CN117037021A - Distribution network operation supervision method and system based on intelligent video analysis - Google Patents
Distribution network operation supervision method and system based on intelligent video analysis Download PDFInfo
- Publication number
- CN117037021A CN117037021A CN202310760086.7A CN202310760086A CN117037021A CN 117037021 A CN117037021 A CN 117037021A CN 202310760086 A CN202310760086 A CN 202310760086A CN 117037021 A CN117037021 A CN 117037021A
- Authority
- CN
- China
- Prior art keywords
- distribution network
- video data
- network operation
- image
- video
- 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
- 238000009826 distribution Methods 0.000 title claims abstract description 77
- 238000000034 method Methods 0.000 title claims abstract description 63
- 238000004458 analytical method Methods 0.000 title claims abstract description 45
- 238000001514 detection method Methods 0.000 claims abstract description 46
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 21
- 238000012544 monitoring process Methods 0.000 claims abstract description 20
- 238000012545 processing Methods 0.000 claims abstract description 17
- 238000013135 deep learning Methods 0.000 claims abstract description 13
- 238000005516 engineering process Methods 0.000 claims abstract description 10
- 238000007781 pre-processing Methods 0.000 claims abstract description 7
- 230000008447 perception Effects 0.000 claims abstract description 4
- 230000001502 supplementing effect Effects 0.000 claims abstract description 4
- 239000013598 vector Substances 0.000 claims description 25
- 230000005540 biological transmission Effects 0.000 claims description 24
- 230000006870 function Effects 0.000 claims description 15
- 239000011159 matrix material Substances 0.000 claims description 13
- 230000008569 process Effects 0.000 claims description 13
- 238000004590 computer program Methods 0.000 claims description 11
- 238000004364 calculation method Methods 0.000 claims description 6
- 238000003860 storage Methods 0.000 claims description 6
- 230000009467 reduction Effects 0.000 claims description 4
- 238000013500 data storage Methods 0.000 claims description 3
- 238000005070 sampling Methods 0.000 claims description 3
- 230000009469 supplementation Effects 0.000 claims description 3
- 230000001629 suppression Effects 0.000 claims description 3
- 238000004519 manufacturing process Methods 0.000 claims description 2
- 239000004065 semiconductor Substances 0.000 claims 1
- 238000010586 diagram Methods 0.000 description 5
- 101150010567 caiT gene Proteins 0.000 description 4
- 238000004891 communication Methods 0.000 description 4
- 238000013527 convolutional neural network Methods 0.000 description 4
- 230000006399 behavior Effects 0.000 description 3
- 239000000047 product Substances 0.000 description 3
- 238000012549 training Methods 0.000 description 3
- 238000013461 design Methods 0.000 description 2
- 238000005315 distribution function Methods 0.000 description 2
- 238000007689 inspection Methods 0.000 description 2
- 230000007246 mechanism Effects 0.000 description 2
- 238000005457 optimization Methods 0.000 description 2
- 238000012805 post-processing Methods 0.000 description 2
- 238000013459 approach Methods 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 210000000988 bone and bone Anatomy 0.000 description 1
- 238000010276 construction Methods 0.000 description 1
- 238000002790 cross-validation Methods 0.000 description 1
- 238000007418 data mining Methods 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000002474 experimental method Methods 0.000 description 1
- 239000004973 liquid crystal related substance Substances 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000007726 management method Methods 0.000 description 1
- 238000011160 research Methods 0.000 description 1
- 239000013589 supplement Substances 0.000 description 1
- 230000009466 transformation Effects 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V20/00—Scenes; Scene-specific elements
- G06V20/40—Scenes; Scene-specific elements in video content
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/045—Combinations of networks
- G06N3/0455—Auto-encoder networks; Encoder-decoder networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/0464—Convolutional networks [CNN, ConvNet]
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
- G06N3/0499—Feedforward networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/40—Extraction of image or video features
- G06V10/42—Global feature extraction by analysis of the whole pattern, e.g. using frequency domain transformations or autocorrelation
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V10/00—Arrangements for image or video recognition or understanding
- G06V10/70—Arrangements for image or video recognition or understanding using pattern recognition or machine learning
- G06V10/82—Arrangements for image or video recognition or understanding using pattern recognition or machine learning using neural networks
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06V—IMAGE OR VIDEO RECOGNITION OR UNDERSTANDING
- G06V2201/00—Indexing scheme relating to image or video recognition or understanding
- G06V2201/07—Target detection
Abstract
The application discloses a distribution network operation supervision method and system based on intelligent video analysis, comprising the steps of collecting video data by using a perception layer; image recognition is carried out on the collected video data, and the recognized video data is stored, transmitted and distributed; supplementing the video data to form multi-format type data of video resources and monitoring data; based on a target detection technology, performing secondary processing on video data through a deep learning algorithm and storing the video data; and feeding back the video data processing result. According to the application, by utilizing the characteristic that a deep learning algorithm does not need manual intervention, video monitoring is carried out on a distribution network operation site, pictures are intercepted by taking frames as a unit, an intelligent analysis system is utilized for analysis, and the pictures can be monitored in a target manner without preprocessing of manually designing context scene characteristics and data, so that the safety monitoring problem of wearing personnel safety helmets on the distribution network operation site is realized.
Description
Technical Field
The application relates to the field of power distribution network operation safety management and control, in particular to a power distribution network operation supervision method and system based on video intelligent analysis.
Background
At present, in the aspects of security behavior supervision, illegal responsibility tracking and the like of distribution network operation, the video image information is usually called, consulted and analyzed by manually calling a video monitoring system after video monitoring recording; target detection in video monitoring is one of the key problems of computer vision research, is an important basis for understanding high-level semantic information of images, and along with the vigorous development of deep learning and artificial intelligence technology, the application of the deep learning technology to the field of image processing has become a trend of modern scientific computer technology.
However, in actual working, the number of monitoring videos is too large, the workload of the mode of acquiring information only by manpower is too large, the working time and the working efficiency of people are limited, misjudgment and omission of information are easy to generate in a fatigue state under a long-time working state, and some methods, such as a learning NMS (network system) method or a relational network, can be utilized to treat the relation among objects by a self-attention-like method, so that unique predictions are obtained, and no post-processing step is needed, but the performance of the methods is often lower, manual intervention, such as manual design of scene features to help model learning, is needed to improve the performance, but the manual design step not only increases the labor cost and complexity, but also has a certain error, so that the target detection step is as simple as possible, and does not need to use excessive manual knowledge, so that a deep learning-based method is selected, and compared with machine learning algorithm, the deep learning algorithm does not need to manually perform feature selection.
Disclosure of Invention
This section is intended to outline some aspects of embodiments of the application and to briefly introduce some preferred embodiments. Some simplifications or omissions may be made in this section as well as in the description of the application and in the title of the application, which may not be used to limit the scope of the application.
The application is provided in view of the problems of asynchronous cooperative treatment, active early warning, untimely overall supervision and insufficient decision support in the conventional distribution network operation supervision process.
Therefore, the problem to be solved by the application is how to provide a method for solving the problem of monitoring the safety helmet wearing safety of personnel on the operation site of the distribution network.
In order to solve the technical problems, the application provides the following technical scheme:
in a first aspect, an embodiment of the present application provides a method for supervising a distribution network operation based on intelligent video analysis, where the method includes collecting video data by using a sensing layer; image recognition is carried out on the collected video data, and the recognized video data is stored, transmitted and distributed; supplementing the video data to form multi-format type data of video resources and monitoring data; based on a target detection technology, performing secondary processing on video data through a deep learning algorithm and storing the video data; and feeding back the video data processing result.
As a preferable scheme of the distribution network operation supervision method based on video intelligent analysis, the application comprises the following steps: image recognition of the acquired video data comprises the following steps: preprocessing an image; extracting feature vectors of the image; performing dimension reduction on the feature vector of the image; inputting the feature vector into a DETR model to predict a target detection task; the DETR model eliminates redundant detection frames through non-maximum suppression to obtain a final detection result.
As a preferable scheme of the distribution network operation supervision method based on video intelligent analysis, the application comprises the following steps: the specific steps of preprocessing the image are as follows: scaling and calculating the picture; determining a probability distribution of the image; and calculating the optimal transmission distance.
As a preferable scheme of the distribution network operation supervision method based on video intelligent analysis, the application comprises the following steps: the judging process of scaling and calculating the picture is as follows: setting a fixed n multiplied by n size, and respectively scaling the length and the width of the image to n; let the original width be w 1 Height is h 1 After image scaling of the current image width w 2 Height is h 2 The method comprises the steps of carrying out a first treatment on the surface of the If w 1 >h 1 W is then 2 ＝n，h 2 ＝[h 1 ]The method comprises the steps of carrying out a first treatment on the surface of the If w 1 <h 1 W is then 2 ＝[w1]，h 2 =n; if w 1 ＝h 1 W is then 2 ＝h 2 =n; wherein [ among others ]]Representing an upward rounding.
As a preferable scheme of the distribution network operation supervision method based on video intelligent analysis, the application comprises the following steps: the probability distribution of the image is determined as follows: taking all pixels of the preprocessed image I as samples obtained by independent sampling from unknown distribution, wherein the specific formula of a probability density function is as follows:
where n is the number of pixels in the image, x i ∈R d Feature vector, K, representing the ith pixel h (. Cndot.) a Gaussian kernel function denoted as h; wherein K is h The formula of (-) is:
where h is the bandwidth of the gaussian kernel function.
As a preferable scheme of the distribution network operation supervision method based on video intelligent analysis, the application comprises the following steps: the optimal transmission distance comprises a calculation transmission plan and an optimal transmission distance; the calculating a transmission plan includes the transmission requirement meeting the following constraints:
wherein n and m are the dimensions of the distributions P and Q, respectively, and P (i) and Q (j) are the values of the ith and jth elements in the distributions P and Q, respectively; under this constraint, minimizing the total cost is expressed as the following linear programming problem:
minimize∑i,j T(i,j)*C(i,j)
wherein C (i, j) is a slave distributionThe distance from the ith element in P to the jth element in the distribution Q, the minimum being the objective function and the subject being the constraint function; the calculation of the optimal transmission distance comprises setting two probability distributions as mu and v, setting their support sets as X and Y respectively, setting the number of elements as n and m respectively, setting C as an n×m cost matrix, wherein the (i, j) th element C ij Representing the cost required to transport the ith sample point in μ to the jth sample point in ν; the minimum cost for converting μ to ν is:
where u (μ, v) represents the set of all matrices T satisfying the constraint, < C, T > represents the sum of the element-wise products of the cost matrix C and the matrix T.
As a preferable scheme of the distribution network operation supervision method based on video intelligent analysis, the application comprises the following steps: the constraint judgment comprises T epsilon R n×m The method comprises the steps of carrying out a first treatment on the surface of the The sum of each row of T is equal to the probability mass corresponding to μ, i.eThe sum of each column of T is equal to the probability mass corresponding to v, i.e. +.>The element in T being non-negative, i.e. T ij ≥0。
In the second aspect, in order to further solve the problems of asynchronous cooperative treatment, active early warning, untimely overall supervision and insufficient decision support in the distribution network operation supervision process, the embodiment of the application provides a distribution network operation supervision system based on video intelligent analysis, which comprises an acquisition module, a control module and a control module, wherein the acquisition module is used for acquiring and uploading video data in real time; the data processing module is used for carrying out image recognition, processing, supplementation and uploading on the collected video data; the data storage module is used for storing and inquiring the processed uploaded images and analysis results; and the data feedback module is used for feeding the identification result back to the controller, and the controller adjusts the supervision system by adopting manual or automatic adjustment according to the result.
In a third aspect, embodiments of the present application provide a computer apparatus comprising a memory and a processor, the memory storing a computer program, wherein: and the processor realizes any step of a distribution network operation supervision method based on intelligent video analysis when executing the computer program.
In a fourth aspect, embodiments of the present application provide a computer-readable storage medium having a computer program stored thereon, wherein: and the computer program, when executed by the processor, realizes any step of the distribution network operation supervision method based on the intelligent video analysis.
The application has the beneficial effects that the detection is carried out by using the DETR, the manual priori process is deleted by directly predicting the detection set by a bipartite graph matching method, the system performance is improved, the whole target detection process is simplified, and the complexity and the use threshold of the system are reduced; by utilizing the characteristic that a deep learning algorithm does not need manual intervention, the video monitoring is carried out on the distribution network operation site, the images are intercepted by taking the video as a unit, the intelligent analysis system is utilized for analysis, and the images can be subjected to target monitoring without preprocessing of manually designing the context scene characteristics and the data, so that the safety monitoring problem of personnel safety helmet wearing on the distribution network operation site is realized.
Drawings
In order to more clearly illustrate the technical solutions of the embodiments of the present application, the drawings that are needed in the description of the embodiments will be briefly described below, it being obvious that the drawings in the following description are only some embodiments of the present application, and that other drawings may be obtained according to these drawings without inventive effort for a person skilled in the art. Wherein:
fig. 1 is a diagram showing the overall configuration of a video analysis system for distribution network operation in embodiment 1.
FIG. 2 is a diagram showing the overall structure of self-intent in example 1.
FIG. 3 is a diagram showing the overall construction of the transducer in example 1.
Fig. 4 is a diagram of the overall network framework of VTOR in embodiment 1.
Fig. 5 is an overall network frame diagram of DETR in example 1.
Fig. 6 is a graph of model training loss for example 2.
Detailed Description
In order that the above-recited objects, features and advantages of the present application will become more readily apparent, a more particular description of the application will be rendered by reference to specific embodiments thereof which are illustrated in the appended drawings.
In the following description, numerous specific details are set forth in order to provide a thorough understanding of the present application, but the present application may be practiced in other ways other than those described herein, and persons skilled in the art will readily appreciate that the present application is not limited to the specific embodiments disclosed below.
Further, reference herein to "one embodiment" or "an embodiment" means that a particular feature, structure, or characteristic can be included in at least one implementation of the application. The appearances of the phrase "in one embodiment" in various places in the specification are not necessarily all referring to the same embodiment, nor are separate or alternative embodiments mutually exclusive of other embodiments.
Example 1
Referring to fig. 1 to 5, in a first embodiment of the present application, the embodiment provides a distribution network operation supervision method based on intelligent video analysis, which includes the following steps:
s1: video data is acquired using a perception layer.
Preferably, as shown in fig. 1, the sensing layer is first used to acquire video data, and the data acquisition mode is to perform video monitoring on a scene in the operation range of the distribution network, so as to acquire the video data required by the whole system for processing and analysis of the subsequent system.
S2: transmitting the video data acquired by the perception layer into an infrastructure layer, carrying out image recognition on the acquired video data, and storing, transmitting and distributing the recognized video data.
Image recognition of the acquired video data comprises the following steps:
s2.1: the image is preprocessed.
S2.1.1: and scaling and calculating the picture.
When the images are preprocessed, they are usually scaled to the same size so that the similarity between them can be easily compared.
Specifically, scaling the picture and calculating includes setting a fixed n×n size, scaling the length and width of the image to n, respectively; let the original width be w 1 Height is h 1 After image scaling of the current image width w 2 Height is h 2 The method comprises the steps of carrying out a first treatment on the surface of the If w 1 >h 1 W is then 2 ＝n，h 2 ＝[h 1 ]The method comprises the steps of carrying out a first treatment on the surface of the If w 1 <h 1 W is then 2 ＝[w1]，h 2 =n; if w 1 ＝h 1 W is then 2 ＝h 2 =n; wherein [ among others ]]Representing an upward rounding.
By this formula we can scale the two images to the same size for comparison, after scaling we need to normalize the gray or RGB value of each pixel to within the range of 0,1 by dividing the value of all pixels by 255.
S2.1.2: a probability distribution of the image is determined.
The probability distribution of the image is constructed by using a Gaussian kernel density estimation method, and for an image I after pretreatment, all pixels of the image I can be regarded as samples obtained by independently sampling from unknown distribution, and then the probability distribution is estimated by using the Gaussian kernel density estimation method.
Specifically, the specific formula of the probability density function is:
where n is the number of pixels in the image, x i ∈R d Feature vector, K, representing the ith pixel h (. Cndot.) a Gaussian kernel function denoted as h;
wherein K is h (·)The formula of (2) is:
where h is the bandwidth of the gaussian kernel, controlling the degree of smoothness of the estimate, it is often necessary to determine the optimal bandwidth size by cross-validation or the like. For an image, all pixels of the image can be substituted into the formula to obtain a continuous probability distribution function, the continuous probability distribution function can be discretized, data can be conveniently processed and stored, the image is divided into a plurality of intervals by using a method similar to a histogram, and the number of samples falling in each interval is counted, so that a discretized probability distribution can be obtained and can be regarded as a vector.
S2.1.3: and calculating the optimal transmission distance.
The optimal transmission distance includes calculating a transmission plan and calculating an optimal transmission distance.
In particular, the transmission plan T may be expressed as a matrix, where T (i, j) represents the mass transmission amount from the point i in the distribution P to the point j in the distribution Q. The transmission plan needs to meet the following constraints:
wherein n and m are the dimensions of the distributions P and Q, respectively, and P (i) and Q (j) are the values of the ith and jth elements in the distributions P and Q, respectively;
under this constraint, minimizing the total cost is expressed as the following linear programming problem:
minimize∑i,j T(i,j)*C(i,j)
wherein C (i, j) is the distance from the i-th element in the distribution P to the j-th element in the distribution Q, and minimize is the objective function, i.e. the optimization objective mathematically, i.e. the sum of Σt (). Times.c () reaches a minimum; the subject is a constraint that the minimize optimization process needs to satisfy a constraint of a condition (mathematical condition after subject to).
Calculating the optimal transmission distance includes setting two probability distributions as mu and v, setting their support sets as X and Y, their element numbers as n and m, setting C as a cost matrix of n×m, wherein the (i, j) th element C ij Representing the cost required to transport the ith sample point in μ to the jth sample point in ν, the minimum cost required to convert μ to ν is:
where u (μ, v) represents the set of all matrices T satisfying the constraint, < C, T > represents the sum of the element-wise products of the cost matrix C and the matrix T.
Where u (μ, v) represents the set of all matrices T satisfying the following constraints:
T∈R n×m the method comprises the steps of carrying out a first treatment on the surface of the The sum of each row of T is equal to the probability mass corresponding to μ, i.eThe sum of each column of T is equal to the probability mass corresponding to v, i.e. +.>The element in T being non-negative, i.e. T ij ≥0。
Where the sum of the element-wise products of the cost matrix C and the matrix T is represented, this optimal transmission problem can be solved using linear programming or the Sinkhorn algorithm.
For two images I 1 And I 2 Their optimal transmission distance can be defined as the Wasserstein distance of their corresponding probability distribution over the pixel space:
where C is a cost matrix between one pixel point, euclidean distance between pixels or difference in color space, etc. can be used as the cost. Mu (mu) 1 Sum mu 2 Respectively represent image I 1 And I 2 Probability distribution over pixel space.
S2.2: feature vectors of the image are extracted.
Specifically, the image to be detected needs to be divided into a plurality of image blocks, and each image block is regarded as a sequence and is transmitted to the CaiT model for processing; for each tile, the CaiT model outputs a feature vector that can be pooled or combined to obtain a feature representation of the entire image.
S2.3: and reducing the dimension of the feature vector of the image.
The output eigenvectors of the CaiT model are fed into a dimension reduction layer, typically using a linear transform or convolution layer, which serves two purposes: one aspect is to reduce the length of the feature vector to make it more suitable for the input of the DETR encoder; on the other hand, by compressing the feature vector, the calculation efficiency is improved.
S2.4: and inputting the feature vector into the DETR model to predict the target detection task.
Further, the feature vector subjected to dimension reduction is input into a DETR model, prediction of a target detection task is performed, a target detection head of DETR is applied to the output feature vector of CaiT, the head contains a set of multi-layer perceptrons, and an attention mechanism is used to predict an object in an image and position information thereof.
Preferably, the DETR model combines a bipartite graph matching approach for prediction set, transformer encoder-decoder architecture, and object detection method for parallel decoding.
A multi-head attention layer is calculated to acquire the correlation of the current characteristics, and then a multi-head attention layer is calculated for each of category attention and regression attention and is subjected to post-processing through a feedforward network.
Specifically, the DETR converts the target detection problem into a set prediction problem, and each time a picture is input, a set of loss functions are used for end-to-end training; the training process is as follows: extracting picture features by using a convolutional neural network; preliminary learning global features with transformer encoder; further learning finer local features with transformer decoder to better distinguish objects from occlusion edges while generating a large number of prediction frames; and carrying out bipartite graph matching on the prediction frame and the Gound Truth object, selecting the prediction frame which finally meets the requirements through calculating the matchingloss between the prediction frame and the Gound Truth frame, marking the prediction frame as an object, and marking the rest frames as background classes to obtain the final prediction output.
As shown in fig. 5, which is an overall network framework of DETR, in fig. 5, the overall VTOR architecture mainly includes three components, respectively: a CaiT back bone backbone for extracting compact feature representations, an encoder-decoder converter, and a simple feed forward network (FFN (feed forward network)) for making final detection predictions.
Further, the whole data processing flow of the CaiT backhaul backbone comprises the following steps:
firstly, caiT divides input image data into a plurality of h×w patches (blocks) of equal size, each patch is mapped into a d-dimensional vector by a trainable projection layer, and the vectors form an input sequence; next, the input patch vector is mapped to a higher dimensional vector representation through the embedding layer. The embedded layer comprises a linear transformation and a regularization layer; in the Encoder, caiT employs a self-attention mechanism based transform module, and CaiT employs multiple Encoder layers to process an input sequence. By performing attention calculations between all vectors in the sequence, caiT is able to capture global dependencies and local similarities between features. Wherein each encoder layer comprises two sublayers: an attention layer and a feedforward neural network layer; after passing through several encoders, caiT outputs a sequence representing the entire image, typically taking the output vector of the first input vector as a representation of the entire image.
S2.5: the DETR model eliminates redundant detection frames through non-maximum suppression to obtain a final detection result.
Preferably, in order to improve the performance, we choose DETR as the method of target detection herein, and in this model, the manual prior process can be deleted by directly predicting the detection set by using bipartite graph matching method, so that the system performance is improved, and meanwhile, the whole target detection process can be simplified, and the complexity and the use threshold of the system are reduced.
S3: and supplementing the video data at the data layer to form multi-format type data of video resources and monitoring data.
The main task of the layer is to collect data of the distribution network operation video supervision system, and meanwhile, the data are used as supplements of video data to provide data support for an application layer.
S4: based on the target detection technology, the video data is processed for the second time through a deep learning algorithm and stored.
Further, video information of the infrastructure layer is transmitted to the platform layer as input, each field picture is captured by taking a frame of the video as a unit through a deep learning algorithm based on a target detection technology, an alarm prompt is carried out according to a judging result, a current video image is stored, and time is recorded, wherein the time corresponds to operation time in basic data of the data layer one by one.
Preferably, the platform layer utilizes an intelligent analysis model to detect targets of specific scenes around data mining, business analysis and the like, builds an intelligent recognition model based on a target detection technology through a deep learning algorithm, grabs each field picture by taking a frame of a video as a unit, carries out alarm prompt according to a judging result, stores a current video image, records time and is convenient for tracing historical images.
S5: and feeding back the video data processing result at the application layer.
Preferably, a worker monitors the safety of the distribution network operation site according to the requirement, and can inquire the intelligent monitoring analysis result of the platform layer through basic data (operators, operation time and operation place); the method and the system realize the monitoring of the safety behavior of the distribution network operation site, the analysis and the tracing of the illegal responsibility of the personnel operation behavior after the operation, and the functions of safety inspection and safety production inspection of the distribution network operation.
The embodiment also provides a distribution network operation supervision system based on intelligent video analysis, which comprises an acquisition module, a control module and a control module, wherein the acquisition module is used for acquiring video data in real time and uploading the video data; the data processing module is used for carrying out image recognition, processing, supplementation and uploading on the collected video data; the data storage module is used for storing and inquiring the processed uploaded images and analysis results; and the data feedback module is used for feeding the identification result back to the controller, and the controller adjusts the supervision system by adopting manual or automatic adjustment according to the result.
The embodiment also provides a computer device, which is suitable for the situation of the distribution network operation supervision method based on the video intelligent analysis, and comprises the following steps:
a memory and a processor; the memory is used for storing computer executable instructions, and the processor is used for executing the computer executable instructions to realize the distribution network operation supervision method based on the intelligent video analysis as proposed in the embodiment.
The computer device may be a terminal comprising a processor, a memory, a communication interface, a display screen and input means connected by a system bus. Wherein the processor of the computer device is configured to provide computing and control capabilities. The memory of the computer device includes a non-volatile storage medium and an internal memory. The non-volatile storage medium stores an operating system and a computer program. The internal memory provides an environment for the operation of the operating system and computer programs in the non-volatile storage media. The communication interface of the computer device is used for carrying out wired or wireless communication with an external terminal, and the wireless mode can be realized through WIFI, an operator network, NFC (near field communication) or other technologies. The display screen of the computer equipment can be a liquid crystal display screen or an electronic ink display screen, and the input device of the computer equipment can be a touch layer covered on the display screen, can also be keys, a track ball or a touch pad arranged on the shell of the computer equipment, and can also be an external keyboard, a touch pad or a mouse and the like.
The present embodiment also provides a storage medium, on which a computer program is stored, which when executed by a processor, implements a distribution network job supervision method based on intelligent video analysis as proposed in the above embodiment.
In summary, the application uses DETR to detect, and directly predicts the detection set to delete the manual priori process by a bipartite graph matching method, so that the system performance is improved, the whole target detection process is simplified, and the complexity and the use threshold of the system are reduced; by utilizing the characteristic that a deep learning algorithm does not need manual intervention, the video monitoring is carried out on the distribution network operation site, the images are intercepted by taking the video as a unit, the intelligent analysis system is utilized for analysis, and the images can be subjected to target monitoring without preprocessing of manually designing the context scene characteristics and the data, so that the safety monitoring problem of personnel safety helmet wearing on the distribution network operation site is realized.
Example 2
Referring to fig. 6, experimental data of the present experiment are provided for verifying the advantageous effects of the second embodiment of the present application on the basis of the first embodiment.
Conventional target detection algorithms are classified into one stage target detection algorithm and two types of two stage target detection algorithms, and specifically are shown in table 1:
table 1 conventional target detection algorithm
| one stage target detection algorithm | two stage target detection algorithm |
| YOLO | R-CNN |
| SSD | SPP-Net |
| SQUeezeDet | Fast-CNN |
| DeceteDet | Faster-CNN |
Compared with the traditional detection method, the detection prediction speed is improved rapidly by outputting the prediction result through the DETR model, and the detection comparison of the detection prediction speed and the detection method of the YOLO is as follows:
TABLE 2 accuracy and mAP for different detection methods
| Parameters (parameters) | YOLOv4 | SSD | The application is that |
| Accuracy rate of | 0.872 | 0.876 | 0.897 |
| mAP | 0.894 | 0.904 | 0.964 |
From the table, the improved algorithm mAP is highest in detection precision and is higher than 0.07 of the YOLOv4 algorithm, and other algorithms are not greatly different in detection precision, so that compared with the existing method for detecting the helmet, the method for supervising the distribution network operation based on the intelligent video analysis is higher in accuracy and higher in detection efficiency, and the detection precision can be improved to a certain extent.
It should be noted that the above embodiments are only for illustrating the technical solution of the present application and not for limiting the same, and although the present application has been described in detail with reference to the preferred embodiments, it should be understood by those skilled in the art that the technical solution of the present application may be modified or substituted without departing from the spirit and scope of the technical solution of the present application, which is intended to be covered in the scope of the claims of the present application.
Claims (10)
1. A distribution network operation supervision method based on intelligent video analysis is characterized by comprising the following steps of: comprising the steps of (a) a step of,
collecting video data by using a perception layer;
image recognition is carried out on the collected video data, and the recognized video data is stored, transmitted and distributed;
supplementing the video data to form multi-format type data of video resources and monitoring data;
based on a target detection technology, performing secondary processing on video data through a deep learning algorithm and storing the video data;
and feeding back the video data processing result.
2. The distribution network operation supervision method based on intelligent video analysis according to claim 1, wherein: the specific steps of image recognition of the collected video data are,
preprocessing an image;
extracting feature vectors of the image;
performing dimension reduction on the feature vector of the image;
inputting the feature vector into a DETR model to predict a target detection task;
the DETR model eliminates redundant detection frames through non-maximum suppression to obtain a final detection result.
3. The distribution network operation supervision method based on intelligent video analysis as set forth in claim 2, wherein: the specific steps of preprocessing the image are,
scaling and calculating the picture;
determining a probability distribution of the image;
and calculating the optimal transmission distance.
4. A distribution network operation supervision method based on intelligent video analysis as recited in claim 3, wherein: the judging process of zooming and calculating the picture comprises the following steps:
setting a fixed n multiplied by n size, and respectively scaling the length and the width of the image to n;
let the original width be w 1 Height is h 1 After image scaling of the current image width w 2 Height is h 2 ：
If w 1 >h 1 W is then 2 ＝n，h 2 ＝[h 1 ]；
If w 1 <h 1 W is then 2 ＝[w1]，h 2 ＝n；
If w 1 ＝h 1 W is then 2 ＝h 2 ＝n；
Wherein [ ] represents an upward rounding.
5. The distribution network operation supervision method based on intelligent video analysis according to claim 4, wherein: the probability distribution of the determined image is that the preprocessed image I is treated as a sample obtained by independently sampling all pixels of the preprocessed image I from unknown distribution, and the specific formula of the probability density function is as follows:
where n is the number of pixels in the image, x i ∈R d Feature vector, K, representing the ith pixel h (. Cndot.) a Gaussian kernel function denoted as h;
wherein K is h The formula of (-) is:
where h is the bandwidth of the gaussian kernel function.
6. The distribution network operation supervision method based on intelligent video analysis according to claim 5, wherein: the optimal transmission distance comprises a calculation transmission plan and an optimal transmission distance;
the calculating a transmission plan includes the transmission requirement meeting the following constraints:
wherein n and m are the dimensions of the distributions P and Q, respectively, and P (i) and Q (j) are the values of the ith and jth elements in the distributions P and Q, respectively;
under this constraint, minimizing the total cost is expressed as the following linear programming problem:
minimize∑i,j T(i,j)*C(i,j)
wherein C (i, j) is the distance from the ith element in the distribution P to the jth element in the distribution Q, minimize is the objective function, and subject is the constraint function;
the calculation of the optimal transmission distance comprises setting two probability distributions as mu and v, setting their support sets as X and Y respectively, setting the number of elements as n and m respectively, setting C as an n×m cost matrix, wherein the (i, j) th element C ij Representing the cost required to transport the ith sample point in μ to the jth sample point in ν; the minimum cost for converting μ to ν is:
where u (μ, v) represents the set of all matrices T satisfying the constraint, < C, T > represents the sum of the element-wise products of the cost matrix C and the matrix T.
7. The distribution network operation supervision method based on intelligent video analysis according to claim 6, wherein: the determination of the constraint includes that,
T∈R n×m ；
the sum of each row of T is equal to the probability mass corresponding to μ, i.e
The sum of each column of T is equal to the probability mass corresponding to v, i.e
The element in T being non-negative, i.e. T ij ≥0。
8. A distribution network operation supervision system adopting the distribution network operation supervision method based on intelligent video analysis as claimed in any one of claims 1 to 7, characterized in that: also included is a method of manufacturing a semiconductor device,
the acquisition module is used for acquiring video data in real time and uploading the video data;
the data processing module is used for carrying out image recognition, processing, supplementation and uploading on the collected video data;
the data storage module is used for storing and inquiring the processed uploaded images and analysis results;
and the data feedback module is used for feeding the identification result back to the controller, and the controller adjusts the supervision system by adopting manual or automatic adjustment according to the result.
9. A computer device comprising a memory and a processor, the memory storing a computer program, characterized in that: the steps of the distribution network operation supervision method based on video intelligent analysis according to any one of claims 1 to 7 are realized when the processor executes the computer program.
10. A computer-readable storage medium having stored thereon a computer program, characterized by: the steps of the distribution network operation supervision method based on video intelligent analysis according to any one of claims 1 to 7 are realized when the computer program is executed by a processor.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310760086.7A CN117037021A (en) | 2023-06-26 | 2023-06-26 | Distribution network operation supervision method and system based on intelligent video analysis |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202310760086.7A CN117037021A (en) | 2023-06-26 | 2023-06-26 | Distribution network operation supervision method and system based on intelligent video analysis |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN117037021A true CN117037021A (en) | 2023-11-10 |
Family
ID=88626907
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202310760086.7A Pending CN117037021A (en) | 2023-06-26 | 2023-06-26 | Distribution network operation supervision method and system based on intelligent video analysis |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117037021A (en) |
-
2023
- 2023-06-26 CN CN202310760086.7A patent/CN117037021A/en active Pending
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Bargshady et al. | Enhanced deep learning algorithm development to detect pain intensity from facial expression images | |
| WO2021093468A1 (en) | Video classification method and apparatus, model training method and apparatus, device and storage medium | |
| Lu et al. | A real-time object detection algorithm for video | |
| US11195067B2 (en) | Systems and methods for machine learning-based site-specific threat modeling and threat detection | |
| US20180253640A1 (en) | Hybrid architecture system and method for high-dimensional sequence processing | |
| US20180060691A1 (en) | Active visual attention models for computer vision tasks | |
| JP2020524348A (en) | Face image retrieval method and system, photographing device, and computer storage medium | |
| CN115331172A (en) | Workshop dangerous behavior recognition alarm method and system based on monitoring video | |
| CN113888514A (en) | Method and device for detecting defects of ground wire, edge computing equipment and storage medium | |
| CN117392604A (en) | Real-time information monitoring and management system and method for Internet of things | |
| CN116129426A (en) | Fine granularity classification method for cervical cell smear 18 category | |
| CN115909336A (en) | Text recognition method and device, computer equipment and computer-readable storage medium | |
| CN117037021A (en) | Distribution network operation supervision method and system based on intelligent video analysis | |
| KR102540175B1 (en) | Ai based pig disease management system | |
| CN114445875A (en) | Deep learning-based identity recognition and face comparison system and training method | |
| CN112616043A (en) | PYNQ-based neural network identification video monitoring alarm system and method | |
| CN117636908B (en) | Digital mine production management and control system | |
| CN113837173A (en) | Target object detection method and device, computer equipment and storage medium | |
| Bhavani | Automated Attendance System and Voice Assistance using Face Recognition | |
| RU2779408C1 (en) | Method for creating combined neural network cascades with common feature extraction layers and with multiple outputs, trained on different datasets simultaneously | |
| CN115239708B (en) | Plant leaf disease detection model training method and plant leaf disease detection method | |
| Harini et al. | A novel static and dynamic hand gesture recognition using self organizing map with deep convolutional neural network | |
| Ding et al. | A learnable end-edge-cloud cooperative network for driving emotion sensing | |
| WO2024080231A1 (en) | Server device, terminal device, information processing method, and information processing system | |
| US20230360402A1 (en) | Video-based public safety incident prediction system and method therefor |
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 |