Insulation pull rod defect detection method and system based on ultrasonic standard atlas

Abstract

The invention discloses an insulation pull rod defect detection method and system based on an ultrasonic standard atlas, and the method comprises the steps: obtaining an ultrasonic original signal of each position point on a two-dimensional expansion surface of an insulation pull rod, and carrying out the preprocessing of each ultrasonic original signal, and obtaining an ultrasonic target signal; performing inversion operation on each ultrasonic target signal to obtain ultrasonic standard maps of the insulating pull rod under different depths on the two-dimensional expansion surface; and inputting the ultrasonic standard map into the improved Faster R-CNN model, and outputting a defect detection result corresponding to the ultrasonic standard map by the improved Faster R-CNN model. Compared with the traditional nearest neighbor interpolation filling mode, the bilinear interpolation data filling mode provided by the invention has the advantages that the distribution of color bands in the result is relatively uniform, obvious'patches' do not exist, and an inversion map better conforms to the actual pull rod morphology; compared with bicubic interpolation, the calculation amount is small, the calculation speed is higher, excessive smoothness and overshoot phenomena can be avoided, and the method is more suitable for processing a large number of maps.

Description

Technical Field

[0001] The invention belongs to the technical field of acoustic nondestructive testing of insulating rods, and in particular relates to a method and system for detecting defects of insulating rods based on ultrasonic standard spectra. Background Art

[0002] Insulating rods are an important part of power equipment. They have the key function of connecting and supporting conductors. Their performance and quality directly affect the safety and operational reliability of power equipment. Insulating rods are subjected to complex electrical, mechanical and environmental stresses during operation. If there are tiny defects inside, they may cause problems such as partial discharge, insulation aging and even breakdown, which seriously threaten the operational safety of the equipment. Comprehensive testing of insulating rods can effectively identify potential hidden dangers, prevent equipment failures, and improve the reliability and life of power equipment, which is of great significance to ensuring the safe and stable operation of the power system.

[0003] Tie rod defects are difficult to detect with the naked eye or conventional means, so accurate defect detection technology is particularly important. Traditional ultrasonic nondestructive testing mainly uses the propagation characteristics of ultrasonic waves in materials to detect equipment defects such as cracks, pores or delamination. It has good detection effects on deep internal defects and is non-destructive to equipment. However, when ultrasound is used for nondestructive testing of insulating tie rods, it is easy to cause scattering and attenuation of ultrasonic waves due to the influence of the multi-layer structure and composite materials of the insulating tie rods, resulting in interference with defect signals and difficulty in accurately assessing the location and size of defects. In addition, traditional ultrasonic testing results mostly rely on experience and visual judgment, lack of automated and quantitative evaluation systems, and may cause inconsistent test results due to subjective errors. Summary of the invention

[0004] The present invention provides an insulating rod defect detection method and system based on ultrasonic standard atlas, which are used to solve the technical problem that it is difficult to accurately evaluate the defect position and size.

[0005] In a first aspect, the present invention provides a method for preparing an ultrasonic standard spectrum for insulating rod defect detection, comprising: Acquire the ultrasonic original signal of each position point on the two-dimensional unfolded surface of the insulating pull rod, and pre-process each ultrasonic original signal to obtain an ultrasonic target signal; Perform inversion operation on each ultrasonic target signal to obtain ultrasonic standard spectra of the insulating rod at different depths on the two-dimensional development surface; The ultrasonic standard atlas is input into the improved Faster R-CNN model, and the improved Faster R-CNN model outputs the defect detection result corresponding to the ultrasonic standard atlas, wherein the loss function of the Faster R-CNN model is expressed as: , , In the formula, is the total loss of the model, is the number of anchor boxes, is the classification loss, is the correction factor, is the feature map size. When the anchor box is a positive sample, Take 1, otherwise take 0, is the regression loss, is the probability of the i-th anchor box being predicted as the true label, is the bounding box regression offset when predicting the i-th anchor box, is the maximum overlap true bounding box regression offset corresponding to the i-th anchor box, x is the central horizontal coordinate of the predicted anchor box, is the central horizontal coordinate of the overlapping box between the predicted anchor box and the true bounding box, and y is the central vertical coordinate of the predicted anchor box. is the central ordinate of the overlapping box between the predicted anchor box and the true bounding box, w is the width of the predicted anchor box, is the width of the overlapping box between the predicted anchor box and the true bounding box, h is the height of the predicted anchor box, is the width of the overlapping box between the predicted anchor box and the true bounding box, is the central horizontal coordinate of the real bounding box, is the center ordinate of the true bounding box, is the width of the true bounding box, are the height of the true bounding box, is the normalized representation of the horizontal coordinate of the predicted anchor box, is the normalized representation of the vertical coordinate of the predicted anchor box, is the normalized representation of the horizontal coordinates of the overlapping boxes of the predicted anchor box and the true bounding box, is the normalized representation of the ordinate of the overlapping box between the predicted anchor box and the true bounding box, is the logarithmic representation of the width of the predicted anchor box, is the logarithmic representation of the overlap width between the predicted anchor box and the true bounding box, is the logarithmic representation of the predicted anchor box height, It is the logarithmic representation of the overlap box height between the predicted anchor box and the true bounding box.

[0006] In a second aspect, the present invention provides an ultrasonic standard atlas production system for insulating rod defect detection, comprising: An acquisition module is configured to acquire an ultrasonic original signal of each position point on the two-dimensional unfolded surface of the insulating pull rod, and pre-process each ultrasonic original signal to obtain an ultrasonic target signal; An inversion module is configured to perform an inversion operation on each ultrasonic target signal to obtain ultrasonic standard spectra of the insulating rod at different depths on a two-dimensional unfolded surface; The output module is configured to input the ultrasonic standard atlas into the improved Faster R-CNN model, and the improved Faster R-CNN model outputs the defect detection result corresponding to the ultrasonic standard atlas, wherein the loss function of the Faster R-CNN model is expressed as: , , In the formula, is the total loss of the model, is the number of anchor boxes, is the classification loss, is the correction factor, is the feature map size. When the anchor box is a positive sample, Take 1, otherwise take 0, is the regression loss, is the probability of the i-th anchor box being predicted as the true label, is the bounding box regression offset when predicting the i-th anchor box, is the maximum overlap true bounding box regression offset corresponding to the i-th anchor box, x is the central horizontal coordinate of the predicted anchor box, is the central horizontal coordinate of the overlapping box between the predicted anchor box and the true bounding box, and y is the central vertical coordinate of the predicted anchor box. is the central ordinate of the overlapping box between the predicted anchor box and the true bounding box, w is the width of the predicted anchor box, is the width of the overlapping box between the predicted anchor box and the true bounding box, h is the height of the predicted anchor box, is the width of the overlapping box between the predicted anchor box and the true bounding box, is the central horizontal coordinate of the real bounding box, is the center ordinate of the true bounding box, is the width of the true bounding box, are the height of the true bounding box, is the normalized representation of the horizontal coordinate of the predicted anchor box, is the normalized representation of the vertical coordinate of the predicted anchor box, is the normalized representation of the horizontal coordinates of the overlapping boxes of the predicted anchor box and the true bounding box, is the normalized representation of the ordinate of the overlapping box between the predicted anchor box and the true bounding box, is the logarithmic representation of the width of the predicted anchor box, is the logarithmic representation of the overlap width between the predicted anchor box and the true bounding box, is the logarithmic representation of the predicted anchor box height, It is the logarithmic representation of the overlap box height between the predicted anchor box and the true bounding box.

[0007] According to a third aspect, an electronic device is provided, comprising: at least one processor, and a memory communicatively connected to the at least one processor, wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can perform the steps of the insulating rod defect detection method based on ultrasonic standard atlas according to any embodiment of the present invention.

[0008] In a fourth aspect, the present invention further provides a computer-readable storage medium having a computer program stored thereon, wherein when the program instructions are executed by a processor, the processor executes the steps of the insulating rod defect detection method based on ultrasonic standard atlas of any embodiment of the present invention.

[0009] The insulating rod defect detection method and system based on ultrasonic standard atlas of the present application provide a bilinear interpolation data filling method in the preparation of ultrasonic standard atlas. Compared with the traditional nearest neighbor interpolation filling method, the color band distribution in the result is relatively uniform, there is no obvious "plaque", and the inversion atlas is more in line with the actual rod morphology; compared with bicubic interpolation, the calculation amount is small, the calculation speed is faster, and excessive smoothing and overshoot can be avoided, which is more suitable for the processing of a large number of atlases. The improved Faster R-CNN model adopted uses the regional candidate network to generate candidate boxes, which improves the detection accuracy and speeds up the detection. BRIEF DESCRIPTION OF THE DRAWINGS

[0010] In order to more clearly illustrate the technical solutions of the embodiments of the present invention, the following briefly introduces the drawings required for use in the description of the embodiments. Obviously, the drawings described below are some embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on these drawings without creative work.

[0011] Figure 1 A flow chart of a method for detecting defects of insulating pull rods based on ultrasonic standard spectra provided by one embodiment of the present invention; Figure 2 A schematic diagram of the principle of producing an ultrasonic standard spectrum using bilinear interpolation provided by an embodiment of the present invention; Figure 3 A structural block diagram of an insulating rod defect detection system based on ultrasonic standard atlas provided by an embodiment of the present invention; Figure 4 It is a schematic diagram of the structure of an electronic device provided by an embodiment of the present invention. DETAILED DESCRIPTION

[0012] In order to make the purpose, technical solution and advantages of the embodiments of the present invention clearer, the technical solution in the embodiments of the present invention will be clearly and completely described below in conjunction with the drawings in the embodiments of the present invention. Obviously, the described embodiments are part of the embodiments of the present invention, not all of the embodiments. Based on the embodiments of the present invention, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of the present invention.

[0013] See also Figure 1 , which shows a flow chart of an insulating rod defect detection method based on ultrasonic standard atlas of the present application.

[0014] like Figure 1 As shown, the insulating rod defect detection method based on ultrasonic standard atlas specifically includes the following steps: Step S101, obtaining the ultrasonic original signal of each position point on the two-dimensional unfolded surface of the insulating pull rod, and preprocessing each ultrasonic original signal to obtain an ultrasonic target signal.

[0015] In this step, the standard tubular insulating rod specimen with good quality is assembled on the mechanical rotating part of the water immersion ultrasonic device, and the ultrasonic detection preparation work such as installation and fixing is done; The ultrasonic probe is precisely controlled to move along the axial direction of the tubular insulating rod to ensure that every sampling position on the axial direction can be scanned; at the same time, the insulating rod is controlled to rotate around its own axis to ensure that every sampling angle on the circumference can be scanned; through the cooperation of the two movements, full coverage scanning of the standard tubular insulating rod can be achieved; The ultrasonic echo data is recorded in real time to complete the ultrasonic signal acquisition of all points on the two-dimensional unfolded surface of the insulating tubular pull rod.

[0016] It should be noted that the original ultrasonic signal collected at each position point on the two-dimensional unfolded surface of the tubular insulating pull rod is subjected to denoising, gain compensation, signal smoothing and other processing to remove interference signals from the environment or equipment, ensure that the signal strength remains consistent throughout the entire collection process, and improve the clarity and stability of the signal.

[0017] The ultrasonic signal of each acquisition point is analyzed, and the acoustic characteristics such as echo intensity and echo time at the same sound path position in the ultrasonic signal of each position point are extracted to obtain the acoustic wave reflection characteristics of the insulating rod material at different depths.

[0018] Specifically, the original ultrasonic signal is , the original ultrasonic signal Decomposed and expanded on the j scale, the expression is: , In the formula, is the roughness coefficient on the j-1 scale, is the detail coefficient on the j-1 scale, is the scale function, For time, are the translation coefficients that constitute a set of orthogonal scaling functions; , , In the formula, is the roughness coefficient on scale j, is the detail coefficient at scale j, is the high-pass filter impulse response, is the low-pass filter impulse response, As a whole, it is the parameter of the height h of the predicted anchor box. , They are all translation coefficients that constitute a set of orthogonal scaling functions, and their values ​​are different; The coefficients in the high-frequency sub-signal whose absolute values ​​are less than the threshold are set to 0, and the coefficients whose absolute values ​​are greater than the threshold are retained to obtain a new set of wavelet coefficients. The new set of wavelet coefficients is used for wavelet reconstruction to obtain the ultrasonic target signal.

[0019] The signal acoustic feature extraction technology is a sliding window energy detection method, which uses a counting sliding window. The sliding window stores the m most recently arrived elements that meet the following conditions: The energy of the signal in the window is used as the basis for judgment, that is, the signal and noise are distinguished from each other from the perspective of the energy difference. When the energy reaches a certain size, we consider the signal to be an ultrasonic echo signal, otherwise it is considered to be a noise signal. The specific implementation process is as follows: Assume that the sampling period of the signal is T, and the envelope duration of an ultrasonic echo signal is t, then its data window width w=t / T. The signal energy in the window is used as the display variable at the start moment of the window, and then it slides backward at a certain step size to calculate the energy of the signal in the window at the next moment as the display variable at that moment. Repeat the above steps until the energy of the signal in the window at all moments is calculated, thus constructing the sliding window energy envelope curve of the signal.

[0020] Step S102, performing an inversion operation on each ultrasonic target signal to obtain ultrasonic standard spectra of the insulating rod at different depths on a two-dimensional unfolded surface.

[0021] In this step, by preprocessing the ultrasonic echo signal, extracting the echo features, and analyzing the ultrasonic reflection intensity at different depths, after inversion algorithms such as interpolation filling, mapping, and coordinate transformation, the ultrasonic standard map of the insulating rod at different depths on the two-dimensional unfolded surface can be obtained.

[0022] The data interpolation filling method is bilinear interpolation, such as Figure 2 As shown in the figure, the weighted average of the grayscale values ​​of the four sampling points adjacent to the pixel point is taken as the output pixel value. When locating the sampling point, the coordinates of point K are converted from the image coordinate system to the scanning coordinate system. In the scanning coordinate system, the polar coordinates of point K are marked as , select the two nearest scanning lines at both ends of point K, and set the angle of the left scanning line to , the right scan line angle is ; Select two adjacent sampling points on each scanning line so that the polar radius of point K is between the polar radius of these two points. Suppose the polar radius of the two sampling points is and , get the coordinates of the 4 sampling points , , as well as , use H to represent the gray value, then the gray value of point k is: , In the formula, , , , are the grayscale values ​​of the four sampling points G11, G12, G21, and G22 closest to the pixel K; A grayscale image is obtained after bilinear interpolation, and the grayscale image is subjected to pseudo-color mapping to obtain a complete ultrasound standard atlas.

[0023] Step S103: input the ultrasonic standard atlas into an improved Faster R-CNN model, and the improved Faster R-CNN model outputs a defect detection result corresponding to the ultrasonic standard atlas.

[0024] In this step, the improved Faster R-CNN model includes: a feature extraction network, a region candidate network, a region of interest pooling layer, and classification and position regression; Specifically, the ultrasonic standard spectrum is input into the feature extraction network to obtain the feature map of the insulating rod ultrasonic spectrum; using the region candidate network, a one-dimensional vector containing the target probability and the bounding box regression parameter is generated at each position using the sliding window method, and a potential candidate region is generated on the feature map, and a candidate box is generated, and the generated candidate box is projected onto the feature map to generate a corresponding feature matrix; the feature matrix is ​​scaled to a fixed size of 7 pixels * 7 pixels through the region of interest pooling layer, and a classification operation is performed on the fixed size to determine the category to which the candidate region belongs, and regression calculation is performed by optimizing the position and size of the box to output the defect detection result.

[0025] It should be noted that the loss function of the Faster R-CNN model is expressed as: , , In the formula, is the total loss of the model, is the number of anchor boxes, is the classification loss, is the correction factor, is the feature map size. When the anchor box is a positive sample, Take 1, otherwise take 0, is the regression loss, is the probability that the i-th anchor box is predicted as the true label, is the bounding box regression offset when predicting the i-th anchor box, is the maximum overlap rate of the real bounding box regression offset corresponding to the i-th anchor box, x is the central horizontal coordinate of the predicted anchor box, is the central horizontal coordinate of the overlapping box between the predicted anchor box and the true bounding box, and y is the central vertical coordinate of the predicted anchor box. is the central ordinate of the overlapping box between the predicted anchor box and the true bounding box, w is the width of the predicted anchor box, is the width of the overlapping box between the predicted anchor box and the true bounding box, h is the height of the predicted anchor box, is the width of the overlapping box between the predicted anchor box and the true bounding box, is the central horizontal coordinate of the real bounding box, is the center ordinate of the true bounding box, is the width of the true bounding box, are the height of the true bounding box, is the normalized representation of the horizontal coordinate of the predicted anchor box, is the normalized representation of the vertical coordinate of the predicted anchor box, is the normalized representation of the horizontal coordinates of the overlapping boxes of the predicted anchor box and the true bounding box, is the normalized representation of the ordinate of the overlapping box between the predicted anchor box and the true bounding box, is the logarithmic representation of the width of the predicted anchor box, is the logarithmic representation of the overlap width between the predicted anchor box and the true bounding box, is the logarithmic representation of the predicted anchor box height, It is the logarithmic representation of the overlap box height between the predicted anchor box and the true bounding box.

[0026] In summary, in the preparation of ultrasonic standard atlases, the bilinear interpolation data filling method provided by the present invention, compared with the traditional nearest neighbor interpolation filling method, has a relatively uniform distribution of color bands in the results, without obvious "plaques", and the inversion atlas is more in line with the actual pull rod morphology; compared with bicubic interpolation, the calculation amount is small, the calculation speed is faster, and excessive smoothing and overshoot can be avoided, and it is more suitable for the processing of a large number of atlases, and the improved Faster R-CNN model adopted uses the regional candidate network to generate candidate boxes, which improves the detection accuracy and speeds up the detection.

[0027] See also Figure 3 , which shows a structural block diagram of an insulating pull rod defect detection system based on ultrasonic standard atlas of the present application.

[0028] like Figure 3 As shown, the insulating tie rod defect detection system 200 includes an acquisition module 210 , an inversion module 220 and an output module 230 .

[0029] The acquisition module 210 is configured to acquire the ultrasonic original signal of each position point on the two-dimensional unfolded surface of the insulating pull rod, and pre-process each ultrasonic original signal to obtain an ultrasonic target signal; An inversion module 220 is configured to perform an inversion operation on each ultrasonic target signal to obtain ultrasonic standard spectra of the insulating rod at different depths on a two-dimensional development surface; The output module 230 is configured to input the ultrasonic standard atlas into the improved Faster R-CNN model, and the improved Faster R-CNN model outputs the defect detection result corresponding to the ultrasonic standard atlas, wherein the loss function of the Faster R-CNN model is expressed as: , , In the formula, is the total loss of the model, is the number of anchor boxes, is the classification loss, is the correction factor, is the feature map size. When the anchor box is a positive sample, Take 1, otherwise take 0, is the regression loss, is the probability of the i-th anchor box being predicted as the true label, is the bounding box regression offset when predicting the i-th anchor box, is the maximum overlap true bounding box regression offset corresponding to the i-th anchor box, x is the central horizontal coordinate of the predicted anchor box, is the central horizontal coordinate of the overlapping box between the predicted anchor box and the true bounding box, and y is the central vertical coordinate of the predicted anchor box. is the central ordinate of the overlapping box between the predicted anchor box and the true bounding box, w is the width of the predicted anchor box, is the width of the overlapping box between the predicted anchor box and the true bounding box, h is the height of the predicted anchor box, is the width of the overlapping box between the predicted anchor box and the true bounding box, is the central horizontal coordinate of the real bounding box, is the center ordinate of the true bounding box, is the width of the true bounding box, are the height of the true bounding box, is the normalized representation of the horizontal coordinate of the predicted anchor box, is the normalized representation of the vertical coordinate of the predicted anchor box, is the normalized representation of the horizontal coordinates of the overlapping boxes of the predicted anchor box and the true bounding box, is the normalized representation of the ordinate of the overlapping box between the predicted anchor box and the true bounding box, is the logarithmic representation of the width of the predicted anchor box, is the logarithmic representation of the overlap width between the predicted anchor box and the true bounding box, is the logarithmic representation of the predicted anchor box height, It is the logarithmic representation of the overlap box height between the predicted anchor box and the true bounding box.

[0030] It should be understood that Figure 3 Modules and references documented in Figure 1 Therefore, the operations and features described above for the method and the corresponding technical effects are also applicable to Figure 3 The modules in it will not be described in detail here.

[0031] In some other embodiments, the present invention also provides a computer-readable storage medium on which a computer program is stored. When the program instructions are executed by a processor, the processor executes the insulating rod defect detection method based on ultrasonic standard atlas in any of the above method embodiments; As an implementation mode, the computer-readable storage medium of the present invention stores computer-executable instructions, and the computer-executable instructions are configured as follows: Acquire the ultrasonic original signal of each position point on the two-dimensional unfolded surface of the insulating pull rod, and pre-process each ultrasonic original signal to obtain an ultrasonic target signal; Perform inversion operation on each ultrasonic target signal to obtain ultrasonic standard spectra of the insulating rod at different depths on the two-dimensional development surface; The ultrasonic standard atlas is input into an improved Faster R-CNN model, and the improved Faster R-CNN model outputs a defect detection result corresponding to the ultrasonic standard atlas.

[0032] The computer-readable storage medium may include a program storage area and a data storage area, wherein the program storage area may store an operating system and an application required by at least one function; the data storage area may store data created according to the use of the insulating rod defect detection system based on the ultrasonic standard atlas, etc. In addition, the computer-readable storage medium may include a high-speed random access memory, and may also include a memory, such as at least one disk storage device, a flash memory device, or other non-volatile solid-state storage device. In some embodiments, the computer-readable storage medium may optionally include a memory remotely arranged relative to the processor, and these remote memories may be connected to the insulating rod defect detection system based on the ultrasonic standard atlas via a network. Examples of the above-mentioned network include, but are not limited to, the Internet, an intranet, a local area network, a mobile communication network, and combinations thereof.

[0033] Figure 4 is a schematic diagram of the structure of an electronic device provided by an embodiment of the present invention, such as Figure 4 As shown, the device includes: a processor 310 and a memory 320. The electronic device may also include: an input device 330 and an output device 340. The processor 310, the memory 320, the input device 330 and the output device 340 may be connected via a bus or other means. Figure 4 In the example, the bus connection is used. The memory 320 is the above-mentioned computer-readable storage medium. The processor 310 executes various functional applications and data processing of the server by running the non-volatile software programs, instructions and modules stored in the memory 320, that is, the insulating rod defect detection method based on the ultrasonic standard spectrum of the above-mentioned method embodiment is realized. The input device 330 can receive input digital or character information, and generate key signal input related to user settings and function control of the insulating rod defect detection system based on the ultrasonic standard spectrum. The output device 340 may include a display device such as a display screen.

[0034] The electronic device can execute the method provided by the embodiment of the present invention, and has the functional modules and beneficial effects corresponding to the execution method. For technical details not described in detail in this embodiment, please refer to the method provided by the embodiment of the present invention.

[0035] As an implementation mode, the electronic device is applied to an insulating rod defect detection system based on ultrasonic standard atlas, and is used for a client, and includes: at least one processor; and a memory connected to the at least one processor in communication; wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor so that the at least one processor can: Acquire the ultrasonic original signal of each position point on the two-dimensional unfolded surface of the insulating pull rod, and pre-process each ultrasonic original signal to obtain an ultrasonic target signal; Perform inversion operation on each ultrasonic target signal to obtain ultrasonic standard spectra of the insulating rod at different depths on the two-dimensional development surface; The ultrasonic standard atlas is input into an improved Faster R-CNN model, and the improved Faster R-CNN model outputs a defect detection result corresponding to the ultrasonic standard atlas.

[0036] Through the description of the above implementation modes, those skilled in the art can clearly understand that each implementation mode can be implemented by means of software plus a necessary general hardware platform, or of course by hardware. Based on such an understanding, the above technical solution can essentially or in other words be embodied in the form of a software product that contributes to the prior art. The computer software product can be stored in a computer-readable storage medium, such as ROM / RAM, a magnetic disk, an optical disk, etc., and includes a number of instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to execute the methods of each embodiment or some parts of the embodiment.

[0037] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention, rather than to limit it. Although the present invention has been described in detail with reference to the aforementioned embodiments, those skilled in the art should understand that they can still modify the technical solutions described in the aforementioned embodiments, or make equivalent replacements for some of the technical features therein. However, these modifications or replacements do not deviate the essence of the corresponding technical solutions from the spirit and scope of the technical solutions of the embodiments of the present invention.

Claims

1. A method for detecting defects in insulating rods based on ultrasonic standard spectra, characterized in that: include: Acquire the ultrasonic original signal of each position point on the two-dimensional unfolded surface of the insulating pull rod, and pre-process each ultrasonic original signal to obtain an ultrasonic target signal; Perform inversion operation on each ultrasonic target signal to obtain ultrasonic standard spectra of the insulating rod at different depths on the two-dimensional development surface; The ultrasonic standard atlas is input into the improved Faster R-CNN model, and the improved Faster R-CNN model outputs the defect detection result corresponding to the ultrasonic standard atlas, wherein the loss function of the Faster R-CNN model is expressed as: ， ， In the formula, is the total loss of the model, is the number of anchor boxes, is the classification loss, is the correction factor, is the feature map size. When the anchor box is a positive sample, Take 1, otherwise take 0, is the regression loss, is the probability that the i-th anchor box is predicted as the true label, is the bounding box regression offset when predicting the i-th anchor box, is the maximum overlap rate of the real bounding box regression offset corresponding to the i-th anchor box, x is the central horizontal coordinate of the predicted anchor box, is the central horizontal coordinate of the overlapping box between the predicted anchor box and the true bounding box, and y is the central vertical coordinate of the predicted anchor box. is the central ordinate of the overlapping box between the predicted anchor box and the true bounding box, w is the width of the predicted anchor box, is the width of the overlapping box between the predicted anchor box and the true bounding box, h is the height of the predicted anchor box, is the width of the overlapping box between the predicted anchor box and the true bounding box, is the central horizontal coordinate of the real bounding box, is the center ordinate of the true bounding box, is the width of the true bounding box, are the height of the true bounding box, is the normalized representation of the horizontal coordinate of the predicted anchor box, is the normalized representation of the vertical coordinate of the predicted anchor box, is the normalized representation of the horizontal coordinates of the overlapping boxes of the predicted anchor box and the true bounding box, is the normalized representation of the ordinate of the overlapping box between the predicted anchor box and the true bounding box, is the logarithmic representation of the width of the predicted anchor box, is the logarithmic representation of the overlap width between the predicted anchor box and the true bounding box, is the logarithmic representation of the predicted anchor box height, It is the logarithmic representation of the overlap box height between the predicted anchor box and the true bounding box.

2. The method for detecting defects of insulating pull rods based on ultrasonic standard atlas according to claim 1 is characterized in that: The preprocessing of each ultrasonic original signal to obtain an ultrasonic target signal comprises: The original ultrasonic signal is , the original ultrasonic signal Decomposed and expanded on the j scale, the expression is: ， In the formula, is the roughness coefficient on the j-1 scale, is the detail coefficient on the j-1 scale, is the scale function, For time, are the translation coefficients that constitute a set of orthogonal scaling functions; ， ， In the formula, is the roughness coefficient on scale j, is the detail coefficient at scale j, is the high-pass filter impulse response, is the low-pass filter impulse response, As a whole, it is the parameter of the height h of the predicted anchor box. , They are all translation coefficients that constitute a set of orthogonal scaling functions, and their values ​​are different; The coefficients in the high-frequency sub-signal whose absolute values ​​are less than the threshold are set to 0, and the coefficients whose absolute values ​​are greater than the threshold are retained to obtain a new set of wavelet coefficients. The new set of wavelet coefficients is used for wavelet reconstruction to obtain the ultrasonic target signal.

3. The method for detecting defects of insulating pull rods based on ultrasonic standard atlas according to claim 1 is characterized in that: The inversion operation is performed on each ultrasonic target signal to obtain ultrasonic standard spectra of the insulating pull rod at different depths on the two-dimensional development surface, including: Take the weighted average of the grayscale values ​​of the four sampling points adjacent to the pixel as the output pixel value. When locating the sampling point, convert the coordinates of point K from the image coordinate system to the scanning coordinate system. In the scanning coordinate system, the polar coordinates of point K are marked as , select the two nearest scanning lines at both ends of point K, and set the angle of the left scanning line to , the right scan line angle is ; Select two adjacent sampling points on each scanning line so that the polar radius of point K is between the polar radius of these two points. Suppose the polar radius of the two sampling points is and , get the coordinates of the 4 sampling points , , as well as , use H to represent the gray value, then the gray value of point k is: , In the formula, , , , are the grayscale values ​​of the four sampling points G11, G12, G21, and G22 closest to the pixel K; A grayscale image is obtained after bilinear interpolation, and the grayscale image is subjected to pseudo-color mapping to obtain a complete ultrasound standard atlas.

4. The method for detecting defects of insulating pull rods based on ultrasonic standard atlas according to claim 1, characterized in that: The improved Faster R-CNN model includes: a feature extraction network, a region candidate network, a region of interest pooling layer, and classification and position regression; The step of inputting the ultrasonic standard atlas into the improved Faster R-CNN model, wherein the improved Faster R-CNN model outputs a defect detection result corresponding to the ultrasonic standard atlas, comprises: Inputting the ultrasonic standard spectrum into a feature extraction network to obtain a feature graph of the insulating rod ultrasonic spectrum; Using the region proposal network, a one-dimensional vector containing target probability and bounding box regression parameters is generated at each position using a sliding window method, potential candidate regions are generated on the feature map, candidate boxes are generated, and the generated candidate boxes are projected onto the feature map to generate the corresponding feature matrix; The feature matrix is ​​scaled to a fixed size of 7 pixels * 7 pixels through the region of interest pooling layer, and classification operations are performed on the fixed size to determine the category to which the candidate region belongs. Regression calculations are performed by optimizing the position and size of the box to output the defect detection results.

5. An insulating rod defect detection system based on ultrasonic standard atlas, characterized in that: include: An acquisition module is configured to acquire an ultrasonic original signal of each position point on the two-dimensional unfolded surface of the insulating pull rod, and pre-process each ultrasonic original signal to obtain an ultrasonic target signal; An inversion module is configured to perform an inversion operation on each ultrasonic target signal to obtain ultrasonic standard spectra of the insulating rod at different depths on a two-dimensional unfolded surface; The output module is configured to input the ultrasonic standard atlas into the improved Faster R-CNN model, and the improved Faster R-CNN model outputs the defect detection result corresponding to the ultrasonic standard atlas, wherein the loss function of the Faster R-CNN model is expressed as: ， ， In the formula, is the total loss of the model, is the number of anchor boxes, is the classification loss, is the correction factor, is the feature map size. When the anchor box is a positive sample, Take 1, otherwise take 0, is the regression loss, is the probability that the i-th anchor box is predicted as the true label, is the bounding box regression offset when predicting the i-th anchor box, is the maximum overlap rate of the real bounding box regression offset corresponding to the i-th anchor box, x is the central horizontal coordinate of the predicted anchor box, is the central horizontal coordinate of the overlapping box between the predicted anchor box and the true bounding box, and y is the central vertical coordinate of the predicted anchor box. is the central ordinate of the overlapping box between the predicted anchor box and the true bounding box, w is the width of the predicted anchor box, is the width of the overlapping box between the predicted anchor box and the true bounding box, h is the height of the predicted anchor box, is the width of the overlapping box between the predicted anchor box and the true bounding box, is the central horizontal coordinate of the real bounding box, is the center ordinate of the true bounding box, is the width of the true bounding box, are the height of the true bounding box, is the normalized representation of the horizontal coordinate of the predicted anchor box, is the normalized representation of the vertical coordinate of the predicted anchor box, is the normalized representation of the horizontal coordinates of the overlapping boxes of the predicted anchor box and the true bounding box, is the normalized representation of the ordinate of the overlapping box between the predicted anchor box and the true bounding box, is the logarithmic representation of the width of the predicted anchor box, is the logarithmic representation of the overlap width between the predicted anchor box and the true bounding box, is the logarithmic representation of the predicted anchor box height, It is the logarithmic representation of the overlap box height between the predicted anchor box and the true bounding box.

6. An electronic device, characterized in that: include: At least one processor, and a memory communicatively connected to the at least one processor, wherein the memory stores instructions executable by the at least one processor, and the instructions are executed by the at least one processor to enable the at least one processor to perform the method described in any one of claims 1 to 4.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that: When the program is executed by a processor, the method according to any one of claims 1 to 4 is implemented.

Images