CN-121982472-A - Wind power blade damage intelligent identification method and system based on image identification

CN121982472ACN 121982472 ACN121982472 ACN 121982472ACN-121982472-A

Abstract

The application provides an intelligent identification method and system for wind power blade damage based on image identification, and belongs to the technical field of image identification. According to the method, the image acquisition parameters are dynamically determined through the ambient illumination intensity and the local weather data, the surface image of the wind power blade is obtained based on the image acquisition parameters, the surface image is preprocessed and enhanced by means of the convolutional neural network, the preset pixel density threshold value judgment and the edge detection algorithm are combined, interference factors can be effectively filtered, potential damage edges can be accurately distinguished and positioned, the real damage area is determined according to the potential damage edges, and the accuracy of damage identification is remarkably improved. The accurate classification of the damage types is realized through the K-means clustering algorithm, the damage depth is quantified by matching with the depth estimation algorithm, so that the damage severity score is determined, and the repair priorities of the real damage areas are determined by sequentially sequencing the damage severity score from high to low, so that the accurate quantitative evaluation of the damage severity is realized, and serious damage caused by missed judgment and misjudgment is avoided.

Inventors

WANG FENG
WANG RONGGANG
LING RAN
LUO SHUAI
LUO QINGWEI
WU KUN
LIU FENGXIANG

Assignees

三峡能源(册亨)发电有限公司
中国长江三峡集团有限公司贵州分公司

Dates

Publication Date: 20260505
Application Date: 20251205

Claims (9)

1. An intelligent identification method for wind power blade damage based on image identification is characterized by comprising the following steps: acquiring illumination change indexes of a damaged area of the wind power blade, wherein the illumination change indexes comprise environmental illumination intensity and local weather data; Determining image acquisition parameters according to the ambient illumination intensity and the local weather data; Acquiring a surface image of the wind power blade according to the image acquisition parameters; Preprocessing the surface image based on a convolutional neural network to determine an enhanced wind power blade image; judging whether the pixel density of the enhanced wind power blade image exceeds a preset pixel density threshold value; if the pixel density threshold value exceeds the preset pixel density threshold value, identifying potential damaged edges of the wind power blade based on an edge detection algorithm; determining a candidate damage region according to the potential damage edge; determining a real damage area according to the candidate damage area; grouping the real damage areas based on a K-means clustering algorithm to determine different damage type groupings, wherein the damage type groupings comprise blade surface cracks, blade surface wear and blade surface pits; Calculating the real damage area corresponding to the damage type group; Determining a depth estimation value corresponding to the damage type group according to the real damage area based on a depth estimation algorithm; Judging whether the depth estimation value exceeds a preset depth estimation value threshold corresponding to the damage type; if the damage type is judged to be more than the preset depth estimation value threshold value corresponding to the damage type, determining a damage severity score corresponding to the damage type group according to the real damage area and the depth estimation value; And sequentially sorting the damage severity scores according to the sequence from high to low, and determining the repair priority of the real damage area according to the sorting result.
2. The intelligent identification method for wind power blade damage based on image identification according to claim 1, wherein the determining the image acquisition parameters according to the ambient illumination intensity and the local weather data comprises: determining an illumination variance according to the ambient illumination intensity based on a sliding window algorithm; and based on a fuzzy control algorithm, determining exposure time and focal length according to the local weather data and the illumination variance, and taking the exposure time and the focal length as image acquisition parameters.
3. The intelligent identification method for wind power blade damage based on image identification according to claim 1, wherein preprocessing the surface image based on a convolutional neural network to determine an enhanced wind power blade image comprises: Denoising the surface image based on an adaptive filtering algorithm to generate a first surface image; acquiring illumination brightness data of the surface of the wind power blade, wherein the illumination brightness data comprises brightness of a strong light area and brightness of a shadow area; Determining a global brightness average value and a local brightness standard deviation according to the brightness of the strong light area and the brightness of the shadow area; determining an illumination change coefficient according to the global brightness mean value and the local brightness standard deviation; determining the size of a filtering kernel corresponding to the illumination change coefficient according to the illumination change coefficient; Based on a convolutional neural network, performing convolutional operation on the first surface image according to the size of the filter kernel to generate a second surface image; And performing contrast enhancement processing on the second surface image to generate a third surface image, and taking the third surface image as an enhanced wind power blade image.
4. The intelligent identification method for wind power blade damage based on image identification according to claim 1, wherein the determining a candidate damage region according to the potential damage edge comprises: Determining a first edge set according to the potential damage edges; Screening edges which do not exceed a preset Euclidean distance threshold between adjacent edges in the first edge set, and generating a second edge set; screening edges exceeding a preset gradient strength threshold value in the second edge set, and generating a third edge set; Performing local histogram equalization processing on the third edge set based on an illumination self-adaptive adjustment algorithm to generate a fourth edge set; performing noise filtering processing on the fourth edge set to generate a fifth edge set; and determining candidate damage areas according to the fifth edge set based on an area growth algorithm.
5. The intelligent identification method for wind power blade damage based on image identification according to claim 1, wherein the determining a real damage area according to the candidate damage area comprises: Determining texture feature vectors of the corresponding candidate damage areas according to the candidate damage areas based on a gray level co-occurrence matrix algorithm; Determining a color distribution histogram corresponding to the candidate damage region according to the texture feature vector; judging whether the color distribution similarity between the color distribution histogram and a natural texture color histogram of a preset healthy wind power blade exceeds a preset similarity threshold value; If the candidate damage area exceeds the preset similarity threshold, determining the current candidate damage area as a real damage area; and if the preset similarity threshold is not exceeded, eliminating the current candidate damage area.
6. The method for intelligently identifying damage to a wind turbine blade based on image recognition of claim 1, wherein the determining a damage severity score corresponding to the damage type group comprises: respectively obtaining a weight corresponding to the real damage area and a weight corresponding to the depth estimation value; Weighting the real damage area and the depth estimation value based on the weight corresponding to the real damage area and the weight corresponding to the depth estimation value; and scoring the weighted operation result as the damage severity degree corresponding to the damage type group.
7. The intelligent identification method for wind power blade damage based on image identification according to claim 1, wherein after the damage severity score is ranked in order from high to low, and the repair priority of the real damage area is determined according to the ranking result, the method further comprises: judging whether the damage severity score exceeds a preset score threshold; if the judgment result exceeds the preset scoring threshold, determining the boundary coordinates of the real damage area according to the real damage area; Based on a coordinate mapping algorithm, projecting the boundary coordinates of the real damage area to a three-dimensional model of the wind power blade to determine projection point coordinates, wherein the origin of a three-dimensional model coordinate system is the root of the wind power blade; Respectively calculating Euclidean distance between the projection point coordinates and the root coordinates and tip coordinates of the wind power blade; And determining a damage positioning report of the wind power blade according to the boundary coordinates of the real damage area, the Euclidean distance between the projection point coordinates and the root coordinates of the wind power blade and the Euclidean distance between the projection point coordinates and the tip coordinates of the wind power blade.
8. The intelligent identification method for wind power blade damage based on image identification according to claim 1, wherein after the damage severity score is ranked in order from high to low, and the repair priority of the real damage area is determined according to the ranking result, the method further comprises: Determining a damaged area with the highest repair priority according to the repair priority of the real damaged area; Determining the shortest repair path according to the damage area with the highest repair priority based on a path optimization algorithm; and determining a repairing scheme of the wind power blade according to the repairing path.
9. Wind-powered electricity generation blade damage intelligent identification system based on image recognition, characterized by comprising: the first acquisition module is used for acquiring illumination change indexes of the damaged areas of the wind power blades, wherein the illumination change indexes comprise environmental illumination intensity and local weather data; the first determining module is used for determining image acquisition parameters according to the ambient illumination intensity and the local weather data; The second acquisition module is used for acquiring the surface image of the wind power blade according to the image acquisition parameters; The second determining module is used for preprocessing the surface image based on the convolutional neural network to determine an enhanced wind power blade image; The first judging module is used for judging whether the pixel density of the enhanced wind power blade image exceeds a preset pixel density threshold value; The identification module is used for identifying the potential damaged edge of the wind power blade based on an edge detection algorithm if the preset pixel density threshold value is judged to be exceeded; A third determining module, configured to determine a candidate damage area according to the potential damage edge; a fourth determining module, configured to determine a real damage area according to the candidate damage area; The grouping module is used for grouping the real damage areas based on a K-means clustering algorithm to determine different damage type groupings, wherein the damage type groupings comprise blade surface cracks, blade surface abrasion and blade surface depressions; the operation module is used for calculating the real damage area corresponding to the damage type group; A fifth determining module, configured to determine, based on a depth estimation algorithm, a depth estimation value corresponding to the damage type group according to the real damage area; the second judging module is used for judging whether the depth estimation value exceeds a preset depth estimation value threshold corresponding to the damage type; A sixth determining module, configured to determine, if the predetermined depth estimation value threshold value corresponding to the damage type is exceeded, a damage severity score corresponding to the damage type packet according to the real damage area and the depth estimation value; and a seventh determining module, configured to sequentially sort the damage severity scores according to a sequence from high to low, and determine a repair priority of the real damage area according to a sorting result.

Description

Wind power blade damage intelligent identification method and system based on image identification Technical Field The application relates to the technical field of image recognition, in particular to an intelligent wind power blade damage recognition method and system based on image recognition. Background At present, wind power generation is used as an important strut of clean energy, and stable operation of the wind power generation is important for energy transformation and environmental protection in China. The wind power blade is used as a core component of wind power generation and directly bears wind load and climate influence in a complex environment, and timely discovery and treatment of surface damage of the wind power blade have decisive significance for guaranteeing service life and power generation efficiency of wind power equipment. The existing wind power blade damage detection method relies on manual inspection or fixed equipment to collect images, is difficult to adapt to complex and changeable illumination conditions, so that image collection parameters cannot be dynamically optimized, and the wind power blade image collection quality is reduced. In addition, in the collected wind power blade image, the damage area of the wind power blade image cannot be accurately distinguished and positioned from the complex background, and accuracy of wind power blade damage identification is reduced. The above information disclosed in the background section is only for enhancement of understanding of the background of the disclosure and therefore it may include information that does not form the prior art that is already known to a person of ordinary skill in the art. Disclosure of Invention In view of the above, the application provides an intelligent identification method for wind power blade damage based on image identification, which can improve the identification precision of wind power blade damage. In a first aspect, an embodiment of the application provides an intelligent identification method for wind power blade damage based on image identification, which comprises the steps of acquiring illumination change indexes of a wind power blade damage area, wherein the illumination change indexes comprise environment illumination intensity and local weather data, determining image acquisition parameters according to the environment illumination intensity and the local weather data, acquiring a surface image of a wind power blade according to the image acquisition parameters, preprocessing the surface image based on a convolutional neural network to determine an enhanced wind power blade image, judging whether pixel density of the enhanced wind power blade image exceeds a preset pixel density threshold value, if so, identifying potential damage edges of the wind power blade based on an edge detection algorithm, determining candidate damage areas according to the potential damage edges, determining real damage areas according to the candidate damage areas, grouping the real damage areas based on a K-means clustering algorithm to determine different damage type groups, wherein the damage type groups comprise blade surfaces, blade surface pits and blade surface pits, calculating real damage area areas corresponding to the damage type groups, estimating depth estimation algorithm based on depth estimation algorithm, judging whether the depth estimation algorithm corresponds to the estimated damage area to the estimated depth estimation damage area or not to the estimated depth estimation type crack area, and judging whether the depth estimation algorithm corresponds to the estimated depth estimation type damage area to the estimated depth estimation depth value exceeds the estimated depth type damage threshold value or not to the estimated depth estimation value, and sequentially sorting the damage severity scores according to the order from high to low, and determining the repair priority of the real damage region according to the sorting result. In a second aspect, the embodiment of the application provides an intelligent wind power blade damage identification system based on image identification, which comprises a first determination module, a second acquisition module, a second determination module, a first judgment module, an identification module, a third determination module, a fourth determination module, a grouping module, an operation module, a fifth determination module, a second judgment module, a sixth determination module and a seventh determination module. The first acquisition module is used for acquiring illumination change indexes of the damaged areas of the wind power blades, wherein the illumination change indexes comprise environment illumination intensity and local weather data; the first determining module is used for determining image acquisition parameters according to the ambient illumination intensity and the local weather data; the second acquisition module is used for acquiri