Insulator filthy component analysis method and analysis system based on deep learning

Abstract

The invention relates to an insulator contamination component analysis method and analysis system based on deep learning. The method comprises the following steps: extracting an insulator object image in an image; performing gray processing on the insulator object image to obtain an insulator object gray image; cutting the insulator spectral image by using the sub-insulator object grayscale image to obtain a sub-insulator spectral image; establishing an analysis reference line on the sub-insulator spectral image, and establishing an analysis curve by using pixel points on the analysis reference line; and processing the analysis curve by using autocorrelation calculation and a fast Fourier transform mode to obtain a power spectrum, selecting a filthy component identification model obtained based on deep learning according to the power spectrum, and determining filthy components included in the sub-insulator object grayscale image. The invention discloses an insulator contamination component analysis method and analysis system based on deep learning. According to the method and the system, contamination and contamination components on an insulator are determined through intelligent recognition of a contamination area on the insulator and model analysis of contamination area components.

Description

Technical Field

[0001] The present invention relates to the field of data processing technology, and in particular to an insulator pollution component analysis method and analysis system based on deep learning. Background Art

[0002] The insulator contamination degree refers to the degree of contamination on the surface of the insulator. The two commonly used evaluation indicators are equivalent salt density and equivalent ash density. The equivalent salt density refers to the amount of NaCl (mg / cm2) equivalent to the content of conductive substances in the contamination attached to the surface of the insulator per square centimeter. The equivalent salt density characterizes the conductivity of the insulator surface after the contamination is fully dissolved. The increase in the equivalent salt density will significantly reduce the insulator pollution flashover voltage.

[0003] For these two evaluation indicators, the current implementation methods are mostly to directly sample insulators or collect insulator model samples. These methods are lacking in ease of implementation, sample accuracy and timeliness, and technical updates are needed.

[0004] A suitable method is to use spectral images combined with machine processing to determine the pollution components on insulators. This method can achieve long-distance non-contact collection and is better in terms of ease of implementation, sample accuracy and timeliness.

[0005] However, spectral images have the problem of large data volume and interference, and the contamination on insulators also has the characteristics of not having a predetermined shape, so further research is needed on how to analyze it. Summary of the invention

[0006] The present invention provides an insulator contamination component analysis method and analysis system based on deep learning, which determines the contamination and its components on the insulator by intelligently identifying the contamination area on the insulator and performing model analysis on the components of the contamination area.

[0007] The above object of the present invention is achieved through the following technical solutions: In a first aspect, the present invention provides a method for analyzing insulator contamination components based on deep learning, comprising: In response to the received image containing the insulator, extracting the insulator object image in the image; Performing grayscale processing on the insulator object image to obtain an insulator object grayscale image; Decomposing and grouping the insulator object grayscale image to obtain an insulator object grayscale image group, each insulator object grayscale image group includes a class of sub-insulator object grayscale images, and the area of ​​the sub-insulator object grayscale image is greater than or equal to an area threshold; The insulator spectrum image is cropped using the sub-insulator object grayscale image to obtain the sub-insulator spectrum image; Establishing an analysis reference line on the sub-insulator spectrum image and establishing an analysis curve using pixel points on the analysis reference line; Use autocorrelation calculation and fast Fourier transform to process the analysis curve and obtain the power spectrum; According to the power spectrum, a pollution component recognition model obtained based on deep learning is selected and the pollution components included in the grayscale image of the sub-insulator object are determined.

[0008] In a possible implementation manner of the first aspect, decomposing the grayscale image of the insulator object includes: Using the grayscale interval to perform content extraction on the grayscale image of the insulator object and establishing a change trend line on the grayscale image of the insulator object according to the content extraction result; Count the convergence area of ​​the changing trend lines and determine the center position of the convergence area; Establish a direction line with the center position as the starting point; Calculate the speed of value change on the direction line and determine the boundary point according to the speed of value change; The obtained boundary points are connected sequentially to obtain the grayscale image of the sub-insulator object.

[0009] In a possible implementation manner of the first aspect, performing content extraction on the grayscale image of the insulator object using the grayscale interval and establishing a change trend line on the grayscale image of the insulator object according to the content extraction result includes: The content extraction results are graded, and each level of content extraction results corresponds to a grayscale interval; Dividing the content extraction results to obtain content extraction result regions, each of which has the same area; Connect multiple content extraction result areas according to the trend to obtain a change trend line; Count the aggregation areas of the change trend line and adjust the content extraction result areas participating in the same change trend line to minimize the number of connected aggregation areas; The content extraction result regions participating in the same change trend line come from different levels, and any two content extraction result regions in the same level do not participate in the same change trend line.

[0010] In a possible implementation manner of the first aspect, calculating a value change speed on the direction line and determining a boundary point according to the value change speed includes: Set reference points at intervals on the direction line, and the distances between adjacent reference points are equal; Calculate the average of the reference point and the pixels around the reference point and use the average as the value of the reference point; Count the quadratic difference series of numerical values ​​and use the quadratic difference series to draw the quadratic difference series curve; Determine similar areas on all obtained quadratic difference series curves; Use the starting points of the similar area as boundary points.

[0011] In a possible implementation manner of the first aspect, selecting, according to the power spectrum, a pollution component identification model obtained based on deep learning includes: Obtaining model features of the pollution component identification model based on deep learning, the model features include frequency quantity, frequency intensity, frequency proportion and frequency interval; Use the model features to match in the power spectrum to determine the pollution component identification model; Among them, the frequency intensity is less than or equal to the intensity of the corresponding frequency in the power spectrum.

[0012] In a possible implementation manner of the first aspect, after the analysis curve is established using the pixel points on the analysis reference line, the method further includes removing interference on the analysis curve. The removing interference on the analysis curve includes: Process the analysis curve by wavelet transform to obtain multiple sub-analysis curves, each of which has a starting point and a cutoff point. The sub-analysis curve is screened once using the transformation length to obtain a grayscale image of a processed sub-insulator object, and a clarity value of the grayscale image of the processed sub-insulator object is calculated; When the clarity value of the grayscale image of the sub-insulator object processed once meets the requirement, the sub-analysis curve is screened once using the transformation length.

[0013] In a possible implementation manner of the first aspect, there is at least one endpoint of a sub-analysis curve that has not been deleted between the starting position point and the ending position point of the deleted sub-analysis curve; Delete other sub-analysis curves with the same frequency as the deleted sub-analysis curve; When the content loss amount of the grayscale image of the sub-insulator object exceeds the allowable range, the other sub-analysis curves to be deleted are adjusted, and the adjustment includes: Setting a deletion length and retaining the other sub-analysis curves whose length is greater than the deletion length; The content loss amount of the grayscale image of the sub-insulator object is recalculated, and the deletion length is stopped from being modified when the content loss amount is within the allowable range.

[0014] In a second aspect, the present invention provides an insulator contamination component analysis device based on deep learning, comprising: A content extraction unit, configured to extract an insulator object image in response to a received image containing an insulator; A grayscale processing unit, used for performing grayscale processing on the insulator object image to obtain an insulator object grayscale image; A first processing unit is used to decompose and group the insulator object grayscale image to obtain an insulator object grayscale image group, each of which includes a type of sub-insulator object grayscale image, and the area of ​​the sub-insulator object grayscale image is greater than or equal to an area threshold; The second processing unit is used to cut the insulator spectrum image using the sub-insulator object grayscale image to obtain the sub-insulator spectrum image; A third processing unit is used to establish an analysis reference line on the sub-insulator spectrum image and to establish an analysis curve using pixel points on the analysis reference line; A spectrum processing unit, used for processing the analysis curve using autocorrelation calculation and fast Fourier transform to obtain a power spectrum; The component determination unit is used to select a pollution component identification model obtained based on deep learning according to the power spectrum and determine the pollution components included in the grayscale image of the sub-insulator object.

[0015] In a third aspect, the present invention provides an insulator pollution component analysis system based on deep learning, the system comprising: one or more memories for storing instructions; and One or more processors, used to call and run the instructions from the memory to execute the method as described in the first aspect and any possible implementation of the first aspect.

[0016] In a fourth aspect, the present invention provides a computer-readable storage medium, the computer-readable storage medium comprising: Program, when the program is executed by a processor, the method described in the first aspect and any possible implementation of the first aspect is executed.

[0017] In a fifth aspect, the present invention provides a computer program product, comprising program instructions, and when the program instructions are executed by a computing device, the method described in the first aspect and any possible implementation of the first aspect is executed.

[0018] In a sixth aspect, the present invention provides a chip system comprising a processor for implementing the functions involved in the above aspects, for example, generating, receiving, sending, or processing the data and / or information involved in the above methods.

[0019] The chip system may be composed of chips, or may include chips and other discrete devices.

[0020] In a possible design, the chip system also includes a memory, which is used to store necessary program instructions and data. The processor and the memory can be decoupled and respectively set on different devices, connected by wired or wireless means, or the processor and the memory can also be coupled on the same device. BRIEF DESCRIPTION OF THE DRAWINGS

[0021] Figure 1 It is a schematic block diagram of the steps of an insulator pollution component analysis method based on deep learning provided by the present invention.

[0022] Figure 2 It is a schematic diagram of decomposing and grouping a grayscale image of an insulator object provided by the present invention.

[0023] Figure 3 It is a schematic diagram of a power spectrum provided by the present invention.

[0024] Figure 4 It is a schematic diagram of establishing a change trend line on a grayscale image of an insulator object provided by the present invention.

[0025] Figure 5 It is a schematic diagram of a changing trend line provided by the present invention.

[0026] Figure 6 It is a schematic diagram of a direction line provided by the present invention.

[0027] Figure 7 It is a schematic diagram of a similar area on a quadratic difference series curve provided by the present invention. DETAILED DESCRIPTION

[0028] The technical solution of the present invention is further described in detail below in conjunction with the accompanying drawings.

[0029] The present invention discloses a method for analyzing the pollution components of insulators based on deep learning. Figure 1 In some examples, the insulator pollution component analysis method based on deep learning disclosed in the present invention includes the following steps: S101, in response to a received image containing an insulator, extracting an insulator object image in the image; S102, performing grayscale processing on the insulator object image to obtain an insulator object grayscale image; S103, decomposing and grouping the insulator object grayscale image to obtain an insulator object grayscale image group, each of which includes a type of sub-insulator object grayscale image, and the area of ​​the sub-insulator object grayscale image is greater than or equal to an area threshold; S104, using the sub-insulator object grayscale image to crop the insulator spectrum image to obtain the sub-insulator spectrum image; S105, establishing an analysis reference line on the sub-insulator spectrum image and establishing an analysis curve using pixel points on the analysis reference line; S106, using autocorrelation calculation and fast Fourier transform to process the analysis curve and obtain a power spectrum; S107, selecting a pollution component identification model obtained based on deep learning according to the power spectrum and determining the pollution components included in the grayscale image of the sub-insulator object.

[0030] In general, the insulator pollution component analysis method based on deep learning provided by the present invention mainly solves two problems, namely, solving the pollution identification problem and the pollution component analysis problem. The specific method is to identify pollution through ordinary images and then analyze the pollution components through spectral images.

[0031] In step S101, first, an insulator object image is extracted from the received image containing the insulator. The insulator object image can be extracted using an image feature extraction process, since the type, shape and color of the insulator can be fixed, and can be extracted using a neural network that has been specifically trained.

[0032] In step S102, the insulator object image is gray-scale processed to obtain an insulator object gray-scale image, and then in step S103, the insulator object gray-scale image is decomposed and grouped. Figure 2 , get the grayscale image group of insulator objects, decomposition is to get the grayscale image part of the insulator object containing dirt, grouping is to put the same type of dirt into a group, Figure 2 The two solid rectangles at the middle arrow are a group, and the two dotted rectangles are a group.

[0033] That is, each insulator object grayscale image group includes a class of sub-insulator object grayscale images. Of course, the area of ​​the sub-insulator object grayscale image is also required to be greater than or equal to the area threshold. The purpose is to avoid falling into a local optimal solution or a no-solution cycle during the processing process.

[0034] In step S104, the insulator spectral image is cropped using the sub-insulator object grayscale image to obtain a sub-insulator spectral image. The sub-insulator object grayscale image corresponds to the position where contamination exists on the insulator. The sub-insulator spectral image obtained by cropping the insulator spectral image using the sub-insulator object grayscale image also corresponds to the position where contamination exists on the insulator.

[0035] By using this processing method, only the part of the insulator spectrum image containing dirt can be processed, thereby avoiding the overall processing of the insulator spectrum image.

[0036] In step S105, an analysis reference line is established on the sub-insulator spectrum image and an analysis curve is established using the pixels on the analysis reference line. Then, in step S106, the analysis curve is processed using autocorrelation calculation and fast Fourier transform to obtain a power spectrum ( Figure 3As shown), finally in step S107, a pollution component recognition model obtained based on deep learning is selected according to the power spectrum and the pollution components included in the grayscale image of the sub-insulator object are determined.

[0037] Specifically, in the aforementioned steps, the area where contamination exists on the insulator is first determined and the sub-insulator spectral image corresponding to the contamination area is obtained. The sub-insulator spectral image includes contamination component information. The content of steps S105 to S107 is to analyze the contamination component information.

[0038] When performing component analysis, firstly, an analysis reference line is established on the sub-insulator spectrum image and an analysis curve is established using the pixel points on the analysis reference line. The sub-insulator spectrum image here can be one or more.

[0039] It should be understood that the spectral image is an image obtained in the order of frequency windows, and each frequency window corresponds to a frequency range. Based on this, the sub-insulator spectral image can be regarded as being caused by multiple separate spectral images, or it can be formed by merging these separate spectral images.

[0040] After establishing an analysis reference line on the sub-insulator spectral image and using the pixels on the analysis reference line to establish an analysis curve, the analysis curve is processed using autocorrelation calculation and fast Fourier transform to obtain the power spectrum. After obtaining the power spectrum through autocorrelation calculation and fast Fourier transform, it is necessary to determine the power spectrum to determine the contamination components.

[0041] The specific method is to first determine the possible components based on the power spectrum. This method is a fuzzy judgment method. For example, assuming that the power spectrum corresponding to a certain contamination includes three components A, B, and C, if the power spectrum at this time also contains the three components A, B, and C, then it will be suspected that this type of contamination exists on the insulator.

[0042] For the three components A, B, and C, they are generally the main components of a certain pollution rather than all the components. Based on this method, the pollution component identification model obtained based on deep learning can be selected first. At this time, the number of pollution component identification models obtained is multiple.

[0043] Then the power spectrum is given to the pollution component recognition model, which determines whether the corresponding pollution component exists.

[0044] In the above manner, the pollution component recognition model is trained separately and only recognizes one pollution component. Here, a recognition model composed of multiple pollution component recognition models that can simultaneously recognize multiple pollution components is also a variant of the pollution component recognition model mentioned in the present invention and should be included in the scope of the present invention.

[0045] In some possible implementations, the method of establishing an analysis reference line on the sub-insulator spectral image is to establish a straight line as long as possible on the sub-insulator spectral image. The number of analysis reference lines can be one or more. When the number of analysis reference lines is multiple, after obtaining the results, the type of results with the largest proportion in the results is used as the final result.

[0046] In some examples, the specific way of decomposing the grayscale image of the insulator object is as follows: S201, performing content extraction on the grayscale image of the insulator object using the grayscale interval and establishing a change trend line on the grayscale image of the insulator object according to the content extraction result; S202, counting the convergence area of ​​the change trend lines and determining the center position of the convergence area; S203, establishing a direction line with the center position as the starting point; S204, calculating the speed of value change on the direction line and determining the boundary point according to the speed of value change; S205, sequentially connect the obtained boundary points to obtain a grayscale image of the sub-insulator object.

[0047] In step S201 to step S205, firstly, the grayscale interval is used to perform content extraction on the grayscale image of the insulator object and a change trend line is established on the grayscale image of the insulator object according to the content extraction result. Figure 4 As shown, the grayscale interval is a numerical interval. The grayscale image of the insulator object is processed using the grayscale interval, and a content extraction result can be obtained for each grayscale interval.

[0048] The specific method of using the grayscale interval to perform content extraction on the grayscale image of the insulator object and establishing a change trend line on the grayscale image of the insulator object according to the content extraction result is as follows: S301, grading the content extraction results, each level of content extraction results corresponds to a grayscale interval; S302, dividing the content extraction result to obtain content extraction result regions, where the area of ​​each content extraction result region is the same; S303, connecting multiple content extraction result areas according to the trend to obtain a change trend line; S304, counting the collection areas of the change trend lines and adjusting the content extraction result areas participating in the same change trend line to minimize the number of connected collection areas; The content extraction result regions participating in the same change trend line come from different levels, and any two content extraction result regions in the same level do not participate in the same change trend line.

[0049] After obtaining the content extraction result, the content extraction result is first classified, and each level of the content extraction result corresponds to a grayscale interval. A suitable classification result is to first sort the grayscale intervals, and the higher the sequence number, the higher the level, and the higher the classification level.

[0050] Then, the content extraction results are divided into content extraction result regions. The area of ​​each content extraction result region is the same. Division is to divide the content extraction results into independent regions one by one. After the region division is completed, multiple content extraction result regions are connected according to the trend to obtain the change trend line ( Figure 5 ).

[0051] Here, it is required that the content extraction result regions participating in the same change trend line come from different levels, and any two content extraction result regions in the same level do not participate in the same change trend line.

[0052] Following the trend means that the color becomes darker or lighter, both methods are possible and there is no restriction here.

[0053] Then, the convergence area of ​​the change trend line is counted and the content extraction result area participating in the same change trend line is adjusted to minimize the number of connected convergence areas. The convergence area of ​​the change trend line obtained here is generally the center area of ​​the filth or an area close to the center area.

[0054] This method can ignore the influence of uncertain shapes and more accurately obtain the center area or the area close to the center area of ​​the contamination.

[0055] In some examples, the specific method of calculating the speed of value change on the direction line and determining the boundary point according to the speed of value change is as follows: S401, setting reference points on the direction line at intervals, with the distances between adjacent reference points being equal; S402, calculating the average of the reference point and the pixels around the reference point and taking the average as the value of the reference point; S403, counting the quadratic difference series of numerical values ​​and using the quadratic difference series to draw a quadratic difference series curve; S404, determining similar regions on all obtained quadratic difference series curves; S405, using the starting point of the similar area as a boundary point.

[0056] In step S401 to step S405, first, the direction line ( Figure 6The reference points are set at intervals on the dotted line in the figure, and then the average of the pixels around the reference point is calculated and the average is used as the value of the reference point. The purpose of calculating the average of the pixels around the reference point is to ensure the stability of the pixel average feedback, because the value of a single pixel has a certain error, and when selecting the reference point, there may be a selection deviation problem due to inappropriate position, and the use of the average calculation method can avoid the above problems.

[0057] After obtaining the value of the reference point, the quadratic difference series of the values ​​is counted and the quadratic difference series curve is drawn using the quadratic difference series ( Figure 7 Then determine the similar areas on all the obtained quadratic difference series curves, and finally use the starting points of the similar areas as boundary points.

[0058] Here, for the similar area on the quadratic difference series curve ( Figure 7 As shown in FIG. 1 ), it refers to the area where the slope changes suddenly, and the sudden change in slope is due to the appearance of an edge. In the present invention, the starting points of similar areas are used as boundary points, and then these boundary points are sequentially connected together to obtain the boundary.

[0059] In some examples, after establishing the analysis curve using the pixel points on the analysis reference line, the interference on the analysis curve is removed. The specific method of removing the interference on the analysis curve is: S501, using wavelet transform to process the analysis curve to obtain multiple sub-analysis curves, each of which has a starting point and a cutoff point; S502, using the transformation length to perform a primary screening on the sub-analysis curve, obtaining a primary processed sub-insulator object grayscale image, and calculating a clarity value of the primary processed sub-insulator object grayscale image; S503, when the clarity value of the grayscale image of the sub-insulator object processed once meets the requirement, stop using the transformation length to perform a screening on the sub-analysis curve.

[0060] In the above steps, the analysis curve is first processed using wavelet transform, and multiple sub-analysis curves will be obtained. Then, the sub-analysis curves are screened using the transformation length. The transformation length means that each time the sub-analysis curves are screened, the sub-analysis curves that are less than a certain set length will be deleted. Each time they are deleted, the clarity value of the grayscale image of the processed sub-insulator object is calculated.

[0061] When the clarity value of the grayscale image of the sub-insulator object processed once meets the requirement, stop using the transformation length to screen the sub-analysis curve once. If the clarity value does not meet the requirement, adjust the set length for the next screening.

[0062] The methods for judging clarity include Brenner gradient function, Tenengrad gradient function, Laplacian gradient function, SMD (grayscale variance) function and SMD2 (grayscale variance product) function.

[0063] During the processing, there are also the following requirements: There is at least one endpoint of a sub-analysis curve that has not been deleted between the starting point and the ending point of the deleted sub-analysis curve. This processing method can avoid deleting isolated points, and the purpose is to retain content within the allowed range.

[0064] This is because the present invention believes that the interference that occurs should be mixed with non-interference data. When a data exists alone, it may be interference data or non-interference data. If too much data is deleted, it will affect the subsequent result judgment. Therefore, the present invention requires that there is at least one endpoint of a sub-analysis curve that has not been deleted between the starting position point and the end position point of the deleted sub-analysis curve.

[0065] In addition, it is also necessary to delete other sub-analysis curves with the same frequency as the deleted sub-analysis curve. This is a simplified processing method, that is, all sub-analysis curves of a certain frequency are treated as interference data.

[0066] However, when the content loss amount of the grayscale image of the sub-insulator object exceeds the allowable range, the other sub-analysis curves are adjusted and deleted, and the adjustment includes: Setting a deletion length and retaining the other sub-analysis curves whose length is greater than the deletion length; The content loss amount of the grayscale image of the sub-insulator object is recalculated, and the deletion length is stopped from being modified when the content loss amount is within the allowable range.

[0067] In some examples, the specific method of selecting the pollution component identification model obtained by deep learning based on the power spectrum is as follows: Obtaining model features of the pollution component identification model based on deep learning, the model features include frequency quantity, frequency intensity, frequency proportion and frequency interval; Use the model features to match in the power spectrum to determine the pollution component identification model; Among them, the frequency intensity is less than or equal to the intensity of the corresponding frequency in the power spectrum.

[0068] In the above method, the model features are first determined, and then the model features are used to match in the power spectrum to determine the pollution component identification model. The model features include four feature quantities: frequency number, frequency intensity, frequency proportion and frequency interval.

[0069] When the model features are matched in the power spectrum, if one or more of the frequency quantity, frequency intensity, frequency proportion and frequency interval are consistent, the pollution component identification model will be retained. After this step is completed, multiple pollution component identification models can be screened from all the pollution component identification models, and each pollution component identification model can be used to identify a pollution component.

[0070] The present invention also provides an insulator pollution component analysis device based on deep learning, comprising: A content extraction unit, configured to extract an insulator object image in response to a received image containing an insulator; A grayscale processing unit, used for performing grayscale processing on the insulator object image to obtain an insulator object grayscale image; A first processing unit is used to decompose and group the insulator object grayscale image to obtain an insulator object grayscale image group, each of which includes a type of sub-insulator object grayscale image, and the area of ​​the sub-insulator object grayscale image is greater than or equal to an area threshold; The second processing unit is used to cut the insulator spectrum image using the sub-insulator object grayscale image to obtain the sub-insulator spectrum image; A third processing unit is used to establish an analysis reference line on the sub-insulator spectrum image and to establish an analysis curve using pixel points on the analysis reference line; A spectrum processing unit, used for processing the analysis curve using autocorrelation calculation and fast Fourier transform to obtain a power spectrum; The component determination unit is used to select a pollution component identification model obtained based on deep learning according to the power spectrum and determine the pollution components included in the grayscale image of the sub-insulator object.

[0071] In a possible implementation manner of the first aspect, decomposing the grayscale image of the insulator object includes: Using the grayscale interval to perform content extraction on the grayscale image of the insulator object and establishing a change trend line on the grayscale image of the insulator object according to the content extraction result; Count the convergence area of ​​the changing trend lines and determine the center position of the convergence area; Establish a direction line with the center position as the starting point; Calculate the speed of value change on the direction line and determine the boundary point according to the speed of value change; The obtained boundary points are connected sequentially to obtain the grayscale image of the sub-insulator object.

[0072] In a possible implementation manner of the first aspect, performing content extraction on the grayscale image of the insulator object using the grayscale interval and establishing a change trend line on the grayscale image of the insulator object according to the content extraction result includes: The content extraction results are graded, and each level of content extraction results corresponds to a grayscale interval; Dividing the content extraction results to obtain content extraction result regions, each of which has the same area; Connect multiple content extraction result areas according to the trend to obtain a change trend line; Count the aggregation areas of the change trend line and adjust the content extraction result areas participating in the same change trend line to minimize the number of connected aggregation areas; The content extraction result regions participating in the same change trend line come from different levels, and any two content extraction result regions in the same level do not participate in the same change trend line.

[0073] In a possible implementation manner of the first aspect, calculating a value change speed on the direction line and determining a boundary point according to the value change speed includes: Set reference points at intervals on the direction line, and the distances between adjacent reference points are equal; Calculate the average of the reference point and the pixels around the reference point and use the average as the value of the reference point; Count the quadratic difference series of numerical values ​​and use the quadratic difference series to draw the quadratic difference series curve; Determine similar areas on all obtained quadratic difference series curves; Use the starting points of the similar area as boundary points.

[0074] In a possible implementation manner of the first aspect, selecting, according to the power spectrum, a pollution component identification model obtained based on deep learning includes: Obtaining model features of the pollution component identification model based on deep learning, the model features include frequency quantity, frequency intensity, frequency proportion and frequency interval; Use the model features to match in the power spectrum to determine the pollution component identification model; Among them, the frequency intensity is less than or equal to the intensity of the corresponding frequency in the power spectrum.

[0075] In a possible implementation manner of the first aspect, after the analysis curve is established using the pixel points on the analysis reference line, the method further includes removing interference on the analysis curve. The removing interference on the analysis curve includes: Process the analysis curve by wavelet transform to obtain multiple sub-analysis curves, each of which has a starting point and a cutoff point. The sub-analysis curve is screened once using the transformation length to obtain a grayscale image of a processed sub-insulator object, and a clarity value of the grayscale image of the processed sub-insulator object is calculated; When the clarity value of the grayscale image of the sub-insulator object processed once meets the requirement, the sub-analysis curve is screened once using the transformation length.

[0076] In a possible implementation manner of the first aspect, there is at least one endpoint of a sub-analysis curve that has not been deleted between the starting position point and the ending position point of the deleted sub-analysis curve; Delete other sub-analysis curves with the same frequency as the deleted sub-analysis curve; When the content loss amount of the grayscale image of the sub-insulator object exceeds the allowable range, the other sub-analysis curves to be deleted are adjusted, and the adjustment includes: Setting a deletion length and retaining the other sub-analysis curves whose length is greater than the deletion length; The content loss amount of the grayscale image of the sub-insulator object is recalculated, and the deletion length is stopped from being modified when the content loss amount is within the allowable range.

[0077] In one example, the unit in any of the above devices can be one or more integrated circuits configured to implement the above methods, such as: one or more application specific integrated circuits (ASICs), or, one or more digital signal processors (DSPs), or, one or more field programmable gate arrays (FPGAs), or a combination of at least two of these integrated circuit forms.

[0078] For another example, when the units in the device can be implemented in the form of a processing element scheduling program, the processing element can be a general-purpose processor, such as a central processing unit (CPU) or other processor that can call a program. For another example, these units can be integrated together and implemented in the form of a system-on-a-chip (SOC).

[0079] Various objects such as various messages / information / equipment / network elements / systems / devices / actions / operations / processes / concepts that may appear in the present invention are named. It can be understood that these specific names do not constitute a limitation on the relevant objects. The names assigned may change with factors such as scenarios, contexts or usage habits. The understanding of the technical meaning of the technical terms in the present invention should be mainly determined from the functions and technical effects embodied / executed in the technical scheme.

[0080] Those skilled in the art can clearly understand that, for the convenience and brevity of description, the specific working processes of the systems, devices and units described above can refer to the corresponding processes in the aforementioned method embodiments and will not be repeated here.

[0081] In the several embodiments provided by the present invention, it should be understood that the disclosed systems, devices and methods can be implemented in other ways. For example, the device embodiments described above are only schematic. For example, the division of the units is only a logical function division. There may be other division methods in actual implementation, such as multiple units or components can be combined or integrated into another system, or some features can be ignored or not executed. Another point is that the mutual coupling or direct coupling or communication connection shown or discussed can be through some interfaces, indirect coupling or communication connection of devices or units, which can be electrical, mechanical or other forms.

[0082] The units described as separate components may or may not be physically separated, and the components shown as units may or may not be physical units, that is, they may be located in one place or distributed on multiple network units. Some or all of the units may be selected according to actual needs to achieve the purpose of the solution of this embodiment.

[0083] Those skilled in the art will appreciate that the units and algorithm steps of each example described in conjunction with the embodiments disclosed herein can be implemented in electronic hardware, or a combination of computer software and electronic hardware. Whether these functions are performed in hardware or software depends on the specific application and design constraints of the technical solution. Professional and technical personnel can use different methods to implement the described functions for each specific application, but such implementation should not be considered to be beyond the scope of the present invention.

[0084] It should also be understood that in various embodiments of the present invention, the first, second, etc. are only used to indicate that multiple objects are different. For example, the first time window and the second time window are only used to indicate different time windows. They should not have any impact on the time window itself, and the first, second, etc. mentioned above should not impose any limitations on the embodiments of the present invention.

[0085] It should also be understood that in the various embodiments of the present invention, unless otherwise specified or there is any logical conflict, the terms and / or descriptions between the different embodiments are consistent and can be referenced to each other, and the technical features in the different embodiments can be combined to form new embodiments according to their internal logical relationships.

[0086] If the functions are implemented in the form of software functional units and sold or used as independent products, they can be stored in a computer-readable storage medium. Based on this understanding, the technical solution of the present invention, or the part that contributes to the prior art or the part of the technical solution, can be embodied in the form of a software product, which is stored in a computer-readable storage medium and includes several instructions for a computer device (which can be a personal computer, a server, or a network device, etc.) to perform all or part of the steps of the methods described in each embodiment of the present invention. The aforementioned computer-readable storage medium includes: various media that can store program codes, such as a USB flash drive, a mobile hard disk, a read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0087] The present invention also provides an insulator pollution component analysis system based on deep learning, the system comprising: one or more memories for storing instructions; and One or more processors are used to call and run the instructions from the memory to execute the method as described above.

[0088] The present invention also provides a computer program product, which includes instructions. When the instructions are executed, the terminal device and the network device perform operations of the terminal device and the network device corresponding to the above method.

[0089] The present invention also provides a chip system, which includes a processor for implementing the functions involved in the above content, such as generating, receiving, sending, or processing the data and / or information involved in the above method.

[0090] The chip system may be composed of chips, or may include chips and other discrete devices.

[0091] The processor mentioned in any of the above places can be a CPU, a microprocessor, an ASIC, or one or more integrated circuits for executing programs for controlling the above-mentioned feedback information transmission method.

[0092] In a possible design, the chip system also includes a memory, which is used to store necessary program instructions and data. The processor and the memory can be decoupled and respectively set on different devices, connected by wire or wireless means to support the chip system to implement various functions in the above embodiments. Alternatively, the processor and the memory can also be coupled on the same device.

[0093] Optionally, the computer instructions are stored in a memory.

[0094] Optionally, the memory is a storage unit within the chip, such as a register, a cache, etc. The memory can also be a storage unit within the terminal located outside the chip, such as a ROM or other types of static storage devices that can store static information and instructions, RAM, etc.

[0095] It can be understood that the memory in the present invention can be a volatile memory or a non-volatile memory, or can include both volatile and non-volatile memories.

[0096] The non-volatile memory may be a ROM, a programmable ROM (PROM), an erasable PROM (EPROM), an electrically erasable PROM (EEPROM), or a flash memory.

[0097] The volatile memory may be a RAM, which is used as an external cache. There are many different types of RAM, such as static RAM (SRAM), dynamic RAM (DRAM), synchronous DRAM (SDRAM), double data rate SDRAM (DDR SDRAM), enhanced SDRAM (ESDRAM), synchronous link DRAM (SLDRAM), and direct memory bus RAM.

[0098] The embodiments of this specific implementation method are all preferred embodiments of the present invention, and are not intended to limit the protection scope of the present invention. Therefore, any equivalent changes made based on the structure, shape, and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A method for analyzing insulator contamination components based on deep learning, characterized in that: include: In response to the received image containing the insulator, extracting the insulator object image in the image; Performing grayscale processing on the insulator object image to obtain an insulator object grayscale image; Decomposing and grouping the insulator object grayscale image to obtain an insulator object grayscale image group, each insulator object grayscale image group includes a class of sub-insulator object grayscale images, and the area of ​​the sub-insulator object grayscale image is greater than or equal to an area threshold; The insulator spectrum image is cropped using the sub-insulator object grayscale image to obtain the sub-insulator spectrum image; Establishing an analysis reference line on the sub-insulator spectrum image and establishing an analysis curve using pixel points on the analysis reference line; Use autocorrelation calculation and fast Fourier transform to process the analysis curve and obtain the power spectrum; According to the power spectrum, a pollution component recognition model obtained based on deep learning is selected and the pollution components included in the grayscale image of the sub-insulator object are determined.

2. The insulator contamination component analysis method based on deep learning according to claim 1 is characterized in that: Decomposing the grayscale image of the insulator object includes: Using the grayscale interval to perform content extraction on the grayscale image of the insulator object and establishing a change trend line on the grayscale image of the insulator object according to the content extraction result; Count the convergence area of ​​the changing trend lines and determine the center position of the convergence area; Establish a direction line with the center position as the starting point; Calculate the speed of value change on the direction line and determine the boundary point according to the speed of value change; The obtained boundary points are connected sequentially to obtain the grayscale image of the sub-insulator object.

3. The insulator contamination component analysis method based on deep learning according to claim 2 is characterized in that: Using the grayscale interval to perform content extraction on the grayscale image of the insulator object and establishing a change trend line on the grayscale image of the insulator object according to the content extraction result includes: The content extraction results are graded, and each level of content extraction results corresponds to a grayscale interval; Dividing the content extraction results to obtain content extraction result regions, each of which has the same area; Connect multiple content extraction result areas according to the trend to obtain a change trend line; Count the aggregation areas of the change trend line and adjust the content extraction result areas participating in the same change trend line to minimize the number of connected aggregation areas; The content extraction result regions participating in the same change trend line come from different levels, and any two content extraction result regions in the same level do not participate in the same change trend line.

4. The method for analyzing insulator contamination components based on deep learning according to claim 2 is characterized in that: Calculating the speed of value change on the direction line and determining the boundary point according to the speed of value change includes: Set reference points at intervals on the direction line, and the distances between adjacent reference points are equal; Calculate the average of the reference point and the pixels around the reference point and use the average as the value of the reference point; Count the quadratic difference series of numerical values ​​and use the quadratic difference series to draw the quadratic difference series curve; Determine the similarity regions on all obtained quadratic difference series curves; Use the starting points of the similar area as boundary points.

5. The insulator contamination component analysis method based on deep learning according to claim 1 is characterized in that: The pollution component identification models based on deep learning selected according to the power spectrum include: Obtaining model features of the pollution component identification model based on deep learning, the model features include frequency quantity, frequency intensity, frequency proportion and frequency interval; Use the model features to match in the power spectrum to determine the pollution component identification model; Among them, the frequency intensity is less than or equal to the intensity of the corresponding frequency in the power spectrum.

6. The insulator contamination component analysis method based on deep learning according to claim 1 or 5, characterized in that: After establishing the analysis curve using the pixels on the analysis reference line, the interference on the analysis curve is removed. The interference on the analysis curve is removed by: Process the analysis curve by wavelet transform to obtain multiple sub-analysis curves, each of which has a starting point and a cutoff point. The sub-analysis curve is screened once using the transformation length to obtain a grayscale image of a processed sub-insulator object, and a clarity value of the grayscale image of the processed sub-insulator object is calculated; When the clarity value of the grayscale image of the sub-insulator object processed once meets the requirement, the sub-analysis curve is screened once using the transformation length.

7. The insulator contamination component analysis method based on deep learning according to claim 6 is characterized in that: There is at least one endpoint of a sub-analysis curve that has not been deleted between the starting position point and the ending position point of the deleted sub-analysis curve; Delete other sub-analysis curves with the same frequency as the deleted sub-analysis curve; When the content loss amount of the grayscale image of the sub-insulator object exceeds the allowable range, the other sub-analysis curves to be deleted are adjusted, and the adjustment includes: Setting a deletion length and retaining the other sub-analysis curves whose length is greater than the deletion length; The content loss amount of the grayscale image of the sub-insulator object is recalculated, and the deletion length is stopped from being modified when the content loss amount is within the allowable range.

8. An insulator contamination component analysis device based on deep learning, characterized in that: include: A content extraction unit, configured to extract an insulator object image in response to a received image containing an insulator; A grayscale processing unit, used for performing grayscale processing on the insulator object image to obtain an insulator object grayscale image; A first processing unit is used to decompose and group the insulator object grayscale image to obtain an insulator object grayscale image group, each of which includes a type of sub-insulator object grayscale image, and the area of ​​the sub-insulator object grayscale image is greater than or equal to an area threshold; The second processing unit is used to cut the insulator spectrum image using the sub-insulator object grayscale image to obtain the sub-insulator spectrum image; A third processing unit is used to establish an analysis reference line on the sub-insulator spectrum image and to establish an analysis curve using pixel points on the analysis reference line; A spectrum processing unit, used for processing the analysis curve using autocorrelation calculation and fast Fourier transform to obtain a power spectrum; The component determination unit is used to select a pollution component identification model obtained based on deep learning according to the power spectrum and determine the pollution components included in the grayscale image of the sub-insulator object.

9. An insulator pollution component analysis system based on deep learning, characterized in that: The system comprises: one or more memories for storing instructions; and One or more processors, configured to call and execute the instructions from the memory to perform the method according to any one of claims 1 to 7.

10. A computer-readable storage medium, characterized in that: The computer readable storage medium comprises: The program, when the program is executed by a processor, the method according to any one of claims 1 to 7 is executed.

Images