CN111650204A - Transmission line hardware defect detection method and system based on cascade target detection - Google Patents
Transmission line hardware defect detection method and system based on cascade target detection Download PDFInfo
- Publication number
- CN111650204A CN111650204A CN202010393940.7A CN202010393940A CN111650204A CN 111650204 A CN111650204 A CN 111650204A CN 202010393940 A CN202010393940 A CN 202010393940A CN 111650204 A CN111650204 A CN 111650204A
- Authority
- CN
- China
- Prior art keywords
- target detection
- network
- detection model
- transmission line
- image
- 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.)
- Granted
Links
Images
Classifications
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/84—Systems specially adapted for particular applications
- G01N21/88—Investigating the presence of flaws or contamination
- G01N21/8851—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges
-
- G—PHYSICS
- G01—MEASURING; TESTING
- G01N—INVESTIGATING OR ANALYSING MATERIALS BY DETERMINING THEIR CHEMICAL OR PHYSICAL PROPERTIES
- G01N21/00—Investigating or analysing materials by the use of optical means, i.e. using sub-millimetre waves, infrared, visible or ultraviolet light
- G01N21/84—Systems specially adapted for particular applications
- G01N21/88—Investigating the presence of flaws or contamination
- G01N21/8851—Scan or image signal processing specially adapted therefor, e.g. for scan signal adjustment, for detecting different kinds of defects, for compensating for structures, markings, edges
- G01N2021/8854—Grading and classifying of flaws
- G01N2021/8861—Determining coordinates of flaws
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F17/00—Digital computing or data processing equipment or methods, specially adapted for specific functions
- G06F17/10—Complex mathematical operations
- G06F17/15—Correlation function computation including computation of convolution operations
-
- 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 power transmission line hardware defect detection method and a system based on cascade target detection, which comprises the following steps: detecting a connection area of the image of the power transmission line by using the trained first target detection model, and cutting out the detected connection area; taking the n connection regions with the area size meeting the preset condition as an image to be identified; using the trained second target detection model to detect the tiny hardware defects of the image to be recognized, and acquiring coordinates of the tiny hardware defects on the image to be recognized; according to the method, the small hardware defect is displayed on the original image according to the mapping relation between the coordinate of the image to be identified and the original image coordinate.
Description
Technical Field
The invention relates to the field of power transmission line defect detection, in particular to a power transmission line hardware defect detection method and system based on cascade target detection.
Background
When the small hardware such as the power bolts and the like are applied to common power transmission and distribution lines, the small hardware needs to be subjected to field operation corrosion for a long time and strong collision friction, has a huge number in a power grid, and plays a role in stabilizing a base, line equipment and the like. However, the environment of the fine and small hardware is complex and harsh, and the fine and small hardware is also an element which is easy to break. Once broken, it causes a power interruption that affects the safe operation of the entire grid. At present, the image recognition processing is carried out on the defective pictures such as bolt pin lack, bolt nut lack and the like through the deep learning image recognition technology, so that defect diagnosis is formed.
When the deep learning network model is used for identifying and processing images, a large number of training samples are needed, but defect detection of small hardware mainly comprises defect categories such as pin missing, pin separating and the like, and no matter which category, the collected samples are too few, so that overfitting is easily generated when the deep learning is carried out, the effect is not ideal when the defect detection is carried out, and the traditional image augmentation modes such as random cutting, random translation, random rotation and the like cannot provide more low-level characteristics of the defects of the small hardware.
And because the fault defect target of the small hardware fittings is too small and accounts for less than 5% of the original image, the feature extraction capability of the traditional deep learning image identification method is insufficient, so that the defect detection effect of the small hardware fittings can not meet the actual service requirement.
Disclosure of Invention
Aiming at the problems in the prior art, the invention provides a power transmission line hardware defect detection method based on cascade target detection, which adopts a generation countermeasure network to enlarge shot images, increases the number of training samples of a deep learning network, provides a two-stage detection strategy, can obviously improve the detection precision of small hardware defects, has popularization and application values, and specifically comprises the following steps:
(11) acquiring a polling shot image and preprocessing the polling shot image;
(12) detecting a connecting area of the polling shot image by using a trained first target detection model, and cutting out the detected rectangular connecting area;
(13) acquiring the area size of the connection region, and taking the n connection regions with the area size meeting a preset condition as an image to be identified;
(14) using the trained second target detection model to detect the tiny hardware defects of the image to be recognized, and acquiring coordinates of the tiny hardware defects on the image to be recognized;
(15) and displaying the tiny hardware defects on the original image according to the mapping relation between the coordinates of the image to be recognized and the coordinates of the original image.
As a further optimization of the scheme, before the first target detection model and the second target detection model are trained, data amplification processing is carried out on the collected images of the power transmission line, and the images generated by the amplification processing and the collected images of the original power transmission line are jointly used as data sets for training the first target detection model and the second target detection model.
As a further optimization of the above scheme, the data augmentation process adopts a trained generative confrontation network model.
As a further optimization of the above scheme, the step of generating the training of the confrontation network model comprises:
(31) establishing a generator and a discriminator, setting a loss function of the discriminator and generating an objective function of the countermeasure network;
(32) inputting a random noise signal z into a generator to obtain a generated sample, labeling the acquired power transmission line image and the generated sample, inputting the labeled power transmission line image and the generated sample into a discriminator to discriminate a real sample from the generated sample, and adjusting network parameters of a discrimination network by using a back propagation algorithm to maximize a generated countermeasure network objective function to obtain an optimized discrimination network;
(33) substituting the obtained optimized network parameters of the judgment network into a generated confrontation network objective function, and adjusting the network parameters of the generated network by using a back propagation algorithm to minimize the generated confrontation network objective function to obtain the optimized network parameters of the judgment network so as to obtain an optimized generated network;
(34) and (5) judging whether the iteration times reach a preset maximum iteration time, if so, repeating the steps (32) - (34), otherwise, stopping training, and storing the generated countermeasure model after training.
As a further optimization of the above scheme, the first target detection model and the second target detection model use a deep neural network based on a target detection algorithm.
As a further optimization of the above scheme, the first target detection model and the second target detection model adopt a deep neural network based on a fastern algorithm, and a training process of the fastern network model includes:
(61) performing nesting labeling on a data set used for training a model, labeling a power transmission line connection area in an image, and labeling whether the small hardware has defects or not based on the connection area;
(62) data division: dividing a training set and a verification set according to a proportion;
(63) establishing a first target detection model and a second target detection model, wherein the forward networks adopt basic FasterRCNN networks, and setting corresponding loss functions and weight updating methods;
(64) respectively inputting a first target detection model and a second target detection model by taking a training set carrying nested marking data as input data, calculating a loss function value of output data of a forward network and marking data of a connection region by using a loss function preset by the first target detection model, and calculating a loss function value of the output data of the forward network and marking data of the tiny hardware defects by using a loss function preset by the second target detection model;
(65) when the loss function value is converged or the iteration times reach the preset maximum iteration times, stopping the corresponding network model training, performing step (66), otherwise, performing weight updating on the FasterRCNN network by using a corresponding weight updating method, and repeating steps (64) and (65);
(66) and after the training is stopped, inputting verification set data into a FasterRCNN network to obtain an output result, counting the recall ratio and the precision ratio, finishing the training and storing the trained target detection model if the recall ratio and the precision ratio meet preset conditions, and otherwise, re-training.
As a further optimization of the above solution, the loss function of the second object detection model comprises a second class loss and a second location loss,
the second category loss employs a modified binary cross entropy loss function, as follows:
where N represents the number of prediction frames output by the network, P0iTo predict the probability value of the frame network output as a normal pin, P1iProbability of outputting as defective pin for prediction box networkValue, y0iThe prediction box is denoted as normal pin, y ═ 10i0 denotes an abnormal pin, y1iThe prediction box is denoted as defective pin, y ═ 11i0 denotes a non-defective pin, g0And g1Respectively representing the Loss weight distribution of the normal pins and the defect pins;
the second position loss equation is as follows:
where M represents the number of positive samples of the network output, i.e., the tag value y0iOr y1iNumber of prediction boxes of 1, (x)i,yi) Indicates the position of the center point of the prediction frame, (w)i,hi) Indicates the prediction frame size, (x)i gt,yi gt) Indicates the position of the center point of the label frame (w)i gt,hi gt) Representing the size of the label box; l is1(x) Using the SmoothL1 function, the formula is as follows:
githe weight assignment in the presentation location loss is consistent with the weight assignment in the category loss, as follows:
as a further optimization of the above scheme, the weight updating methods of the first target detection model and the second target detection model both adopt a random gradient descent method to optimize the loss function value.
The invention also provides a power transmission line hardware defect detection system based on the cascade target detection, which comprises the following steps:
the data acquisition module is used for acquiring the inspection shot image and preprocessing the inspection shot image;
the connection area detection module is used for detecting the connection area of the polling shot image by using the trained first target detection model and cutting out the detected rectangular connection area;
the connection region screening module is used for acquiring the area size of the connection region and taking the n connection regions with the area size meeting the preset condition as images to be identified;
the fine hardware defect detection module is used for detecting fine hardware defects of the image to be recognized by using the trained second target detection model to obtain coordinates of the fine hardware defects on the image to be recognized;
and the fine hardware defect original image display module is used for displaying the fine hardware defects on the original image according to the mapping relation between the coordinates of the image to be identified and the original image coordinates.
As a further optimization of the above scheme, the system further comprises a data augmentation processing module, which is used for performing data augmentation processing on the acquired images of the power transmission line before training the first target detection model and the second target detection model, and using the images generated by the augmentation processing and the acquired images of the original power transmission line as a data set for training the first target detection model and the second target detection model.
As a further optimization of the above scheme, the first target detection model and the second target detection model use a deep neural network based on a target detection algorithm.
The power transmission line hardware defect detection method based on the cascade target detection has the following beneficial effects:
1. according to the method for detecting the defects of the power transmission line hardware based on the cascade target detection, for the identification and detection of the small targets of the fine and small power transmission line hardware, due to the fact that the samples which can be collected by the defect detection of the fine and small power transmission line hardware are too few, the method adopts the steps that firstly, the shot images are subjected to generation of the countermeasure network, the number of the samples is increased, and the situation that the results are not ideal during testing due to the fact that overfitting is easy to occur when deep learning is directly used for.
2. According to the method for detecting the defects of the power transmission line hardware fittings based on the cascade target detection, the cascade target detection algorithm deep convolution neural network is adopted for the identification and detection of the small targets of the power transmission line small hardware fittings, the large targets, namely the connection areas, in the power transmission line images shot in the inspection mode are firstly identified by the first target detection model, based on the identification result of the first target detection model, the defect conditions of the small hardware fittings in the connection areas are identified by the second target detection model, the feature extraction work of a large amount of invalid image information is reduced, the network is concentrated in the key areas, the loss of effective image information is greatly reduced, the quality of key pixel information is guaranteed, and the defect detection precision of the small hardware fittings is obviously improved.
Drawings
Fig. 1 is an overall flow diagram of fine hardware defect detection of a to-be-detected inspection image in the cascaded target detection-based transmission line hardware defect detection method of the present invention;
fig. 2 is an overall flow diagram of a method for detecting hardware defects of a power transmission line based on cascade target detection, which is disclosed by the invention, for performing countermeasure generation on sample data to generate a network model, a first target detection model and a second target detection model for performing model parameter training;
FIG. 3 is a block diagram of a process of performing countermeasure generation network model parameter training on sample data in the power transmission line hardware defect detection method based on cascade target detection according to the present invention;
FIG. 4 is an overall flow chart of model parameter training of a first target detection model and a second target detection model performed on sample data in the power transmission line hardware defect detection method based on cascade target detection of the present invention;
FIG. 5 is a block diagram of the internal structure of the system for detecting the hardware defect of the power transmission line based on the cascade target detection according to the present invention;
fig. 6 is a diagram illustrating nesting and labeling of images in the power transmission line hardware defect detection method based on cascade target detection according to the present invention;
fig. 7 is a diagram illustrating a small hardware defect detected by using a second target detection model in the cascaded target detection-based transmission line hardware defect detection method of the present invention;
fig. 8 is a diagram showing the small hardware defects identified in fig. 7 displayed in the original image in the power transmission line hardware defect detection method based on the cascade target detection.
Detailed description of the preferred embodiments
The invention is described in detail below with reference to the figures and specific embodiments.
Referring to fig. 1-8, the invention provides a power transmission line hardware defect detection method based on cascade target detection, considering that a small hardware fault defect target in a power transmission line is too small and accounts for less than 5% of an original image, and a traditional deep learning image identification method has insufficient feature extraction capability, so that a small hardware defect detection effect cannot meet actual service requirements, the scheme provides a two-stage detection strategy for small hardware fault defect identification, can obviously improve small hardware defect detection accuracy, and specifically comprises the following steps:
before the cascaded target detection model is adopted to detect the hardware defects of the transmission line, the training process of the target detection model used in two-stage detection is firstly carried out, in addition, considering that when the deep learning network model is used for image recognition processing, a large number of training samples are required, but the defect detection of the tiny hardware mainly comprises defect categories such as pin missing, pin falling and the like, but no matter which category, the collected samples are too few, when the method is used for deep learning, overfitting is easy to occur, so that the effect is not ideal when testing is carried out, in the embodiment, before the first target detection model and the second target detection model are trained, and carrying out data amplification processing on the acquired images of the power transmission line, and using the images generated by the amplification processing and the acquired images of the original power transmission line as data sets for training the first target detection model and the second target detection model together.
In the scheme, the data augmentation processing adopts a trained generation confrontation network model, and the steps of training the generation confrontation network model are as follows:
(31) establishing a generator and a discriminator, setting a loss function of the discriminator and generating an objective function of the countermeasure network; the penalty function for the discriminator is:
l (p) — y × (n) (p) + (y-1) × ln (1-p), where p is the probability value for discriminating the network output; y represents a label value, and the value of y is 0 or 1;
the generation of the antagonistic network objective function is:
V(D,G)=EX~pd(x)lgD(x)+EX~pz(x)lg (1-D (x)), wherein,
x represents a judgment network input, G represents a generation network, D represents a judgment network, X-Pd (X) represents the distribution Pd (X) of the transmission line image data of which X obeys the acquisition, namely X is from the transmission line image which is really acquired, X-Pz (X) represents the distribution Pz (X) of the noise signal z of which X obeys the randomness, namely X is a generation sample, and E [ · ] represents the mathematical expectation.
(32) Inputting a random noise signal z into a generator to obtain a generated sample, labeling the acquired power transmission line image and the generated sample, inputting the labeled power transmission line image and the generated sample into a discriminator to discriminate a real sample from the generated sample, and adjusting network parameters of a discrimination network by using a back propagation algorithm to maximize a generated countermeasure network objective function to obtain an optimized discrimination network;
(33) substituting the obtained optimized network parameters of the judgment network into a generated confrontation network objective function, and adjusting the network parameters of the generated network by using a back propagation algorithm to minimize the generated confrontation network objective function to obtain the optimized network parameters of the judgment network so as to obtain an optimized generated network;
(34) and (5) judging whether the iteration times reach a preset maximum iteration time, if so, repeating the steps (32) - (34), otherwise, stopping training, and storing the generated countermeasure model after training.
And then calling the trained generation countermeasure network, performing data amplification processing on the acquired transmission line image by using a generator in the trained generation countermeasure network to obtain a generation sample, and then taking the generation sample generated by the amplification processing and the acquired original transmission line image as a data set for training a first target detection model and a second target detection model together.
In this embodiment, the first target detection model and the second target detection model adopt a deep neural network based on a target detection algorithm, the target detection algorithm includes, but is not limited to, fasternn, FPN, YoloV3, and the deep neural network based on the fasternn algorithm is adopted in the present solution, and the training process for the first target detection model and the second target detection model includes:
(61) performing nesting labeling on a data set used for training a model, referring to fig. 6, labeling a power transmission line connection area in an image, and labeling whether the small hardware has defects or not based on the connection area;
(62) data division: dividing a training set and a verification set according to a proportion;
(63) establishing a first target detection model and a second target detection model, wherein the forward networks adopt basic FasterRCNN networks, and setting corresponding loss functions and weight updating methods;
(64) respectively inputting a first target detection model and a second target detection model by taking a training set carrying nested marking data as input data, calculating a loss function value of output data of a forward network and marking data of a connection region by using a loss function preset by the first target detection model, and calculating a loss function value of the output data of the forward network and marking data of the tiny hardware defects by using a loss function preset by the second target detection model;
(65) when the loss function value is converged or the iteration times reach the preset maximum iteration times, stopping the corresponding network model training, performing step (66), otherwise, performing weight updating on the FasterRCNN network by using a corresponding weight updating method, and repeating steps (64) and (65);
(66) and after the training is stopped, inputting verification set data into a FasterRCNN network to obtain an output result, counting the recall ratio and the precision ratio, finishing the training and storing the trained target detection model if the recall ratio and the precision ratio meet preset conditions, and otherwise, re-training.
Wherein the loss function of the first target detection model includes a first class loss and a first location loss,
the first class loss employs a cross-entropy loss function, as follows:
wherein, N represents the number of prediction frames output by the network, and the default is 2000, piOutputting probability values, y, for the network of the connection regioniIndicating a label value, 1 indicates that the region is marked as a connected region, and 0 indicates a background region.
The first position loss equation is as follows:
where M represents the number of positive samples of the network output, i.e., the tag value yiNumber of prediction boxes of 1, (x)i,yi) Indicates the position of the center point of the prediction frame, (w)i,hi) Indicates the prediction frame size, (x)i gt,yi gt) Indicates the position of the center point of the label frame (w)i gt,hi gt) Representing the size of the label box; l is1(x) Using the SmoothL1 function, the formula is as follows:
the loss function of the second object detection model includes a second class loss and a second location loss,
the second category loss employs a modified binary cross entropy loss function, as follows:
wherein, N represents the number of prediction frames output by the network, and the default is 2000, P0iTo predict the probability value of the frame network output as a normal pin, P1iOutput probability value, y, for the prediction frame network as a defective pin0iThe prediction box is denoted as normal pin, y ═ 10i0 denotes an abnormal pin, y1i1 is expressed asThe prediction box is labeled as the defective pin, y1i0 denotes a non-defective pin, g0And g1Respectively representing the Loss weight distribution of normal pins and defective pins, and g is because the model focuses on the defective pins1Greater than g0Default g0=0.7,g1=1.4。
The second position loss equation is as follows:
where M represents the number of positive samples of the network output, i.e., the tag value y0iOr y1iNumber of prediction boxes of 1, (x)i,yi) Indicates the position of the center point of the prediction frame, (w)i,hi) Indicates the prediction frame size, (x)i gt,yi gt) Indicates the position of the center point of the label frame (w)i gt,hi gt) Representing the size of the label box; l is1(x) Using the SmoothL1 function, the formula is as follows:
githe weight assignment in the presentation location loss is consistent with the weight assignment in the category loss, as follows:
and the weight updating methods of the first target detection model and the second target detection model both adopt a random gradient descent method to optimize the loss function value.
When the method for detecting the hardware defect of the power transmission line based on the cascade target detection is adopted to identify and detect the tiny hardware defect of the power transmission line, the following steps are adopted:
(11) firstly, acquiring a patrol shot image and preprocessing the patrol shot image, and performing normalization and denoising operations on the patrol shot image in the embodiment in consideration of inputting the patrol shot image into a neural network model for identification;
(12) then, using the trained first target detection model to detect a connection area of the polling shot image, and cutting out the detected rectangular connection area;
(13) acquiring the area size of the connection region, and taking the n connection regions with the area size meeting a preset condition as an image to be identified;
specifically, in consideration of the research on the distribution characteristics of the defects of the inspection image, the defective connection regions generally appear in the part with a large relative area of the foreground, and other regions can be ignored as the background, so that the area size of the cut rectangular connection regions is calculated through the acquired coordinates, the detected connection regions are sorted from large to small according to the area size, the first n connection regions are selected as the images to be identified, and in this embodiment, n is 3, that is, the 3 connection regions with the largest area are used for identifying the small hardware defects.
(14) Using the trained second target detection model to detect the tiny hardware defects of the image to be recognized, and acquiring coordinates of the tiny hardware defects on the image to be recognized, wherein the coordinates are shown in fig. 7;
(15) and displaying the small hardware defects on the original image according to the mapping relation between the coordinates of the image to be recognized and the coordinates of the original image, and referring to fig. 8.
Referring to fig. 5, the invention further provides a system for detecting the hardware defect of the power transmission line based on the detection of the cascade target, which includes:
the data acquisition module is used for acquiring the inspection shot image and preprocessing the inspection shot image;
the connection area detection module is used for detecting the connection area of the polling shot image by using the trained first target detection model and cutting out the detected rectangular connection area; the training data sets of the first target detection model and the second target detection model are based on data sets generated by the countermeasure network and subjected to data amplification processing, the number of training samples of the network models is increased, the occurrence of an overfitting phenomenon during deep learning of the target detection models is reduced, and the target detection accuracy of the network models is improved.
The connection region screening module is used for acquiring the area size of the connection region and taking the n connection regions with the area size meeting the preset condition as images to be identified;
the fine hardware defect detection module is used for detecting fine hardware defects of the image to be recognized by using the trained second target detection model to obtain coordinates of the fine hardware defects on the image to be recognized;
and the fine hardware defect original image display module is used for displaying the fine hardware defects on the original image according to the mapping relation between the coordinates of the image to be identified and the original image coordinates.
And the data augmentation processing module is used for performing data augmentation processing on the acquired images of the power transmission line before the first target detection model and the second target detection model are trained by adopting the generation countermeasure network, and the images generated by the augmentation processing and the acquired images of the original power transmission line are jointly used as data sets for training the first target detection model and the second target detection model.
The present invention is not limited to the above-described embodiments, and those skilled in the art will be able to make various modifications without creative efforts from the above-described conception, and fall within the scope of the present invention.
Claims (10)
1. The method for detecting the hardware defect of the power transmission line based on the cascade target detection is characterized by comprising the following steps of: the method comprises the following steps:
(11) acquiring a polling shot image and preprocessing the polling shot image;
(12) detecting a connecting area of the polling shot image by using a trained first target detection model, and cutting out the detected rectangular connecting area;
(13) acquiring the area size of the connection region, and taking the n connection regions with the area size meeting a preset condition as an image to be identified;
(14) using the trained second target detection model to detect the tiny hardware defects of the image to be recognized, and acquiring coordinates of the tiny hardware defects on the image to be recognized;
(15) and displaying the tiny hardware defects on the original image according to the mapping relation between the coordinates of the image to be recognized and the coordinates of the original image.
2. The method for detecting the hardware defect of the power transmission line based on the cascade target detection according to claim 1, wherein the method comprises the following steps: before the first target detection model and the second target detection model are trained, data augmentation processing is carried out on the collected images of the power transmission line, and the images generated by the augmentation processing and the collected images of the original power transmission line are jointly used as data sets for training the first target detection model and the second target detection model.
3. The power transmission line hardware defect detection method based on cascade target detection according to claim 2, characterized in that: the data augmentation processing adopts a trained generation confrontation network model, and the generation confrontation network model training comprises the following steps:
(31) establishing a generator and a discriminator, setting a loss function of the discriminator and generating an objective function of the countermeasure network;
(32) inputting a random noise signal z into a generator to obtain a generated sample, labeling the acquired power transmission line image and the generated sample, inputting the labeled power transmission line image and the generated sample into a discriminator to discriminate a real sample from the generated sample, and adjusting network parameters of a discrimination network by using a back propagation algorithm to maximize a generated countermeasure network objective function to obtain an optimized discrimination network;
(33) substituting the obtained optimized network parameters of the judgment network into a generated confrontation network objective function, and adjusting the network parameters of the generated network by using a back propagation algorithm to minimize the generated confrontation network objective function to obtain the optimized network parameters of the judgment network so as to obtain an optimized generated network;
(34) and (5) judging whether the iteration times reach a preset maximum iteration time, if so, repeating the steps (32) - (34), otherwise, stopping training, and storing the generated countermeasure model after training.
4. The power transmission line hardware defect detection method based on cascade target detection according to claim 2, characterized in that: the first target detection model and the second target detection model adopt a deep neural network based on a target detection algorithm.
5. The power transmission line hardware defect detection method based on cascade target detection according to claim 4, characterized in that: the first target detection model and the second target detection model adopt a depth neural network based on a FasterRCNN algorithm, and the training process of the FasterRCNN network model comprises the following steps:
(51) performing nesting labeling on a data set used for training a model, labeling a power transmission line connection area in an image, and labeling whether the small hardware has defects or not based on the connection area;
(52) data division: dividing a training set and a verification set according to a proportion;
(53) establishing a first target detection model and a second target detection model, wherein the forward networks adopt basic FasterRCNN networks, and setting corresponding loss functions and weight updating methods;
(54) respectively inputting a first target detection model and a second target detection model by taking a training set carrying nested marking data as input data, calculating a loss function value of output data of a forward network and marking data of a connection region by using a loss function preset by the first target detection model, and calculating a loss function value of the output data of the forward network and marking data of the tiny hardware defects by using a loss function preset by the second target detection model;
(55) when the loss function value is converged or the iteration times reach the preset maximum iteration times, stopping the corresponding network model training, performing step (56), otherwise, performing weight updating on the FasterRCNN network by using a corresponding weight updating method, and repeating the steps (54) and (55);
(56) and after the training is stopped, inputting verification set data into a FasterRCNN network to obtain an output result, counting the recall ratio and the precision ratio, finishing the training and storing the trained target detection model if the recall ratio and the precision ratio meet preset conditions, and otherwise, re-training.
6. The method for detecting the hardware defect of the power transmission line based on the cascade target detection according to claim 5, wherein the method comprises the following steps: the loss function of the second object detection model comprises a second class loss and a second location loss,
the second category loss employs a modified binary cross entropy loss function, as follows:
where N represents the number of prediction frames output by the network, P0iTo predict the probability value of the frame network output as a normal pin, P1iOutput probability value, y, for the prediction frame network as a defective pin0iThe prediction box is denoted as normal pin, y ═ 10i0 denotes an abnormal pin, y1iThe prediction box is denoted as defective pin, y ═ 11i0 denotes a non-defective pin, g0And g1Respectively representing the Loss weight distribution of the normal pins and the defect pins;
the second position loss equation is as follows:
where M represents the number of positive samples of the network output, i.e., the tag value y0iOr y1iNumber of prediction boxes of 1, (x)i,yi) Indicates the position of the center point of the prediction frame, (w)i,hi) Indicates the prediction frame size, (x)i gt,yi gt) Indicates the position of the center point of the label frame (w)i gt,hi gt) Representing the size of the label box; l is1(x) Using the SmoothL1 function, the formula is as follows:
githe weight assignment in the presentation location loss is consistent with the weight assignment in the category loss, as follows:
7. the method for detecting the hardware defect of the power transmission line based on the cascade target detection according to claim 5, wherein the method comprises the following steps: and the weight updating methods of the first target detection model and the second target detection model both adopt a random gradient descent method to optimize the loss function value.
8. Transmission line gold utensil defect detecting system based on cascade target detection, its characterized in that: the method comprises the following steps:
the data acquisition module is used for acquiring the inspection shot image and preprocessing the inspection shot image;
the connection area detection module is used for detecting the connection area of the polling shot image by using the trained first target detection model and cutting out the detected rectangular connection area;
the connection region screening module is used for acquiring the area size of the connection region and taking the n connection regions with the area size meeting the preset condition as images to be identified;
the fine hardware defect detection module is used for detecting fine hardware defects of the image to be recognized by using the trained second target detection model to obtain coordinates of the fine hardware defects on the image to be recognized;
and the fine hardware defect original image display module is used for displaying the fine hardware defects on the original image according to the mapping relation between the coordinates of the image to be identified and the original image coordinates.
9. The power transmission line hardware defect detection system based on cascade target detection according to claim 8, wherein: the data amplification processing module is used for carrying out data amplification processing on the collected images of the power transmission line before the first target detection model and the second target detection model are trained, and the images generated by the amplification processing and the collected images of the original power transmission line are jointly used as data sets for training the first target detection model and the second target detection model.
10. The power transmission line hardware defect detection system based on cascade target detection according to claim 7, characterized in that: the first target detection model and the second target detection model adopt a deep neural network based on a target detection algorithm.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010393940.7A CN111650204B (en) | 2020-05-11 | 2020-05-11 | Power transmission line hardware defect detection method and system based on cascade target detection |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN202010393940.7A CN111650204B (en) | 2020-05-11 | 2020-05-11 | Power transmission line hardware defect detection method and system based on cascade target detection |
Publications (2)
Publication Number | Publication Date |
---|---|
CN111650204A true CN111650204A (en) | 2020-09-11 |
CN111650204B CN111650204B (en) | 2023-05-12 |
Family
ID=72346054
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN202010393940.7A Active CN111650204B (en) | 2020-05-11 | 2020-05-11 | Power transmission line hardware defect detection method and system based on cascade target detection |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN111650204B (en) |
Cited By (22)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112419680A (en) * | 2020-11-19 | 2021-02-26 | 中国南方电网有限责任公司超高压输电公司检修试验中心 | Power transmission line potential safety hazard classification and identification method and system |
CN112508905A (en) * | 2020-12-01 | 2021-03-16 | 广东电网有限责任公司 | Hardware rust image detection method and computer readable storage medium |
CN112669264A (en) * | 2020-12-17 | 2021-04-16 | 国网山西省电力公司运城供电公司 | Artificial intelligence defect identification method and system for unmanned aerial vehicle routing inspection of distribution network line |
CN112683916A (en) * | 2020-12-17 | 2021-04-20 | 华能新能源股份有限公司云南分公司 | Method and device for identifying missing or mounting error of small hardware fittings of current collecting line tower |
CN112712060A (en) * | 2021-01-15 | 2021-04-27 | 国网浙江省电力有限公司电力科学研究院 | Automatic high-voltage post porcelain insulator acoustic signal identification method |
CN112837281A (en) * | 2021-01-27 | 2021-05-25 | 湘潭大学 | Pin defect identification method, device and equipment based on cascade convolutional neural network |
CN112926401A (en) * | 2021-01-29 | 2021-06-08 | 广州中科智巡科技有限公司 | Transmission line hardware corrosion detection method and system |
CN112975953A (en) * | 2021-01-28 | 2021-06-18 | 珠海迪沃航空工程有限公司 | Manipulator motion trajectory planning method and bolt grabbing detection system |
CN113012107A (en) * | 2021-02-22 | 2021-06-22 | 江苏方天电力技术有限公司 | Power grid defect detection method and system |
CN113052125A (en) * | 2021-04-09 | 2021-06-29 | 内蒙古科电数据服务有限公司 | Construction site violation image recognition and alarm method |
CN113052809A (en) * | 2021-03-18 | 2021-06-29 | 中科海拓(无锡)科技有限公司 | EfficientNet-based nut surface defect classification method |
CN113327255A (en) * | 2021-05-28 | 2021-08-31 | 宁波新胜中压电器有限公司 | Power transmission line inspection image processing method based on YOLOv3 detection, positioning and cutting and fine-tune |
CN113469951A (en) * | 2021-06-08 | 2021-10-01 | 燕山大学 | Hub defect detection method based on cascade region convolutional neural network |
CN113537341A (en) * | 2021-07-14 | 2021-10-22 | 安徽炬视科技有限公司 | Online monitoring device and identification method for line hidden danger based on big data and self-learning |
CN113903009A (en) * | 2021-12-10 | 2022-01-07 | 华东交通大学 | Railway foreign matter detection method and system based on improved YOLOv3 network |
CN114155246A (en) * | 2022-02-10 | 2022-03-08 | 国网江西省电力有限公司电力科学研究院 | Deformable convolution-based power transmission tower pin defect detection method |
CN114202526A (en) * | 2021-12-10 | 2022-03-18 | 北京百度网讯科技有限公司 | Quality detection method, system, apparatus, electronic device, and medium |
CN114397306A (en) * | 2022-03-25 | 2022-04-26 | 南方电网数字电网研究院有限公司 | Power grid grading ring hypercomplex category defect multi-stage model joint detection method |
CN115619763A (en) * | 2022-10-31 | 2023-01-17 | 国网湖北省电力有限公司超高压公司 | Power transmission line small part image target detection method based on multistage reasoning detection |
CN116228778A (en) * | 2023-05-10 | 2023-06-06 | 国网山东省电力公司菏泽供电公司 | Insulator rupture detection method and system based on multi-mode information fusion |
CN116843686A (en) * | 2023-08-31 | 2023-10-03 | 成都考拉悠然科技有限公司 | Method and device for detecting defects of wire clamps and nuts of contact net locator |
CN116883391A (en) * | 2023-09-05 | 2023-10-13 | 中国科学技术大学 | Two-stage distribution line defect detection method based on multi-scale sliding window |
Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2016078081A1 (en) * | 2014-11-21 | 2016-05-26 | 河南送变电工程公司 | Multi-rotor-wing inspection aircraft and power transmission line inspection system |
CN109344753A (en) * | 2018-09-21 | 2019-02-15 | 福州大学 | A kind of tiny fitting recognition methods of Aerial Images transmission line of electricity based on deep learning |
-
2020
- 2020-05-11 CN CN202010393940.7A patent/CN111650204B/en active Active
Patent Citations (2)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
WO2016078081A1 (en) * | 2014-11-21 | 2016-05-26 | 河南送变电工程公司 | Multi-rotor-wing inspection aircraft and power transmission line inspection system |
CN109344753A (en) * | 2018-09-21 | 2019-02-15 | 福州大学 | A kind of tiny fitting recognition methods of Aerial Images transmission line of electricity based on deep learning |
Non-Patent Citations (1)
Title |
---|
邹捷;许瑞庆;习雨同;: "深度学习在输电线路工程验收智能缺陷识别中的应用" * |
Cited By (30)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN112419680B (en) * | 2020-11-19 | 2022-09-27 | 中国南方电网有限责任公司超高压输电公司检修试验中心 | Power transmission line potential safety hazard classification and identification method and system |
CN112419680A (en) * | 2020-11-19 | 2021-02-26 | 中国南方电网有限责任公司超高压输电公司检修试验中心 | Power transmission line potential safety hazard classification and identification method and system |
CN112508905A (en) * | 2020-12-01 | 2021-03-16 | 广东电网有限责任公司 | Hardware rust image detection method and computer readable storage medium |
CN112669264A (en) * | 2020-12-17 | 2021-04-16 | 国网山西省电力公司运城供电公司 | Artificial intelligence defect identification method and system for unmanned aerial vehicle routing inspection of distribution network line |
CN112683916A (en) * | 2020-12-17 | 2021-04-20 | 华能新能源股份有限公司云南分公司 | Method and device for identifying missing or mounting error of small hardware fittings of current collecting line tower |
CN112712060A (en) * | 2021-01-15 | 2021-04-27 | 国网浙江省电力有限公司电力科学研究院 | Automatic high-voltage post porcelain insulator acoustic signal identification method |
CN112837281A (en) * | 2021-01-27 | 2021-05-25 | 湘潭大学 | Pin defect identification method, device and equipment based on cascade convolutional neural network |
CN112837281B (en) * | 2021-01-27 | 2022-10-28 | 湘潭大学 | Pin defect identification method, device and equipment based on cascade convolution neural network |
CN112975953A (en) * | 2021-01-28 | 2021-06-18 | 珠海迪沃航空工程有限公司 | Manipulator motion trajectory planning method and bolt grabbing detection system |
CN112926401A (en) * | 2021-01-29 | 2021-06-08 | 广州中科智巡科技有限公司 | Transmission line hardware corrosion detection method and system |
CN113012107A (en) * | 2021-02-22 | 2021-06-22 | 江苏方天电力技术有限公司 | Power grid defect detection method and system |
CN113012107B (en) * | 2021-02-22 | 2022-07-08 | 江苏方天电力技术有限公司 | Power grid defect detection method and system |
CN113052809A (en) * | 2021-03-18 | 2021-06-29 | 中科海拓(无锡)科技有限公司 | EfficientNet-based nut surface defect classification method |
CN113052809B (en) * | 2021-03-18 | 2021-12-10 | 中科海拓(无锡)科技有限公司 | EfficientNet-based nut surface defect classification method |
CN113052125A (en) * | 2021-04-09 | 2021-06-29 | 内蒙古科电数据服务有限公司 | Construction site violation image recognition and alarm method |
CN113327255A (en) * | 2021-05-28 | 2021-08-31 | 宁波新胜中压电器有限公司 | Power transmission line inspection image processing method based on YOLOv3 detection, positioning and cutting and fine-tune |
CN113469951A (en) * | 2021-06-08 | 2021-10-01 | 燕山大学 | Hub defect detection method based on cascade region convolutional neural network |
CN113537341A (en) * | 2021-07-14 | 2021-10-22 | 安徽炬视科技有限公司 | Online monitoring device and identification method for line hidden danger based on big data and self-learning |
CN113903009B (en) * | 2021-12-10 | 2022-07-05 | 华东交通大学 | Railway foreign matter detection method and system based on improved YOLOv3 network |
CN114202526A (en) * | 2021-12-10 | 2022-03-18 | 北京百度网讯科技有限公司 | Quality detection method, system, apparatus, electronic device, and medium |
CN113903009A (en) * | 2021-12-10 | 2022-01-07 | 华东交通大学 | Railway foreign matter detection method and system based on improved YOLOv3 network |
CN114155246A (en) * | 2022-02-10 | 2022-03-08 | 国网江西省电力有限公司电力科学研究院 | Deformable convolution-based power transmission tower pin defect detection method |
CN114397306A (en) * | 2022-03-25 | 2022-04-26 | 南方电网数字电网研究院有限公司 | Power grid grading ring hypercomplex category defect multi-stage model joint detection method |
CN115619763A (en) * | 2022-10-31 | 2023-01-17 | 国网湖北省电力有限公司超高压公司 | Power transmission line small part image target detection method based on multistage reasoning detection |
CN116228778A (en) * | 2023-05-10 | 2023-06-06 | 国网山东省电力公司菏泽供电公司 | Insulator rupture detection method and system based on multi-mode information fusion |
CN116228778B (en) * | 2023-05-10 | 2023-09-08 | 国网山东省电力公司菏泽供电公司 | Insulator rupture detection method and system based on multi-mode information fusion |
CN116843686A (en) * | 2023-08-31 | 2023-10-03 | 成都考拉悠然科技有限公司 | Method and device for detecting defects of wire clamps and nuts of contact net locator |
CN116843686B (en) * | 2023-08-31 | 2023-11-10 | 成都考拉悠然科技有限公司 | Method and device for detecting defects of wire clamps and nuts of contact net locator |
CN116883391A (en) * | 2023-09-05 | 2023-10-13 | 中国科学技术大学 | Two-stage distribution line defect detection method based on multi-scale sliding window |
CN116883391B (en) * | 2023-09-05 | 2023-12-19 | 中国科学技术大学 | Two-stage distribution line defect detection method based on multi-scale sliding window |
Also Published As
Publication number | Publication date |
---|---|
CN111650204B (en) | 2023-05-12 |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN111650204A (en) | Transmission line hardware defect detection method and system based on cascade target detection | |
CN111339882B (en) | Power transmission line hidden danger detection method based on example segmentation | |
CN110148130B (en) | Method and device for detecting part defects | |
CN110084165B (en) | Intelligent identification and early warning method for abnormal events in open scene of power field based on edge calculation | |
CN111431986B (en) | Industrial intelligent quality inspection system based on 5G and AI cloud edge cooperation | |
CN113095438B (en) | Wafer defect classification method, device and system thereof, electronic equipment and storage medium | |
CN107966454A (en) | A kind of end plug defect detecting device and detection method based on FPGA | |
CN111797890A (en) | Method and system for detecting defects of power transmission line equipment | |
CN111695493B (en) | Method and system for detecting hidden danger of power transmission line | |
CN111862065B (en) | Power transmission line diagnosis method and system based on multitask deep convolutional neural network | |
WO2022262054A1 (en) | Image processing method, apparatus and device, and storage medium | |
CN111179262A (en) | Electric power inspection image hardware fitting detection method combined with shape attribute | |
CN115439458A (en) | Industrial image defect target detection algorithm based on depth map attention | |
Wah et al. | Analysis on feature extraction and classification of rice kernels for Myanmar rice using image processing techniques | |
KR100868884B1 (en) | Flat glass defect information system and classification method | |
CN115170529A (en) | Multi-scale tiny flaw detection method based on attention mechanism | |
CN110334775B (en) | Unmanned aerial vehicle line fault identification method and device based on width learning | |
CN111507249A (en) | Transformer substation nest identification method based on target detection | |
CN113421223B (en) | Industrial product surface defect detection method based on deep learning and Gaussian mixture | |
Zhang et al. | Zero-DD: Zero-sample defect detection for industrial products | |
CN112381794B (en) | Printing defect detection method based on deep convolution generation network | |
CN116895030B (en) | Insulator detection method based on target detection algorithm and attention mechanism | |
CN113506281A (en) | Bridge crack detection method based on deep learning framework | |
CN117635565A (en) | Semiconductor surface defect detection system based on image recognition | |
CN117152161A (en) | Shaving board quality detection method and system based on image recognition |
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 |