: num con >M con ornum p >M p orL canny <L N Output L canny =L canny +L gap Otherwise L canny =L canny -L gap And count num again con and num p Until condition 1 is met. Where, T canny and L canny The high and low threshold parameters required for the Canny operator; L M For parameter L canny The initial value, L gap L is the step value. N M is the termination value. p For num p The threshold, M con For num con The threshold values ​​for the above parameters are related to the size of the 3D image.

[0009] Furthermore, the fault visible light color screening rule in S2 is as follows: First, based on the brightness V of the suspected fault area... fault Make a judgment if V fault >V t If the area is considered non-faulty, then the suspected faulty area is filtered out; secondly, based on the faulty chromaticity H... fault Make a judgment if H fault Satisfies: 30 - 0.5 * temp ≤ H fault ≤90+1.5*tempor210-0.5*temp≤H fault If V ≤ 270 + 1.5 * temp, then the region is considered a fault region. t is the brightness threshold, and temp is the chromaticity adjustment parameter. The value of temp is related to the brightness of the 3D image.

[0010] Beneficial Effects: This invention provides a method for locating and filtering fault areas in wind turbines using 3D image recognition. The fault area location method adjusts the parameters of the Canny operator based on the results of edge detection, enabling the identification of edges formed by color differences between the fault and surrounding areas while avoiding excessive noise, thus locating the fault area. The fault area filtering method uses HSV-based color filtering rules to determine whether a fault exists within a suspected fault area, and the temp parameter eliminates the influence of image brightness, thereby improving the accuracy of image recognition. The proposed method utilizes existing 3D image information to locate and filter fault areas on the surface of wind turbines, helping to construct typical fault models for deep-sea wind farms, improving the accuracy of the models, and thus increasing the efficiency of inspections of deep-sea wind farms. Attached Figure Description

[0011] Figure 1 A flowchart for locating the fault area.

[0012] Figure 2 A flowchart for filtering faulty areas.

[0013] Figure 3 This is a two-dimensional representation of the results of implementing the present invention.

[0014] Figure 4 This is a flowchart of the steps of the present invention. Detailed Implementation

[0015] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, other embodiments obtained by those skilled in the art without creative effort are all within the scope of protection of the present invention.

[0016] like Figure 1-4 As shown, this invention provides a method for locating and filtering wind turbine fault areas for three-dimensional image recognition. The complete steps of this method for three-dimensional image recognition are as follows:

[0017] Step 1: Fault Area Location: Use the Canny operator to detect the edges of the fan surface, and calculate the change in the number of detected edges (num). con and the change in the number of edge points num p The method for determining the parameters required for the Canny operator and the edge detection algorithm is as follows: Let L canny =L M T canny =2*L canny Compare the results with the previous edge detection, and calculate the num.con and num p If condition 1 is satisfied: num con >M con ornum p >M p orL canny <L N Output L canny =L canny +L gap Otherwise L canny =L canny -L gap And count num again con and num p Until condition 1 is met, the specific steps are as follows: Figure 1 As shown. Based on the center coordinates of each edge, the location of all edges is determined, and the area enclosed by these edges and their vicinity is considered a suspected fault area;

[0018] Step Two: Fault Area Filtering: Based on the HSV color space, construct fault visible light color filtering rules to filter out some suspected fault areas. The fault visible light color filtering rules are as follows: First, based on the brightness V of the suspected fault area... fault Make a judgment if V fault >V t If the area is considered non-faulty, then the suspected faulty area is filtered out; secondly, based on the faulty chromaticity H... fault Make a judgment if H fault Satisfies: 30 - 0.5 * temp ≤ H fault ≤90+1.5*tempor210-0.5*temp≤H fault If the value is ≤270+1.5*temp, then the area is considered a fault area. The specific steps are as follows: Figure 2 As shown. All suspected fault areas identified by thermal imaging colors are retained;

[0019] Step 3: Fault Shape Judgment and Screening: Extract the remaining suspected fault areas in the 3D state space. Use the shape factor and the ratio of major to minor axis to determine the fault shape. Fault shapes are categorized into three types: circular, elongated, and regular. The shape factor P... region The calculation formula is: The ratio of major to minor axis is the ratio of the longest line segment in the suspected fault area to the length of its perpendicular line segment. Regularly shaped suspected fault areas are filtered out, and the remaining areas are the fault areas. The area of ​​each fault area is recorded.

[0020] Among them, C region The perimeter of the suspected fault area, S region The area is the region suspected of being faulty.

[0021] Step 4: Fault Category and Severity Determination: Determine the fault category based on the visible light color and shape of the fault area. The determination method is as follows: The severity of the fault is determined based on the thermal imaging color and area of ​​the fault region. The determination method is as follows:

[0022] Among them, V dif V represents the difference in visible light brightness between the fault area and the surrounding area. g V is the threshold for visible light brightness difference. g The value of is related to the brightness of the three-dimensional state space; R dif S represents the difference in thermal imaging brightness between the fault area and the surrounding area. R R is the area of ​​the fault region. g R is the threshold for thermal imaging brightness difference. g The value of S is related to the brightness of the three-dimensional state space. g S is the threshold value for the area of ​​the fault region. g The value of is related to the size of the three-dimensional state space.

[0023] Finally, the fault category determination result will be marked in the area where the fault is located in the 3D image. Figure 3 The results of the fault category labeling are displayed in a two-dimensional format.

[0024] This invention, based on existing 3D images, utilizes information such as point coordinates, visible light color, thermal imaging color, and area to locate and filter suspected fault areas. This helps construct typical fault models for deep-sea wind farms, intelligently identifying surface faults in wind turbines and promoting the intelligent development of deep-sea wind farm inspections. The fault area location method uses the results of Canny operator edge detection to adjust the parameters of the Canny operator, finding edges formed by color differences between the fault and surrounding areas while avoiding excessive noise, thus locating the fault area. The fault area filtering method is based on HSV-designed color filtering rules. These rules determine whether a fault exists within a suspected fault area by color, and the temp parameter eliminates the influence of image brightness, thereby improving image recognition accuracy. The proposed method utilizes existing 3D image information to locate and filter surface fault areas of wind turbines, helping to construct typical fault models for deep-sea wind farms, improving model accuracy, and thus increasing the efficiency of deep-sea wind farm inspections.

[0025] The above provides a detailed description of the wind turbine fault area location and screening method for three-dimensional image recognition provided by the present invention. The description of the above embodiments is only for the purpose of helping to understand the method and core ideas of the present invention. At the same time, for those skilled in the art, there will be changes in the actual implementation and application scope according to the ideas of the present invention. Therefore, the content of this specification should not be construed as a limitation of the present invention.

Claims

1. A method for locating and screening fault areas in wind turbines using three-dimensional image recognition, characterized in that, The three-dimensional image is constructed from laser point clouds, visible light images, and infrared thermal images. The three-dimensional image contains the coordinates, visible light color, and thermal imaging color information of various points on the wind turbine surface. The specific steps are as follows: S1. Fault Area Location: The edges of the fan surface are detected using the Canny operator, and the change in the number of detected edges is calculated based on the variable num. con and the change in the number of edge points num p Determine the parameters required for the edge detection algorithm, and define the area enclosed by these edges and its vicinity as the suspected fault area; S2. Fault Area Filtering: Based on the HSV color space, construct fault visible light color filtering rules to filter out some suspected fault areas and retain all suspected fault areas formed by thermal imaging colors. S3. Fault shape judgment and screening: Extract the remaining suspected fault areas in the three-dimensional state space, use the shape factor and the ratio of major and minor axes to judge the shape of the fault, screen out the suspected fault areas with regular shapes, and finally the remaining areas are the fault areas, and record the area of ​​each fault area. S4. Fault Category and Severity Judgment: Determine the fault category based on the visible light color and shape of the fault area; The severity of the fault is determined by the color and area of ​​the thermal imaging of the fault area; S5. The final fault category determination result is marked in the area where the fault is located in the 3D image; The edge detection algorithm in S1 is the Canny operator. The method for determining the parameters required for the edge detection algorithm is as follows: Let L canny =L M T canny =2*L canny Compare the results with the previous edge detection, and calculate the num. con and num p If condition 1 is satisfied: num con >M con or num p >M p or L canny <L N Output L canny =L canny +L gap Otherwise L canny =L canny -L gap And count num again con and num p Until condition 1 is met; where T canny and L canny The high and low threshold parameters required for the Canny operator; L M For parameter L canny The initial value, L gap L is the step value. N M is the termination value. p For num p The threshold, M con For num con The threshold values ​​of the above parameters are related to the size of the 3D image. The visible light color selection rule is as follows: First, based on the brightness V of the suspected fault area... fault Make a judgment if V fault >V t Then discard the suspected faulty area; secondly, based on the fault's chromaticity H... fault Make a judgment if H fault Satisfies: 30 - 0.5 * temp ≤ H fault ≤90+1.5*temp or 210-0.5*temp≤H fault If V ≤ 270 + 1.5 * temp, then the suspected faulty region is retained; where V t Here, temp is the brightness threshold, and temp is the chromaticity adjustment parameter; The shapes of the faults include three categories: circular, elongated, and regular, with a shape factor P. region The calculation formula is: The ratio of the major to the minor diameter is the ratio of the length of the longest line segment within the suspected fault area to the length of the line segment perpendicular to it; where C region The perimeter of the suspected fault area, S region The area is the region suspected of being faulty.

2. The method for locating and screening wind turbine fault areas for three-dimensional image recognition according to claim 1, characterized in that, The fault area localization algorithm detects the visible light and infrared thermal colors of the three-dimensional image separately each time the parameters are adjusted, and then merges the detection results of the two.

3. The method for locating and screening wind turbine fault areas for three-dimensional image recognition according to claim 1, characterized in that, The type of fault is determined based on the visible light color and shape of the fault area. The determination method is as follows: Among them, V dif V represents the difference in visible light brightness between the fault area and the surrounding area. g The threshold value is the visible light brightness difference threshold.

4. The method for locating and screening wind turbine fault areas for three-dimensional image recognition according to claim 3, characterized in that, The severity of the fault is determined based on the thermal imaging color and area of ​​the fault region. The determination method is as follows: Among them, R dif S represents the difference in thermal imaging brightness between the fault area and the surrounding area. R R is the area of ​​the fault region. g S is the threshold for thermal imaging brightness difference. g This is the threshold value for the area of ​​the fault region.

Images