CN111581409A - Damage image feature database construction method and system and engine - Google Patents

Abstract

The invention belongs to the technical field of database construction, and discloses a damage image feature database construction method, a construction system and a civil aviation engine. Damage images of different structures inside the civil aviation engine are collected through non-destructive testing technology; digital image processing technology is used to analyze the collected engine damage images Perform preprocessing operations of noise reduction, enhancement and segmentation; extract digital features of damage images based on color moment features and grayscale co-occurrence matrix texture features; build engine damage image feature database based on digital features of damage images; use engine damage image feature database to identify unknown damage images, and display the recognition results; at the same time, the new damage features are included in the database. The feature extraction method based on the color moment and gray level co-occurrence matrix texture features proposed by the present invention is more conducive to describe the damage image of the engine, prepares to express the damage feature of the engine, and provides a reasonable and effective damage image feature database.

Description

A damage image feature database construction method, construction system and engine

technical field

The invention belongs to the technical field of database construction, and in particular relates to a construction method, construction system and engine of a damage image feature database.

Background technique

At present, the safe operation of civil aviation aircraft is directly related to the safety of life and property of passengers, and ensuring the safe flight of civil aviation aircraft is the lifeline of the civil aviation industry. As a highly integrated and sophisticated complex industrial product, the engine provides sufficient power for aircraft operation and is a key system to ensure flight safety. According to the statistics of the global civil aviation industry, flight accidents caused by engines account for about 50%; and engine maintenance expenditures account for about 40% of all expenditure costs. Therefore, it is of great significance to carry out efficient and accurate decision-making research on civil aviation engine maintenance to ensure flight safety, reduce maintenance costs, and improve operational efficiency.

Civil aviation engines have been operating in a harsh environment of high temperature, high pressure and high load for a long time. Key components such as wheel discs, blades, turbines, and fuel nozzles at all levels are easily subjected to various impact loads, resulting in cracks, corrosion, tearing, burns, and falling off. Blocks and other types of damage. This not only causes the performance of the engine to decline, but also easily causes the engine to fail, and even threatens flight safety in severe cases, resulting in serious consequences. With the help of digital image processing technology, operations such as noise reduction, enhancement, and segmentation are performed on the damaged image, which can enhance the image contrast and improve the accuracy of damage identification. However, the existing damage identification based on digital image processing technology still relies on expert experience. However, the engine structure is complex, and the damage types and features are numerous. Traditional methods relying on expert experience are increasingly difficult to accurately identify damage types. At the same time, there is no report on constructing a feature database of engine damage images in the prior art.

Through the above analysis, the problems and defects of the existing technology are as follows: (1) The existing damage identification based on digital image processing technology still relies on expert experience, but the engine structure is complex, and the damage types and characteristics are numerous. Methods are increasingly incapable of accurate identification of injury types.

(2) In the prior art, there is no report on constructing a feature database of engine damage images.

The difficulty of solving the above problems and defects is as follows:

The engine's graphic loss type and the cause of the loss are responsible, and the graphic image features are diverse, so it is difficult to accurately extract the graphic features of the loss mode.

The significance of solving the above problems and defects is that the problems existing in the extraction of engine damage graphic features, such as vehicle number plate recognition, are well solved.

SUMMARY OF THE INVENTION

Aiming at the problems existing in the prior art, the present invention provides a damage image feature database construction method, construction system and engine, in particular to a damage image feature database construction method based on color features and texture features.

The present invention is achieved in this way, a method for constructing a damage image feature database, comprising the following steps:

The first step is to collect damage images of different structures inside the civil aviation engine through non-destructive testing technology.

Step 2, using digital image processing technology to perform preprocessing operations of noise reduction, enhancement and segmentation on the collected engine damage images.

The third step is to extract the digital features of the damaged image based on the color moment feature and the gray level co-occurrence matrix texture feature.

Step 4: Build an engine damage image feature database according to the digital features of the damage image.

Step 5: Identify the unknown damage image by using the engine damage image feature database, and display the identification result; at the same time, classify the new damage feature into the database.

Further, in step 3, the method for extracting the damage image features includes:

(1) Grayscale processing is performed on the damaged image to be detected.

(2) Perform color moment feature extraction to obtain first-order moment, second-order moment and third-order moment.

(3) Extract the features of the gray-level co-occurrence matrix, and characterize the texture features reflected by the gray-level co-occurrence matrix through energy, contrast, correlation, entropy and inverse difference moment.

Further, the step (1) of performing grayscale processing on the damaged image to be detected includes:

Let R, G, and B be the red, green, and blue component matrices of the engine damage image. According to the sensitivity of the human eye to different color systems, the grayscale of the image is obtained by the weighted average of the three components of R, G, and B. Its calculation expression is as follows:

f(i,j)=0.30·R(i,j)+0.59·G(i,j)+0.11·B(i,j) (1)

where f represents the grayscale matrix of the image.

Further, the method for performing color moment feature extraction in the step (2) includes:

After obtaining the grayscale damage image, the statistical characteristics of the color are described by the calculated moments. Usually, the color distribution information is mainly concentrated in the low-order moments, so the first three-order moments are used to indicate the color distribution of the image. The calculation formula is as follows:

Among them, N is the number of pixels of the matrix; p _ij is the pixel point of the i-th row and j column; μ is the first-order moment, also called the mean value of the matrix, which represents the average intensity of the grayscale image; σ is the second-order moment, also called the matrix The variance of , reflects the inhomogeneity of the grayscale image; while ζ is the third-order moment, also called the skewness of the matrix, which defines the asymmetry of the grayscale image.

Further, the method for performing gray level co-occurrence matrix feature extraction in the step (3) includes:

The gray-level co-occurrence matrix GLCM is obtained by statistics on the gray-level image of two pixels with a distance d and a direction of θ respectively having a certain gray level, which reflects the comprehensive information about the direction, interval and amplitude changes of the gray level image. , which is the basis for analyzing local patterns in images. Energy (angular second moment, ASM), contrast (contrast), correlation (correlation), entropy (entropy), inverse different moment (inverse different moment, IDM) are used to characterize the texture features reflected by GLCM. The specific calculation formula is as follows:

in,

The energy reflects the uniformity and texture thickness of the grayscale distribution of the image. When the elements in the GLCM are concentratedly distributed, the ASM has a relatively large value, indicating that the grayscale image is a relatively uniform and regular texture pattern. Contrast reflects the sharpness of the grayscale image. Generally, the deeper the image groove, the larger the CON, and the clearer the visual effect. Correlation measures the similarity of row elements or column elements in GLCM. When the elements are uniform and equal, the COR is larger, reflecting the local grayscale correlation in the image. Entropy is a measure of the random information carried by the image. When the elements in GLCM are dispersedly distributed, the ENT is larger, which indicates the degree of non-uniformity and complexity of the image. The inverse difference moment reflects the roughness of the image texture, the IDM of the coarse texture is larger, and vice versa. So far, 8 color moment features and GLCM texture features of grayscale images have been extracted to describe engine damage images, and then an engine damage image database is constructed to provide sample support for automatic identification.

Another object of the present invention is to provide a damage image feature database construction system including:

The structural damage image acquisition module collects different structural damage images inside the civil aviation engine through non-destructive testing technology;

The engine damage image processing module uses digital image processing technology to perform noise reduction, enhancement and segmentation preprocessing on the collected engine damage images;

Damage image digital feature extraction module, extracts damage image digital features based on color moment feature and gray level co-occurrence matrix texture feature;

The engine damage image feature database building module, which builds the engine damage image feature database according to the digital features of the damage image;

The identification result display module uses the engine damage image feature database to identify unknown damage images, and displays the identification results; at the same time, the new damage features are included in the database.

Another object of the present invention is to provide a program storage medium for receiving user input, and the stored computer program enables an electronic device to execute the damage image feature database construction method.

Another object of the present invention is to provide a computer program product stored on a computer-readable medium, including a computer-readable program that, when executed on an electronic device, provides a user input interface to implement the method for constructing the damage image feature database .

Another object of the present invention is to provide a civil aviation engine that uses the damage image feature database construction method to analyze and process damage information of different internal structures.

Combined with all the above technical solutions, the advantages and positive effects of the present invention are as follows: the present invention uses the color moment and the gray level co-occurrence matrix (GLCM) to extract the digital features of the engine damage image, and according to a certain type Four kinds of damage type images of the engine are constructed according to different feature extraction methods to construct a feature database based on the non-destructive testing images of the engine. The feature extraction method based on color moments and GLCM texture features proposed by the present invention is more beneficial to describe the damage image of the engine, prepare to express the damage feature of the engine, and provide a reasonable and effective damage image feature database.

By means of the non-destructive testing technology, the invention can accurately detect the structural damage generated inside the engine and form image information. Determining the type of engine damage is the key link of engine image analysis technology, which has a guiding role in further judging the damage mechanism, determining the damaged parts, and evaluating the damage degree. With the help of digital image processing technology, operations such as noise reduction, enhancement, and segmentation are performed on the damaged image, which can enhance the image contrast and improve the accuracy of damage identification.

Description of drawings

FIG. 1 is a flowchart of a method for constructing a damage image feature database based on color features and texture features provided by an embodiment of the present invention.

FIG. 2 is a schematic diagram of an engine damage image recognition process provided by an embodiment of the present invention.

FIG. 3 is a flowchart of image feature extraction provided by an embodiment of the present invention.

FIG. 4 is a schematic diagram of a damage image of a certain type of engine provided by an embodiment of the present invention.

Detailed ways

In order to make the objectives, technical solutions and advantages of the present invention clearer, the present invention will be further described in detail below with reference to the embodiments. It should be understood that the specific embodiments described herein are only used to explain the present invention, but not to limit the present invention.

In view of the problems existing in the prior art, the present invention provides a method for constructing a damage image feature database based on color features and texture features. The present invention is described in detail below with reference to the accompanying drawings.

As shown in FIG. 1 , the method for constructing a damage image feature database based on color features and texture features provided by an embodiment of the present invention includes the following steps:

In S101, the damage images of different structures inside the civil aviation engine are collected through non-destructive testing technology.

S102, using digital image processing technology to perform preprocessing operations of noise reduction, enhancement and segmentation on the collected engine damage image.

S103, extracting digital features of the damaged image based on the color moment feature and the gray level co-occurrence matrix texture feature.

S104, build an engine damage image feature database according to the digital features of the damage image.

S105 , use the engine damage image feature database to identify the unknown damage image, and display the identification result; at the same time, classify the new damage feature into the database.

As shown in Figure 2, the present invention provides a damage image feature database construction system including:

The structural damage image acquisition module collects different structural damage images inside the civil aviation engine through non-destructive testing technology.

The engine damage image processing module uses digital image processing technology to perform noise reduction, enhancement and segmentation preprocessing on the collected engine damage images.

Damage image digital feature extraction module, based on color moment feature and gray level co-occurrence matrix texture feature to extract digital feature of damage image.

The engine damage image feature database building module builds the engine damage image feature database according to the digital features of the damage image.

The identification result display module uses the engine damage image feature database to identify unknown damage images, and displays the identification results; at the same time, the new damage features are included in the database.

Figure 2 shows the identification process for the type of engine damage. After obtaining different damage images using NDT, the key is to form a digital image feature representation to build an engine damage database and provide original samples for training SVM.

The present invention is further described below in conjunction with a damage database based on color features and texture features.

1 Feature extraction based on color moment and gray level co-occurrence matrix GLCM

The analysis and extraction of low-level features such as color, texture, and shape of images has become a widely used method in industry and academia, and has formed relatively rich research results. Literature [1] proposed a color feature extraction method based on HSV space. , the literature [2] uses the gray level co-occurrence matrix (gray level co-occurrence matrix, GLCM) statistics to describe the texture features of the image, the literature [3] fuses the GLCM feature and the Tamura feature to obtain a new graphic digital feature. Usually, the color feature of the image has strong robustness, and is the most intuitive and obvious feature to reflect the damaged part of the engine. The texture feature reflects the structural characteristics of the surface of the object and the surrounding environment information, which is the attribute of image grayscale. According to the damage image characteristics of non-destructive testing, the present invention proposes an image feature extraction method based on color moment feature and GLCM texture feature. Figure 3 shows a flowchart of image feature extraction.

Let R, G, and B be the red, green, and blue component matrices of the engine damage image. According to the sensitivity of the human eye to different color systems, the grayscale of the image is obtained by the weighted average of the three components of R, G, and B. Its calculation expression is as follows:

f(i,j)=0.30·R(i,j)+0.59·G(i,j)+0.11·B(i,j) (1)

where f represents the grayscale matrix of the image. After obtaining the grayscale damage image, the statistical characteristics of the color are described by the calculated moments. Usually, the color distribution information is mainly concentrated in the low-order moments, so the first three-order moments are used to indicate the color distribution of the image. The calculation formula is as follows:

Among them, N is the number of pixels of the matrix; p _ij is the pixel point of the i-th row and j column; μ is the first-order moment, also called the mean value of the matrix, which represents the average intensity of the grayscale image; σ is the second-order moment, also called the matrix The variance of , reflects the inhomogeneity of the grayscale image; while ζ is the third-order moment, also called the skewness of the matrix, which defines the asymmetry of the grayscale image. Gray level co-occurrence matrix (GLCM) refers to a common method to describe texture by studying the spatial correlation characteristics of gray level, which was proposed by R. Haralick et al. in the early 1970s. GLCM is obtained by statistics on the grayscale image of two pixels with a distance d and a direction of θ respectively having a certain grayscale. It reflects the comprehensive information about the direction, interval and amplitude changes of the grayscale image. The basis of local mode. For the definition and calculation process of GLCM, please refer to the literature [3,4]. In order to describe the image texture features with GLCM more intuitively, the present invention uses energy (angular second moment, ASM), contrast (contrast), correlation (correlation), entropy (entropy), inverse difference moment (inverse different moment, IDM) to To characterize the texture features reflected by GLCM, the specific calculation formula is as follows:

in,

The energy reflects the uniformity and texture thickness of the grayscale distribution of the image. When the elements in the GLCM are concentratedly distributed, the ASM has a relatively large value, indicating that the grayscale image is a relatively uniform and regular texture pattern. Contrast reflects the sharpness of the grayscale image. Generally, the deeper the image groove, the larger the CON, and the clearer the visual effect. Correlation measures the similarity of row elements or column elements in GLCM. When the elements are uniform and equal, the COR is larger, reflecting the local grayscale correlation in the image. Entropy is a measure of the random information carried by the image. When the elements in GLCM are dispersedly distributed, the ENT is larger, which indicates the degree of non-uniformity and complexity of the image. The inverse difference moment reflects the roughness of the image texture, the IDM of the coarse texture is larger, and vice versa. So far, 8 color moment features and GLCM texture features of grayscale images have been extracted to describe engine damage images, and then an engine damage image database is constructed to provide sample support for automatic identification.

2 Construction of engine damage image feature database

Collect 4 types of damage images of a certain type of civil aviation engine in the process of non-destructive testing. Figure 4 shows one of the damage images in each type.

Figure 4(a) shows the burn-through damage of the fuel nozzle baffle of the engine, Figure 4(b) shows the ablation damage of the high pressure turbine blade (HPT blade), and Figure 4(c) shows the high pressure turbine blade (HPT blade). The trailing edge of the compressor blade (high pressure compressor blade trailing edge, HPC blade TE) is agglomerated and cracked. Figure 4(d) shows the perforation of the combustion chamber. According to the aforementioned feature extraction method, a test database of civil aviation engine damage image features is constructed. The basic features of the database are shown in the table. 339 samples are randomly selected as training samples, and the remaining 86 samples are used as test samples, so as to test the identification proposed by the present invention. performance of the algorithm. As shown in Table 1 and 2.

Table 1 Civil Aviation Engine Damage Image Feature Database

part damage type number of training samples Test sample data fuel nozzle baffle burn through 78 20 HPT blade ablation 90 twenty three HPC blade TE drop block 78 20 combustion chamber perforation 93 twenty three

Table 2 Partial image feature templates

The feature extraction method based on color moments and GLCM texture features proposed by the present invention is more beneficial to describe the damage image of the engine, prepare to express the damage feature of the engine, and provide a reasonable and effective damage image feature database.

The present invention will be further described below in conjunction with the verification of the damage image recognition performance.

In order to verify the feature extraction effect based on color moment and GLCM proposed by the present invention, and at the same time verify the damage identification performance of SVM (AMPSO-SVM) based on AMPSO optimization, the present invention will do the following comparative verification.

1 Influence of different feature extraction methods on recognition accuracy

There are many kinds of image feature extraction methods. According to the characteristics of the engine damage image, the feature extraction method proposed by the present invention is compared with the method in the literature [1-3, 5, 6]. The HSV space color proposed in the literature [1] is compared. The feature extraction method, the texture feature extraction method based on GLCM statistics proposed by the literature [2], the feature extraction method based on Tamura was proposed in the literature [5], the feature extraction method based on the fusion of GLCM and Tamura was proposed in the literature [3], and the literature [ 6] A feature extraction method based on Tamura feature and local gray color (GC) feature fusion is proposed. Using AMPSO-SVM as the identification algorithm, the population of PSO is set to 40, the maximum number of iterations is set to 100, the particle search range is [0, 100], and the speed is randomly initialized between [-1, 1], and the k of cross-validation is set to 5 . Using civil aviation engine damage images, image features are extracted according to different methods, and the algorithm is trained and tested. At the same time, common intelligent algorithms based on knowledge learning are introduced, such as BP (backpropagation) network, ELM (extreme learning machines) network, and k-NN (k-nearest neighborhood) algorithm to test the impact of various feature extraction methods on recognition accuracy. The error target of the BP network is set to 0.005, while the number of iterations is set to 300. The ELM network is optimized by the method proposed in [7]. The recognition accuracy of K-NN when k is 1. For the specific calculation process of each algorithm, please refer to the relevant references. Affected by the random initialization of weights, the output accuracy of BP network and ELM network presents uncertainty. Therefore, these two methods are run 50 times continuously under the same computing environment, and the average accuracy is taken as the final output. Table 2 shows the recognition results based on different feature extraction methods, where (C, γ)best represents the output solution after optimization using AMPSO, and the fitness value is the average precision of CV. From the longitudinal observation of the recognition accuracy in Table 2, it can be seen that for different recognition algorithms, the feature extraction method proposed in the present invention has a better recognition effect than other feature extraction methods. Analysis of various feature extraction algorithms shows that: Literatures [1], [2] and [5] belong to a single feature extraction method, and the description of damage images is not comprehensive enough, resulting in relatively poor recognition accuracy; Literatures [3] and [6] The method belongs to the fusion feature extraction method, which depicts the damage image features from multiple dimensions. Therefore, compared with the single feature extraction method, it is relatively more conducive to the damage image recognition, and the reference [6] also considers the color feature, which can identify the damage image. also at relatively high precision. The method based on color moment and GLCM feature extraction proposed by the present invention comprehensively considers color features and texture statistical features, and can extract engine damage image features more objectively and comprehensively. The experimental results prove that the feature extraction method proposed by the present invention is more reasonable and effective. .

2. Comparison of recognition performance of different recognition algorithms

The invention proposes an SVM optimized based on AMPSO, and obtains the optimal (C, γ) through cross-validation, thereby ensuring stable and reliable output of the SVM. In order to verify the advantages of the proposed algorithm in damage image recognition, the above four recognition algorithms are compared with AMPSO-SVM in recognition performance. Looking at the recognition accuracy in Table 2, it is obvious that AMPSO-SVM is basically the best for different feature data recognition.

Table 3 Recognition performance comparison table of feature extraction methods of recognition algorithms

From Table 3, it can be concluded that the recognition ability of AMPSO_SVM recognition method is better than BP, ELM and K-nn. And longitudinal comparison, it can be concluded that the feature extraction method of the present invention is superior to HSV, GLCM, Tamura, GLCM-Tamura and GC-Tamura.

Since the recognition principles of various recognition algorithms are different, it is difficult to ensure that one method is effective for all data distribution types. For example, using the data extracted in literature [5], the recognition accuracy of AMPSO-SVM is not as good as that of ELM network. However, AMPSO-SVM effectively overcomes the influence of randomness, and does not have the defects of output uncertainty such as BP network and ELM network, and at the same time, it is not as sensitive to k value as k-NN. Therefore, AMPSO-SVM not only has good recognition performance, but also provides stable and accurate output, which can provide reliable technical support for the identification of damage types of civil aviation engines.

references:

[1] Yang Aobo, Sheng Jiachuan, Li Yuzhi, Liu Shang, Zhao Kunyuan. Color Feature Extraction Based on HSV Space [J]. Computer Knowledge and Technology, 2017(18):193-195.

[2] Elevation Cheng, Hui Xiaowei. Texture feature extraction based on gray level co-occurrence matrix [J]. Computer System Application, 2010, 19(6): 195-198.

[3] Geng Yanping, Gao Hongbin, Ren Zhiying. Image retrieval algorithm combining color features and texture features [J]. Wireless Internet Technology, 2017, 24: 113-116.

[4]R.M.Haralick,K.Shanmugam,I.Dinstein.Texture features for imageclassification[J].IEEE Transactions on Systems,Man and Cybernetics,1973,SMC-3(6):610-621.

[5] Y. Liu, Z. Li, Z. M. Gao. An Improved Texture Feature Extraction Method for Tyre Tread Patterns [C]. International Conference on Intelligent Science and Big Data Engineering. Springer, Berlin, Heidelberg, 2013: 705-713.

[6] Li Na, Xiong Zhiyong, Xie Jin, Peng Chuan, Ren Kai. Multimodal brain tumor MR image segmentation based on Tamura texture feature extraction and SVM [J]. Journal of South Central University for Nationalities (Natural Science Edition), 2018, 37( 03):148-153.

[7]S.U.Hongjun,S.Tian,Y.Cai,et al.Optimized extreme learning machinefor urban land cover classification using hyperspectral imagery[J].Frontiersof Earth Science,2017,11(4):765-773.

From the description of the above embodiments, those skilled in the art can clearly understand that the present invention can be implemented by means of software plus a necessary hardware platform, and certainly can also be implemented entirely by hardware. Based on this understanding, all or part of the technical solutions of the present invention can be embodied in the form of software products, and the computer software products can be stored in storage media, such as ROM/RAM, magnetic disks, optical disks, etc. , including several instructions for causing a computer device (which may be a personal computer, a server, or a network device, etc.) to perform the methods described in various embodiments or some parts of the embodiments of the present invention.

The above are only specific embodiments of the present invention, but the protection scope of the present invention is not limited to this. Any person skilled in the art is within the technical scope disclosed by the present invention, and all within the spirit and principle of the present invention Any modifications, equivalent replacements and improvements made within the scope of the present invention should be included within the protection scope of the present invention.

Claims (10)

1. A method for constructing a damage image feature database is characterized by comprising the following steps: constructing an engine damage image characteristic database according to the acquired damage image digital characteristics;

identifying unknown damage images by using an engine damage image characteristic database, and displaying identification results; at the same time, the new damage characteristics are included in the database.

2. The method for constructing the database of characteristics of the damage image according to claim 1, wherein the method for extracting the characteristics of the damage image comprises:

(1) carrying out gray processing on a damage image to be detected;

(2) performing color moment feature extraction to obtain a first moment, a second moment and a third moment;

(3) and (4) extracting the characteristics of the gray level co-occurrence matrix, and representing the texture characteristics reflected by the gray level co-occurrence matrix through energy, contrast, correlation, entropy and inverse difference moment.

3. The method for constructing the damage image feature database according to claim 2, wherein in the step (1), the method for performing graying processing on the damage image to be detected comprises:

let R, G, B be the red, green, blue component matrix of the engine damage image, and according to the sensitivity of human eyes to different color systems, the graying of the image is obtained by performing weighted average on R, G, B three components, and the calculation expression is as follows:

f(i,j)＝0.30·R(i,j)+0.59·G(i,j)+0.11·B(i,j)

where f represents the gray matrix of the image.

4. The method for constructing a database of lesion image features according to claim 2, wherein in the step (2), the method for extracting the color moment features comprises:

after obtaining a grayed damage image, describing the statistical characteristics of the color through the calculated moment; the color distribution information is mainly concentrated in low-order moments, the first three-order moments are adopted to indicate the color distribution of the image, and the calculation formula is as follows:

wherein N is the number of pixels of the matrix; p is a radical of_ijThe pixel points of the ith row and the j column are set; mu is a first moment, also called the mean value of the matrix, and represents the average intensity of the gray level image; sigma is a second moment; ζ is the third moment.

5. The method for constructing a damage image feature database according to claim 2, wherein in the step (3), the method for extracting the gray level co-occurrence matrix features comprises:

the gray level co-occurrence matrix GLCM is obtained by counting the condition that two pixels with the distance d and the direction theta on a gray level image respectively have certain gray levels, reflects the comprehensive information of the gray level image about the direction, interval and amplitude change, and is the basis for analyzing the local mode of the image;

the texture characteristics reflected by the GLCM are represented by adopting energy, contrast, correlation, entropy and inverse difference moment, and the specific calculation formula is as follows:

wherein,

6. the method for constructing a database of lesion image features according to claim 1, wherein the method for constructing a database of lesion image features comprises the steps of:

acquiring damage images of different structures in a civil aircraft engine by a nondestructive testing technology;

secondly, preprocessing the acquired engine damage image by noise reduction, enhancement and segmentation by using a digital image processing technology;

extracting digital features of the damaged image based on the color moment features and the gray level co-occurrence matrix texture features;

fourthly, constructing an engine damage image characteristic database according to the digital characteristics of the damage image;

identifying unknown damage images by using an engine damage image characteristic database, and displaying identification results; at the same time, the new damage characteristics are included in the database.

7. A damage image feature database construction system for implementing the damage image feature database construction method according to any one of claims 1 to 6, wherein the damage image feature database construction system comprises:

the structure damage image acquisition module acquires different structure damage images in the civil aircraft engine through a nondestructive testing technology;

the engine damage image processing module is used for carrying out preprocessing of noise reduction, enhancement and segmentation on the acquired engine damage image by using a digital image processing technology;

the damage image digital feature extraction module is used for extracting damage image digital features based on the color moment features and the gray level co-occurrence matrix texture features;

the engine damage image characteristic database construction module is used for constructing an engine damage image characteristic database according to the damage image digital characteristics;

the recognition result display module is used for recognizing the unknown damage image by using the engine damage image characteristic database and displaying the recognition result; at the same time, the new damage characteristics are included in the database.

8. A program storage medium for receiving a user input, the stored computer program causing an electronic device to execute the method for constructing a database of characteristics of a lesion image according to any one of claims 1 to 6.

9. A computer program product stored on a computer readable medium, comprising a computer readable program for providing a user input interface for implementing a method of constructing a database of image characteristics of lesions according to any one of claims 1 to 6 when executed on an electronic device.

10. A civil aviation engine for analyzing and processing damage information of different internal structures by using the damage image feature database construction method of any one of claims 1 to 6.

Images