CN117197700A - Intelligent unmanned inspection contact net defect identification system - Google Patents
Intelligent unmanned inspection contact net defect identification system Download PDFInfo
- Publication number
- CN117197700A CN117197700A CN202311466053.8A CN202311466053A CN117197700A CN 117197700 A CN117197700 A CN 117197700A CN 202311466053 A CN202311466053 A CN 202311466053A CN 117197700 A CN117197700 A CN 117197700A
- Authority
- CN
- China
- Prior art keywords
- contact net
- catenary
- defect recognition
- calculated
- rigid
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Granted
Links
- 230000007547 defect Effects 0.000 title claims abstract description 84
- 238000007689 inspection Methods 0.000 title claims abstract description 36
- 238000000034 method Methods 0.000 claims abstract description 50
- 238000012545 processing Methods 0.000 claims abstract description 29
- 230000004927 fusion Effects 0.000 claims abstract description 16
- 239000011159 matrix material Substances 0.000 claims description 38
- 238000004364 calculation method Methods 0.000 claims description 12
- 230000008859 change Effects 0.000 claims description 10
- 238000001914 filtration Methods 0.000 claims description 6
- 238000005516 engineering process Methods 0.000 abstract description 9
- 238000013473 artificial intelligence Methods 0.000 abstract description 2
- 238000009826 distribution Methods 0.000 description 8
- 238000000605 extraction Methods 0.000 description 5
- 238000004458 analytical method Methods 0.000 description 4
- 230000008569 process Effects 0.000 description 4
- 230000004044 response Effects 0.000 description 4
- 230000008901 benefit Effects 0.000 description 3
- 230000000694 effects Effects 0.000 description 3
- 102000002274 Matrix Metalloproteinases Human genes 0.000 description 2
- 108010000684 Matrix Metalloproteinases Proteins 0.000 description 2
- 230000009286 beneficial effect Effects 0.000 description 2
- 230000005540 biological transmission Effects 0.000 description 2
- 230000002950 deficient Effects 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 230000002708 enhancing effect Effects 0.000 description 2
- 238000009499 grossing Methods 0.000 description 2
- 238000010801 machine learning Methods 0.000 description 2
- 238000010606 normalization Methods 0.000 description 2
- 230000035945 sensitivity Effects 0.000 description 2
- 238000005299 abrasion Methods 0.000 description 1
- 238000013459 approach Methods 0.000 description 1
- 239000002131 composite material Substances 0.000 description 1
- 239000012141 concentrate Substances 0.000 description 1
- 238000013135 deep learning Methods 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000003708 edge detection Methods 0.000 description 1
- 230000006870 function Effects 0.000 description 1
- JEIPFZHSYJVQDO-UHFFFAOYSA-N iron(III) oxide Inorganic materials O=[Fe]O[Fe]=O JEIPFZHSYJVQDO-UHFFFAOYSA-N 0.000 description 1
- 238000012423 maintenance Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 238000006467 substitution reaction Methods 0.000 description 1
- 239000000725 suspension Substances 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
- 238000013519 translation Methods 0.000 description 1
- 238000011179 visual inspection Methods 0.000 description 1
Landscapes
- Image Analysis (AREA)
Abstract
The invention belongs to the technical field of artificial intelligence, and particularly relates to an intelligent unmanned inspection contact net defect identification system. The system comprises: an image acquisition device, an image processing device and a defect recognition device; the image acquisition device is used for acquiring an original image of the target area; the image processing device is used for identifying the outline of the contact net supporting structure from the original image to obtain the category of the contact net; the defect recognition device is configured to perform defect recognition of the contact network based on the shape descriptor of the contact network and the category of the contact network, and obtain a defect recognition result. According to the method, the technology of unmanned aerial vehicle image acquisition, feature fusion, differential classification and the like are combined, so that the efficient and accurate railway contact net defect identification is realized.
Description
Technical Field
The invention belongs to the technical field of artificial intelligence, and particularly relates to an intelligent unmanned inspection contact net defect identification system.
Background
In the railway traffic field, the contact net is used as a key component of railway power supply and plays an important role in maintaining the normal operation and safety of railway traffic. However, the operation state of the overhead line system is affected by various factors including weather, wear, fatigue, etc., which may cause various defects and faults of the overhead line system, thereby affecting the normal operation of railway traffic.
In the past decades, with the development of technology, the inspection method of the railway contact network is gradually changed from manual inspection to automatic inspection, so that inspection efficiency and accuracy are improved. The initial inspection of the overhead line system mainly depends on manual visual inspection, however, the method has the problems of low efficiency, high omission ratio and the like. With the progress of technologies such as image processing and machine learning, some automated inspection systems have been introduced that collect and analyze image data of overhead lines to detect and identify defects.
However, the current automatic overhead line inspection technology still has some problems: the false alarm rate is high: some automated inspection systems may produce false positives during image processing that incorrectly identify normal structures as defective, resulting in unnecessary maintenance and downtime. The feature extraction is difficult: current inspection techniques still present challenges in extracting valid features from large amounts of image data. The conventional feature extraction method may not accurately capture complex features of the catenary. Multi-category identification problem: the shape and structure of the catenary is affected by a number of factors, which may lead to different types of catenary variations. However, the prior art may not perform well in classifying and identifying multiple categories of catenary.
Disclosure of Invention
The intelligent unmanned inspection contact network defect recognition system provided by the invention has the main purposes that the efficient and accurate railway contact network defect recognition is realized by combining the technologies of unmanned aerial vehicle image acquisition, feature fusion, differential classification and the like, and the safety and inspection efficiency of railway traffic are improved.
In order to solve the technical problems, the invention provides an intelligent unmanned inspection contact net defect identification system, which comprises: an image acquisition device, an image processing device and a defect recognition device; the image acquisition device is used for acquiring an original image of the target area; the image processing device is used for identifying the outline of the contact net supporting structure from the original image, then using the shape descriptor to represent the geometric shape characteristic of the contact net supporting structure, and using the texture characteristic to capture the wire characteristic of the contact net; carrying out feature fusion on geometric features of the contact net supporting structure and contact net wire features to obtain a feature matrix, and classifying the contact net based on the feature matrix to obtain the category of the contact net; the defect recognition device is configured to perform defect recognition of the contact network based on the shape descriptor of the contact network and the category of the contact network, and obtain a defect recognition result.
Further, the image acquisition device is an unmanned plane; the unmanned aerial vehicle is provided with a camera or a video camera; the unmanned aerial vehicle flies according to a preset path, and an original image of a target area is automatically acquired.
Further, the method for identifying the outline of the contact net supporting structure from the original image by the image acquisition device comprises the following steps: the method for removing noise in the original image by using the invariant differential filter specifically comprises the following steps: convolving the original image with the invariant differential filter kernel using the following formula to obtain a denoised image of the original image:
；
wherein,is an original image; />Is a constant differential filtering kernel; />Representing a convolution operation;
calculating the gradient strength and the gradient direction of the denoising image in the directions of the horizontal axis and the vertical axis by using a Sobel filter; checking the pixels of the denoised image in the gradient direction, if the pixels are the maximum of their neighbors, then preserving, otherwise setting to 0; defining two thresholds, namely a high threshold and a low threshold; pixels with pixel values above the high threshold are marked directly as contours, pixels with pixel values below the low threshold are lost, pixels with pixel values between the high and low thresholds are also considered contours if they are connected to strong contour pixels, otherwise lost; and (5) finishing the identification of the outline of the contact net supporting structure from the original image.
Further, the invariant differential filtering kernel is expressed using the following formula:
；
wherein,for a constant differential filter kernel +.>Is the standard deviation of the Gaussian kernel, +.>Is the abscissa of the pixel, +.>Is the ordinate of the pixel.
Further, the method for representing the geometric shape characteristic of the contact net support structure by the image processing device by using the shape descriptor comprises the following steps:
based on the identified profile of the catenary support structure, a normalized center moment of the catenary support structure is calculated using the following formula:
；
wherein,is the center moment, calculated using the following formula:
and->Is the centroid coordinate of the profile of the contact net support structure, < + >>And->The order is used for describing the complexity of the shape, and the value ranges are 0 to 3; />Is a pixel value;
then using the following formula, the 9 invariant moments are calculated and used as shape descriptors:
；
wherein,is a first invariant moment; />Is a second invariant moment; />Is a third invariant moment; />The fourth torque invariant moment;is a fifth invariant moment; />A sixth torque converter; />A seventh torque converter; />Eight invariant moment; />Is a ninth invariant moment;is the normalized central moment of the contact net supporting structure, wherein +.>And->The number of the steps is 0 to 3.
Further, the method for capturing the characteristics of the contact network wire by using the texture characteristics by the image processing device comprises the following steps: dividing the denoised image into non-overlapping 3x3 window regions; for the central pixel in each window area, carrying out difference calculation on the central pixel and surrounding pixels, and generating a binary code from the result of the difference calculation; connecting binary codes of each window area together to serve as sub-texture features, and forming a matrix by the sub-texture features to serve as texture features; the texture features are represented using the following formula:the method comprises the steps of carrying out a first treatment on the surface of the Wherein (1)>Is a texture feature matrix; />For sub-texture feature->For subscripts, the values range from 1 to 9.
Further, the image processing device performs feature fusion on geometric feature of the contact net supporting structure and contact net wire feature, and the obtained feature matrix is expressed by using the following formula:
；
wherein,is a feature matrix.
Further, the method for classifying the contact network based on the feature matrix to obtain the classification of the contact network comprises the following steps: the classification value is calculated using the following formula:
；
wherein,is a classification value; />The operation of the trace for the matrix; if the calculated classification value is smaller than 15, obtaining the classification of the contact net as a rigid contact net; if the calculated classification value is greater than 45, obtaining the classification of the contact net as a flexible contact net; if the calculated classification value is more than 15 and less than 45, obtaining a rigid-flexible mixed contact net with the contact net type; if the calculated classification value is less than 15, if the calculated classification value is less than10, obtaining a rigid knife-type contact net of the contact net type; and if the calculated classification value is smaller than 3, obtaining the rigid trapezoid contact net with the contact net type.
Further, if the contact net type is a rigid contact net, a rigid knife contact net or a rigid trapezoid contact net, the defect recognition device performs defect recognition of the contact net based on a shape descriptor of the contact net and the contact net type, and the method for obtaining the defect recognition result includes: the first discrimination value is calculated using the following formula:
；
wherein,for the first discrimination value, < >>As a first category factor, when the catenary category is rigid catenary,when the contact net type is a rigid knife type contact net, the +.>When the contact net type is a rigid trapezoid contact net,the method comprises the steps of carrying out a first treatment on the surface of the And comparing the calculated first discrimination value with a first discrimination threshold, and if the calculated first discrimination value is larger than the first discrimination threshold, judging that the contact net has defects, so as to obtain a defect identification result.
Further, if the contact network type is a rigid-flexible mixed contact network or a flexible contact network, the defect recognition device performs defect recognition of the contact network based on the shape descriptor of the contact network and the contact network type, and the method for obtaining the defect recognition result comprises the following steps: calculating a second discrimination value using the following formula:
；
wherein,for the second discrimination value, < >>As the second category factor, when the category of the contact net is a rigid-flexible mixed contact net,when the contact net type is a flexible contact net, the +.>The method comprises the steps of carrying out a first treatment on the surface of the And comparing the calculated second discrimination value with a second discrimination threshold, and if the calculated second discrimination value is smaller than the first discrimination threshold, judging that the contact net has defects, so as to obtain a defect identification result.
The intelligent unmanned inspection contact net defect identification system has the following beneficial effects: firstly, the unmanned aerial vehicle technology is introduced as an image acquisition device, so that the contact net inspection can realize omnibearing and high-efficiency coverage. The unmanned aerial vehicle can fly autonomously under the preset path, and the original image of the target area is automatically acquired, so that the inspection efficiency and coverage area are greatly improved compared with the traditional manual inspection mode, and the safe operation of railway traffic is better ensured. Secondly, the invention adopts the non-differential filter to identify the outline of the contact net supporting structure, calculates the gradient strength and the gradient direction by means of the Sobel filter, effectively eliminates noise interference in the image, and improves the accuracy of the outline. This way of integrated application enables the geometrical features of the catenary to be extracted and described more precisely. The invention comprehensively uses the shape descriptor and the texture feature, and fully combines the geometric shape and the wire feature of the contact net. By calculating the central moment and the invariant moment, complex geometric features of the contact net supporting structure can be captured. And by utilizing the difference value calculation and binary coding of the window area, the texture information of the wire characteristics can be effectively captured. The feature fusion mode enables the system to acquire the feature information of the contact net more comprehensively, and accuracy and reliability of defect identification are improved. In addition, the invention designs a discrimination value calculation method aiming at different types of contact networks. According to the category characteristics of the contact net, different category coefficients are formulated, and the classification and identification of the defects are realized by calculating the discrimination value. The differentiated classification mode enables the system to be more suitable for the characteristics of different types of contact networks, and therefore the multi-type recognition effect is improved. In conclusion, the intelligent unmanned inspection contact net defect identification system has obvious beneficial effects. The method realizes the efficient and accurate identification of the defects of the contact network by introducing unmanned plane technology, image processing, feature fusion, differential classification and other methods. The safety and stability of railway traffic are improved, the inspection efficiency is effectively improved, the labor and time cost is reduced, and positive influence is brought to railway operation management. Meanwhile, the innovation and the technical leadership of the invention also open up a new direction for the development of the railway inspection field, and have important practical application value and market potential.
Drawings
In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings that are required to be used in the embodiments or the description of the prior art will be briefly described below, and it is obvious that the drawings in the following description are only embodiments of the present invention, and that other drawings can be obtained according to the provided drawings without inventive effort for a person skilled in the art.
Fig. 1 is a schematic system structure diagram of an intelligent unmanned inspection catenary defect recognition system provided by an embodiment of the invention.
Detailed Description
The method of the present invention will be described in further detail with reference to the accompanying drawings.
Example 1: intelligent unmanned inspection contact net defect identification system, the system includes: an image acquisition device, an image processing device and a defect recognition device; the image acquisition device is used for acquiring an original image of the target area; the image processing device is used for identifying the outline of the contact net supporting structure from the original image, then using the shape descriptor to represent the geometric shape characteristic of the contact net supporting structure, and using the texture characteristic to capture the wire characteristic of the contact net; carrying out feature fusion on geometric features of the contact net supporting structure and contact net wire features to obtain a feature matrix, and classifying the contact net based on the feature matrix to obtain the category of the contact net; the defect recognition device is configured to perform defect recognition of the contact network based on the shape descriptor of the contact network and the category of the contact network, and obtain a defect recognition result.
Specifically, the obtained image is passed through an image processing device. First, the outline of the catenary support structure is identified by an image processing algorithm, which is to find the appearance of the catenary by detecting lines and changes. The system then uses shape descriptors, such as invariant moments, to convert the geometry of the catenary support structure to digitized features for machine understanding. Meanwhile, the surface texture characteristics of the contact net wire, such as abrasion, rust and the like of the wire surface, are captured through texture analysis.
The image acquisition device is used for capturing an original image of the contact net. Unmanned inspection requires a large amount of image data as input for subsequent processing and analysis. These images may be obtained by drones, cameras, etc. The original image is the initial data of the system, which contains the structure and characteristics of the catenary that need to be analyzed. The geometrical characteristics of the contact net support structure are very important for defect identification. By identifying contours and calculating features such as area, perimeter, aspect ratio, etc., the system can learn the shape of each support structure. These geometric features provide important information in subsequent classification and identification. The texture features of the catenary wires also help identify defects. The texture may include color of the wire, density of lines, variation in texture, and the like. These features can help the system distinguish normal wires from wires that may be defective. The purpose of feature fusion is to combine geometric features and texture features to form a comprehensive feature matrix. This has the advantage that combining the features of multiple dimensions can provide more comprehensive, differentiated data, thereby enhancing the accuracy of defect identification.
The fused feature matrix is used as input, and the system classifies the contact network based on machine learning or deep learning and other technologies. The purpose of the classification is to divide the catenary into different categories for further analysis and processing. Different types of catenary may have different defect patterns, so classification is a prerequisite step for identifying defects.
Example 2: on the basis of the above embodiment, the image acquisition device is an unmanned plane; the unmanned aerial vehicle is provided with a camera or a video camera; the unmanned aerial vehicle flies according to a preset path, and an original image of a target area is automatically acquired.
Specifically, the autonomous flight of the unmanned aerial vehicle is realized through a preset path. This means that the system has planned in advance the route that the drone should fly, ensuring that it can cover all parts of the target area. The presetting of the flight path can be realized through technologies such as GPS, inertial navigation system and the like, so that the unmanned aerial vehicle is ensured to be stable and accurate in the flight process.
Example 3: on the basis of the above embodiment, the method for identifying the outline of the contact net support structure from the original image by the image acquisition device includes: the method for removing noise in the original image by using the invariant differential filter specifically comprises the following steps: convolving the original image with the invariant differential filter kernel using the following formula to obtain a denoised image of the original image:
；
wherein,is an original image; />Is a constant differential filtering kernel; />Representing a convolution operation;
calculating the gradient strength and the gradient direction of the denoising image in the directions of the horizontal axis and the vertical axis by using a Sobel filter; checking the pixels of the denoised image in the gradient direction, if the pixels are the maximum of their neighbors, then preserving, otherwise setting to 0; defining two thresholds, namely a high threshold and a low threshold; pixels with pixel values above the high threshold are marked directly as contours, pixels with pixel values below the low threshold are lost, pixels with pixel values between the high and low thresholds are also considered contours if they are connected to strong contour pixels, otherwise lost; and (5) finishing the identification of the outline of the contact net supporting structure from the original image.
Specifically, the non-variance filter denoises:
a constant differential filter (invariant differential filter) is a filter commonly used for image denoising. The filter can reduce noise in the image by convolution operation. The convolution operation is a process of multiplying and summing the filter kernel with each pixel point of the image to obtain a processed pixel value. Convolution operation in a formulaIndicating (I)>For the original image +.>Is a constant differential filter kernel.
The Sobel filter calculates the gradient: the Sobel filter is a filter commonly used for edge detection for detecting intensity gradients in an image. Here, the gradient intensity and gradient direction of the denoised image in the horizontal and vertical axis directions are calculated using Sobel filters. Gradient intensity indicates the degree of color change of the pixel, and gradient direction indicates the direction in which the change is largest.
Contour detection in gradient direction: the pixels of the denoised image are examined in the gradient direction and remain if they are the maximum of their neighbors, otherwise set to 0. This step helps to highlight the contour areas in the image, i.e. where the color changes strongly in the image, since the edge areas will produce significant maxima in the gradient direction.
Thresholding and contour extraction: at this stage, two thresholds are defined: a high threshold and a low threshold. Pixels with pixel values above the high threshold are marked directly as contours, which are distinct edge points. Pixels with pixel values below the low threshold are discarded and these points are considered noise or flat areas. And pixels with pixel values between the high and low thresholds require further judgment. If these pixels are connected to strong contour pixels, they are also considered part of the contour, otherwise they are discarded.
And (3) contour recognition: through the above thresholding step, the system will identify the outline of the catenary support structure from the original image. These contours are where the color changes drastically in the image, representing the shape of the catenary.
In summary, this step involves a plurality of image processing techniques, from denoising to gradient computation, to thresholding and contour extraction, to finally achieve the identification of the contour of the catenary support structure from the original image. This process helps to extract key features, which lay the foundation for subsequent defect identification.
Example 4: on the basis of the above embodiment, the invariant differential filtering kernel is expressed using the following formula:
；
wherein,for a constant differential filter kernel +.>Is the standard deviation of the invariant differential filter kernel, < >>Is the abscissa of the pixel, +.>Is a pixelIs defined by the vertical coordinate of (c).
In particular, the method comprises the steps of,: this is the invariant differential filter kernel +.>And a response value thereon. In an image convolution operation, the filter kernel will convolve at each pixel location to smooth the image or extract features. />: this is the standard deviation of the invariant differential filter kernel, which determines the scale of the filter kernel. Less->The values will result in sharp filter kernels, larger +.>The value will produce a smoother kernel. />: this is the exponential part of the Gaussian distribution, representing the pixel +.>Distance from the center of the nucleus. />The larger the value of (c), the closer to 0 the exponential portion, resulting in a broader smoothing effect during the convolution process. />: this is a normalization factor of the gaussian distribution, ensuring that the sum of the filter kernels is 1 to keep the image brightness unchanged. />: this is in part to center the filter kernel +.>And obtaining the maximum response value. By taking the logarithm, the multiplication operation of the exponent part is converted into addition operation, and the maximum filter kernel response of the central position is ensured.
In summary, this formula defines a constant difference filter kernel for image processing. In the convolution operation, it can be based on different positionsAnd standard deviation->Different filter responses are generated to achieve smoothing and denoising of the image. In the image processing, the appropriate +.>The value may control the degree of filtering effect according to the application requirements.
Example 5: on the basis of the above embodiment, the method for representing the geometric feature of the catenary support structure by the image processing device using the shape descriptor includes:
based on the identified profile of the catenary support structure, a normalized center moment of the catenary support structure is calculated using the following formula:
；
wherein,is the center moment, calculated using the following formula:
and->Is the outline of the contact net supporting structureCentroid coordinates of>And->The order is used for describing the complexity of the shape, and the value ranges are 0 to 3; />Is a pixel value;
then using the following formula, the 9 invariant moments are calculated and used as shape descriptors:
；
wherein,is a first invariant moment; />Is a second invariant moment; />Is a third invariant moment; />The fourth torque invariant moment;is a fifth invariant moment; />A sixth torque converter; />A seventh torque converter; />Eight invariant moment; />Is a ninth invariant moment;is the normalized central moment of the contact net supporting structure, wherein +.>And->The number of the steps is 0 to 3.
Specifically, first, the center moment describes the geometric feature of the shape by calculating the relationship between the pixel position on the contour and the centroid coordinates of the contour. In the formulaIn (I)>Represents the central moment>And->Is the coordinates of the pixel, ">And->Is the centroid coordinates of the contour, +.>And->Is an order for describing shape complexity, +.>Is the value of the pixel. From these calculations, the distribution of the shape and the degree of decentration can be quantified.
In the formulaIn, normalized center moment +.>By dividing the central moment by the zero order central moment +.>To some degree, normalization is achieved. This has the advantage that shapes of different dimensions and sizes can be compared and classified, since these normalized central moments have dimensional invariance.
The calculation of the set of invariant moments is based on normalized central moments, with invariant properties extracted from the geometry by different combinations and operations. For example, a first invariant momentArea corresponding to shape, second invariant moment +.>The distribution of the shape is measured, third and fourth invariant moment +.>And->The degree of eccentricity of the shape is shown. As the order increases, the invariant moment may describe the features of the shape in more detail.
These invariant moment calculations are based on geometric properties of the shape such that they remain unchanged under translational, rotational and dimensional changes. These invariant properties make them very useful in image processing, and can be used for shape recognition, classification, feature extraction, and other tasks. In general, the purpose of these formulas is to extract invariant shape descriptors from the outline of the catenary support structure by computing different moments and combinations for representing and analyzing geometric features in the image.
In the first invariant moment, the total area of the shape is described.Indicating that the contact net supporting structure is +.>Distribution in the direction, and->Is indicated at->Distribution in the direction. By adding these two distributions, the total area of the shape can be obtained without being affected by the translation and rotation of the shape.
The eccentricity of the shape and the direction of the principal axis are measured. />Square representation +.>And->Difference in direction, ++>Indicating the difference in direction of the principal axis. The larger the value of this invariant moment, the more prone the shape to concentrate in one direction.
For measuring eccentricity and rotational invariance of the shape. />And->Respectively expressed in->And->The degree of deviation of the distribution in the direction. By calculating their sum of squares, the change in shape in different directions can be measured.
Example 6: on the basis of the above embodiment, the method for capturing the characteristics of the contact line wire by using the texture characteristics by the image processing device includes: dividing the denoised image into non-overlapping 3x3 window regions; for the central pixel in each window area, carrying out difference calculation on the central pixel and surrounding pixels, and generating a binary code from the result of the difference calculation; connecting binary codes of each window area together to serve as sub-texture features, and forming a matrix by the sub-texture features to serve as texture features; the texture features are represented using the following formula:the method comprises the steps of carrying out a first treatment on the surface of the Wherein (1)>Is a texture feature matrix; />For sub-texture feature->For subscripts, the values range from 1 to 9.
In particular, the key to this approach is to divide the image into small windows and then calculate and encode the difference of the pixels within each window. By concatenating and arranging these encodings, a texture feature matrix is formed, which contains texture information for the entire image. The method has the advantages that the method can effectively capture the texture characteristics of different areas in the image, and is helpful for distinguishing different texture modes of the contact net wire, so that the accuracy of defect identification is improved.
Example 7: on the basis of the above embodiment, the image processing device performs feature fusion on the geometric feature of the overhead line system supporting structure and the overhead line system wire feature, and the obtained feature matrix is represented by using the following formula:
；
wherein,is a feature matrix.
In particular, in this matrix, the matrix,texture feature submatrix representing contact net wires, +.>Representing the geometric characteristics of the catenary support structure. The products of them are placed in the feature matrix at corresponding locations to perform feature fusion.
The purpose of this feature fusion step is to combine the two types of features to obtain a more comprehensive and accurate representation of the features. This will help to improve the performance of the catenary defect recognition system, making it more able to recognize and classify different types of catenary problems. The geometrical features of the catenary support structure and the textural features of the catenary wires are extracted in the previous steps. These features capture different aspects of the catenary. The goal of feature fusion is to combine these different types of features together to obtain a more comprehensive, richer representation of the features, thereby enhancing the accuracy of recognition and classification. The feature matrix is constructed by combining different types of features together in a manner to form a composite matrix representation. This feature matrix will be used in further analysis and processing.
Example 8: on the basis of the above embodiment, the method for classifying the catenary based on the feature matrix to obtain the classification of the catenary includes: the classification value is calculated using the following formula:
；
wherein,is a classification value; />The operation of the trace for the matrix; if the calculated classification value is smaller than 15, obtaining the classification of the contact net as a rigid contact net; if the calculated classification value is greater than 45, obtaining the classification of the contact net as a flexible contact net; if the calculated classification value is more than 15 and less than 45, obtaining a rigid-flexible mixed contact net with the contact net type; when the calculated classification value is smaller than 15, if the calculated classification value is smaller than 10, obtaining the contact net type as a rigid knife type contact net; and if the calculated classification value is smaller than 3, obtaining the rigid trapezoid contact net with the contact net type.
In particular, the basic feature of a rigid catenary is its stability in geometry. This means that the geometry of the rigid catenary changes less even in different images. Thus, geometric features upon feature fusionIs relatively limited, without applying to the feature matrix +.>A significant impact is produced. Thus, the +.>The value is smaller, and the characteristics of the rigid contact net are met.
For example, in railway traffic, rigid catenary systems may be standardized suspension systems, in which the support structure is relatively fixed and is not prone to distortion or deformation.
The flexible contact net has the main characteristics of higher shape variability and can be influenced by external factors to generate distortion and deformation. This means that the geometric features are at the time of feature fusionIs greatly varied, will be about the feature matrix>A significant impact is produced. Thus, the +.>The value is larger, and the characteristics of the flexible contact net are reflected.
For example, in an electrical transmission line, the flexible contact net may be an electric wire, and its shape may be significantly distorted due to wind force or the like.
Rigid-flexible hybrid catenary combines rigid and flexible characteristics. Geometric features during feature fusionAnd texture feature->Is in the feature matrix->But are relatively balanced. Thus, calculated by matrix operation +.>The value is in the middle range, and reflects the characteristics of the rigid-flexible mixed contact net.
The rigid blade contact net is characterized in that the geometric shape of the rigid blade contact net presents the characteristics of a blade, and the shape is generally uniform and stable. The characteristic fusion of the knife-type contact net can highlight geometric shape characteristics, and the geometric shape characteristics are formed in a characteristic matrixIn each item->The variation in matrix operation is small. This results in a calculated +.>The value is relatively small, and is matched with the characteristics of the rigid knife type contact net.
For example, in urban rail transit, a rigid blade contact net may be a contact with a uniform geometry.
The rigid trapezoid contact net features similar geometry to trapezoid, parallel upper and lower boundaries and less shape change. The feature fusion of the contact net also focuses more on geometric features, but in a feature matrixIn each item->The variation in matrix operations is still small. This also results in a calculated +.>The value is smaller, and the characteristics of the contact net are identical to those of the rigid trapezoid contact net.
For example, in power transmission, a rigid trapezoidal catenary may be a power pole having a trapezoidal-like cross section.
For example, a bridge or the like may have both rigid support and flexible connection, and thus may vary in form over a range.
Example 9: on the basis of the above embodiment, if the type of the contact net is a rigid contact net, a rigid blade contact net or a rigid trapezoid contact net, the defect recognition device performs defect recognition of the contact net based on a shape descriptor of the contact net and the type of the contact net, and the method for obtaining the defect recognition result includes: the first discrimination value is calculated using the following formula:
；
wherein,for the first discrimination value, < >>As a first category factor, when the catenary category is rigid catenary,when the contact net type is a rigid knife type contact net, the +.>When the contact net type is a rigid trapezoid contact net,the method comprises the steps of carrying out a first treatment on the surface of the And comparing the calculated first discrimination value with a first discrimination threshold, and if the calculated first discrimination value is larger than the first discrimination threshold, judging that the contact net has defects, so as to obtain a defect identification result.
In particular, for a rigid catenary,. This suggests that we are more sensitive to shape descriptors of rigid catenaries. In the weighted sum calculation of shape descriptors, each +.>Will be calculating->Higher weights are obtained. The basic feature of rigid catenary is that the shape is relatively stable, so any small shape change may mean a possible defect. Thus, if->The value exceeds a preset first judging threshold value, so that whether the rigid contact net has defects can be judged more sensitively.
For a rigid blade-type catenary,the method comprises the steps of carrying out a first treatment on the surface of the For rigid trapezoid contact net->. Both types of contact networks are more flexible in shape characteristics than rigid contact networks, but still maintain a certain stability. Due to the shape change of these contact linesThe degree is slightly larger than that of a rigid contact net, and we calculate +.>More shape descriptor weights are considered. If->The value exceeding the threshold value indicates that the change in shape may exceed the normal range, so we can infer that a defect may be present.
Example 10: on the basis of the above embodiment, if the contact network type is a rigid-flexible mixed contact network or a flexible contact network, the defect recognition device performs defect recognition of the contact network based on the shape descriptor of the contact network and the contact network type, and the method for obtaining the defect recognition result includes: calculating a second discrimination value using the following formula:
；
wherein,for the second discrimination value, < >>As the second category factor, when the category of the contact net is a rigid-flexible mixed contact net,when the contact net type is a flexible contact net, the +.>The method comprises the steps of carrying out a first treatment on the surface of the And comparing the calculated second discrimination value with a second discrimination threshold, and if the calculated second discrimination value is smaller than the first discrimination threshold, judging that the contact net has defects, so as to obtain a defect identification result.
In particular, for a rigid-flexible hybrid catenary,the method comprises the steps of carrying out a first treatment on the surface of the Flexible contact net>。/>The coefficients represent the sensitivity to different classes of catenary. Different->The values take into account the characteristics of the contact network in terms of flexibility and shape variation. In calculating->In case of value, the->The variation amplitude of the whole discrimination value is determined.
Representing a change in shape descriptor. Here, the shape descriptor is changed +>The method is based on the characteristics and the shape change amplitude of different types of contact networks. In calculating->In case of value, the->The sensitivity degree of different types of contact networks to shape changes is shown.
By passing throughFor->The values are log-scale transformed, mapping the range of variation magnitudes to log space. This transformation preserves the order of magnitude of the changes but reduces the gap between the larger and smaller values. The modulus of absolute value ensures symmetry in the positive and negative directions.
Final endThe value is by->Calculated. +.>The factor is to adjust the weights in the formula. The combination of this formula takes into account the trade-off of shape descriptor variations and class characteristics.
While specific embodiments of the present invention have been described above, it will be understood by those skilled in the art that these specific embodiments are by way of example only, and that various omissions, substitutions, and changes in the form and details of the methods and systems described above may be made by those skilled in the art without departing from the spirit and scope of the invention. For example, it is within the scope of the present invention to combine the above-described method steps to perform substantially the same function in substantially the same way to achieve substantially the same result. Accordingly, the scope of the invention is limited only by the following claims.
Claims (10)
1. Intelligent unmanned inspection contact net defect identification system, its characterized in that, the system includes: an image acquisition device, an image processing device and a defect recognition device; the image acquisition device is used for acquiring an original image of the target area; the image processing device is used for identifying the outline of the contact net supporting structure from the original image, then using the shape descriptor to represent the geometric shape characteristic of the contact net supporting structure, and using the texture characteristic to capture the wire characteristic of the contact net; carrying out feature fusion on geometric features of the contact net supporting structure and contact net wire features to obtain a feature matrix, and classifying the contact net based on the feature matrix to obtain the category of the contact net; the defect recognition device is configured to perform defect recognition of the contact network based on the shape descriptor of the contact network and the category of the contact network, and obtain a defect recognition result.
2. The intelligent unmanned inspection catenary defect recognition system according to claim 1, wherein the image acquisition device is an unmanned plane; the unmanned aerial vehicle is provided with a camera or a video camera; the unmanned aerial vehicle flies according to a preset path, and an original image of a target area is automatically acquired.
3. The intelligent unmanned inspection catenary defect recognition system according to claim 2, wherein the method for recognizing the outline of the catenary support structure from the original image by the image acquisition device comprises: the method for removing noise in the original image by using the invariant differential filter specifically comprises the following steps: convolving the original image with the invariant differential filter kernel using the following formula to obtain a denoised image of the original image:
；
wherein,is an original image; />Is a constant differential filtering kernel; />Representing a convolution operation;
calculating the gradient strength and the gradient direction of the denoising image in the directions of the horizontal axis and the vertical axis by using a Sobel filter; checking the pixels of the denoised image in the gradient direction, if the pixels are the maximum of their neighbors, then preserving, otherwise setting to 0; defining two thresholds, namely a high threshold and a low threshold; pixels with pixel values above the high threshold are marked directly as contours, pixels with pixel values below the low threshold are lost, pixels with pixel values between the high and low thresholds are also considered contours if they are connected to strong contour pixels, otherwise lost; and (5) finishing the identification of the outline of the contact net supporting structure from the original image.
4. The intelligent unmanned inspection catenary defect recognition system of claim 3, wherein the invariant differential filter kernel is represented using the formula:
；
wherein,for a constant differential filter kernel +.>Is the standard deviation of the Gaussian kernel, +.>Is the abscissa of the pixel, +.>Is the ordinate of the pixel.
5. The intelligent unmanned inspection catenary defect recognition system of claim 4, wherein the method for the image processing device to represent geometric features of the catenary support structure using shape descriptors comprises:
based on the identified profile of the catenary support structure, a normalized center moment of the catenary support structure is calculated using the following formula:
；
wherein,is the center moment, calculated using the following formula:
and->Is the centroid coordinate of the profile of the contact net support structure, < + >>And->The order is used for describing the complexity of the shape, and the value ranges are 0 to 3; />Is a pixel value;
then using the following formula, the 9 invariant moments are calculated and used as shape descriptors:
；
wherein,is a first invariant moment; />Is a second invariant moment; />Is a third invariant moment; />The fourth torque invariant moment; />Is fifth oneTorque invariant; />A sixth torque converter; />A seventh torque converter; />Eight invariant moment; />Is a ninth invariant moment; />Is the normalized central moment of the contact net supporting structure, wherein +.>And->The number of the steps is 0 to 3.
6. The intelligent unmanned inspection catenary defect recognition system of claim 5, wherein the method for capturing catenary wire characteristics by the image processing device using texture characteristics comprises: dividing the denoised image into non-overlapping 3x3 window regions; for the central pixel in each window area, carrying out difference calculation on the central pixel and surrounding pixels, and generating a binary code from the result of the difference calculation; connecting binary codes of each window area together to serve as sub-texture features, and forming a matrix by the sub-texture features to serve as texture features; the texture features are represented using the following formula:the method comprises the steps of carrying out a first treatment on the surface of the Wherein (1)>Is a texture feature matrix; />For sub-texture feature->For subscripts, the values range from 1 to 9.
7. The intelligent unmanned inspection catenary defect recognition system according to claim 6, wherein the image processing device performs feature fusion on geometric features of a catenary support structure and catenary wire features, and the obtained feature matrix is represented by using the following formula:
；
wherein,is a feature matrix.
8. The intelligent unmanned inspection catenary defect recognition system according to claim 7, wherein the catenary classification method based on the feature matrix comprises: the classification value is calculated using the following formula:
；
wherein,is a classification value; />The operation of the trace for the matrix; if the calculated classification value is smaller than 15, obtaining the classification of the contact net as a rigid contact net; if the calculated classification value is greater than45, obtaining the flexible contact net of the contact net type; if the calculated classification value is more than 15 and less than 45, obtaining a rigid-flexible mixed contact net with the contact net type; when the calculated classification value is smaller than 15, if the calculated classification value is smaller than 10, obtaining the contact net type as a rigid knife type contact net; and if the calculated classification value is smaller than 3, obtaining the rigid trapezoid contact net with the contact net type.
9. The intelligent unmanned inspection catenary defect recognition system of claim 8, wherein if the catenary type is a rigid catenary, a rigid blade catenary, or a rigid trapezoid catenary, the defect recognition device performs defect recognition of the catenary based on a shape descriptor of the catenary and the type of the catenary, and the method for obtaining the defect recognition result comprises: the first discrimination value is calculated using the following formula:
；
wherein,for the first discrimination value, < >>As a first category factor, when the catenary category is rigid catenary, the +.>When the contact net type is a rigid knife type contact net, the +.>When the contact net type is a rigid trapezoid contact net, < + >>The method comprises the steps of carrying out a first treatment on the surface of the Comparing the calculated first discrimination value with a first discrimination threshold, and if the calculated first discrimination value is larger than the first discrimination threshold, judging that the contact net has defects to obtainTo the defect recognition result.
10. The intelligent unmanned inspection catenary defect recognition system of claim 8, wherein if the catenary type is a rigid-flexible hybrid catenary or a flexible catenary, the defect recognition device performs defect recognition of the catenary based on a shape descriptor of the catenary and the type of the catenary, and the method for obtaining the defect recognition result comprises: calculating a second discrimination value using the following formula:
；
wherein,for the second discrimination value, < >>As the second category factor, when the category of the contact net is a rigid-flexible mixed contact net,when the contact net type is a flexible contact net, the +.>The method comprises the steps of carrying out a first treatment on the surface of the Comparing the calculated second discrimination value with a second discrimination threshold, and if the calculated second discrimination value is smaller than the first discrimination threshold, judging that the contact net has defects, so as to obtain a defect identification result; />Representing a change in shape descriptor.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311466053.8A CN117197700B (en) | 2023-11-07 | 2023-11-07 | Intelligent unmanned inspection contact net defect identification system |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311466053.8A CN117197700B (en) | 2023-11-07 | 2023-11-07 | Intelligent unmanned inspection contact net defect identification system |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN117197700A true CN117197700A (en) | 2023-12-08 |
| CN117197700B CN117197700B (en) | 2024-01-26 |
Family
ID=89001994
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311466053.8A Active CN117197700B (en) | 2023-11-07 | 2023-11-07 | Intelligent unmanned inspection contact net defect identification system |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117197700B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117690096A (en) * | 2024-02-04 | 2024-03-12 | 成都中轨轨道设备有限公司 | Contact net safety inspection system adapting to different scenes |
Citations (17)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP0789258A1 (en) * | 1996-02-09 | 1997-08-13 | Red Nacional De Los Ferrocarriles Espanoles (Renfe) | Automatic measuring system of the wear of the overhead distribution contact wires |
| CN102903080A (en) * | 2012-09-06 | 2013-01-30 | 西安工程大学 | Non-supervision estimation method on speckle noise suppression performance of synthetic aperture radar image |
| CN107633266A (en) * | 2017-09-07 | 2018-01-26 | 西安交通大学 | A kind of electric locomotive OCS and pantograph arc method for measuring |
| CN108508023A (en) * | 2018-03-30 | 2018-09-07 | 深圳市益鑫智能科技有限公司 | The defect detecting system of end puller bolt is contacted in a kind of railway contact line |
| CN109710983A (en) * | 2018-12-04 | 2019-05-03 | 北京大学 | A kind of cylinder of diesel engine hierarchical fault diagnosis method based on Key Performance Indicator |
| JP2019111942A (en) * | 2017-12-25 | 2019-07-11 | 株式会社明電舎 | Overhead power line metal fitting detection device and method |
| CN110197173A (en) * | 2019-06-13 | 2019-09-03 | 重庆邮电大学 | A kind of curb detection method based on binocular vision |
| CN110264448A (en) * | 2019-06-06 | 2019-09-20 | 中南大学 | A kind of insulator breakdown detection method based on machine vision |
| CN111127382A (en) * | 2018-10-15 | 2020-05-08 | 西南交通大学 | Contact net suspension assembly U-shaped ring spare cap defect detection method |
| CN111125416A (en) * | 2019-12-27 | 2020-05-08 | 郑州轻工业大学 | Image retrieval method based on multi-feature fusion |
| CN111579560A (en) * | 2020-05-28 | 2020-08-25 | 成都国铁电气设备有限公司 | Defect positioning device and method used in tunnel |
| CN111951212A (en) * | 2020-04-08 | 2020-11-17 | 北京交通大学 | Method for identifying defects of contact network image of railway |
| CN112862770A (en) * | 2021-01-29 | 2021-05-28 | 珠海迪沃航空工程有限公司 | Defect analysis and diagnosis system, method and device based on artificial intelligence |
| CN115184922A (en) * | 2022-06-24 | 2022-10-14 | 中国铁道科学研究院集团有限公司 | Vehicle-mounted tunnel lining defect detection device and method |
| CN115205398A (en) * | 2022-07-14 | 2022-10-18 | 安徽工程大学 | Camera calibration method and system under uneven illumination condition |
| CN116645620A (en) * | 2023-05-04 | 2023-08-25 | 国网电力空间技术有限公司 | High-precision power grid unmanned aerial vehicle inspection image defect identification method and system |
| CN116757990A (en) * | 2023-01-09 | 2023-09-15 | 河南省科学院应用物理研究所有限公司 | Railway fastener defect online detection and identification method based on machine vision |
-
2023
- 2023-11-07 CN CN202311466053.8A patent/CN117197700B/en active Active
Patent Citations (17)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| EP0789258A1 (en) * | 1996-02-09 | 1997-08-13 | Red Nacional De Los Ferrocarriles Espanoles (Renfe) | Automatic measuring system of the wear of the overhead distribution contact wires |
| CN102903080A (en) * | 2012-09-06 | 2013-01-30 | 西安工程大学 | Non-supervision estimation method on speckle noise suppression performance of synthetic aperture radar image |
| CN107633266A (en) * | 2017-09-07 | 2018-01-26 | 西安交通大学 | A kind of electric locomotive OCS and pantograph arc method for measuring |
| JP2019111942A (en) * | 2017-12-25 | 2019-07-11 | 株式会社明電舎 | Overhead power line metal fitting detection device and method |
| CN108508023A (en) * | 2018-03-30 | 2018-09-07 | 深圳市益鑫智能科技有限公司 | The defect detecting system of end puller bolt is contacted in a kind of railway contact line |
| CN111127382A (en) * | 2018-10-15 | 2020-05-08 | 西南交通大学 | Contact net suspension assembly U-shaped ring spare cap defect detection method |
| CN109710983A (en) * | 2018-12-04 | 2019-05-03 | 北京大学 | A kind of cylinder of diesel engine hierarchical fault diagnosis method based on Key Performance Indicator |
| CN110264448A (en) * | 2019-06-06 | 2019-09-20 | 中南大学 | A kind of insulator breakdown detection method based on machine vision |
| CN110197173A (en) * | 2019-06-13 | 2019-09-03 | 重庆邮电大学 | A kind of curb detection method based on binocular vision |
| CN111125416A (en) * | 2019-12-27 | 2020-05-08 | 郑州轻工业大学 | Image retrieval method based on multi-feature fusion |
| CN111951212A (en) * | 2020-04-08 | 2020-11-17 | 北京交通大学 | Method for identifying defects of contact network image of railway |
| CN111579560A (en) * | 2020-05-28 | 2020-08-25 | 成都国铁电气设备有限公司 | Defect positioning device and method used in tunnel |
| CN112862770A (en) * | 2021-01-29 | 2021-05-28 | 珠海迪沃航空工程有限公司 | Defect analysis and diagnosis system, method and device based on artificial intelligence |
| CN115184922A (en) * | 2022-06-24 | 2022-10-14 | 中国铁道科学研究院集团有限公司 | Vehicle-mounted tunnel lining defect detection device and method |
| CN115205398A (en) * | 2022-07-14 | 2022-10-18 | 安徽工程大学 | Camera calibration method and system under uneven illumination condition |
| CN116757990A (en) * | 2023-01-09 | 2023-09-15 | 河南省科学院应用物理研究所有限公司 | Railway fastener defect online detection and identification method based on machine vision |
| CN116645620A (en) * | 2023-05-04 | 2023-08-25 | 国网电力空间技术有限公司 | High-precision power grid unmanned aerial vehicle inspection image defect identification method and system |
Non-Patent Citations (5)
| Title |
|---|
| HAO QIANG 等: "Transmission Line Fault Detection and Classification Based on Improved YOLOv8s", 《ELECTRONICS》, pages 1 - 18 * |
| 刘子仪 等: "基于机器视觉的列车部件故障诊断方法综述", 《计算机应用与软件》, pages 1 - 9 * |
| 张子健 等: "基于深度学习与Hu不变矩的绝缘子故障检测", 《铁道学报》, pages 71 - 77 * |
| 朱振坤 等: "基于特征融合的减振垫缺陷分类", 《组合机床与自动化加工技术》, pages 142 - 145 * |
| 马敏: "基于图像处理的绝缘子破损识别", 《中国优秀硕士学位论文全文数据库 工程科技II辑》, pages 042 - 163 * |
Cited By (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117690096A (en) * | 2024-02-04 | 2024-03-12 | 成都中轨轨道设备有限公司 | Contact net safety inspection system adapting to different scenes |
| CN117690096B (en) * | 2024-02-04 | 2024-04-12 | 成都中轨轨道设备有限公司 | Contact net safety inspection system adapting to different scenes |
Also Published As
| Publication number | Publication date |
|---|---|
| CN117197700B (en) | 2024-01-26 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107610114B (en) | optical satellite remote sensing image cloud and snow fog detection method based on support vector machine | |
| CN111428748B (en) | HOG feature and SVM-based infrared image insulator identification detection method | |
| CN111784633B (en) | Insulator defect automatic detection algorithm for electric power inspection video | |
| Bu et al. | Crack detection using a texture analysis-based technique for visual bridge inspection | |
| CN110837768B (en) | Online detection and identification method for rare animal protection | |
| CN111814686A (en) | Vision-based power transmission line identification and foreign matter invasion online detection method | |
| CN108985170A (en) | Transmission line of electricity hanger recognition methods based on Three image difference and deep learning | |
| CN106485694B (en) | A kind of high iron catenary double-jacket tube connector six-sided nut based on cascade classifier falls off defective mode detection method | |
| CN108564814A (en) | A kind of parking position detection method and device based on image | |
| CN117197700B (en) | Intelligent unmanned inspection contact net defect identification system | |
| CN106530281A (en) | Edge feature-based unmanned aerial vehicle image blur judgment method and system | |
| CN110659649A (en) | Image processing and character recognition algorithm based on near infrared light imaging | |
| CN108710909B (en) | Counting method for deformable, rotary and invariant boxed objects | |
| CN109376740A (en) | A kind of water gauge reading detection method based on video | |
| CN111008961A (en) | Transmission line equipment defect detection method and system, equipment and medium thereof | |
| Farag | Real-time detection of road lane-lines for autonomous driving | |
| CN110222661B (en) | Feature extraction method for moving target identification and tracking | |
| CN109063669B (en) | Bridge area ship navigation situation analysis method and device based on image recognition | |
| Ye et al. | Overhead ground wire detection by fusion global and local features and supervised learning method for a cable inspection robot | |
| CN110458064B (en) | Low-altitude target detection and identification method combining data driving type and knowledge driving type | |
| CN115841633A (en) | Power tower and power line associated correction power tower and power line detection method | |
| CN110348307B (en) | Path edge identification method and system for crane metal structure climbing robot | |
| CN111667473A (en) | Insulator hydrophobicity grade judging method based on improved Canny algorithm | |
| CN110276747B (en) | Insulator fault detection and fault rating method based on image analysis | |
| CN108520252B (en) | Road sign identification method based on generalized Hough transform and wavelet transform |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |