Feature extraction method of PRPD atlas, related device and storage medium
By performing unsupervised feature extraction on the PRPD map, including superpixel segmentation and semantic segmentation value calculation, the problems of low recognition rate and difficult training of feature extractors in the prior art are solved, and higher recognition accuracy and lower manual labeling cost are achieved.
Patent Information
- Application Number
- CN202510083270.1
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-01-20
- Publication Date
- 2025-05-06
AI Technical Summary
The existing local discharge pattern recognition method based on PRPD map features is generally low in recognition rate, mainly due to the poor generalization performance of the training data and the difficulty of training feature extractors.
The training difficulty of the graph feature extractor is reduced by performing unsupervised feature extraction on the original PRPD map information, including superpixel segmentation, background grayscale range determination, semantic segmentation value calculation and feature annotation.
This method greatly reduces the training difficulty of the graph feature extractor, improves the recognition rate, reduces the cost of manual feature annotation, and pays attention to significant features in the subsequent recognition process, improving the accuracy of discharge type recognition.
Smart Images
Figure CN119942138A_ABST
Abstract
Description
Technical Field
[0001] The present application relates to the field of image processing technology, and in particular to a feature extraction method, related device and storage medium for a PRPD spectrum. Background Art
[0002] UHF detection has become the main means to identify the types of partial discharge defects inside GIS equipment. The PRPD spectrum accumulated during the UHF detection process can obtain the statistical information of the phase-discharge amplitude-discharge number of partial discharge signals, and is widely used in the process of identifying GIS partial discharge types.
[0003] The existing partial discharge pattern recognition methods based on PRPD spectrum features generally have a low recognition rate, which is mainly caused by two reasons. First, the PRPD measured spectrum data is limited, and the data used to train the recognition network are mostly laboratory measurements, showing strong sample repeatability and no noise interference, resulting in poor generalization performance of the trained network; second, the original data input for training has high feature redundancy. When the training data set is limited, the original image may cause "dimensionality disaster". In addition, the existing machine learning and deep learning recognition networks, the process of training network parameters is actually training feature extractors. When the data set is very limited, it makes the training of spectrum feature extractors extremely difficult. Summary of the invention
[0004] In view of this, the present application provides a feature extraction method, related device and storage medium for PRPD atlas, which greatly reduces the training difficulty of the atlas feature extractor by performing unsupervised feature extraction on the original PRPD atlas information.
[0005] The first aspect of the present application provides a method for extracting features of a PRPD spectrum, comprising:
[0006] The PRPD spectrum is segmented by superpixels to obtain a segmentation result of the PRPD spectrum; wherein the segmentation result of the PRPD spectrum includes N superpixels; N is the number of pre-segmented superpixels;
[0007] Determining the background grayscale range of the PRPD spectrum according to the segmentation result of the PRPD spectrum;
[0008] For each superpixel, determining the semantic segmentation value of the superpixel according to the background grayscale range of the PRPD atlas;
[0009] The PRPD atlas is feature-annotated according to the semantic segmentation value of the superpixel, the first segmentation threshold, and the second segmentation threshold to obtain a feature-annotated map of the PRPD atlas.
[0010] Optionally, performing superpixel segmentation on the PRPD map to obtain a segmentation result of the PRPD map includes:
[0011] According to the number of pre-segmented superpixels, the superpixel center position is initialized in the PRPD atlas to obtain the initial center seed;
[0012] For each initial center seed, reselect a seed point in the first target neighborhood of the initial center seed to obtain a target center seed;
[0013] For each pixel point in the PRPD map, determine the target center seed to which the pixel point belongs;
[0014] Generate superpixels from all pixels under the target center seed;
[0015] The segmentation result of the PRPD atlas is generated based on all superpixels.
[0016] Optionally, the process of initializing the superpixel center position in the PRPD atlas according to the number of pre-segmented superpixels to obtain an initial center seed further includes:
[0017] If the sizes of pixels in the PRPD spectrum are inconsistent, the pixels in the PRPD spectrum are normalized to obtain a normalized PRPD spectrum.
[0018] Optionally, for each initial center seed, reselecting a seed point in a first target neighborhood of the initial center seed to obtain a target center seed includes:
[0019] For each first neighborhood pixel point in the first target neighborhood of each initial central seed, calculating the gradient value of the first neighborhood pixel point;
[0020] The first neighborhood pixel with the smallest gradient value is used as the target center seed.
[0021] Optionally, for each pixel point in the PRPD atlas, determining the target center seed to which the pixel point belongs includes:
[0022] For each target center seed, a neighborhood search is performed with the target center seed as the center, and a label is assigned to each pixel point in the second target neighborhood of the target center seed;
[0023] For each pixel, determine the target center seed to which the pixel belongs based on the assigned label.
[0024] Optionally, determining the background grayscale range of the PRPD spectrum according to the segmentation result of the PRPD spectrum includes:
[0025] Selecting superpixels at the boundary of the PRPD map as a boundary superpixel set;
[0026] The superpixel with the most pixels in the boundary superpixel set is used as the background superpixel;
[0027] Taking the average grayscale value of all pixels in the background superpixel as the target grayscale;
[0028] According to the target grayscale and floating threshold, the background grayscale range of the PRPD spectrum is determined.
[0029] Optionally, for each superpixel, determining the semantic segmentation value of the superpixel according to the background grayscale range of the PRPD atlas includes:
[0030] For each superpixel, counting the number of pixels in the superpixel whose grayscale value is within the background grayscale range of the PRPD spectrum as the first number of pixels;
[0031] Determine a semantic segmentation value of the superpixel based on the first number of pixels and the total number of pixels in the superpixel.
[0032] The second aspect of the present application provides a feature extraction device for a PRPD spectrum, comprising:
[0033] A segmentation unit, used to perform superpixel segmentation on the PRPD spectrum to obtain a segmentation result of the PRPD spectrum; wherein the segmentation result of the PRPD spectrum includes N superpixels; N is the number of pre-segmented superpixels;
[0034] A background grayscale range determination unit, used to determine the background grayscale range of the PRPD spectrum according to the segmentation result of the PRPD spectrum;
[0035] A semantic segmentation value determination unit, configured to determine, for each superpixel, a semantic segmentation value of the superpixel according to a background grayscale range of the PRPD atlas;
[0036] The labeling unit is used to perform feature labeling on the PRPD map according to the semantic segmentation value of the superpixel, the first segmentation threshold and the second segmentation threshold to obtain a feature labeling map of the PRPD map.
[0037] Optionally, the segmentation unit includes:
[0038] An initialization unit, used to initialize the superpixel center position in the PRPD atlas according to the number of pre-segmented superpixels to obtain an initial center seed;
[0039] A selection unit is used for reselecting a seed point in a first target neighborhood of each initial center seed to obtain a target center seed;
[0040] A first determination unit is used to determine, for each pixel point in the PRPD map, a target center seed to which the pixel point belongs;
[0041] A superpixel generation unit, used to generate superpixels from all pixel points under the target center seed;
[0042] The segmentation result generating unit is used to generate the segmentation result of the PRPD atlas according to all superpixels.
[0043] Optionally, the feature extraction device of the PRPD spectrum further includes:
[0044] The normalization unit is used to normalize the pixel points in the PRPD spectrum if the pixel points in the PRPD spectrum are inconsistent, so as to obtain a normalized PRPD spectrum.
[0045] Optionally, the selection unit includes:
[0046] A gradient value calculation unit, used for calculating the gradient value of each first neighborhood pixel point in the first target neighborhood of each initial central seed;
[0047] The selection subunit is used to use the first neighborhood pixel point with the smallest gradient value as the target center seed.
[0048] Optionally, the first determining unit includes:
[0049] A search unit is used to perform a neighborhood search with each target center seed as the center, and assign a label to each pixel point in the second target neighborhood of the target center seed;
[0050] The first determination subunit is used to determine, for each pixel point, the target center seed to which the pixel point belongs according to the assigned label.
[0051] Optionally, the background grayscale range determining unit includes:
[0052] A selection unit, configured to select superpixels located at the boundary in the PRPD spectrum as a boundary superpixel set;
[0053] A second determining unit is used to select the superpixel with the most pixels in the boundary superpixel set as the background superpixel;
[0054] A third determining unit, configured to take the average grayscale value of all pixels in the background superpixel as the target grayscale;
[0055] The fourth determining unit is used to determine the background grayscale range of the PRPD spectrum according to the target grayscale and the floating threshold.
[0056] Optionally, the semantic segmentation value determining unit includes:
[0057] A fifth determining unit is used to count, for each superpixel, the number of pixels whose grayscale values in the superpixel are within the background grayscale range of the PRPD spectrum as the first number of pixels;
[0058] The sixth determination unit is used to determine the semantic segmentation value of the superpixel according to the first number of pixels and the total number of pixels in the superpixel.
[0059] A third aspect of the present application provides an electronic device, including:
[0060] one or more processors;
[0061] a storage device having one or more programs stored thereon;
[0062] When the one or more programs are executed by the one or more processors, the one or more processors implement the feature extraction method of the PRPD spectrum as described in any one of the first aspects.
[0063] The fourth aspect of the present application provides a storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the feature extraction method of the PRPD spectrum as described in any one of the first aspects is implemented.
[0064] It can be seen from the above scheme that the present application provides a feature extraction method, related device and storage medium for a PRPD spectrum, which performs superpixel segmentation on the PRPD spectrum to obtain the segmentation result of the PRPD spectrum, and then determines the background grayscale range of the PRPD spectrum according to the segmentation result of the PRPD spectrum; then, for each superpixel, determines the semantic segmentation value of the superpixel according to the background grayscale range of the PRPD spectrum; then, performs feature annotation on the PRPD spectrum according to the semantic segmentation value of the superpixel, the first segmentation threshold and the second segmentation threshold, and obtains a feature annotation map of the PRPD spectrum. Unsupervised feature extraction is performed on the original PRPD spectrum information, thereby greatly reducing the training difficulty of the spectrum feature extractor. BRIEF DESCRIPTION OF THE DRAWINGS
[0065] In order to more clearly illustrate the embodiments of the present invention or the technical solutions in the prior art, the drawings required for use in the embodiments or the description of the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For ordinary technicians in this field, other drawings can be obtained based on the provided drawings without paying creative work.
[0066] Figure 1 A specific flow chart of a feature extraction method of a PRPD spectrum provided in an embodiment of the present application;
[0067] Figure 2 A specific flow chart of a superpixel segmentation method provided in another embodiment of the present application;
[0068] Figure 3 A specific flow chart of a method for reselecting a seed point provided in another embodiment of the present application;
[0069] Figure 4 A specific flow chart of a method for determining a target center seed to which a pixel point belongs provided by another embodiment of the present application;
[0070] Figure 5 A specific flow chart of a method for determining the background grayscale range of a PRPD spectrum provided in another embodiment of the present application;
[0071] Figure 6 A specific flow chart of a method for determining a semantic segmentation value of a superpixel provided in another embodiment of the present application;
[0072] Figure 7 A schematic diagram of an original PRPD two-dimensional spectrum collected by a partial discharge monitoring device provided in another embodiment of the present application;
[0073] Figure 8 A schematic diagram of an original PRPD two-dimensional spectrum collected by a partial discharge monitoring device provided in another embodiment of the present application;
[0074] Fig. 9 A schematic diagram of semantic segmentation of an original PRPD two-dimensional map provided in another embodiment of the present application;
[0075] Fig.10 A schematic diagram of semantic segmentation of an original PRPD two-dimensional map provided in another embodiment of the present application;
[0076] Fig.11 A schematic diagram after feature annotation provided in another embodiment of the present application;
[0077] Fig.12 A schematic diagram after feature annotation provided in another embodiment of the present application;
[0078] Fig.13 A schematic diagram of a feature extraction device for a PRPD spectrum provided in another embodiment of the present application;
[0079] Fig.14A schematic diagram of an electronic device for implementing a feature extraction method of a PRPD spectrum provided in another embodiment of the present application. DETAILED DESCRIPTION
[0080] The following will be combined with the drawings in the embodiments of the present application to clearly and completely describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are only part of the embodiments of the present application, not all of the embodiments. Based on the embodiments in the present application, all other embodiments obtained by ordinary technicians in this field without creative work are within the scope of protection of this application.
[0081] The term "including" and its variations used herein are open inclusions, i.e., "including but not limited to". The term "based on" means "based at least in part on". The term "one embodiment" means "at least one embodiment"; the term "another embodiment" means "at least one additional embodiment"; the term "some embodiments" means "at least some embodiments". The relevant definitions of other terms will be given in the following description.
[0082] It should be noted that the information (including but not limited to user device information, user personal information, etc.) and data (including but not limited to data used for analysis, stored data, displayed data, etc.) involved in this application are all information and data authorized by the user or fully authorized by all parties, and the collection, use and processing of relevant data must comply with relevant laws, regulations and standards of relevant countries and regions.
[0083] It should be noted that the concepts such as "first" and "second" mentioned in this application are only used to distinguish different devices, modules or units, and are not used to limit the order or interdependence of the functions performed by these devices, modules or units.
[0084] It should be noted that the modifications of "one" and "plurality" mentioned in the present application are illustrative rather than restrictive, and those skilled in the art should understand that unless otherwise clearly indicated in the context, it should be understood as "one or more".
[0085] First, the technical terms used in this application are explained:
[0086] GIS: Gas Insulated Switcher, gas insulated metal enclosed switchgear.
[0087] PRPD spectrum: Phase Resolved Partial Discharge, phase resolved partial discharge spectrum.
[0088] The present application embodiment provides a method for extracting features of a PRPD spectrum, such as Figure 1 As shown, the specific steps include:
[0089] S101, performing superpixel segmentation on the PRPD map to obtain a segmentation result of the PRPD map.
[0090] The segmentation result of the PRPD atlas includes N superpixels, where N is the number of pre-segmented superpixels.
[0091] Optionally, in another embodiment of the present application, an implementation of step S101 is as follows: Figure 2 As shown, including:
[0092] S201. Initialize the superpixel center position in the PRPD atlas according to the number of pre-segmented superpixels to obtain an initial center seed.
[0093] Specifically, assuming that the number of pre-segmented superpixels N=100, the PRPD map includes 440*570 pixels, and the central seeds are evenly distributed in the high pixels in the PRPD map according to the number of pre-segmented superpixels, then the initial superpixel consists of 2508 pixels on average, and the step size of adjacent seed points is approximately D=50. Then the initial central seed is expressed as C={c1, c2, …, c 100}.
[0094] It is understandable that, since the pixel size in the PRPD map may be inconsistent, in another embodiment of the present application, according to the number of pre-segmented superpixels, the superpixel center position is initialized in the PRPD map, and before obtaining the initial center seed, the following is also included:
[0095] If the sizes of pixels in the PRPD spectrum are inconsistent, the pixels in the PRPD spectrum are normalized to obtain a normalized PRPD spectrum.
[0096] Specifically, the pixels in the PRPD map may be standardized to the size of the target, which is not limited here.
[0097] S202: for each initial center seed, reselect a seed point in the first target neighborhood of the initial center seed to obtain a target center seed.
[0098] Among them, the first target neighborhood includes but is not limited to a 3*3 neighborhood, which is pre-set and changed by technicians, experts, etc. and is not limited here.
[0099] Optionally, in another embodiment of the present application, an implementation of step S202 is as follows: Figure 3 As shown, including:
[0100] S301. For each first neighborhood pixel point in a first target neighborhood of each initial central seed, calculate the gradient value of the first neighborhood pixel point.
[0101] In the specific implementation process of this application, the first neighborhood pixel point The gradient value The calculation method can be shown as follows:
[0102] ;in, is the first neighboring pixel The gradient value of the horizontal axis, is the first neighboring pixel The gradient value of the vertical axis.
[0103] in, ; ; It is the PRPD map.
[0104] S302: Use the first neighborhood pixel point with the smallest gradient value as the target center seed.
[0105] By using the first neighborhood pixel with the smallest gradient value as the target center seed, the initial center seed is prevented from falling on the image edge contour line with a larger gradient, thereby affecting the subsequent superpixel clustering results.
[0106] S203: For each pixel point in the PRPD atlas, determine the target center seed to which the pixel point belongs.
[0107] Optionally, in another embodiment of the present application, an implementation of step S203 is as follows: Figure 4 As shown, including:
[0108] S401 . For each target center seed, a neighborhood search is performed with the target center seed as the center, and a label is assigned to each pixel point in the second target neighborhood of the target center seed.
[0109] Among them, the neighborhood range of the second target neighborhood is S×S, and S can be but is not limited to 20. It is pre-set and changed by technicians, experts, etc. and is not limited here.
[0110] In the specific implementation process of the present application, the actual search range can also be 2S×2S, and the similarity between the pixels is calculated by using the distance metric. Therefore, the label at least includes the distance metric between the pixel and the target center seed. The specific calculation method can be as follows:
[0111] ;
[0112] ;
[0113] ;
[0114] in, Represents the color distance, Represents spatial distance; , , Respectively Grayscale values of red, blue and green channels; , is the horizontal and vertical spatial coordinates of the pixel point. It represents a measurement method that comprehensively considers the spatial distance and color distance between two pixels. This application adopts As the final distance metric, , is the maximum distance within the class, that is, the maximum color distance and maximum spatial distance under the target center seed.
[0115] S402: For each pixel point, determine the target center seed to which the pixel point belongs according to the assigned label.
[0116] If a pixel has only one label, then the target center seed corresponding to the distance metric in the label is the target center seed to which the pixel belongs. If the pixel has multiple labels, then the target center seed corresponding to the minimum distance metric is taken as the target center seed to which the pixel belongs.
[0117] In the specific implementation process of the present application, a convergence condition may also be set, that is, step S401 to step S402 are repeated until the convergence condition is met.
[0118] The convergence condition may be, but is not limited to, the number of convergences, such as 10 times, which is not limited here.
[0119] It is understandable that the convergence condition is set and changed in advance by technicians, experts, etc. and is not limited here.
[0120] In the specific implementation process of this application, From left to right and from top to bottom, we retrieve discontinuous superpixels and superpixels with too small sizes one by one and reallocate them to neighboring superpixels, thereby enhancing connectivity.
[0121] S204: Generate superpixels from all pixels under the target center seed.
[0122] S205: Generate a segmentation result of the PRPD atlas based on all superpixels.
[0123] The present invention uses scale transformation to normalize the size to a uniform size. Superpixel segmentation can define points in the image with similar distances and consistent grayscale features as the same superpixel. This step can segment background pixels at different positions of the PRPD spectrum into the same superpixel, and segment the spectrum's significant partial discharge features into the same superpixel, which can preliminarily achieve background and significant feature segmentation, and help to further extract significant features in the future.
[0124] S102: Determine the background grayscale range of the PRPD spectrum according to the segmentation result of the PRPD spectrum.
[0125] Optionally, in another embodiment of the present application, an implementation of step S102 is as follows: Figure 5 As shown, including:
[0126] S501, selecting superpixels at the boundary in the PRPD map as a boundary superpixel set.
[0127] S502: The superpixel with the most pixels in the boundary superpixel set is used as the background superpixel.
[0128] S503: Taking the average grayscale value of all pixels in the background superpixel as the target grayscale.
[0129] S504: Determine the background grayscale range of the PRPD spectrum according to the target grayscale and the floating threshold.
[0130] Among them, the floating threshold It is set and modified in advance by technicians, experts, etc. and is not limited here.
[0131] Specifically, the background grayscale range of the PRPD spectrum can be expressed as , is the target grayscale.
[0132] Since the local discharge characteristic pixels in the original PRPD spectrum are generally distributed in the internal position of the spectrum, and the background pixels account for a very high proportion of the peripheral pixels of the spectrum, the present invention selects the periphery of the spectrum as the estimation position of the background grayscale value of the spectrum. Secondly, the background pixels of the spectrum are distributed more concentratedly but with a large area, and the grayscale value fluctuation range is small. Based on this prior knowledge, the superpixel with the most pixel values in the superpixel set SPb is selected as the background superpixel, so as to estimate the background grayscale value interval. Wherein, SPb is the boundary superpixel set, SPb={spb1, spb2, spb3,…, spb i}.
[0133] S103: For each superpixel, determine the semantic segmentation value of the superpixel according to the background grayscale range of the PRPD map.
[0134] Optionally, in another embodiment of the present application, an implementation of step S103 is as follows: Figure 6 As shown, including:
[0135] S601. For each superpixel, count the number of pixels in the superpixel whose grayscale value is within the background grayscale range of the PRPD spectrum as the first pixel number.
[0136] S602: Determine a semantic segmentation value of the superpixel according to the first number of pixels and the total number of pixels in the superpixel.
[0137] In the specific implementation process of this application, the semantic segmentation value of the superpixel can be calculated using but not limited to the following calculation formula:
[0138] ;in, for Middle Superpixels The semantic segmentation value of is the segmentation result of PRPD atlas, , for The total number of pixels in for Medium gray values belong to The number of pixels.
[0139] Due to the irregular spatial distribution of characteristic pixels in the PRPD spectrum, it is difficult to ensure that all pixels in the superpixel belong to the data background or partial discharge feature pixels after superpixel segmentation. Therefore, the semantic segmentation result is characterized by statistically analyzing the distribution probability of partial discharge feature pixels in the superpixel. This step of the present invention statistics the proportion P of partial discharge feature pixels in the superpixel to the total superpixels. The value of P is close to 1, which means that the more partial discharge feature pixels there are in the superpixel unit, the closer it is to the background sample.
[0140] S104, feature-labeling the PRPD atlas according to the semantic segmentation value of the superpixel, the first segmentation threshold, and the second segmentation threshold to obtain a feature-labeled map of the PRPD atlas.
[0141] The original image has background noise, uneven lighting, low spatial resolution and other factors that cause the background gray value distribution to deviate. The background pixels are directly defined as meeting the background gray threshold. The remaining positions are partial discharge feature pixels, which will lead to a high false detection rate and poor robustness in the partial discharge feature position detection results. Therefore, this patent performs step-by-step partial discharge feature detection based on the semantic segmentation results after superpixel calculation. First, based on step S103, a rough detection is implemented to preliminarily screen out superpixels belonging to background and partial discharge features in the map. Finally, the semantic segmentation value is within the maximum threshold. and minimum threshold Fine PD feature pixel positioning is performed on individual pixels within the superpixel between the pixels.
[0142] Specifically, when Greater than , the local discharge feature calibration in the superpixel is shown in the following formula; when Less than , the superpixel is the background as a whole; when > > , it is necessary to determine whether the superpixel satisfies > If the superpixel is connected to the pixel block that satisfies > If the superpixel is not connected to the pixel block that satisfies > If the pixel blocks are connected, all the pixels in the superpixel are background pixels.
[0143] ;
[0144] in, is the local discharge pixel in the PRPD spectrum; are the background pixels in the PRPD spectrum; For the map The gray value of each pixel; Superpixel characteristics.
[0145] like Figure 7 and Figure 8 As shown in the figure, it is the original PRPD two-dimensional spectrum collected by the partial discharge monitoring device input by the present invention. The four spectra, sizes, and corresponding positions of the coordinate axes are all different, and the partial discharge pixel features in the spectrum are also quite different. Based on the aggregated partial discharge pixel blocks, there are also scattered partial discharge pixels.
[0146] Fig. 9 and Fig.10 It is the result of semantic segmentation after superpixel segmentation. Fig. 9 and Fig.10 It can be seen that for pixel areas that are relatively concentrated with the partial discharge features of the spectrum, the superpixel clustering method can preliminarily realize the effective segmentation of features and background. For pixel points where the discharge features of the spectrum are relatively scattered, the performance of this method needs to be improved. However, the present invention focuses on significant spectrum features. When scattered events occur at a low frequency, they are not significant statistical features of discharge. The more concentrated the partial discharge features of the spectrum are, the higher the brightness value after semantic segmentation. Conversely, the closer the background is, the darker the superpixel value after semantic segmentation.
[0147] Fig.11 and Fig.12 It is a typical graph feature extraction result. Fig.11 and Fig.12 The black area in the figure is the redundant information location of the two-dimensional PRPD spectrum feature, and the white location is the true value map corresponding to the significant feature. Fig.11 and Fig.12 It can be seen from the figure that, for locations with high discharge frequency and concentrated discharge, the method proposed in the present invention can effectively locate the locations. Most of the redundant information in the input image is effectively eliminated.
[0148] It can be seen from the above scheme that the present application provides a feature extraction method for PRPD spectra, which can effectively extract the PRPD spectra of different partial discharge monitoring equipment, replacing the traditional manual feature labeling process, greatly saving labor costs. Secondly, the PRPD spectrum significant feature extraction method proposed by the present invention can achieve a large amount of redundant information elimination, and unsupervised PRPD significant feature extraction, without the need to provide any training data, and only focus on significant features in the subsequent recognition process, which greatly improves the accuracy of future discharge type recognition.
[0149] The present application embodiment provides a feature extraction device for a PRPD spectrum, such as Fig.13 As shown, specifically including:
[0150] The segmentation unit 1301 is used to perform superpixel segmentation on the PRPD spectrum to obtain a segmentation result of the PRPD spectrum.
[0151] The segmentation result of the PRPD atlas includes N superpixels, where N is the number of pre-segmented superpixels.
[0152] Optionally, in another embodiment of the present application, an implementation of the segmentation unit 1301 includes:
[0153] The initialization unit is used to initialize the superpixel center position in the PRPD atlas according to the number of pre-segmented superpixels to obtain an initial center seed.
[0154] The selection unit is used to reselect a seed point within a first target neighborhood of the initial center seed for each initial center seed to obtain a target center seed.
[0155] Optionally, in another embodiment of the present application, an implementation of the selection unit includes:
[0156] The gradient value calculation unit is used to calculate the gradient value of each first neighborhood pixel point in the first target neighborhood of each initial central seed.
[0157] The selection subunit is used to use the first neighborhood pixel point with the smallest gradient value as the target center seed.
[0158] For the specific working process of the units disclosed in the above embodiments of the present application, please refer to the corresponding method embodiments, such as Figure 3 As shown, no further description is given here.
[0159] The first determination unit is used to determine, for each pixel point in the PRPD atlas, a target center seed to which the pixel point belongs.
[0160] Optionally, in another embodiment of the present application, an implementation of the first determining unit includes:
[0161] The search unit is used to perform a neighborhood search with the target center seed as the center for each target center seed, and assign a label to each pixel point in the second target neighborhood of the target center seed.
[0162] The first determination subunit is used to determine, for each pixel point, the target center seed to which the pixel point belongs according to the assigned label.
[0163] For the specific working process of the units disclosed in the above embodiments of the present application, please refer to the corresponding method embodiments, such as Figure 4 As shown, no further description is given here.
[0164] The superpixel generation unit is used to generate superpixels from all pixel points under the target center seed.
[0165] The segmentation result generating unit is used to generate the segmentation result of the PRPD atlas according to all superpixels.
[0166] For the specific working process of the units disclosed in the above embodiments of the present application, please refer to the corresponding method embodiments, such as Figure 2 As shown, no further description is given here.
[0167] Optionally, in another embodiment of the present application, an implementation of the feature extraction device of the PRPD spectrum further includes:
[0168] The normalization unit is used to normalize the pixel points in the PRPD spectrum if the pixel points in the PRPD spectrum are inconsistent, so as to obtain a normalized PRPD spectrum.
[0169] The specific working process of the units disclosed in the above embodiments of the present application can be found in the corresponding method embodiments, which will not be repeated here.
[0170] The background grayscale range determining unit 1302 is used to determine the background grayscale range of the PRPD spectrum according to the segmentation result of the PRPD spectrum.
[0171] Optionally, in another embodiment of the present application, an implementation of the background grayscale range determining unit 1302 includes:
[0172] The selection unit is used to select superpixels located at the boundary in the PRPD map as a boundary superpixel set.
[0173] The second determining unit is used to take the superpixel with the most pixels in the boundary superpixel set as the background superpixel.
[0174] The third determining unit is used to take the average grayscale value of all pixels in the background superpixel as the target grayscale.
[0175] The fourth determining unit is used to determine the background grayscale range of the PRPD spectrum according to the target grayscale and the floating threshold.
[0176] For the specific working process of the units disclosed in the above embodiments of the present application, please refer to the corresponding method embodiments, such as Figure 5 As shown, no further description is given here.
[0177] The semantic segmentation value determination unit 1303 is used to determine the semantic segmentation value of each superpixel according to the background grayscale range of the PRPD spectrum.
[0178] Optionally, in another embodiment of the present application, an implementation of the semantic segmentation value determining unit 1303 includes:
[0179] The fifth determining unit is used to count, for each superpixel, the number of pixels whose grayscale values in the superpixel are within the background grayscale range of the PRPD spectrum as the first pixel number.
[0180] The sixth determination unit is used to determine the semantic segmentation value of the superpixel according to the first number of pixels and the total number of pixels in the superpixel.
[0181] For the specific working process of the units disclosed in the above embodiments of the present application, please refer to the corresponding method embodiments, such as Figure 6 As shown, no further description is given here.
[0182] The labeling unit 1304 is used to perform feature labeling on the PRPD spectrum according to the semantic segmentation value of the superpixel, the first segmentation threshold and the second segmentation threshold to obtain a feature labeling map of the PRPD spectrum.
[0183] For the specific working process of the units disclosed in the above embodiments of the present application, please refer to the corresponding method embodiments, such as Figure 1 As shown, no further description is given here.
[0184] It can be seen from the above scheme that the present application provides a feature extraction device for a PRPD spectrum. The segmentation unit 1301 performs superpixel segmentation on the PRPD spectrum. After obtaining the segmentation result of the PRPD spectrum, the background grayscale range determination unit 1302 determines the background grayscale range of the PRPD spectrum according to the segmentation result of the PRPD spectrum; then, the semantic segmentation value determination unit 1303 determines the semantic segmentation value of the superpixel according to the background grayscale range of the PRPD spectrum for each superpixel; thereafter, the annotation unit 1304 performs feature annotation on the PRPD spectrum according to the semantic segmentation value of the superpixel, the first segmentation threshold and the second segmentation threshold, and obtains a feature annotation map of the PRPD spectrum. Unsupervised feature extraction is performed on the original PRPD spectrum information, thereby greatly reducing the difficulty of training the spectrum feature extractor.
[0185] The functions described above herein may be performed at least in part by one or more hardware logic components. For example, without limitation, exemplary types of hardware logic components that may be used include: field programmable gate arrays (FPGAs), application specific integrated circuits (ASICs), application specific standard products (ASSPs), systems on chips (SOCs), complex programmable logic devices (CPLDs), and the like.
[0186] Another embodiment of the present application provides an electronic device, such as Fig.14 As shown, including:
[0187] One or more processors 1401.
[0188] The storage device 1402 stores one or more programs.
[0189] When the one or more programs are executed by the one or more processors 1401, the one or more processors 1401 implement the feature extraction method of the PRPD spectrum as described in any one of the above embodiments.
[0190] Another embodiment of the present application provides a storage medium having a computer program stored thereon, wherein when the computer program is executed by a processor, the feature extraction method of the PRPD spectrum as described in any one of the above embodiments is implemented.
[0191] In the context of the present application, a machine-readable medium may be a tangible medium that may contain or store a program for use by or in conjunction with an instruction execution system, device, or equipment. A machine-readable medium may be a machine-readable signal medium or a machine-readable storage medium. A machine-readable medium may include, but is not limited to, an electronic, magnetic, optical, electromagnetic, infrared, or semiconductor system, device, or device, or any suitable combination of the foregoing. A more specific example of a machine-readable storage medium may include an electrical connection based on one or more lines, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the foregoing.
[0192] It should be noted that the computer-readable medium mentioned above in the present application may be a computer-readable signal medium or a computer-readable storage medium or any combination of the above two. The computer-readable storage medium may be, for example, but not limited to, an electrical, magnetic, optical, electromagnetic, infrared, or semiconductor system, device or device, or any combination of the above. More specific examples of computer-readable storage media may include, but are not limited to: an electrical connection with one or more wires, a portable computer disk, a hard disk, a random access memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), an optical fiber, a portable compact disk read-only memory (CD-ROM), an optical storage device, a magnetic storage device, or any suitable combination of the above. In the present application, a computer-readable storage medium may be any tangible medium containing or storing a program that can be used by or in combination with an instruction execution system, device or device. In the present application, a computer-readable signal medium may include a data signal propagated in a baseband or as part of a carrier wave, which carries a computer-readable program code. This propagated data signal may take a variety of forms, including but not limited to electromagnetic signals, optical signals, or any suitable combination of the above. Computer readable signal media may also be any computer readable medium other than computer readable storage media, which may send, propagate or transmit a program for use by or in conjunction with an instruction execution system, apparatus or device. The program code contained on the computer readable medium may be transmitted using any appropriate medium, including but not limited to: wires, optical cables, RF (radio frequency), etc., or any suitable combination of the above.
[0193] The computer-readable medium may be included in the electronic device, or may exist independently without being incorporated into the electronic device.
[0194] Another embodiment of the present application provides a computer program product, which, when executed, is used to perform the above-mentioned feature extraction method of the PRPD spectrum.
[0195] In particular, according to an embodiment of the present application, the process described above with reference to the flowchart can be implemented as a computer software program. For example, an embodiment of the present application includes a computer program product, which includes a computer program carried on a non-transitory computer-readable medium, and the computer program includes a program code for executing the method shown in the flowchart. In such an embodiment, the computer program can be downloaded and installed from a network via a communication device, or installed from a storage device, or installed from a ROM. When the computer program is executed by a processing device, the above-mentioned functions defined in the method of the embodiment of the present application are executed.
[0196] Although the subject matter has been described in language specific to structural features and / or method logic actions, it should be understood that the subject matter defined in this application is not necessarily limited to the specific features or actions described above. On the contrary, the specific features and actions described above are merely example forms of implementing this application.
[0197] Although several specific implementation details are included in the above discussion, these should not be interpreted as limiting the scope of the present application. Certain features described in the context of a separate embodiment can also be implemented in a single embodiment in combination. On the contrary, the various features described in the context of a single embodiment can also be implemented in multiple embodiments individually or in any suitable sub-combination.
[0198] The above description is only a preferred embodiment of the present application and an explanation of the technical principles used. Those skilled in the art should understand that the scope of application involved in the present application is not limited to the technical solution formed by a specific combination of the above technical features, but should also cover other technical solutions formed by any combination of the above technical features or their equivalent features without departing from the above application concept. For example, the above features are replaced with (but not limited to) technical features with similar functions applied in the present application.
Claims
1. A method for extracting features of a PRPD spectrum, characterized in that: include: The PRPD spectrum is segmented by superpixels to obtain a segmentation result of the PRPD spectrum; wherein the segmentation result of the PRPD spectrum includes N superpixels; N is the number of pre-segmented superpixels; Determining the background grayscale range of the PRPD spectrum according to the segmentation result of the PRPD spectrum; For each superpixel, determining the semantic segmentation value of the superpixel according to the background grayscale range of the PRPD atlas; The PRPD atlas is feature-annotated according to the semantic segmentation value of the superpixel, the first segmentation threshold, and the second segmentation threshold to obtain a feature-annotated map of the PRPD atlas.
2. The feature extraction method of PRPD spectrum according to claim 1, characterized in that: The PRPD map is subjected to superpixel segmentation to obtain a segmentation result of the PRPD map, including: According to the number of pre-segmented superpixels, the superpixel center position is initialized in the PRPD atlas to obtain the initial center seed; For each initial center seed, reselect a seed point in the first target neighborhood of the initial center seed to obtain a target center seed; For each pixel point in the PRPD map, determine the target center seed to which the pixel point belongs; Generate superpixels from all pixels under the target center seed; The segmentation result of the PRPD atlas is generated based on all superpixels.
3. The feature extraction method of PRPD spectrum according to claim 2, characterized in that: The method further includes initializing the superpixel center position in the PRPD atlas according to the number of pre-segmented superpixels to obtain the initial center seed: If the sizes of pixels in the PRPD spectrum are inconsistent, the pixels in the PRPD spectrum are normalized to obtain a normalized PRPD spectrum.
4. The feature extraction method of PRPD spectrum according to claim 2, characterized in that: The step of reselecting a seed point within a first target neighborhood of each initial center seed to obtain a target center seed includes: For each first neighborhood pixel point in the first target neighborhood of each initial central seed, calculating the gradient value of the first neighborhood pixel point; The first neighborhood pixel with the smallest gradient value is used as the target center seed.
5. The feature extraction method of PRPD spectrum according to claim 2, characterized in that: The step of determining, for each pixel point in the PRPD atlas, a target center seed to which the pixel point belongs includes: For each target center seed, a neighborhood search is performed with the target center seed as the center, and a label is assigned to each pixel point in the second target neighborhood of the target center seed; For each pixel, determine the target center seed to which the pixel belongs based on the assigned label.
6. The feature extraction method of PRPD spectrum according to claim 1, characterized in that: Determining the background grayscale range of the PRPD spectrum according to the segmentation result of the PRPD spectrum includes: Selecting superpixels at the boundary of the PRPD map as a boundary superpixel set; The superpixel with the most pixels in the boundary superpixel set is used as the background superpixel; Taking the average grayscale value of all pixels in the background superpixel as the target grayscale; According to the target grayscale and floating threshold, the background grayscale range of the PRPD spectrum is determined.
7. The feature extraction method of PRPD spectrum according to claim 1, characterized in that: For each superpixel, determining the semantic segmentation value of the superpixel according to the background grayscale range of the PRPD atlas includes: For each superpixel, counting the number of pixels in the superpixel whose grayscale value is within the background grayscale range of the PRPD spectrum as the first number of pixels; Determine a semantic segmentation value of the superpixel based on the first number of pixels and the total number of pixels in the superpixel.
8. A feature extraction device for a PRPD spectrum, characterized in that: include: A segmentation unit, used to perform superpixel segmentation on the PRPD spectrum to obtain a segmentation result of the PRPD spectrum; wherein the segmentation result of the PRPD spectrum includes N superpixels; N is the number of pre-segmented superpixels; A background grayscale range determination unit, used to determine the background grayscale range of the PRPD spectrum according to the segmentation result of the PRPD spectrum; A semantic segmentation value determination unit, configured to determine, for each superpixel, a semantic segmentation value of the superpixel according to a background grayscale range of the PRPD atlas; The labeling unit is used to perform feature labeling on the PRPD map according to the semantic segmentation value of the superpixel, the first segmentation threshold and the second segmentation threshold to obtain a feature labeling map of the PRPD map.
9. An electronic device, characterized in that: include: one or more processors; a storage device having one or more programs stored thereon; When the one or more programs are executed by the one or more processors, the one or more processors implement the feature extraction method of the PRPD spectrum as described in any one of claims 1 to 7.
10. A storage medium, characterized in that: A computer program is stored thereon, wherein when the computer program is executed by a processor, the feature extraction method of the PRPD spectrum as described in any one of claims 1 to 7 is implemented.