CN115908786A - Electrical cabinet grounding cable abnormity identification method and system based on deep learning - Google Patents
Electrical cabinet grounding cable abnormity identification method and system based on deep learning Download PDFInfo
- Publication number
- CN115908786A CN115908786A CN202211439915.3A CN202211439915A CN115908786A CN 115908786 A CN115908786 A CN 115908786A CN 202211439915 A CN202211439915 A CN 202211439915A CN 115908786 A CN115908786 A CN 115908786A
- Authority
- CN
- China
- Prior art keywords
- grounding cable
- network
- module
- picture data
- abnormity
- 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.)
- Pending
Links
Images
Classifications
-
- 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
Landscapes
- Image Analysis (AREA)
Abstract
An electrical cabinet grounding cable abnormity identification method and system based on deep learning are disclosed, wherein the identification method comprises the following steps: step 1, acquiring original picture data of a grounding cable of an electric power protection cabinet; step 2, preprocessing original picture data of the grounding cable of the power protection cabinet to obtain sample picture data; step 3, constructing a convolutional neural network grounding cable appearance abnormity detection model, inputting sample picture data to train the abnormity detection model, inputting the trained abnormity detection model into the sample picture data in the step 2 to obtain a detection frame of the sample picture, and fusing the output sample picture detection frames to obtain an abnormal region detection frame of the original picture; and 4, further processing the result output in the step 3 by using a digital image processing method to obtain a final abnormal area of the grounding cable of the power protection cabinet. The method has high detection and identification speed, and can overcome the defects of dense ground wire distribution, low analysis speed and the like of the power control cabinet.
Description
Technical Field
The invention belongs to the technical field of electrical equipment, and particularly relates to an electrical cabinet grounding cable abnormity identification method and system based on deep learning.
Background
The electric power protection cabinet has wide application in national power grid construction, wherein the grounding cable is used as an important component of the protection cabinet and plays a vital role in normal operation of equipment. At present, the abnormity of the grounding cable is mainly identified by monitoring the electrical quantity index, however, when the electrical quantity index is abnormal, the equipment is usually failed, and serious consequences are easy to occur, so that the monitoring of the electrical index is difficult to play a good early warning role. The appearance of the grounding cable is used as the most direct performance characteristic of the grounding cable, the abnormal appearance greatly influences the running state of equipment, meanwhile, when the terminal of the appearance of the grounding cable is loosened, serious faults of the equipment cannot be caused immediately, and the appearance of the grounding cable is monitored to play a role in early warning of the faults of the equipment. Therefore, the abnormal detection of the appearance of the grounding cable is significant for the reliable operation of the power equipment.
Because the ground connection winding displacement of electric power protection cabinet is more in quantity, the mode of artifical patrolling and examining is consuming time, hard. In addition, a large potential safety hazard also exists in close-range inspection in the operation process of the equipment, and the problems can be better solved through an image recognition technology. At present, the image recognition technology for the grounding cable generally uses traditional image analysis methods, such as edge feature extraction, support vector machine classification, and the like. The method perhaps has a good recognition result when single-class defect feature analysis is carried out on the grounding cable, but the situation that multiple types of defect features exist in the power protection cabinet is difficult to recognize accurately, and in addition, detection time is greatly increased by comprehensively applying multiple image analysis methods. The target detection method based on deep learning can detect a plurality of types of defects in one picture, has high detection and identification speed, and can overcome the defects of large quantity of grounding cables, low analysis speed and the like when being applied to abnormal identification of the grounding cables of the power protection cabinet.
Prior art document 1 (CN 115222983A) discloses a cable breakage detection method and system. The method comprises the following steps: acquiring an image of a cable to be detected; inputting the image of the cable to be detected into the trained convolutional neural network to obtain the damage category of the cable to be detected; wherein the loss function of the convolutional neural network is the sum of the cross entropy loss function and the ratio of the intra-class distance metric and the inter-class distance metric. The defect of the prior art document 1 is that the dense cable abnormity detection precision is not high, and the invention adopts a multi-target detection algorithm based on deep learning, so that the conditions of dense cable distribution and different types of cables, such as an electric control cabinet, can be accurately identified.
Prior art document 2 (CN 114155435A) discloses a cable interface loose detection method based on deep learning. Before detection, marking lines on cable connectors at two ends of an interface; during detection, a cable interface image with a marking line side is shot and obtained, a neural network model is utilized to detect and identify a cable interface in the image and output a corresponding cable interface image frame selection area, and then the marking line in the cable interface image area is extracted. The prior art document 2 has a disadvantage of low degree of intelligence. According to the invention, identification lines are not needed to be made on the cable joints at two ends of the interface before detection, and the positions of the cable joints are automatically positioned through a deep learning algorithm.
Disclosure of Invention
In order to overcome the defects in the prior art, the electrical cabinet grounding cable abnormity identification method and system based on deep learning are provided, can be embedded into a large-scale resource management platform, can perform intelligent analysis on grounding cables in an electrical protection cabinet, reduce the dependence of manual routing inspection analysis, and improve the accuracy and efficiency of electrical protection cabinet grounding cable abnormity image identification.
The invention adopts the following technical scheme.
An electrical cabinet grounding cable abnormity identification method based on deep learning comprises the following steps:
and 4, further processing the result output in the step 3 by using a digital image processing method according to the abnormal area detection frame obtained in the step 3 to obtain a final abnormal area of the grounding cable of the power protection cabinet.
Preferably, in step 1, the fixed camera photographs and obtains original picture data of the grounding cable of the power protection cabinet through the inspection robot.
Preferably, in step 2, the preprocessing of the original picture data includes: cutting an original picture of the grounding cable of the power protection cabinet in a sliding frame mode, wherein the size of the original picture is (w x h), the size of the sliding frame is set to be (0.5 w x 0.6 h), the step length of the sliding frame is set to be 0.25w in the width direction and 0.4h in the height direction, and finally obtaining 6 sample pictures through cutting of the sliding frame.
Preferably, step 3 specifically comprises:
step 3.1, constructing a convolutional neural network grounding cable appearance abnormity detection model, initializing network weight, designing an activation function of a convolutional layer autonomously, and training the model;
step 3.2, inputting the preprocessed pictures into the trained network model to obtain the detection frame result of each cut picture;
and 3.3, mapping the detection frame of each local clipping sample to the corresponding position of the original picture according to the picture clipping strategy in the step 2 and the detection result in the step 3.2, so that the detection frame is matched with the pixel information of the original picture, and then fusing the obtained detection frame information of all sample pictures through non-maximum value inhibition to finally obtain all abnormal area detection frames of the original picture.
Preferably, in step 3.1, constructing an activation function of a convolutional layer in a convolutional neural network grounding cable appearance anomaly detection model is as follows:
in the formula (I), the compound is shown in the specification,
x i representing the input tensor of the i-th network layer,
γ i represents one activation function coefficient of the i-th layer,
y i representing the i-th network layer output tensor.
In step 3.1, the network weights to be initialized for initializing during training are from an upper bound to a lower bound
Is obtained from the average distribution of (a), the formula is as follows: />
In the formula (I), the compound is shown in the specification,
representing that the variable obeys a certain distribution;
w represents the weight to be initialized in the network;
u represents a uniform distribution;
n represents the number of neurons;
j denotes the current network layer number.
Preferably, in step 3.1, the convolutional neural network grounding cable appearance anomaly detection model includes: 5 network modules, a multi-scale pyramid module and a scale attention mechanism module;
the last layer of each network module is a down-sampling layer, the convolution kernel of each network module convolution layer is 3 x 3, the convolution front edge is expanded by 1 bit each time, and the convolution kernel moving step length is 1; the down sampling uses convolution with the step length of 2, and gathers the characteristic information input by the upper layer while reducing the calculated amount; the characteristic scale of the output of each network module is different;
the multi-scale pyramid module is used for realizing the network with emphasis fusion of network characteristics of each layer, and the specific process is as follows: the 5 network modules are divided according to the down-sampling layer position, the output characteristic graph scale of the network modules is from large to small in the sequence from network input to network output, when the output of the current network module is calculated, the characteristic graph output by the current module is up-sampled to the same size as the characteristic graph of the previous network module through bilinear interpolation, and then the two characteristic graphs are added to be used as the final output characteristic graph of the current network module;
the scale attention mechanism module is used for calculating the attention weight of each scale by passing the output of each network module through the convolutional layer, the global average pooling layer and the full connection layer.
Preferably, in step 4, the digital image processing method specifically includes: firstly, enhancing original picture data, then carrying out edge extraction on the image by adopting a Sobel operator so as to obtain a binary edge image, and then carrying out characteristic curve extraction on the result image output in the step 3 by adopting Hough transformation so as to obtain an extracted characteristic curve containing the grounding cable in the picture.
An electrical cabinet grounding cable abnormity identification system based on deep learning comprises: collection module, preprocessing module, modeling module, identification module, wherein:
the acquisition module is used for acquiring original picture data of a grounding cable of the power protection cabinet;
the preprocessing module is used for preprocessing the original picture data of the grounding cable of the power protection cabinet to obtain sample picture data;
the modeling module is used for constructing a convolutional neural network grounding cable appearance abnormity detection model, inputting sample picture data to train the abnormity detection model, inputting the sample picture data into the trained abnormity detection model to obtain a detection frame of a sample picture, and fusing the output sample picture detection frames to obtain an abnormal region detection frame of an original picture;
and the identification module is used for further processing the result output by the modeling module by using a digital image processing method according to the obtained abnormal area detection frame to obtain the final abnormal area of the grounding cable of the power protection cabinet.
A terminal comprising a processor and a storage medium; wherein:
the storage medium is used for storing instructions;
the processor is used for operating according to the instruction to execute the steps of the electrical cabinet grounding cable abnormity identification method based on deep learning.
The computer readable storage medium stores a computer program, and the program is executed by a processor to realize the steps of the electrical cabinet grounding cable abnormity identification method based on deep learning.
Compared with the prior art, the method has the advantages that the detection and identification speed is high, the defects that grounding wires of the electric control cabinet are densely distributed and the analysis speed is low can be overcome, the method can be embedded into a large-scale resource management platform to intelligently analyze the dense cables in the electric control cabinet, the dependence of manual routing inspection analysis is reduced, and the accuracy and the efficiency of identifying the abnormal images of the cables of the electric control cabinet are improved.
Drawings
Fig. 1 is a flow chart of an electrical cabinet grounding cable abnormality identification method based on deep learning according to the invention;
fig. 2 is a diagram of an electrical cabinet grounding cable abnormality identification network structure based on deep learning.
Detailed Description
In order to make the objects, technical solutions and advantages of the present invention clearer, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present invention. The embodiments described herein are only some embodiments of the invention, and not all embodiments. All other embodiments obtained by a person skilled in the art without any inventive step based on the spirit of the present invention are within the scope of the present invention.
Example 1.
An electrical cabinet grounding cable abnormity identification method based on deep learning is disclosed, as shown in fig. 1, and comprises the following steps:
Preferably, this embodiment is through patrolling and examining robot, fixed camera, according to presetting program and independently adjusting the camera cloud platform and predetermineeing the angle, and the whole is independently taken a picture, acquires the original picture data of electric power protection cabinet ground connection cable.
because the appearance abnormal area in the grounding cable photo of the electric protection cabinet is usually small, namely, the loose area of one terminal only occupies a very small proportion of the whole picture, in order to improve the detection precision, the grounding cable photo is cut. In order not to influence the detection of the defective area of the grounding cable at the cutting edge, the cutting is carried out in a sliding frame mode. The size of the original picture is (w × h), the size of the slide frame is set to (0.5 w × 0.6 h), the step size of the slide frame is set to 0.25w in the width direction and 0.4h in the height direction, and 6 sample pictures are finally obtained through slide frame cutting. A sample picture with resize 832 × 3 was obtained.
And 3, constructing a convolutional neural network grounding cable appearance abnormity detection model, inputting the sample picture data obtained in the step 2 to train the abnormity detection model, inputting the sample picture data obtained in the step 2 into the trained abnormity detection model to obtain a detection frame of the sample picture, and fusing the output sample picture detection frames to obtain an abnormal area detection frame of the original picture.
Step 3.1, constructing a convolutional neural network grounding cable appearance abnormity detection model, initializing network weight, designing an activation function of a convolutional layer autonomously, and training the model;
the specific structure for constructing the convolutional neural network grounding cable appearance abnormity detection model is formed by stacking a plurality of convolutional layers, the non-linearity of the model is increased, the generalization capability of the model is improved, the detection performance of the abnormal part of the grounding cable is improved, and an autonomously designed activation function is used, and the specific structure is as follows:
in the formula, x i Representing the i-th network layer input tensor, gamma i An activation function coefficient representing the i-th layer for adjusting the smoothness of the activation function between positive regions, preferably 1,y i Representing the i-th network layer output tensor.
The network weight to be initialized is initialized when the model is trained, wherein the network weight is from an upper boundary to a lower boundary
Is obtained from the average distribution of (a), the formula is as follows:
in the formula (I), the compound is shown in the specification,
representing that the variable obeys a certain distribution;
w represents the weight to be initialized in the network;
u represents a uniform distribution;
n represents the number of neurons;
j denotes the current network layer number.
Preferably, in this embodiment, as shown in fig. 2, the convolutional neural network grounding cable appearance anomaly detection model includes: 5 network modules, a multi-scale pyramid module and a scale attention mechanism module;
the last layer of each network module is a down-sampling layer, the convolution kernel of each network module convolution layer is 3 x 3, the convolution front edge is expanded by 1 bit each time, and the convolution kernel moving step length is 1; the down sampling uses convolution with the step length of 2, and the feature information input by the upper layer is aggregated while the calculated amount is reduced; the characteristic scale of the output of each network module is different;
the multi-scale pyramid module is used for realizing the network with emphasis fusion of network characteristics of each layer, and the specific process is as follows: the 5 network modules are divided according to the down-sampling layer position, the output characteristic graph size is from large to small when viewed from the sequence from network input to network output, when the output of the current network module is calculated, the characteristic graph output by the current module is up-sampled to the same size as the characteristic graph of the previous network module through bilinear interpolation, and then the two characteristic graphs are added to be used as the final output characteristic graph of the current network module;
and the scale attention mechanism module is used for calculating the attention weight of each scale by passing the output of each network module through the convolutional layer, the global average pooling layer and the full-connection layer.
Step 3.2, inputting the preprocessed pictures into the trained network model to obtain the detection frame result of each cut picture;
and 3.3, mapping the detection frame of each local clipping sample to the corresponding position of the original picture according to the picture clipping strategy in the step 2 and the detection result in the step 3.2, so that the detection frame is matched with the pixel information of the original picture, and then fusing the obtained detection frame information of all sample pictures through non-maximum value inhibition to finally obtain all abnormal area detection frames of the original picture.
The whole network model is divided into 5 large network modules according to the positions of the down-sampling layers as nodes. The number of the convolution layers of the 5 network modules is respectively 2 layers, 9 layers, 12 layers, 18 layers and 9 layers.
The last layer of each network module is a down-sampling layer, the convolution kernel of the convolution layer is 3 x 3, the convolution front edge is expanded by 1 bit each time, and the convolution kernel moving step length is 1. The down-sampling uses convolution with step size 2, which can gather feature information of the upper input while reducing the amount of computation.
If the feature graph after 5 times of downsampling is regressed and classified according to the conventional method, the features of the abnormal part of the grounding cable are seriously lost, so that the model precision is seriously reduced. On the other hand, the receptive field of the high-level convolutional layer is greatly reduced, the information of the bottom convolutional layer is difficult to aggregate, and the model is difficult to learn the global information of the picture. Aiming at the problem, a multi-scale pyramid structure layer with an attention mechanism is designed to help the network to focus on fusing network characteristics of each layer, so that the model can select proper network layer characteristics, and the final performance of the model is improved. The multi-scale pyramid structure is as shown in fig. 2, 5 times of downsampling is performed, so that 5 inputs with different resolution scales are obtained, convolution, pooling and other processing are performed respectively to obtain feature maps, and finally fusion is performed, so that feature information of targets with different scales is kept as far as possible. Specifically, the feature map output by the downsampling network is upsampled to the same size as the feature map of the previous network block through bilinear interpolation, and then the two feature maps are added to be used as the output feature map of the network block. The design of the scale attention mechanism is that all network block feature maps are respectively subjected to 3 × 3 convolution, then a global average pooling is carried out, then a 5 × 1 weight vector is obtained through a concat feature series layer, an FC full connection layer and an SM nonlinear conversion layer to represent the weight corresponding to each scale, and then the obtained weight is multiplied by the output feature map of the corresponding network block to obtain a final feature map for final regression and classification features.
The conventional a priori anchor block is set using a general aspect ratio of 1:2,2:1,1: the anchor frame of 1 is possibly not matched with the rusted shape and size of the grounding cable terminal, so that the difficulty of model regression is increased, and the precision of the model is reduced.
In the embodiment, a k-means clustering algorithm is used for carrying out clustering analysis on the labeling boundary box of the grounding cable sample, so that the size of the prior frame in the model is more suitable for detecting the abnormal part of the grounding cable. And calculating to obtain prior frames of 6 scales, specifically (15, 16), (25, 22), (35, 42), (53, 51), (63, 55) and (78, 70).
In this embodiment, the loss function during model training is divided into two parts, that is, the classification loss function of the anchor frame and the IoU loss function of the prediction frame and the corresponding label frame. The classification loss function of the anchor block uses a cross-entropy loss function. And the IoU loss functions of the prediction box and the corresponding box are CIoU loss functions.
After the model is trained, in actual use, a series of detection frame information (x, y, w, h, class, confidence) is obtained from each cut sample picture in step 2, wherein x, y, w, h are position information, and class and confidence are label information. And fusing all the obtained detection frame information according to the position relation of the 6 cutting samples, and mapping the detection frame of each local cutting sample back to the corresponding position of the original picture to enable the detection frame to be matched with the pixel information of the original picture. And finally obtaining all abnormal area detection frames of the original picture.
And 4, further confirming the result in the step 3 by utilizing a digital image processing technology according to the primary abnormal area detection frame obtained in the step 3 to obtain a final abnormal area of the node cable.
And 3, acquiring all abnormal area detection frames of the whole electric power protection cabinet grounding cable photo, and firstly, preliminarily removing overlapped redundant detection frame information by a non-maximum inhibition method. In order to further reduce the false recognition of the abnormal region of the non-grounded cable, secondary determination is performed according to the appearance characteristics of the grounded cable. A grounding cable feature extraction step: firstly, enhancing an original shot picture, then carrying out edge extraction on the image by adopting a Sobel operator so as to obtain a binary edge image, and then carrying out characteristic curve extraction on the result image by adopting Hough transformation, wherein the extracted characteristic curve is considered to contain a grounding cable in the picture. And (4) because the abnormal area of the grounding cable is always positioned on the grounding cable, sequentially judging whether the coordinates of all the detection frames are overlapped with the characteristic curve of the picture to a certain extent or not, and removing the error detection frames which do not meet the overlapping requirement. And drawing the rest grounding cable appearance abnormal frames meeting the requirements on the original picture for backup storage to obtain the final grounding cable abnormality detection result of the power protection cabinet.
Example 2.
An electrical cabinet grounding cable abnormity identification system based on deep learning comprises: the system comprises an acquisition module, a preprocessing module, a modeling module and an identification module, wherein:
the acquisition module is used for acquiring original picture data of a grounding cable of the power protection cabinet;
the preprocessing module is used for preprocessing the original picture data of the grounding cable of the power protection cabinet to obtain sample picture data;
the modeling module is used for constructing a convolutional neural network grounding cable appearance abnormity detection model, inputting sample picture data to train the abnormity detection model, inputting the sample picture data into the trained abnormity detection model to obtain a detection frame of a sample picture, and fusing the output sample picture detection frames to obtain an abnormal area detection frame of an original picture;
and the identification module is used for further processing the result output by the modeling module by using a digital image processing method according to the obtained abnormal area detection frame to obtain the final abnormal area of the grounding cable of the power protection cabinet.
Example 3.
A computer-readable storage medium, which stores a program, and when the program is executed by a processor, the program implements the steps in the deep learning-based electrical cabinet grounding cable abnormality identification method according to the first embodiment of the present invention.
The detailed steps are the same as those of the electrical cabinet grounding cable abnormality identification method based on deep learning provided in embodiment 1, and are not described herein again.
Example 4.
An electronic device comprises a memory, a processor and a program which is stored on the memory and can run on the processor, wherein the processor executes the program to realize the steps of the electrical cabinet grounding cable abnormality identification method based on deep learning according to the embodiment of the invention.
The detailed steps are the same as those of the electrical cabinet grounding cable abnormality identification method based on deep learning provided in embodiment 1, and are not described herein again.
The method has the advantages of high detection and identification speed, capability of overcoming the defects of dense ground wire distribution, low analysis speed and the like of the electric control cabinet, capability of being embedded into a large-scale resource management platform to intelligently analyze the dense cables in the electric control cabinet, reduction of dependence of manual routing inspection analysis, and improvement of accuracy and efficiency of abnormal image identification of the grounding cables of the electric control cabinet.
The present disclosure may be systems, methods, and/or computer program products. The computer program product may include a computer-readable storage medium having computer-readable program instructions embodied thereon for causing a processor to implement various aspects of the present disclosure.
The computer-readable storage medium may be a tangible device that can hold and store the instructions for use by the instruction execution device. The computer readable storage medium may be, for example, but not limited to, an electronic, magnetic, optical, electromagnetic, semiconductor memory device, or any suitable combination of the foregoing. More specific examples (a non-exhaustive list) of the computer readable storage medium would include the following: a portable computer diskette, a hard disk, a Random Access Memory (RAM), a read-only memory (ROM), an erasable programmable read-only memory (EPROM or flash memory), a Static Random Access Memory (SRAM), a portable compact disc read-only memory (CD-ROM), a Digital Versatile Disc (DVD), a memory stick, a floppy disk, a mechanical coding device, such as a punch card or an in-groove protruding structure with instructions stored thereon, and any suitable combination of the foregoing. Computer-readable storage media as used herein is not to be interpreted as a transitory signal per se, such as a radio wave or other freely propagating electromagnetic wave, an electromagnetic wave propagating through a waveguide or other transmission medium (e.g., optical pulses through a fiber optic cable), or an electrical signal transmitted through an electrical wire.
The computer-readable program instructions described herein may be downloaded from a computer-readable storage medium to a respective computing/processing device, or to an external computer or external storage device via a network, such as the internet, a local area network, a wide area network, and/or a wireless network. The network may include copper transmission cables, fiber optic transmission, wireless transmission, routers, firewalls, switches, gateway computers and/or edge servers. The network adapter card or network interface in each computing/processing device receives the computer-readable program instructions from the network and forwards the computer-readable program instructions for storage in a computer-readable storage medium in the respective computing/processing device.
The computer program instructions for carrying out operations of the present disclosure may be assembler instructions, instruction Set Architecture (ISA) instructions, machine-related instructions, microcode, firmware instructions, state setting data, or source or object code written in any combination of one or more programming languages, including an object oriented programming language such as Smalltalk, C + + or the like and conventional procedural programming languages, such as the "C" programming language or similar programming languages. The computer readable program instructions may execute entirely on the user's computer, partly on the user's computer, as a stand-alone software package, partly on the user's computer and partly on a remote computer or entirely on the remote computer or server. In the latter scenario, the remote computer may be connected to the user's computer through any type of network, including a Local Area Network (LAN) or a Wide Area Network (WAN), or the connection may be made to an external computer (for example, through the Internet using an Internet service provider). In some embodiments, aspects of the disclosure are implemented by personalizing an electronic circuit, such as a programmable logic circuit, a Field Programmable Gate Array (FPGA), or a Programmable Logic Array (PLA), with state information of computer-readable program instructions, which can execute the computer-readable program instructions.
Finally, it should be noted that the above embodiments are only used for illustrating the technical solutions of the present invention and not for limiting the same, and although the present invention is described in detail with reference to the above embodiments, those of ordinary skill in the art should understand that: modifications and equivalents may be made to the embodiments of the invention without departing from the spirit and scope of the invention, which is to be covered by the claims.
Claims (11)
1. An electrical cabinet grounding cable abnormity identification method based on deep learning is characterized by comprising the following steps:
step 1, acquiring original picture data of a grounding cable of an electric power protection cabinet;
step 2, preprocessing the original picture data of the grounding cable of the power protection cabinet obtained in the step 1 to obtain sample picture data;
step 3, constructing a convolutional neural network grounding cable appearance abnormity detection model, inputting the sample picture data obtained in the step 2 to train the abnormity detection model, inputting the sample picture data obtained in the step 2 into the trained abnormity detection model to obtain a detection frame of the sample picture, and fusing the output sample picture detection frames to obtain an abnormal area detection frame of the original picture;
and 4, further processing the result output in the step 3 by using a digital image processing method according to the abnormal area detection frame obtained in the step 3 to obtain a final abnormal area of the grounding cable of the power protection cabinet.
2. The electrical cabinet grounding cable abnormity identification method based on deep learning of claim 1,
in the step 1, the fixed camera shoots to obtain original picture data of the grounding cable of the power protection cabinet through the inspection robot.
3. The electrical cabinet grounding cable abnormity identification method based on deep learning of claim 1, wherein,
in step 2, the preprocessing of the original picture data includes: cutting an original picture of the grounding cable of the power protection cabinet in a sliding frame mode, wherein the size of the original picture is (w x h), the size of the sliding frame is set to be (0.5 w x 0.6 h), the step length of the sliding frame is set to be 0.25w in the width direction and 0.4h in the height direction, and finally obtaining 6 sample pictures through cutting of the sliding frame.
4. The electrical cabinet grounding cable abnormity identification method based on deep learning of claim 1,
step 3, specifically comprising:
step 3.1, constructing a convolutional neural network grounding cable appearance abnormity detection model, initializing network weight, designing an activation function of a convolutional layer autonomously, and training the model;
step 3.2, inputting the preprocessed pictures into the trained network model to obtain the detection frame result of each cut picture;
and 3.3, mapping the detection frame of each local clipping sample to the corresponding position of the original picture according to the picture clipping strategy in the step 2 and the detection result in the step 3.2, so that the detection frame is matched with the pixel information of the original picture, and then fusing the obtained detection frame information of all sample pictures through non-maximum value inhibition to finally obtain all abnormal area detection frames of the original picture.
5. The electrical cabinet grounding cable abnormity identification method based on deep learning according to claim 4,
in step 3.1, constructing an activation function of a convolution layer in a convolutional neural network grounding cable appearance anomaly detection model as follows:
in the formula (I), the compound is shown in the specification,
x i representing the input tensor of the i-th network layer,
γ i represents one activation function coefficient of the i-th layer,
y i representing the i-th network layer output tensor.
6. The electrical cabinet grounding cable abnormity identification method based on deep learning according to claim 4,
in step 3.1, the network weights to be initialized for initializing during training are from an upper bound to a lower bound
Is obtained from the average distribution of (a), the formula is as follows:
in the formula (I), the compound is shown in the specification,
representing that the variable obeys a certain distribution;
w represents the weight in the network that needs to be initialized;
u represents a uniform distribution;
n represents the number of neurons;
j denotes the current network layer number.
7. The electrical cabinet grounding cable abnormity identification method based on deep learning according to claim 4,
in step 3.1, the convolutional neural network grounding cable appearance anomaly detection model comprises the following steps: 5 network modules, a multi-scale pyramid module and a scale attention mechanism module;
the last layer of each network module is a down-sampling layer, the convolution kernel of each network module convolution layer is 3 x 3, the convolution front edge is expanded by 1 bit each time, and the convolution kernel moving step length is 1; the down sampling uses convolution with the step length of 2, and the feature information input by the upper layer is aggregated while the calculated amount is reduced; the characteristic scale of the output of each network module is different;
the multi-scale pyramid module is used for realizing the network with emphasis fusion of network characteristics of each layer, and the specific process is as follows: the 5 network modules are divided according to the down-sampling layer position, the output characteristic graph scale of the network modules is from large to small in the sequence from network input to network output, when the output of the current network module is calculated, the characteristic graph output by the current module is up-sampled to the same size as the characteristic graph of the previous network module through bilinear interpolation, and then the two characteristic graphs are added to be used as the final output characteristic graph of the current network module;
and the scale attention mechanism module is used for calculating the attention weight of each scale by passing the output of each network module through the convolutional layer, the global average pooling layer and the full-connection layer.
8. The electrical cabinet grounding cable abnormity identification method based on deep learning of claim 1,
in step 4, the digital image processing method specifically comprises: firstly, enhancing original picture data, then carrying out edge extraction on the image by adopting a Sobel operator so as to obtain a binary edge image, and then carrying out characteristic curve extraction on the result image output in the step 3 by adopting Hough transformation so as to obtain an extracted characteristic curve containing the grounding cable in the picture.
9. A deep learning-based electrical cabinet grounding cable anomaly identification system using the method of any one of claims 1-8, comprising: the system comprises an acquisition module, a preprocessing module, a modeling module and an identification module, and is characterized in that:
the acquisition module is used for acquiring original picture data of a grounding cable of the power protection cabinet;
the preprocessing module is used for preprocessing the original picture data of the grounding cable of the power protection cabinet to obtain sample picture data;
the modeling module is used for constructing a convolutional neural network grounding cable appearance abnormity detection model, inputting sample picture data to train the abnormity detection model, inputting the sample picture data into the trained abnormity detection model to obtain a detection frame of a sample picture, and fusing the output sample picture detection frames to obtain an abnormal region detection frame of an original picture;
and the identification module is used for further processing the result output by the modeling module by using a digital image processing method according to the obtained abnormal area detection frame to obtain the final abnormal area of the grounding cable of the power protection cabinet.
10. A terminal comprising a processor and a storage medium; the method is characterized in that:
the storage medium is to store instructions;
the processor is used for operating according to the instructions to execute the steps of the electrical cabinet grounding cable abnormality identification method based on deep learning according to any one of claims 1 to 8.
11. Computer-readable storage medium, on which a computer program is stored, wherein the program, when executed by a processor, implements the steps of the deep learning-based electrical cabinet grounding cable abnormality identification method according to any one of claims 1 to 8.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211439915.3A CN115908786A (en) | 2022-11-17 | 2022-11-17 | Electrical cabinet grounding cable abnormity identification method and system based on deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211439915.3A CN115908786A (en) | 2022-11-17 | 2022-11-17 | Electrical cabinet grounding cable abnormity identification method and system based on deep learning |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN115908786A true CN115908786A (en) | 2023-04-04 |
Family
ID=86470491
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211439915.3A Pending CN115908786A (en) | 2022-11-17 | 2022-11-17 | Electrical cabinet grounding cable abnormity identification method and system based on deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN115908786A (en) |
Cited By (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116405310A (en) * | 2023-04-28 | 2023-07-07 | 北京宏博知微科技有限公司 | Network data security monitoring method and system |
| CN117934481A (en) * | 2024-03-25 | 2024-04-26 | 国网浙江省电力有限公司宁波供电公司 | Power transmission cable state identification processing method and system based on artificial intelligence |
| CN117934481B (en) * | 2024-03-25 | 2024-06-11 | 国网浙江省电力有限公司宁波供电公司 | Power transmission cable state identification processing method and system based on artificial intelligence |
-
2022
- 2022-11-17 CN CN202211439915.3A patent/CN115908786A/en active Pending
Cited By (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN116405310A (en) * | 2023-04-28 | 2023-07-07 | 北京宏博知微科技有限公司 | Network data security monitoring method and system |
| CN116405310B (en) * | 2023-04-28 | 2024-03-15 | 北京宏博知微科技有限公司 | Network data security monitoring method and system |
| CN117934481A (en) * | 2024-03-25 | 2024-04-26 | 国网浙江省电力有限公司宁波供电公司 | Power transmission cable state identification processing method and system based on artificial intelligence |
| CN117934481B (en) * | 2024-03-25 | 2024-06-11 | 国网浙江省电力有限公司宁波供电公司 | Power transmission cable state identification processing method and system based on artificial intelligence |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN111488789B (en) | Pedestrian detection method and device for monitoring based on image analysis | |
| US11475660B2 (en) | Method and system for facilitating recognition of vehicle parts based on a neural network | |
| CN110059694B (en) | Intelligent identification method for character data in complex scene of power industry | |
| US11676365B2 (en) | Explainable artificial intelligence (AI) based image analytic, automatic damage detection and estimation system | |
| US11823443B2 (en) | Segmenting objects by refining shape priors | |
| US10127449B2 (en) | Condition detection using image processing | |
| CN109636848B (en) | Unmanned aerial vehicle-based oil and gas pipeline inspection method | |
| CN110084165B (en) | Intelligent identification and early warning method for abnormal events in open scene of power field based on edge calculation | |
| CN110264444B (en) | Damage detection method and device based on weak segmentation | |
| CN112016467B (en) | Traffic sign recognition model training method, recognition method, system, device and medium | |
| US20220156483A1 (en) | Efficient three-dimensional object detection from point clouds | |
| CN116579616B (en) | Risk identification method based on deep learning | |
| CN114842026A (en) | Real-time fan blade image segmentation method and system | |
| Liu et al. | Review of data analysis in vision inspection of power lines with an in-depth discussion of deep learning technology | |
| CN116385958A (en) | Edge intelligent detection method for power grid inspection and monitoring | |
| US11200455B2 (en) | Generating training data for object detection | |
| CN115908786A (en) | Electrical cabinet grounding cable abnormity identification method and system based on deep learning | |
| CN113393442A (en) | Method and system for detecting abnormality of train parts, electronic device and storage medium | |
| CN112529881B (en) | Power control cabinet cable anomaly identification method and device | |
| CN114842183A (en) | Convolutional neural network-based switch state identification method and system | |
| Silva et al. | Line-based deep learning method for tree branch detection from digital images | |
| CN112749293A (en) | Image classification method and device and storage medium | |
| CN113569600A (en) | Method and device for identifying weight of object, electronic equipment and storage medium | |
| CN117173676A (en) | Method, device, equipment and medium for identifying lane change intention of driver | |
| Nordeng | Dead End Body Component Inspections with Convolutional Neural Networks Using UAS Imagery |
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 |