CN111047598B - Deep learning-based ultraviolet discharge light spot segmentation method and device for power transmission and transformation equipment - Google Patents
Deep learning-based ultraviolet discharge light spot segmentation method and device for power transmission and transformation equipment Download PDFInfo
- Publication number
- CN111047598B CN111047598B CN201911066537.7A CN201911066537A CN111047598B CN 111047598 B CN111047598 B CN 111047598B CN 201911066537 A CN201911066537 A CN 201911066537A CN 111047598 B CN111047598 B CN 111047598B
- Authority
- CN
- China
- Prior art keywords
- feature map
- prediction
- ultraviolet
- layer
- power transmission
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T7/00—Image analysis
- G06T7/10—Segmentation; Edge detection
- G06T7/11—Region-based segmentation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06Q—INFORMATION AND COMMUNICATION TECHNOLOGY [ICT] SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES; SYSTEMS OR METHODS SPECIALLY ADAPTED FOR ADMINISTRATIVE, COMMERCIAL, FINANCIAL, MANAGERIAL OR SUPERVISORY PURPOSES, NOT OTHERWISE PROVIDED FOR
- G06Q50/00—Systems or methods specially adapted for specific business sectors, e.g. utilities or tourism
- G06Q50/06—Electricity, gas or water supply
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20081—Training; Learning
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06T—IMAGE DATA PROCESSING OR GENERATION, IN GENERAL
- G06T2207/00—Indexing scheme for image analysis or image enhancement
- G06T2207/20—Special algorithmic details
- G06T2207/20084—Artificial neural networks [ANN]
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y04—INFORMATION OR COMMUNICATION TECHNOLOGIES HAVING AN IMPACT ON OTHER TECHNOLOGY AREAS
- Y04S—SYSTEMS INTEGRATING TECHNOLOGIES RELATED TO POWER NETWORK OPERATION, COMMUNICATION OR INFORMATION TECHNOLOGIES FOR IMPROVING THE ELECTRICAL POWER GENERATION, TRANSMISSION, DISTRIBUTION, MANAGEMENT OR USAGE, i.e. SMART GRIDS
- Y04S10/00—Systems supporting electrical power generation, transmission or distribution
- Y04S10/50—Systems or methods supporting the power network operation or management, involving a certain degree of interaction with the load-side end user applications
Abstract
The invention discloses a deep learning-based power transmission and transformation equipment ultraviolet discharge facula segmentation method and device, which comprise a full convolution neural network and multi-scale feature fusion, wherein the full convolution neural network model is utilized to automatically extract ultraviolet spots in an ultraviolet map and learn the characteristics of the spots such as textures, shapes and the like, so that the defect that a traditional segmentation model cannot effectively separate a high-brightness white background from scattered small spots is avoided, the complex feature selection process of the traditional segmentation algorithm model is avoided, and the autonomy and the intellectualization of end-to-end feature extraction are realized through an FCN model. The manual power inspection workload is reduced, the inspection efficiency and accuracy are improved, the practicability is high, and the complex field environment is met. The automatic segmentation and diagnosis method greatly reduces the condition of false detection, and makes full-automatic inspection of abnormal discharge of electrical equipment possible.
Description
Technical Field
The invention relates to the technical field of power transmission and transformation equipment, in particular to a method and a device for dividing ultraviolet discharge light spots of power transmission and transformation equipment based on deep learning.
Background
At present, if the external insulation of high-voltage equipment for power transmission and transformation is damaged and defective, a corona discharge phenomenon can be continuously generated in the operation process, and the corona discharge phenomenon not only can bring certain electric energy loss, but also can generate electromagnetic interference and noise interference. As the voltage class of the power system is continuously improved, the corona discharge phenomenon of the power transmission and transformation high-voltage equipment is also emphasized and focused.
A solar blind type ultraviolet imager that can detect ultraviolet light generated by corona discharge has been applied to discharge detection of high-voltage power equipment in recent years. The solar blind ultraviolet imaging technology is utilized to realize that only the ultraviolet light with the wave band of 240-280 nm radiated by the discharge is detected to discharge, and the interference of the ultraviolet light of the sun is shielded. Compared with detection modes such as infrared imaging, ultrasonic partial discharge detection, leakage current and the like, the ultraviolet imaging detection technology has a more visual detection effect on corona discharge, and can rapidly detect the corona discharge state of high-voltage power equipment which runs in a live mode outside a safe distance in a long-distance, real-time and non-contact mode. Practical field application shows that when the solar blind ultraviolet imager detects high-voltage power equipment on site, due to ultraviolet scattering and inherent noise, continuous discrete small-area ultraviolet light spots appear in the display of the ultraviolet imager, and the occurrence of the small-area ultraviolet light spots can be counted into the photon technical area of the ultraviolet imager, so that misjudgment can be generated on discharge intensity by inspection personnel using the ultraviolet imager.
Disclosure of Invention
The invention aims to at least solve one of the technical problems in the prior art, and provides a method and a device for dividing ultraviolet discharge light spots of power transmission and transformation equipment based on deep learning, which not only reduce the workload of manual power inspection, improve the inspection efficiency and accuracy, but also have great practicability and are in very line with complex field environment; and the false detection condition is greatly reduced, so that the full-automatic inspection of the abnormal discharge of the electrical equipment is possible.
The invention provides a power transmission and transformation equipment ultraviolet discharge light spot segmentation method based on deep learning, which comprises the following steps:
acquiring an ultraviolet spectrum of power transmission and transformation equipment;
sequentially carrying out feature extraction and sampling operation on the ultraviolet map through a first convolution layer, a first pooling layer, a second convolution layer, a second pooling layer, a third convolution layer and a third pooling layer to obtain a first prediction feature map;
the first prediction feature map is subjected to feature extraction of a fourth convolution layer and sampling operation of a fourth pooling layer to obtain a second prediction feature map;
the second prediction feature map is subjected to feature extraction of a fifth convolution layer and sampling operation of a fifth pooling layer to obtain a third prediction feature map;
8 times of up-sampling prediction processing is carried out on the first prediction feature map and the third prediction feature map through an FCN-8s model, 16 times of up-sampling prediction processing is carried out on the second prediction feature map and the third prediction feature map through an FCN-16s model, 32 times of up-sampling prediction processing is carried out on the third prediction feature map through an FCN-32s model after the third prediction feature map passes through a first full-connection layer and a second full-connection layer, and segmented images are output through deconvolution during the three types of up-sampling prediction processing, so that segmentation extraction of ultraviolet discharge main light spots is realized.
According to the deep learning-based power transmission and transformation equipment ultraviolet discharge light spot segmentation method provided by the first aspect of the invention, the extracted features in the ultraviolet spectrum comprise texture features and shape features of the main light spot.
According to a second aspect of the invention, a deep learning-based ultraviolet discharge light spot segmentation device for power transmission and transformation equipment is provided, and the device comprises the following units:
the ultraviolet spectrum acquisition unit is used for acquiring an ultraviolet spectrum of the power transmission and transformation equipment;
the first prediction feature map output unit is used for sequentially carrying out feature extraction and sampling operation on the ultraviolet map through a first convolution layer, a first pooling layer, a second convolution layer, a second pooling layer, a third convolution layer and a third pooling layer to obtain a first prediction feature map;
the second prediction feature map output unit is used for extracting the features of the first prediction feature map through a fourth convolution layer and performing sampling operation of a fourth pooling layer to obtain a second prediction feature map;
the third prediction feature map output unit is used for extracting the features of the second prediction feature map through a fifth convolution layer and sampling operation of a fifth pooling layer to obtain a third prediction feature map;
the multi-scale feature fusion segmentation extraction unit is used for carrying out 8 times of upsampling prediction processing on the first prediction feature map and the third prediction feature map through an FCN-8s model, carrying out 16 times of upsampling prediction processing on the second prediction feature map and the third prediction feature map through an FCN-16s model, carrying out 32 times of upsampling prediction processing on the third prediction feature map through a first full-connection layer and a second full-connection layer and then through an FCN-32s model, and outputting segmented images through deconvolution during the three upsampling prediction processing so as to realize segmentation extraction of ultraviolet discharge main light spots.
In a third aspect of the present invention, there is provided a deep learning-based power transmission and transformation device ultraviolet discharge light spot segmentation device, comprising at least one control processor and a memory for communication connection with the at least one control processor; the memory stores instructions executable by the at least one control processor to enable the at least one control processor to perform the deep learning-based power transmission and transformation device ultraviolet discharge spot segmentation method as described above.
In a fourth aspect of the present invention, there is provided a computer-readable storage medium storing computer-executable instructions for causing a computer to perform the deep learning-based power transmission and transformation device ultraviolet discharge spot segmentation method as described above.
In a fifth aspect of the present invention, there is provided a computer program product comprising a computer program stored on a computer readable storage medium, the computer program comprising program instructions which, when executed by a computer, cause the computer to perform a deep learning based power transmission and transformation apparatus uv discharge spot segmentation method as described above.
The ultraviolet discharge light spot segmentation method and device for the power transmission and transformation equipment based on deep learning have the following beneficial effects: the invention can acquire the ultraviolet spectrum of the power transmission and transformation equipment; sequentially carrying out feature extraction and sampling operation on the ultraviolet map through a first convolution layer, a first pooling layer, a second convolution layer, a second pooling layer, a third convolution layer and a third pooling layer to obtain a first prediction feature map; the first prediction feature map is subjected to feature extraction of a fourth convolution layer and sampling operation of a fourth pooling layer to obtain a second prediction feature map; the second prediction feature map is subjected to feature extraction of a fifth convolution layer and sampling operation of a fifth pooling layer to obtain a third prediction feature map; 8 times of up-sampling prediction processing is carried out on the first prediction feature map and the third prediction feature map through an FCN-8s model, 16 times of up-sampling prediction processing is carried out on the second prediction feature map and the third prediction feature map through an FCN-16s model, 32 times of up-sampling prediction processing is carried out on the third prediction feature map through an FCN-32s model after the third prediction feature map passes through a first full-connection layer and a second full-connection layer, and segmented images are output through deconvolution during the three types of up-sampling prediction processing, so that segmentation extraction of ultraviolet discharge main light spots is realized. The method comprises a full convolution neural network and multi-scale feature fusion, the full convolution neural network model is utilized to automatically extract ultraviolet light spots in the ultraviolet spectrum and learn the characteristics of textures, shapes and the like of the light spots, the defect that a traditional segmentation model cannot effectively separate a high-brightness white background and scattered small light spots is avoided, the complex feature selection process of the traditional segmentation algorithm model is avoided, and the autonomy and the intellectualization of end-to-end feature extraction are realized through the FCN model. The manual power inspection workload is reduced, the inspection efficiency and accuracy are improved, the practicability is high, and the complex field environment is met. The automatic segmentation and diagnosis method greatly reduces the condition of false detection, and makes full-automatic inspection of abnormal discharge of electrical equipment possible.
Drawings
The invention is further described below with reference to the drawings and examples.
FIG. 1 is a flow chart of a method for dividing ultraviolet discharge light spots of power transmission and transformation equipment based on deep learning according to one embodiment of the invention;
FIG. 2 is a diagram of a full convolutional neural network structure of a deep learning-based power transmission and transformation equipment ultraviolet discharge facula segmentation method according to one embodiment of the invention;
FIG. 3 is a schematic diagram of a multi-scale feature fusion model of a power transmission and transformation equipment ultraviolet discharge facula segmentation method based on deep learning according to an embodiment of the invention;
fig. 4 is a multi-layer visual diagram of a convolutional neural network model of a power transmission and transformation equipment ultraviolet discharge facula segmentation method based on deep learning according to an embodiment of the invention;
fig. 5 is an ultraviolet discharge spectrum FCN segmentation effect diagram of an ultraviolet discharge light spot segmentation method of a power transmission and transformation device based on deep learning according to an embodiment of the present invention;
fig. 6 is a schematic structural diagram of an ultraviolet discharge light spot segmentation device of a power transmission and transformation device based on deep learning according to an embodiment of the present invention;
fig. 7 is a schematic structural diagram of an ultraviolet discharge light spot segmentation device of a power transmission and transformation device based on deep learning according to an embodiment of the present invention.
Detailed Description
Reference will now be made in detail to the present embodiments of the present invention, examples of which are illustrated in the accompanying drawings, wherein the accompanying drawings are used to supplement the description of the written description so that one can intuitively and intuitively understand each technical feature and overall technical scheme of the present invention, but not to limit the scope of the present invention.
In the description of the present invention, it should be understood that references to orientation descriptions such as upper, lower, front, rear, left, right, etc. are based on the orientation or positional relationship shown in the drawings, are merely for convenience of description of the present invention and to simplify the description, and do not indicate or imply that the apparatus or elements referred to must have a particular orientation, be constructed and operated in a particular orientation, and thus should not be construed as limiting the present invention.
In the description of the present invention, a number means one or more, a number means two or more, and greater than, less than, exceeding, etc. are understood to not include the present number, and above, below, within, etc. are understood to include the present number. The description of the first and second is for the purpose of distinguishing between technical features only and should not be construed as indicating or implying relative importance or implicitly indicating the number of technical features indicated or implicitly indicating the precedence of the technical features indicated.
In the description of the present invention, unless explicitly defined otherwise, terms such as arrangement, installation, connection, etc. should be construed broadly and the specific meaning of the terms in the present invention can be reasonably determined by a person skilled in the art in combination with the specific contents of the technical scheme.
The traditional segmentation and extraction thought of the ultraviolet spectrum light spots at present mainly comprises the following steps: firstly, filtering, noise reduction and graying pretreatment are carried out on an ultraviolet map; then, performing binarization processing on the ultraviolet spectrum by setting a proper threshold value to obtain a segmentation map of the primary ultraviolet spot; finally, the image processing method through the open operation and the close operation eliminates noise and small light spots. The scholars propose to divide the discharge area by using a mathematical morphology method, and the method can effectively remove noise points in the discharge image and make up holes and cracks. However, morphological methods involve the problem of selecting the geometry and dimensions of "structural elements", which are different in shape and dimensions, and the segmentation effect on the same image will not be negligible, and the strategies for the variation of the structural elements remain to be further studied. And a scholars adopt Sobel and Canny edge detection algorithm to detect the edge of the image, so that the method has the advantages of high calculation efficiency, easiness in programming and the like, but the obtained image edge is a discrete point, and the subsequent calculation is difficult to directly carry out. Still further, the image is processed by selecting an appropriate threshold value based on the pixel gray level distribution of the image, i.e. the image element is retained when the gray level value of the image is greater than the selected threshold value, otherwise the image element is eliminated. The method is simple in principle and easy to implement, but an accurate threshold value is difficult to determine, and noise with a high gray value far away from a discharge point is easy to be misrecorded as a discharge area. There are studies to characterize the discharge state by using the sum of gray values of the discharge regions, but the study object is only directed to ultraviolet images containing a single discharge point, whereas in practice corona discharge of high-voltage equipment may have a plurality of discharge points at the same time.
The research results are integrated, and it is easy to find that the current ultraviolet imaging discharge area main light spot extraction method is not ideal enough due to the artificially set multiple steps and different extraction characteristics, and the proposed discharge area segmentation extraction algorithm is insufficient for covering ultraviolet light spot segmentation extraction under multiple conditions of multiple discharge points, multiple discrete interference points, sky white cluster interference and the like. Meanwhile, the method is too dependent on the pretreatment of the ultraviolet spectrum, the pretreatment intuitively influences the ultraviolet discharge light spot segmentation and extraction effect, and a fixed pretreatment mode and a reasonable threshold cannot be effectively set to realize the ultraviolet light spot segmentation and extraction under complex conditions.
Based on the above situation, the invention provides a power transmission and transformation equipment ultraviolet discharge light spot segmentation method and device based on deep learning, which are used for solving the problems.
Referring to fig. 1, in a first aspect of the present invention, a method for dividing ultraviolet discharge light spots of a power transmission and transformation device based on deep learning is provided, including the following steps:
s1: acquiring an ultraviolet spectrum of power transmission and transformation equipment;
s2: sequentially carrying out feature extraction and sampling operation on the ultraviolet map through a first convolution layer, a first pooling layer, a second convolution layer, a second pooling layer, a third convolution layer and a third pooling layer to obtain a first prediction feature map;
s3: the first prediction feature map is subjected to feature extraction of a fourth convolution layer and sampling operation of a fourth pooling layer to obtain a second prediction feature map;
s4: the second prediction feature map is subjected to feature extraction of a fifth convolution layer and sampling operation of a fifth pooling layer to obtain a third prediction feature map;
s5: 8 times of up-sampling prediction processing is carried out on the first prediction feature map and the third prediction feature map through an FCN-8s model, 16 times of up-sampling prediction processing is carried out on the second prediction feature map and the third prediction feature map through an FCN-16s model, 32 times of up-sampling prediction processing is carried out on the third prediction feature map through an FCN-32s model after the third prediction feature map passes through a first full-connection layer and a second full-connection layer, and segmented images are output through deconvolution during the three types of up-sampling prediction processing, so that segmentation extraction of ultraviolet discharge main light spots is realized.
The invention mainly comprises a full convolution neural network and multi-scale feature fusion. When the power transmission and transformation equipment has abnormal discharge, the discharge photon number signal is captured by an ultraviolet imager and converted into an image signal. The invention can process the image signal, extract the discharge area, avoid the interference of complex background noise, and can be effectively applied to intelligent inspection of power transmission and transformation equipment to reduce the complexity of manually processing ultraviolet images.
The full convolution neural network firstly performs abstract feature extraction through convolution layers, each convolution layer comprises a plurality of convolution neurons, feature extraction of different local features is realized through a sliding convolution kernel, and once the local features are extracted, the position relation among other features is also determined. Each output profile contains a convolution with a plurality of input profiles. And then carrying out secondary sampling operation through the pooling layer to obtain a characteristic layer. And finally outputting and inputting the split pictures with the same size through the deconvolution layer.
According to the multi-scale feature fusion method, in order to achieve a better segmentation effect of ultraviolet discharge main light spots, full-convolution neural network FCN-32s, FCN-16s and FCN-8s models are respectively built according to the number of the detail features of a fused shallow convolution layer, and finally an optimal full-convolution network structure model is determined through experimental effects.
The method comprises a full convolution neural network and multi-scale feature fusion, the full convolution neural network model is utilized to automatically extract ultraviolet light spots in the ultraviolet spectrum and learn the characteristics of textures, shapes and the like of the light spots, the defect that a traditional segmentation model cannot effectively separate a high-brightness white background and scattered small light spots is avoided, the complex feature selection process of the traditional segmentation algorithm model is avoided, and the autonomy and the intellectualization of end-to-end feature extraction are realized through the FCN model. The automatic cutting device has the advantages that the workload of manual power inspection is reduced, the inspection efficiency and accuracy are improved, the automatic cutting device has great practicability, interference sources such as small noise light spots, white sky with the same color as the light spots, complex backgrounds and the like are effectively eliminated, and automatic cutting of ultraviolet discharge light spots of power transmission and transformation equipment under the complex backgrounds and interference is realized. The automatic segmentation and diagnosis method greatly reduces the condition of false detection, and makes full-automatic inspection of abnormal discharge of electrical equipment possible.
According to the deep learning-based power transmission and transformation equipment ultraviolet discharge light spot segmentation method provided by the first aspect of the invention, the extracted features in the ultraviolet spectrum comprise texture features and shape features of the main light spot.
Referring to fig. 2, the full connection layer in the convolutional neural network is replaced by the convolutional layer through the full convolutional neural network, and at the same time, up-sampling calculation is performed on the last convolutional layer feature map, so that the feature map is restored to the same size as the input image, and on the premise of keeping the spatial information of the input original image, classification output predicted values of each input pixel are kept, and the segmentation extraction of the ultraviolet main light spots of the input ultraviolet map is realized.
Referring to fig. 3, since the image stored by the solar blind ultraviolet imager contains many detailed characteristic information such as the outline and the shape of the ultraviolet discharge main light spot, if the ultraviolet spectrum with the ultraviolet discharge light spot is directly deconvolved after passing through the full convolution layer, many detailed characteristics are lost due to the dimension reduction of the image calculated by the multi-layer convolution, and the obtained result is rough. Therefore, in order to achieve a better segmentation effect on the ultraviolet discharge main light spots, the invention achieves a multi-scale fusion of two network structure models of FCN-16s and FCN-8s in combination with the detailed characteristics of shallower convolution layers besides a classical full convolution neural network FCN-32s model, and finally determines an optimal full convolution network structure model through experimental effects.
Referring to fig. 4, in order to more fully reveal the model parameters after training of the full convolutional neural network, the model parameters are derived from the model file generated by training to perform visual gray scale coloring so as to more intuitively understand the network model parameter distribution generated by the full convolutional neural network, and a model multi-layer visual effect diagram is shown in fig. 4. In the full convolution visualization network for segmenting and extracting the main light spot of the ultraviolet spectrum in fig. 4, the network of each layer is composed of a plurality of feature images. As the layers of the convolution layer are deep, the feature map of the convolution layer becomes diverse, and the edge features of the feature map gradually appear. The change of the multi-layer convolution characteristic diagram shows that the sensitivity degree of the full convolution neural network to the outline edge of the ultraviolet main light spot in the input ultraviolet picture is gradually improved, and finally the segmented image is output through deconvolution, so that the segmentation and extraction of the ultraviolet discharge main light spot are realized.
Referring to fig. 5, by comparing FCN models of FCN-8s, FCN-16s and FCN-32s, which fuse more detailed features in different degrees, the ultraviolet discharge patterns in the test set are tested by the FCN model parameters after training is finished, and as shown in fig. 5, the test effect diagram of the test set in the three FCN models is shown.
Referring to fig. 6, in a second aspect of the present invention, there is provided a deep learning-based ultraviolet discharge spot segmentation apparatus 100 for power transmission and transformation equipment, including, but not limited to, the following units: an ultraviolet map acquisition unit 110, a first predicted feature map output unit 120, a second predicted feature map output unit 130, a third predicted feature map output unit 140, and a multi-scale feature fusion segmentation extraction unit 150.
The ultraviolet spectrum acquisition unit 110 is used for acquiring an ultraviolet spectrum of the power transmission and transformation equipment;
the first prediction feature map output unit 120 is configured to sequentially perform feature extraction and sampling operations on the ultraviolet map through a first convolution layer, a first pooling layer, a second convolution layer, a second pooling layer, a third convolution layer, and a third pooling layer, so as to obtain a first prediction feature map;
a second prediction feature map output unit 130, configured to obtain a second prediction feature map by performing feature extraction of a fourth convolution layer and sampling operation of a fourth pooling layer on the first prediction feature map;
a third prediction feature map output unit 140, configured to obtain a third prediction feature map by performing feature extraction of a fifth convolution layer and sampling operation of a fifth pooling layer on the second prediction feature map;
the multiscale feature fusion segmentation extraction unit 150 performs 8 times of upsampling prediction processing on the first prediction feature map and the third prediction feature map through an FCN-8s model, performs 16 times of upsampling prediction processing on the second prediction feature map and the third prediction feature map through an FCN-16s model, performs 32 times of upsampling prediction processing on the third prediction feature map through a first full-connection layer and a second full-connection layer and then through an FCN-32s model, and outputs segmented images through deconvolution during the three upsampling prediction processing, so as to realize segmentation extraction of ultraviolet discharge main light spots.
It should be noted that, since the deep learning-based power transmission and transformation equipment ultraviolet discharge light spot segmentation device 100 in the present embodiment and the above-described deep learning-based power transmission and transformation equipment ultraviolet discharge light spot segmentation method are based on the same inventive concept, the corresponding content in the method embodiment is also applicable to the present device embodiment, and will not be described in detail here.
Referring to fig. 7, in a third aspect of the present invention, a deep learning-based power transmission and transformation apparatus ultraviolet discharge light spot segmentation apparatus 200 is provided, and the deep learning-based power transmission and transformation apparatus ultraviolet discharge light spot segmentation apparatus 200 may be any type of intelligent terminal, such as a mobile phone, a tablet computer, a personal computer, and the like.
Specifically, the deep learning-based power transmission and transformation device ultraviolet discharge light spot segmentation device 200 includes: one or more control processors 210 and a memory 220, one control processor 210 being exemplified in fig. 7.
The control processor 210 and the memory 220 may be connected by a bus or otherwise, for example in fig. 7.
The memory 220 is used as a non-transitory computer readable storage medium, and may be used to store a non-transitory software program, a non-transitory computer executable program, and a module, such as program instructions/modules corresponding to the deep learning-based power transmission and transformation device ultraviolet discharge light spot segmentation method in the embodiment of the present invention, for example, the ultraviolet map acquisition unit 110, the first prediction feature map output unit 120, the second prediction feature map output unit 130, the third prediction feature map output unit 140, and the multi-scale feature fusion segmentation extraction unit 150 shown in fig. 6. The control processor 210 executes various functional applications and data processing of the deep learning-based power transmission and transformation equipment ultraviolet discharge light spot segmentation device 100 by running non-transitory software programs, instructions and modules stored in the memory 220, namely, the deep learning-based power transmission and transformation equipment ultraviolet discharge light spot segmentation method in the method embodiment is implemented.
Memory 220 may include a storage program area that may store an operating system, at least one application program required for functionality, and a storage data area; the storage data area may store data or the like created according to the use of the deep learning-based power transmission and transformation device ultraviolet discharge spot segmentation apparatus 100. In addition, the memory 220 may include high-speed random access memory 220, and may also include non-transitory memory 220, such as at least one disk memory 220 piece, flash memory device, or other non-transitory solid state memory 220 piece. In some embodiments, the memory 220 optionally includes a memory 220 remotely located relative to the control processor 210, the remote memories 220 being connectable to the deep learning based power transmission and transformation device ultraviolet discharge spot segmentation device 200 via a network. Examples of such networks include, but are not limited to, the internet, intranets, local area networks, mobile communication networks, and combinations thereof.
The one or more modules are stored in the memory 220, and when executed by the one or more control processors 210, perform the deep learning-based power transmission and transformation device ultraviolet discharge spot segmentation method in the above method embodiment, for example, perform the method steps S1 to S5 in fig. 1 described above, to implement the functions of the units 110 to 150 in fig. 6.
In a fourth aspect of the present invention, a computer-readable storage medium is provided, where computer-executable instructions are stored, where the computer-executable instructions are executed by one or more control processors 210, for example, by one of the control processors 210 in fig. 7, and where the one or more control processors 210 are caused to perform the deep learning-based power transmission and transformation device ultraviolet discharge spot segmentation method in the method embodiment described above, for example, to perform the method steps S1 to S5 in fig. 1 described above, to implement the functions of the units 110-150 in fig. 6.
The above described embodiments of the apparatus are only illustrative, wherein the units described as separate components may or may not be physically separate, i.e. may be located in one place, or may be distributed over a plurality of network units. Some or all of the modules may be selected according to actual needs to achieve the purpose of the solution of this embodiment.
From the above description of embodiments, it will be apparent to those skilled in the art that the embodiments may be implemented in software plus a general purpose hardware platform. Those skilled in the art will appreciate that all or part of the processes implementing the methods of the above embodiments may be implemented by a computer program for instructing relevant hardware, where the program may be stored in a computer readable storage medium, and where the program, when executed, may include the processes of the embodiments of the methods described above. The storage medium may be a magnetic disk, an optical disk, a Read Only Memory (ROM), a random access Memory (Random Access Memory, RAM), or the like.
In a fifth aspect of the present invention, there is provided a computer program product comprising a computer program stored on a computer readable storage medium, the computer program comprising program instructions which, when executed by a computer, cause the computer to perform a deep learning based power transmission and transformation apparatus uv discharge spot segmentation method as described above.
The embodiments of the present invention have been described in detail with reference to the accompanying drawings, but the present invention is not limited to the above embodiments, and various changes can be made within the knowledge of one of ordinary skill in the art without departing from the spirit of the present invention.
Claims (5)
1. The ultraviolet discharge light spot segmentation method for the power transmission and transformation equipment based on deep learning is characterized by comprising the following steps of:
acquiring an ultraviolet spectrum of power transmission and transformation equipment;
sequentially carrying out feature extraction and sampling operation on the ultraviolet map through a first convolution layer, a first pooling layer, a second convolution layer, a second pooling layer, a third convolution layer and a third pooling layer to obtain a first prediction feature map;
the first prediction feature map is subjected to feature extraction of a fourth convolution layer and sampling operation of a fourth pooling layer to obtain a second prediction feature map;
the second prediction feature map is subjected to feature extraction of a fifth convolution layer and sampling operation of a fifth pooling layer to obtain a third prediction feature map;
and 8 times of up-sampling prediction processing is carried out on the first prediction feature map and the third prediction feature map through an FCN-8s model, 16 times of up-sampling prediction processing is carried out on the second prediction feature map and the third prediction feature map through an FCN-16s model, 32 times of up-sampling prediction processing is carried out on the third prediction feature map through an FCN-32s model after the third prediction feature map passes through a first full-connection layer and a second full-connection layer, abstract feature extraction is carried out through a convolution layer firstly, each convolution layer comprises a plurality of convolution neurons, feature extraction of different local features is realized through a sliding convolution kernel, each output feature map comprises a feature map layer obtained through secondary sampling operation through a pooling layer, and finally segmented images with the same size are output and input through a deconvolution layer, so that segmented extraction of ultraviolet discharge main light spots is realized.
2. The deep learning-based power transmission and transformation equipment ultraviolet discharge light spot segmentation method is characterized by comprising the following steps of: the extracted features in the ultraviolet spectrum comprise texture features and shape features of the main light spot.
3. Power transmission and transformation equipment ultraviolet discharge facula segmentation device based on deep learning, its characterized in that includes following unit:
the ultraviolet spectrum acquisition unit is used for acquiring an ultraviolet spectrum of the power transmission and transformation equipment;
the first prediction feature map output unit is used for sequentially carrying out feature extraction and sampling operation on the ultraviolet map through a first convolution layer, a first pooling layer, a second convolution layer, a second pooling layer, a third convolution layer and a third pooling layer to obtain a first prediction feature map;
the second prediction feature map output unit is used for extracting the features of the first prediction feature map through a fourth convolution layer and performing sampling operation of a fourth pooling layer to obtain a second prediction feature map;
the third prediction feature map output unit is used for extracting the features of the second prediction feature map through a fifth convolution layer and sampling operation of a fifth pooling layer to obtain a third prediction feature map;
the multi-scale feature fusion segmentation extraction unit is used for carrying out 8 times of upsampling prediction processing on the first prediction feature map and the third prediction feature map through an FCN-8s model, carrying out 16 times of upsampling prediction processing on the second prediction feature map and the third prediction feature map through an FCN-16s model, carrying out 32 times of upsampling prediction processing on the third prediction feature map through a first full-connection layer and a second full-connection layer and then through an FCN-32s model, and carrying out abstract feature extraction through a convolution layer when the three upsampling prediction processing are carried out.
4. Power transmission and transformation equipment ultraviolet discharge facula segmentation equipment based on deep learning, its characterized in that: comprising at least one control processor and a memory for communication connection with the at least one control processor; the memory stores instructions executable by the at least one control processor to enable the at least one control processor to perform the deep learning based power transmission and transformation device ultraviolet discharge spot segmentation method of claim 1 or 2.
5. A computer-readable storage medium storing computer-executable instructions for causing a computer to perform the deep learning-based power transmission and transformation device ultraviolet discharge spot segmentation method according to claim 1 or 2.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911066537.7A CN111047598B (en) | 2019-11-04 | 2019-11-04 | Deep learning-based ultraviolet discharge light spot segmentation method and device for power transmission and transformation equipment |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201911066537.7A CN111047598B (en) | 2019-11-04 | 2019-11-04 | Deep learning-based ultraviolet discharge light spot segmentation method and device for power transmission and transformation equipment |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN111047598A CN111047598A (en) | 2020-04-21 |
| CN111047598B true CN111047598B (en) | 2023-09-01 |
Family
ID=70232763
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201911066537.7A Active CN111047598B (en) | 2019-11-04 | 2019-11-04 | Deep learning-based ultraviolet discharge light spot segmentation method and device for power transmission and transformation equipment |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN111047598B (en) |
Families Citing this family (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111932493A (en) * | 2020-06-28 | 2020-11-13 | 北京国网富达科技发展有限责任公司 | Power distribution network partial discharge ultrasonic detection method and system |
| CN113065484A (en) * | 2021-04-09 | 2021-07-02 | 华北电力大学（保定） | Insulator contamination state assessment method based on ultraviolet spectrum |
| CN113538351B (en) * | 2021-06-30 | 2024-01-19 | 国网山东省电力公司电力科学研究院 | Method for evaluating defect degree of external insulation equipment by fusing multiparameter electric signals |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108734143A (en) * | 2018-05-28 | 2018-11-02 | 江苏迪伦智能科技有限公司 | A kind of transmission line of electricity online test method based on binocular vision of crusing robot |
| CN109087327A (en) * | 2018-07-13 | 2018-12-25 | 天津大学 | A kind of thyroid nodule ultrasonic image division method cascading full convolutional neural networks |
| CN109118491A (en) * | 2018-07-30 | 2019-01-01 | 深圳先进技术研究院 | A kind of image partition method based on deep learning, system and electronic equipment |
| US10304193B1 (en) * | 2018-08-17 | 2019-05-28 | 12 Sigma Technologies | Image segmentation and object detection using fully convolutional neural network |
-
2019
- 2019-11-04 CN CN201911066537.7A patent/CN111047598B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108734143A (en) * | 2018-05-28 | 2018-11-02 | 江苏迪伦智能科技有限公司 | A kind of transmission line of electricity online test method based on binocular vision of crusing robot |
| CN109087327A (en) * | 2018-07-13 | 2018-12-25 | 天津大学 | A kind of thyroid nodule ultrasonic image division method cascading full convolutional neural networks |
| CN109118491A (en) * | 2018-07-30 | 2019-01-01 | 深圳先进技术研究院 | A kind of image partition method based on deep learning, system and electronic equipment |
| US10304193B1 (en) * | 2018-08-17 | 2019-05-28 | 12 Sigma Technologies | Image segmentation and object detection using fully convolutional neural network |
Non-Patent Citations (1)
| Title |
|---|
| Shaotong Pei et al..UV-flashover evaluation of porcelain insulators based on deep learning.《IET Science, Measurement & Technology》.2018,第第12卷卷(第第12卷期),全文. * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN111047598A (en) | 2020-04-21 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN112734692B (en) | Defect identification method and device for power transformation equipment | |
| CN111047598B (en) | Deep learning-based ultraviolet discharge light spot segmentation method and device for power transmission and transformation equipment | |
| CN107578418B (en) | Indoor scene contour detection method fusing color and depth information | |
| CN110321933B (en) | Fault identification method and device based on deep learning | |
| CN111784633B (en) | Insulator defect automatic detection algorithm for electric power inspection video | |
| CN111046880A (en) | Infrared target image segmentation method and system, electronic device and storage medium | |
| CN107679495B (en) | Detection method for movable engineering vehicles around power transmission line | |
| US11238301B2 (en) | Computer-implemented method of detecting foreign object on background object in an image, apparatus for detecting foreign object on background object in an image, and computer-program product | |
| CN111598889B (en) | Identification method and device for inclination fault of equalizing ring and computer equipment | |
| CN110763685B (en) | Artificial intelligent detection method and device for DFB semiconductor laser chip surface defects | |
| CN104504686A (en) | Hyper-spectral image abnormity detection method adopting local self-adaptive threshold segmentation | |
| CN108318773B (en) | Transmission conductor strand breakage detection method and system | |
| CN112730251B (en) | Device and method for detecting screen color defects | |
| CN106228541A (en) | The method and device of screen location in vision-based detection | |
| CN111833369A (en) | Alum image processing method, system, medium and electronic device | |
| CN112819784A (en) | Method and system for detecting broken strands and scattered strands of wires of distribution line | |
| CN109345586A (en) | Electrical equipment discharge characteristic extracting method based on ultraviolet imagery technology | |
| CN112730454A (en) | Intelligent damage detection method for composite material based on fusion of optics, infrared thermal waves and ultrasonic waves | |
| Luo et al. | A new recognition algorithm for high‐voltage lines based on improved LSD and convolutional neural networks | |
| CN110992306B (en) | Method and device for segmenting deteriorated insulator in infrared image based on deep learning | |
| CN111626104B (en) | Cable hidden trouble point detection method and device based on unmanned aerial vehicle infrared thermal image | |
| CN115965934A (en) | Parking space detection method and device | |
| Wang et al. | Machine vision-based conveyor belt tear detection in a harsh environment | |
| CN114708266A (en) | Tool, method and device for detecting card defects and medium | |
| CN117456371B (en) | Group string hot spot detection method, device, equipment and medium |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |