CN117236520B - Distributed multi-unmanned aerial vehicle cluster cooperative scheduling system and method thereof - Google Patents
Distributed multi-unmanned aerial vehicle cluster cooperative scheduling system and method thereof Download PDFInfo
- Publication number
- CN117236520B CN117236520B CN202311492077.0A CN202311492077A CN117236520B CN 117236520 B CN117236520 B CN 117236520B CN 202311492077 A CN202311492077 A CN 202311492077A CN 117236520 B CN117236520 B CN 117236520B
- Authority
- CN
- China
- Prior art keywords
- inspection
- fault point
- path
- image
- traffic condition
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000034 method Methods 0.000 title claims abstract description 55
- 238000007689 inspection Methods 0.000 claims abstract description 161
- 238000011084 recovery Methods 0.000 claims abstract description 49
- 238000011156 evaluation Methods 0.000 claims abstract description 45
- 238000009826 distribution Methods 0.000 claims abstract description 31
- 238000012549 training Methods 0.000 claims description 78
- 239000013598 vector Substances 0.000 claims description 58
- 230000003993 interaction Effects 0.000 claims description 30
- 238000013527 convolutional neural network Methods 0.000 claims description 21
- 230000006870 function Effects 0.000 claims description 21
- 230000007246 mechanism Effects 0.000 claims description 21
- 238000000605 extraction Methods 0.000 claims description 18
- 239000011159 matrix material Substances 0.000 claims description 18
- 230000008878 coupling Effects 0.000 claims description 16
- 238000010168 coupling process Methods 0.000 claims description 16
- 238000005859 coupling reaction Methods 0.000 claims description 16
- 238000013135 deep learning Methods 0.000 claims description 14
- 238000010586 diagram Methods 0.000 claims description 13
- 238000005457 optimization Methods 0.000 claims description 10
- 238000011176 pooling Methods 0.000 claims description 9
- 238000003709 image segmentation Methods 0.000 claims description 8
- 230000004913 activation Effects 0.000 claims description 6
- 238000003062 neural network model Methods 0.000 claims description 2
- 238000012545 processing Methods 0.000 description 11
- 230000008569 process Effects 0.000 description 9
- 238000004458 analytical method Methods 0.000 description 4
- 238000005516 engineering process Methods 0.000 description 4
- 230000008439 repair process Effects 0.000 description 4
- 238000005065 mining Methods 0.000 description 3
- 238000007477 logistic regression Methods 0.000 description 2
- 238000012544 monitoring process Methods 0.000 description 2
- 210000002569 neuron Anatomy 0.000 description 2
- 230000000750 progressive effect Effects 0.000 description 2
- 230000009466 transformation Effects 0.000 description 2
- 238000009825 accumulation Methods 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 238000004590 computer program Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 230000004927 fusion Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 230000010354 integration Effects 0.000 description 1
- 238000004519 manufacturing process Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 238000010606 normalization Methods 0.000 description 1
- 230000002787 reinforcement Effects 0.000 description 1
- 230000003252 repetitive effect Effects 0.000 description 1
- 238000013468 resource allocation Methods 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 1
Classifications
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02T—CLIMATE CHANGE MITIGATION TECHNOLOGIES RELATED TO TRANSPORTATION
- Y02T10/00—Road transport of goods or passengers
- Y02T10/10—Internal combustion engine [ICE] based vehicles
- Y02T10/40—Engine management systems
Abstract
A distributed multi-unmanned aerial vehicle cluster cooperative scheduling system and a method thereof are disclosed. The method comprises the following steps: acquiring a city digital distribution network topological map; positioning distribution information of fault points; planning a routing inspection path; shooting an inspection path image along the inspection path and an inspection fault point image of a fault point by a distributed multi-unmanned aerial vehicle cluster; extracting path traffic condition characteristics of the inspection path image; extracting fault point disaster condition characteristics of the inspection fault point image; obtaining the evaluation result based on the path traffic condition characteristics and the fault point disaster recovery condition characteristics; based on the evaluation result, an optimal scheduling scheme is obtained; and implementing scheduling according to the optimal scheduling scheme. Thus, accurate assessment of disaster conditions can be achieved.
Description
Technical Field
The present application relates to the field of unmanned aerial vehicles, and more particularly, to a distributed multi-unmanned aerial vehicle cluster collaborative scheduling system and a method thereof.
Background
The emergency repair of the power distribution network after typhoons, waterlogging, storms, ice and snow and other extreme meteorological events is an important link in the related work of the power distribution network. When the power distribution network fails, economic loss can be brought to normal production of users, and safe operation of the power grid is jeopardized. The problems of slow response, uneven rush-repair resource allocation and inadequate monitoring of the rush-repair progress exist in the conventional power distribution network fault rush-repair. In order to improve the power supply quality of the power distribution network and reduce the economic loss caused by extreme meteorological events, the fault point needs to be positioned at the first time, the disaster-stricken condition is quantitatively evaluated, and an optimal emergency scheduling scheme is made.
Patent CN111582697B provides a method and a system for evaluating and scheduling faults of a power distribution network, which are used for rapidly positioning fault points and evaluating disaster conditions after extreme weather events occur, and providing an optimal rush repair scheduling scheme and a rush repair progress monitoring function. However, in the process of actually performing disaster recovery evaluation, the modified AlexNet model is used to quantitatively evaluate the disaster recovery condition, which may have a problem of insufficient stability or insufficient generalization capability. That is, the disaster situation cannot be accurately evaluated when a new or complex scene is encountered. Thus, an optimized solution is desired.
Disclosure of Invention
In view of this, the present application provides a distributed multi-unmanned aerial vehicle cluster collaborative scheduling system and a method thereof, which can utilize an image processing technology based on deep learning to perform feature mining and feature interaction on a patrol path image and a patrol fault point image, and thereby realize accurate assessment of disaster conditions.
According to an aspect of the present application, there is provided a distributed multi-unmanned aerial vehicle cluster cooperative scheduling method, including: acquiring a city digital distribution network topological map; positioning distribution information of fault points; planning a routing inspection path; shooting an inspection path image along the inspection path and an inspection fault point image of a fault point by a distributed multi-unmanned aerial vehicle cluster; analyzing the inspection path image and the inspection fault point image, and evaluating disaster conditions and traffic conditions to obtain evaluation results; based on the evaluation result, an optimal scheduling scheme is obtained; and, implementing scheduling according to the optimal scheduling scheme; analyzing the inspection path image and the inspection fault point image, evaluating disaster-stricken conditions and traffic conditions, and obtaining evaluation results, wherein the method comprises the following steps: extracting path traffic condition characteristics of the inspection path image; extracting fault point disaster condition characteristics of the inspection fault point image; and obtaining the evaluation result based on the path traffic condition characteristics and the fault point disaster recovery condition characteristics.
According to another aspect of the present application, there is provided a distributed multi-unmanned aerial vehicle cluster cooperative scheduling system, including: the map acquisition module is used for acquiring a topological map of the urban digital distribution network; the positioning module is used for positioning the distribution information of the fault points; the planning module is used for planning a routing inspection path; the image shooting module is used for acquiring an inspection path image along the inspection path shot by the distributed multi-unmanned aerial vehicle cluster and an inspection fault point image of the fault point; the evaluation module is used for analyzing the inspection path image and the inspection fault point image, evaluating disaster-stricken conditions and traffic conditions and obtaining evaluation results; the optimal scheme acquisition module is used for acquiring an optimal scheduling scheme based on the evaluation result; the implementation module is used for implementing scheduling according to the optimal scheduling scheme; wherein the evaluation module comprises: the path characteristic extraction unit is used for extracting the path traffic condition characteristics of the inspection path image; the fault point feature extraction unit is used for extracting fault point disaster condition features of the inspection fault point image; and an evaluation result acquisition unit for acquiring the evaluation result based on the path traffic condition characteristics and the fault point disaster-affected condition characteristics.
According to an embodiment of the present application, the method comprises: acquiring a city digital distribution network topological map; positioning distribution information of fault points; planning a routing inspection path; shooting an inspection path image along the inspection path and an inspection fault point image of a fault point by a distributed multi-unmanned aerial vehicle cluster; extracting path traffic condition characteristics of the inspection path image; extracting fault point disaster condition characteristics of the inspection fault point image; obtaining the evaluation result based on the path traffic condition characteristics and the fault point disaster recovery condition characteristics; based on the evaluation result, an optimal scheduling scheme is obtained; and implementing scheduling according to the optimal scheduling scheme. Thus, accurate assessment of disaster conditions can be achieved.
Other features and aspects of the present application will become apparent from the following detailed description of exemplary embodiments, which proceeds with reference to the accompanying drawings.
Drawings
The accompanying drawings, which are incorporated in and constitute a part of this specification, illustrate exemplary embodiments, features and aspects of the present application and together with the description, serve to explain the principles of the present application.
Fig. 1 shows a flowchart of a distributed multi-drone cluster co-scheduling method according to an embodiment of the present application.
Fig. 2 shows a flowchart of sub-step S150 of a distributed multi-drone cluster co-scheduling method according to an embodiment of the present application.
Fig. 3 shows an architectural diagram of sub-step S150 of a distributed multi-drone cluster co-scheduling method according to an embodiment of the present application.
Fig. 4 shows a flowchart of sub-step S151 of a distributed multi-drone cluster co-scheduling method according to an embodiment of the present application.
Fig. 5 shows a flowchart of sub-step S152 of a distributed multi-drone cluster co-scheduling method according to an embodiment of the present application.
Fig. 6 shows a flowchart of sub-step S153 of the distributed multi-drone cluster co-scheduling method according to an embodiment of the present application.
Fig. 7 shows a block diagram of a distributed multi-drone cluster co-scheduling system according to an embodiment of the present application.
Fig. 8 shows an application scenario diagram of a distributed multi-unmanned cluster co-scheduling method according to an embodiment of the present application.
Detailed Description
The following description of the embodiments of the present application will be made clearly and fully with reference to the accompanying drawings, in which it is apparent that the embodiments described are only some, but not all embodiments of the present application. All other embodiments, which can be made by one of ordinary skill in the art based on the embodiments of the present application without making any inventive effort, are also within the scope of the present application.
As used in this application and in the claims, the terms "a," "an," "the," and/or "the" are not specific to the singular, but may include the plural, unless the context clearly dictates otherwise. In general, the terms "comprises" and "comprising" merely indicate that the steps and elements are explicitly identified, and they do not constitute an exclusive list, as other steps or elements may be included in a method or apparatus.
Various exemplary embodiments, features and aspects of the present application will be described in detail below with reference to the accompanying drawings. In the drawings, like reference numbers indicate identical or functionally similar elements. Although various aspects of the embodiments are illustrated in the accompanying drawings, the drawings are not necessarily drawn to scale unless specifically indicated.
In addition, numerous specific details are set forth in the following detailed description in order to provide a better understanding of the present application. It will be understood by those skilled in the art that the present application may be practiced without some of these specific details. In some instances, methods, means, elements, and circuits have not been described in detail as not to unnecessarily obscure the present application.
Aiming at the technical problems, the technical concept of the method is to utilize an image processing technology based on deep learning to perform feature mining and feature interaction on the inspection path image and the inspection fault point image, and accurately evaluate the disaster condition.
Based on this, fig. 1 shows a flowchart of a distributed multi-drone cluster co-scheduling method according to an embodiment of the present application. As shown in fig. 1, the distributed multi-unmanned aerial vehicle cluster cooperative scheduling method according to the embodiment of the application includes the following steps: s110, obtaining a topological map of the urban digital distribution network; s120, positioning distribution information of fault points; s130, planning a routing inspection path; s140, shooting an inspection path image along the inspection path and an inspection fault point image of a fault point by the distributed multi-unmanned aerial vehicle cluster; s150, analyzing the inspection path image and the inspection fault point image, and evaluating disaster-stricken conditions and traffic conditions to obtain evaluation results; s160, obtaining an optimal scheduling scheme based on the evaluation result; and S170, implementing scheduling according to the optimal scheduling scheme.
Further, fig. 2 shows a flowchart of sub-step S150 of the distributed multi-drone cluster co-scheduling method according to an embodiment of the present application. Fig. 3 shows an architectural diagram of sub-step S150 of a distributed multi-drone cluster co-scheduling method according to an embodiment of the present application. As shown in fig. 2 and fig. 3, according to the distributed multi-unmanned aerial vehicle cluster collaborative scheduling method in the embodiment of the present application, analyzing the inspection path image and the inspection fault point image, and evaluating disaster conditions and traffic conditions to obtain evaluation results, including: s151, extracting path traffic condition characteristics of the inspection path image; s152, extracting fault point disaster condition characteristics of the inspection fault point image; and S153, obtaining the evaluation result based on the path traffic condition characteristics and the fault point disaster recovery condition characteristics.
Specifically, in the technical scheme of the application, the route traffic condition features of the inspection route image are extracted first. Here, it should be noted that, although the characteristics of the path traffic condition are not strictly taken as a basis for judging the disaster situation. However, the technical problem of the present application is to accurately evaluate the disaster situation to obtain an evaluation result, and the evaluation result should be used as an important basis for selecting an optimal scheduling scheme in the subsequent application, while the path traffic condition feature information in the routing inspection path can provide important traffic condition information such as vehicle density, driving speed, road surface condition, etc., and can reflect the difficulty level and time cost of the emergency repair personnel to reach the failure point, so in the technical scheme of the present application, the present application is considered as one of the judging basis for the disaster situation.
In a specific example of the present application, as shown in fig. 4, the encoding process for extracting the path traffic condition feature of the patrol path image includes: s1511, image segmentation is carried out on the inspection path image along the extending direction of the inspection path so as to obtain a sequence of inspection path image blocks; and S1512, extracting the path traffic condition characteristics from the sequence of the inspection path image blocks by using a network model based on a deep learning algorithm. It should be understood that the purpose of step S1511 is to break up the entire inspection path image into a plurality of tiles for subsequent feature extraction processing, and by slicing the image along the direction along which the inspection path extends, the image can be segmented into successive tiles, each tile representing a particular region on the path. The objective of step S1512 is to analyze the patrol path image block by using a deep learning algorithm, to extract features on the path traffic condition from it, the deep learning algorithm can learn complex patterns and features in the image, so that information on the path traffic condition can be captured automatically, and by extracting features from the image block sequence, information on the traffic condition at different positions on the path, such as vehicle density, road congestion degree, vehicle speed, etc., can be obtained. In combination, step S1511 is used to segment the inspection path image into small blocks for subsequent processing, while step S1512 uses a deep learning algorithm to extract path traffic condition features from the image blocks for further analysis and application.
Specifically, in step S1512, extracting the path traffic condition feature from the sequence of the patrol path image blocks using a network model based on a deep learning algorithm includes: the sequence of the inspection path image block passes through a path traffic condition feature extractor based on a convolutional neural network model to obtain a sequence of inspection path traffic condition feature vectors; and taking the sequence of the routing inspection path traffic condition feature vector as the path traffic condition feature.
It should be noted that the convolutional neural network (Convolutional Neural Network, CNN) is a deep learning algorithm model, and is mainly used for processing data with a grid structure, such as image data. The main characteristic of the convolutional neural network is that the convolutional neural network can automatically learn the characteristic representation in the image, extract the local characteristics in the image through a plurality of convolutional layers and pooling layers, and combine the characteristics through a full connection layer to carry out classification or regression tasks. The following are the levels and functions that are common in convolutional neural networks: 1. input Layer (Input Layer): input image data is received. 2. Convolution layer (Convolutional Layer): features in the image are extracted by a convolution operation. The convolution layer contains a series of convolution kernels (also called filters), each of which is responsible for extracting a particular feature in the image, such as an edge, texture, etc. 3. Activation Layer (Activation Layer): nonlinear transformation is introduced to increase the expressive power of the model. Common activation functions include ReLU (Rectified Linear Unit), sigmoid, tanh, and the like. 4. Pooling Layer (Pooling Layer): the feature map size is reduced by a downsampling operation, reducing the number of parameters while retaining important features. Common pooling operations include maximum pooling and average pooling. 5. Output Layer (Output Layer): and outputting a prediction result of the model according to the requirements of specific tasks. For example, in the image classification task, the output layer may be a full connection layer, outputting the probability of each category. The convolutional neural network model is widely applied to image processing tasks, and local and global features in an image can be effectively captured by the convolutional neural network through multi-layer convolution and pooling operation, so that high-level understanding and analysis of the image are realized. In the path traffic condition feature extraction, the convolutional neural network may extract features related to the path traffic condition, such as vehicle density, road congestion degree, etc., by learning a feature representation of the patrol path image block. These features can be used for further analysis, prediction and decision making.
Accordingly, the path traffic condition feature extractor based on the convolutional neural network model comprises an input layer, a convolutional layer, an activation layer, a pooling layer and an output layer.
And meanwhile, extracting the disaster condition characteristics of the fault points of the inspection fault point image. Namely, capturing hidden characteristic information about disaster conditions of fault points contained in the inspection fault point image. In a specific example of the present application, the implementation manner of extracting the fault point disaster situation feature of the inspection fault point image is to obtain an inspection fault point disaster situation feature map by using a fault point disaster situation feature extractor of a spatial attention mechanism.
Here, the fault point disaster condition feature extractor using the spatial attention mechanism may adaptively focus on an important region in the fault point image, that is, the spatial attention mechanism may assign different weights according to the importance of a local region of the fault point image, thereby highlighting feature distribution of disaster regions such as collapse, water accumulation regions.
Accordingly, as shown in fig. 5, extracting the disaster situation feature of the fault point of the inspection fault point image includes: s1521, obtaining a patrol fault point disaster recovery condition characteristic diagram by using a fault point disaster recovery condition characteristic extractor of a spatial attention mechanism according to the patrol fault point image; and S1522, taking the patrol fault point disaster-affected condition characteristic map as the fault point disaster-affected condition characteristic.
It is worth mentioning that the spatial attention mechanism (Spatial Attention Mechanism) is a technique used in deep learning to enhance the attention and processing power of the model to spatial information. The method enables the model to pay more attention to important areas or features by adaptively adjusting weights of different spatial positions, thereby improving the performance and the expression capacity of the model. In image processing tasks, spatial attention mechanisms are often used to weight features of different regions in an image. It enables the model to automatically select and focus on the most relevant and useful areas in the image through learned weight assignments. This mechanism can improve the perceptibility of the model to key features and reduce interference to extraneous features. The spatial attention mechanism is typically implemented by the following steps: 1. extracting characteristics: first, a feature representation is extracted from an input image by a convolutional neural network or the like. 2. Generating attention weights: a weight matrix is generated using the attention mechanism, each element of the matrix representing an importance weight for a corresponding location. 3. And (5) weighting feature fusion: the original features are multiplied by the attention weights and the features are fused in a weighted manner. Thus, the model may be more focused on the features of the regions with higher weights. 4. And (3) generating a feature map: the weighted features are further processed, for example, by a convolution operation, into a feature map. Spatial attention mechanisms have a wide range of applications in image processing. In the process of extracting the fault point disaster recovery condition characteristics of the inspection fault point image, the fault point disaster recovery condition characteristic extractor using the spatial attention mechanism can adaptively adjust the weight according to the importance of different positions in the image, so that the model focuses on the fault point and the related area of the disaster recovery condition. In this way, the ability to perceive and understand the disaster condition of the fault point can be improved, so that the characteristics of the disaster condition of the fault point can be extracted more accurately for further analysis and decision making.
And then, the sequence of the routing inspection path traffic condition feature vector and the routing inspection fault point disaster recovery condition feature map are subjected to feature coupling module to obtain routing inspection traffic condition-fault point disaster recovery condition interaction feature vector. Here, the feature coupling module can fuse the traffic condition feature vector of the inspection path and the disaster-stricken condition feature map of the fault point, so that traffic condition feature information and disaster-stricken condition feature information are comprehensively considered to describe nuances among different scenes, and the model can have finer consideration on judgment of the disaster-stricken condition. And then, the inspection traffic condition-fault point disaster-affected condition interaction feature vector is passed through a classifier to obtain a classification result as the evaluation result, wherein the classification result is used for representing a label of disaster-affected degree.
Accordingly, as shown in fig. 6, the evaluation result is obtained based on the path traffic condition feature and the disaster-affected condition feature of the fault point, including: s1531, the sequence of the inspection path traffic condition feature vector and the inspection fault point disaster recovery condition feature map are subjected to feature coupling module to obtain inspection traffic condition-fault point disaster recovery condition interaction feature vector; and S1532, passing the inspection traffic condition-fault point disaster recovery condition interaction feature vector through a classifier to obtain a classification result as the evaluation result, wherein the classification result is used for representing a label of disaster recovery degree.
In step S1532, the inspection traffic condition-fault point disaster recovery condition interaction feature vector is passed through a classifier to obtain a classification result as the evaluation result, where the classification result is used as a label for indicating the disaster recovery degree, and includes: performing full connection coding on the patrol traffic condition-fault point disaster-affected condition interaction feature vector by using a full connection layer of the classifier to obtain a coding classification feature vector; and inputting the coding classification feature vector into a Softmax classification function of the classifier to obtain the classification result.
It should be appreciated that the role of the classifier is to learn the classification rules and classifier using a given class, known training data, and then classify (or predict) the unknown data. Logistic regression (logistics), SVM, etc. are commonly used to solve the classification problem, and for multi-classification problems (multi-class classification), logistic regression or SVM can be used as well, but multiple bi-classifications are required to compose multiple classifications, but this is error-prone and inefficient, and the commonly used multi-classification method is the Softmax classification function.
Here, it should be noted that the full-connection encoding (Fully Connected Encoding) refers to a process of encoding input data through a full-connection layer to obtain an encoded classification feature vector. Fully connected layers are a common layer type in neural networks, where each neuron is connected to all neurons of the previous layer. In the fully connected layer, each input feature is multiplied by a weight, and nonlinear transformation is performed through an activation function to obtain an output feature. Specifically, the process of full-concatenated coding is as follows: 1. input data: and (5) inspecting the traffic condition-fault point disaster-affected condition interaction feature vector. 2. Full tie layer: multiplying the input eigenvector by the weight matrix, and adding bias term to obtain the output of the full connection layer. 3. Encoding the classification feature vector: the output of the fully connected layer is used as a coding classification feature vector to represent the coded representation of the input data in the fully connected layer. Softmax classification function: and inputting the coding classification feature vector into a Softmax function of the classifier, and carrying out normalization processing through the Softmax function to obtain probability distribution of a classification result. The Softmax function may convert each element in the vector into a probability value representing the probability that the element belongs to the respective category. Finally, the classifier selects the category with the highest probability as a classification result according to the output probability distribution of the Softmax function, and the classification result is used for representing the label of the disaster degree. The process of fully-connected encoding may help extract relevant features in the input data and convert them into encoded representations suitable for classification tasks.
Further, in the technical solution of the present application, the distributed multi-unmanned aerial vehicle cluster collaborative scheduling method further includes a training step: and training the path traffic condition feature extractor based on the convolutional neural network model, the fault point disaster condition feature extractor using the spatial attention mechanism, the feature coupling module and the classifier. It should be appreciated that the training step functions in the distributed multi-unmanned cluster co-scheduling method to optimize and adjust the parameters of the model by using the noted training data, so that it can better adapt to task demands and improve performance. Specifically, the training step has several important roles: 1. model parameter optimization: through the training step, parameters in the model can be adjusted by a back propagation algorithm and an optimization algorithm (such as gradient descent), so that the model can better fit training data. Through the iterative optimization process, the parameters of the model are gradually adjusted to the optimal values, so that the performance and generalization capability of the model are improved. 2. Feature extractor training: in the distributed multi-unmanned aerial vehicle cluster collaborative scheduling method, the training step is used for training a path traffic condition feature extractor based on a convolutional neural network model and a fault point disaster condition feature extractor using a spatial attention mechanism. By training these feature extractors, they can be made to automatically learn and extract valid feature representations related to path traffic conditions and fault point disaster conditions, thereby improving feature extraction accuracy and expressive power. 3. Training a characteristic coupling module: the characteristic coupling module is used for coupling the path traffic condition characteristics and the fault point disaster condition characteristics in the distributed multi-unmanned-plane cluster cooperative scheduling method so as to obtain comprehensive characteristic representation. Through training the feature coupling module, the two types of features can be integrated and fused in a self-adaptive mode, and therefore the comprehensive capacity and the expression capacity of the features are improved. 4. Training a classifier: in the distributed multi-unmanned aerial vehicle cluster collaborative scheduling method, a classifier is used for classifying and predicting path traffic conditions and fault point disaster conditions according to feature representation. By training the classifier, the method can learn the mapping relation from the characteristic representation to the specific classification result, thereby realizing accurate classification and prediction of the path traffic condition and the disaster condition of the fault point. Through the training step, the model can be gradually optimized and adjusted to an optimal state, so that the performance and accuracy of the distributed multi-unmanned aerial vehicle cluster cooperative scheduling method are improved, and the method can be better applied to actual scenes.
Wherein, in one example, the training step comprises: acquiring training data, wherein the training data comprises training patrol path images and training patrol fault point images of fault points along the patrol path shot by the distributed multi-unmanned aerial vehicle cluster and true values of labels of disaster degree; image segmentation is carried out on the training inspection path image along the extending direction of the inspection path so as to obtain a sequence of training inspection path image blocks; passing the sequence of the training inspection path image blocks through the path traffic condition feature extractor based on the convolutional neural network model to obtain a sequence of training inspection path traffic condition feature vectors; the training inspection fault point image passes through the fault point disaster recovery condition feature extractor using a spatial attention mechanism to obtain a training inspection fault point disaster recovery condition feature map; the sequence of the traffic condition feature vectors of the training inspection path and the disaster recovery condition feature map of the inspection fault point are used for obtaining training inspection traffic condition-disaster recovery condition interaction feature vectors of the fault point through the feature coupling module; the training inspection traffic condition-fault point disaster recovery condition interaction feature vector passes through a classifier to obtain a classification loss function value; and training the path traffic condition feature extractor based on the convolutional neural network model, the fault point disaster recovery condition feature extractor using a spatial attention mechanism, the feature coupling module and the classifier by using the classification loss function value, wherein in each round of iteration of training, iterative optimization of a weight matrix in a weight space is carried out on the training patrol traffic condition-fault point disaster recovery condition interaction feature vector.
In the technical scheme of the application, each training routing inspection path traffic condition feature vector in the sequence of training routing inspection path traffic condition feature vectors expresses image semantic features of corresponding training routing inspection path image blocks, a training inspection fault point disaster recovery condition feature map expresses image semantic features of spatial distribution reinforcement of the training inspection fault point images, after the sequence of training routing inspection path traffic condition feature vectors and the training inspection fault point disaster recovery condition feature map pass through a feature coupling module, the training inspection traffic condition-fault point disaster recovery condition interaction feature vectors are fused with distribution dimensions corresponding to the spatial distribution dimensions and the channel distribution dimensions corresponding to the image semantic features of the training inspection path image blocks and the training inspection fault point images, and each feature value is also provided with fused image semantic feature dense representation, so that training efficiency of a matrix of the classifier is reduced when the training inspection traffic condition-fault point disaster recovery condition interaction feature vectors pass through the classifier for classification and regression training.
Based on the above, when the applicant performs classification regression training on the training patrol traffic condition-fault point disaster recovery condition interaction feature vector through a classifier, the applicant performs iterative optimization of a weight matrix in a weight space based on the training patrol traffic condition-fault point disaster recovery condition interaction feature vector.
Accordingly, in a specific example, in each iteration of the training, performing iterative optimization on a weight matrix in a weight space on the training patrol traffic condition-fault point disaster recovery condition interaction feature vector includes: in each iteration of the training, carrying out iterative optimization on a weight matrix in a weight space on the training inspection traffic condition-fault point disaster-affected condition interaction feature vector according to the following optimization formula; wherein, the optimization formula is:wherein (1)>And->The weight matrix of the last iteration and the current iteration are respectively adopted, wherein, during the first iteration, different initialization strategies are adopted to set +.>And->(e.g.)>Is arranged as a unit matrixSet as the diagonal matrix of the mean value of the feature vector to be classified),>is the training inspection traffic condition-fault point disaster-affected condition interaction feature vector to be classified, < +.>And->Respectively represent feature vector +>And->Global mean of (2), and->Is a bias matrix, e.g. initially set as a unity matrix, the vectors being in the form of column vectors, +.>Representing vector multiplication, ++>Representing the addition of the matrix,representing multiplication by location +.>Representing a transpose operation->Represents a maximum function>Representing the optimized weight matrix.
That is, consider that the traffic condition-fault point disaster-affected condition interaction feature vector is inspected based on training to be classifiedDuring the intensive prediction task of (1), the high-resolution representation of the weight matrix and the interaction feature vector of the training inspection traffic condition to be classified and the disaster-stricken condition of the fault point are needed to be carried out +.>The image semantic feature multidimensional distribution association context is integrated, so that progressive integration (progressive integrity) is realized based on iteration association representation resource-aware (resource-aware) by maximizing a distribution boundary of a weight space in an iteration process, thereby improving the training effect of a weight matrix and improving the training efficiency of the whole model.
In summary, according to the distributed multi-unmanned aerial vehicle cluster collaborative scheduling method disclosed by the embodiment of the application, feature mining and feature interaction can be performed on the inspection path image and the inspection fault point image by using an image processing technology based on deep learning, and therefore, the disaster recovery condition can be accurately evaluated.
Fig. 7 shows a block diagram of a distributed multi-drone cluster co-scheduling system 100 according to an embodiment of the present application. As shown in fig. 7, a distributed multi-unmanned cluster cooperative scheduling system 100 according to an embodiment of the present application includes: the map acquisition module 110 is configured to acquire a topology map of the urban digital distribution network; the positioning module 120 is used for positioning distribution information of fault points; a planning module 130, configured to plan a routing inspection path; the image shooting module 140 is configured to obtain an inspection path image along an inspection path shot by the distributed multi-unmanned aerial vehicle cluster and an inspection fault point image of a fault point; the evaluation module 150 is configured to analyze the inspection path image and the inspection fault point image, and evaluate disaster conditions and traffic conditions to obtain an evaluation result; an optimal solution obtaining module 160, configured to obtain an optimal scheduling solution based on the evaluation result; and an implementation module 170, configured to implement scheduling according to the optimal scheduling scheme.
In one possible implementation, the evaluation module 150 includes: the path characteristic extraction unit is used for extracting the path traffic condition characteristics of the inspection path image; the fault point feature extraction unit is used for extracting fault point disaster condition features of the inspection fault point image; and an evaluation result acquisition unit for acquiring the evaluation result based on the path traffic condition characteristics and the fault point disaster-affected condition characteristics.
In one possible implementation manner, the path feature extraction unit includes: the image segmentation subunit is used for carrying out image segmentation on the patrol path image along the direction in which the patrol path extends so as to obtain a sequence of patrol path image blocks; and a path traffic condition feature extraction subunit, configured to extract the path traffic condition feature from the sequence of the inspection path image blocks using a network model based on a deep learning algorithm.
In one possible implementation manner, the path traffic condition feature extraction subunit includes: the convolution coding secondary subunit is used for enabling the sequence of the inspection path image block to pass through a path traffic condition feature extractor based on a convolution neural network model to obtain a sequence of inspection path traffic condition feature vectors; and a path traffic condition feature acquisition secondary subunit, configured to take the sequence of the patrol path traffic condition feature vector as the path traffic condition feature.
Here, it will be appreciated by those skilled in the art that the specific functions and operations of the respective units and modules in the above-described distributed multi-unmanned aerial vehicle cluster co-scheduling system 100 have been described in detail in the above description of the distributed multi-unmanned aerial vehicle cluster co-scheduling method with reference to fig. 1 to 6, and thus, repetitive descriptions thereof will be omitted.
As described above, the distributed multi-unmanned cluster cooperative scheduling system 100 according to the embodiment of the present application may be implemented in various wireless terminals, for example, a server or the like having a distributed multi-unmanned cluster cooperative scheduling algorithm. In one possible implementation, the distributed multi-drone cluster co-scheduling system 100 according to embodiments of the present application may be integrated into a wireless terminal as one software module and/or hardware module. For example, the distributed multi-drone cluster co-scheduling system 100 may be a software module in the operating system of the wireless terminal, or may be an application developed for the wireless terminal; of course, the distributed multi-drone cluster co-scheduling system 100 may also be one of the many hardware modules of the wireless terminal.
Alternatively, in another example, the distributed multi-drone cluster co-scheduling system 100 and the wireless terminal may be separate devices, and the distributed multi-drone cluster co-scheduling system 100 may be connected to the wireless terminal through a wired and/or wireless network, and transmit interaction information in a agreed data format.
Fig. 8 shows an application scenario diagram of a distributed multi-unmanned cluster co-scheduling method according to an embodiment of the present application. As shown in fig. 8, in this application scenario, first, a patrol path image (e.g., D1 illustrated in fig. 8) and a patrol fault point image (e.g., D2 illustrated in fig. 8) are acquired, and then the patrol path image and the patrol fault point image are input to a server (e.g., S illustrated in fig. 8) in which a distributed multi-unmanned cluster cooperative scheduling algorithm is deployed, wherein the server can process the patrol path image and the patrol fault point image using the distributed multi-unmanned cluster cooperative scheduling algorithm to obtain a classification result of a tag for representing a disaster degree.
The flowcharts and block diagrams in the figures illustrate the architecture, functionality, and operation of possible implementations of systems, methods and computer program products according to various embodiments of the present application. In this regard, each block in the flowchart or block diagrams may represent a module, segment, or portion of instructions, which comprises one or more executable instructions for implementing the specified logical function(s). In some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the block diagrams and/or flowchart illustration, and combinations of blocks in the block diagrams and/or flowchart illustration, can be implemented by special purpose hardware-based systems which perform the specified functions or acts, or combinations of special purpose hardware and computer instructions.
The embodiments of the present application have been described above, the foregoing description is exemplary, not exhaustive, and not limited to the embodiments disclosed. Many modifications and variations will be apparent to those of ordinary skill in the art without departing from the scope and spirit of the various embodiments described. The terminology used herein was chosen in order to best explain the principles of the embodiments, the practical application, or the improvement of technology in the marketplace, or to enable others of ordinary skill in the art to understand the embodiments disclosed herein.
Claims (4)
1. A distributed multi-unmanned aerial vehicle cluster cooperative scheduling method comprises the following steps: acquiring a city digital distribution network topological map; positioning distribution information of fault points; planning a routing inspection path; shooting an inspection path image along the inspection path and an inspection fault point image of a fault point by a distributed multi-unmanned aerial vehicle cluster; analyzing the inspection path image and the inspection fault point image, and evaluating disaster conditions and traffic conditions to obtain evaluation results; based on the evaluation result, an optimal scheduling scheme is obtained; and, implementing scheduling according to the optimal scheduling scheme; the method is characterized by analyzing the patrol path image and the patrol fault point image, evaluating disaster-stricken conditions and traffic conditions, and obtaining evaluation results, and comprises the following steps: extracting path traffic condition characteristics of the inspection path image; extracting fault point disaster condition characteristics of the inspection fault point image; and obtaining the evaluation result based on the path traffic condition characteristics and the fault point disaster-affected condition characteristics; the extracting the path traffic condition characteristics of the inspection path image comprises the following steps: image segmentation is carried out on the inspection path image along the extending direction of the inspection path so as to obtain a sequence of inspection path image blocks; extracting the path traffic condition characteristics from the sequence of the inspection path image blocks by using a network model based on a deep learning algorithm; the method for extracting the path traffic condition features from the sequence of the inspection path image blocks by using a network model based on a deep learning algorithm comprises the following steps: the sequence of the inspection path image block passes through a path traffic condition feature extractor based on a convolutional neural network model to obtain a sequence of inspection path traffic condition feature vectors; taking the sequence of the routing inspection path traffic condition feature vector as the path traffic condition feature; the extracting the disaster condition features of the fault points of the inspection fault point image comprises the following steps: the inspection fault point image is subjected to fault point disaster recovery condition feature extraction by using a spatial attention mechanism to obtain an inspection fault point disaster recovery condition feature image; taking the patrol fault point disaster-affected condition characteristic diagram as the fault point disaster-affected condition characteristic; the method for obtaining the evaluation result based on the path traffic condition characteristics and the fault point disaster-affected condition characteristics comprises the following steps: the sequence of the routing inspection path traffic condition feature vector and the routing inspection fault point disaster recovery condition feature map are subjected to feature coupling module to obtain routing inspection traffic condition-fault point disaster recovery condition interaction feature vector; and passing the inspection traffic condition-fault point disaster-affected condition interaction feature vector through a classifier to obtain a classification result as the evaluation result, wherein the classification result is used for representing a label of disaster-affected degree.
2. The distributed multi-unmanned aerial vehicle cluster collaborative scheduling method according to claim 1, wherein the convolutional neural network model-based path traffic condition feature extractor comprises an input layer, a convolutional layer, an activation layer, a pooling layer, and an output layer.
3. The distributed multi-unmanned aerial vehicle cluster cooperative scheduling method of claim 2, further comprising the training step of: training the path traffic condition feature extractor based on the convolutional neural network model, the fault point disaster condition feature extractor using the spatial attention mechanism, the feature coupling module and the classifier; wherein the training step comprises: acquiring training data, wherein the training data comprises training patrol path images and training patrol fault point images of fault points along the patrol path shot by the distributed multi-unmanned aerial vehicle cluster and true values of labels of disaster degree; image segmentation is carried out on the training inspection path image along the extending direction of the inspection path so as to obtain a sequence of training inspection path image blocks; passing the sequence of the training inspection path image blocks through the path traffic condition feature extractor based on the convolutional neural network model to obtain a sequence of training inspection path traffic condition feature vectors; the training inspection fault point image passes through the fault point disaster recovery condition feature extractor using a spatial attention mechanism to obtain a training inspection fault point disaster recovery condition feature map; the sequence of the traffic condition feature vectors of the training inspection path and the disaster recovery condition feature map of the inspection fault point are used for obtaining training inspection traffic condition-disaster recovery condition interaction feature vectors of the fault point through the feature coupling module; the training inspection traffic condition-fault point disaster recovery condition interaction feature vector passes through a classifier to obtain a classification loss function value; and training the path traffic condition feature extractor based on the convolutional neural network model, the fault point disaster recovery condition feature extractor using a spatial attention mechanism, the feature coupling module and the classifier by using the classification loss function value, wherein in each round of iteration of training, iterative optimization of a weight matrix in a weight space is carried out on the training patrol traffic condition-fault point disaster recovery condition interaction feature vector.
4. A distributed multi-unmanned cluster cooperative scheduling system, comprising: the map acquisition module is used for acquiring a topological map of the urban digital distribution network; the positioning module is used for positioning the distribution information of the fault points; the planning module is used for planning a routing inspection path; the image shooting module is used for acquiring an inspection path image along the inspection path shot by the distributed multi-unmanned aerial vehicle cluster and an inspection fault point image of the fault point; the evaluation module is used for analyzing the inspection path image and the inspection fault point image, evaluating disaster-stricken conditions and traffic conditions and obtaining evaluation results; the optimal scheme acquisition module is used for acquiring an optimal scheduling scheme based on the evaluation result; the implementation module is used for implementing scheduling according to the optimal scheduling scheme; wherein the evaluation module comprises: the path characteristic extraction unit is used for extracting the path traffic condition characteristics of the inspection path image; the fault point feature extraction unit is used for extracting fault point disaster condition features of the inspection fault point image; the evaluation result acquisition unit is used for acquiring the evaluation result based on the path traffic condition characteristics and the fault point disaster-affected condition characteristics; wherein the path feature extraction unit includes: the image segmentation subunit is used for carrying out image segmentation on the patrol path image along the direction in which the patrol path extends so as to obtain a sequence of patrol path image blocks; the path traffic condition feature extraction subunit is used for extracting the path traffic condition features from the sequence of the inspection path image blocks by utilizing a network model based on a deep learning algorithm; wherein the path traffic condition feature extraction subunit includes: the convolution coding secondary subunit is used for enabling the sequence of the inspection path image block to pass through a path traffic condition feature extractor based on a convolution neural network model to obtain a sequence of inspection path traffic condition feature vectors; the path traffic condition feature acquisition secondary subunit is used for taking the sequence of the routing inspection path traffic condition feature vector as the path traffic condition feature; wherein, the fault point feature extraction unit includes: the inspection fault point image is subjected to fault point disaster recovery condition feature extraction by using a spatial attention mechanism to obtain an inspection fault point disaster recovery condition feature image; taking the patrol fault point disaster-affected condition characteristic diagram as the fault point disaster-affected condition characteristic; wherein the evaluation result acquisition unit includes: the sequence of the routing inspection path traffic condition feature vector and the routing inspection fault point disaster recovery condition feature map are subjected to feature coupling module to obtain routing inspection traffic condition-fault point disaster recovery condition interaction feature vector; and passing the inspection traffic condition-fault point disaster-affected condition interaction feature vector through a classifier to obtain a classification result as the evaluation result, wherein the classification result is used for representing a label of disaster-affected degree.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311492077.0A CN117236520B (en) | 2023-11-10 | 2023-11-10 | Distributed multi-unmanned aerial vehicle cluster cooperative scheduling system and method thereof |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202311492077.0A CN117236520B (en) | 2023-11-10 | 2023-11-10 | Distributed multi-unmanned aerial vehicle cluster cooperative scheduling system and method thereof |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN117236520A CN117236520A (en) | 2023-12-15 |
| CN117236520B true CN117236520B (en) | 2024-01-26 |
Family
ID=89088366
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202311492077.0A Active CN117236520B (en) | 2023-11-10 | 2023-11-10 | Distributed multi-unmanned aerial vehicle cluster cooperative scheduling system and method thereof |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117236520B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117522091B (en) * | 2024-01-08 | 2024-04-16 | 国网四川省电力公司电力科学研究院 | Intelligent scheduling system and method for emergency repair of post-earthquake power equipment |
Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105373134A (en) * | 2015-12-11 | 2016-03-02 | 国网四川省电力公司电力应急中心 | Unmanned aerial vehicle aerial photography technology-based electric power disaster situation acquisition system and method |
| CN205229806U (en) * | 2015-12-11 | 2016-05-11 | 国网四川省电力公司电力应急中心 | System is acquireed to disaster -stricken condition of electric power based on unmanned aerial vehicle technique of taking photo by plane |
| CN111582697A (en) * | 2020-04-29 | 2020-08-25 | 西安交通大学 | Power distribution network fault assessment and scheduling method and system |
| CN113486726A (en) * | 2021-06-10 | 2021-10-08 | 广西大学 | Rail transit obstacle detection method based on improved convolutional neural network |
| CN113936268A (en) * | 2021-12-16 | 2022-01-14 | 比亚迪股份有限公司 | Obstacle detection method for rail vehicle, computer device, and storage medium |
| CN114008685A (en) * | 2019-06-25 | 2022-02-01 | 辉达公司 | Intersection region detection and classification for autonomous machine applications |
| CN115512222A (en) * | 2022-09-26 | 2022-12-23 | 武汉大学 | Method for evaluating damage of ground objects in disaster scene of offline training and online learning |
| CN115619118A (en) * | 2022-08-30 | 2023-01-17 | 浙大城市学院 | Emergency material scheduling method integrating edge-weight attention deep reinforcement learning |
| CN115656731A (en) * | 2022-12-09 | 2023-01-31 | 武汉格蓝若智能技术股份有限公司 | Overhead distribution line arc grounding fault point positioning method |
| CN116307757A (en) * | 2023-01-18 | 2023-06-23 | 辽宁荣科智维云科技有限公司 | Intelligent data interaction method, interaction system, computer equipment and application |
| CN116846059A (en) * | 2023-03-07 | 2023-10-03 | 云南电网有限责任公司玉溪供电局 | Edge detection system for power grid inspection and monitoring |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20230281533A1 (en) * | 2021-04-16 | 2023-09-07 | Strong Force Vcn Portfolio 2019, Llc | Demand-Responsive Raw Material Management System |
-
2023
- 2023-11-10 CN CN202311492077.0A patent/CN117236520B/en active Active
Patent Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN105373134A (en) * | 2015-12-11 | 2016-03-02 | 国网四川省电力公司电力应急中心 | Unmanned aerial vehicle aerial photography technology-based electric power disaster situation acquisition system and method |
| CN205229806U (en) * | 2015-12-11 | 2016-05-11 | 国网四川省电力公司电力应急中心 | System is acquireed to disaster -stricken condition of electric power based on unmanned aerial vehicle technique of taking photo by plane |
| CN114008685A (en) * | 2019-06-25 | 2022-02-01 | 辉达公司 | Intersection region detection and classification for autonomous machine applications |
| CN111582697A (en) * | 2020-04-29 | 2020-08-25 | 西安交通大学 | Power distribution network fault assessment and scheduling method and system |
| CN113486726A (en) * | 2021-06-10 | 2021-10-08 | 广西大学 | Rail transit obstacle detection method based on improved convolutional neural network |
| CN113936268A (en) * | 2021-12-16 | 2022-01-14 | 比亚迪股份有限公司 | Obstacle detection method for rail vehicle, computer device, and storage medium |
| CN115619118A (en) * | 2022-08-30 | 2023-01-17 | 浙大城市学院 | Emergency material scheduling method integrating edge-weight attention deep reinforcement learning |
| CN115512222A (en) * | 2022-09-26 | 2022-12-23 | 武汉大学 | Method for evaluating damage of ground objects in disaster scene of offline training and online learning |
| CN115656731A (en) * | 2022-12-09 | 2023-01-31 | 武汉格蓝若智能技术股份有限公司 | Overhead distribution line arc grounding fault point positioning method |
| CN116307757A (en) * | 2023-01-18 | 2023-06-23 | 辽宁荣科智维云科技有限公司 | Intelligent data interaction method, interaction system, computer equipment and application |
| CN116846059A (en) * | 2023-03-07 | 2023-10-03 | 云南电网有限责任公司玉溪供电局 | Edge detection system for power grid inspection and monitoring |
Non-Patent Citations (5)
| Title |
|---|
| Anomaly Detection in Automated Vehicles Using Multistage Attention-Based Convolutional Neural Network;Abdul Rehman Javed 等;《IEEE Transactions on Intelligent Transportation Systems》;第22卷(第7期);第4291-4300页 * |
| Self-attentive deep learning method for online traffic classification and its interpretability;Guorui Xie 等;《Computer Networks》;第196卷;第1-12页 * |
| 基于Gabor小波变换的人脸识别;李云峰;《中国优秀博硕士学位论文全文数据库 (博士) 信息科技辑》(第8期);第I138-8页 * |
| 基于深度学习的轻量化网络火灾烟雾检测方法研究;李孝钊;《中国优秀硕士学位论文全文数据库 工程科技Ⅰ辑》(第1期);第B026-51页 * |
| 考虑公平性应急物资调配选址-路径优化研究;杜雨;《中国优秀硕士学位论文全文数据库 社会科学Ⅰ辑》(第7期);第G110-118页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN117236520A (en) | 2023-12-15 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Gao et al. | Cnn-based density estimation and crowd counting: A survey | |
| CN109063723B (en) | Weak supervision image semantic segmentation method based on common features of iteratively mined objects | |
| CN110717526B (en) | Unsupervised migration learning method based on graph convolution network | |
| CN113313947B (en) | Road condition evaluation method of short-term traffic prediction graph convolution network | |
| CN117236520B (en) | Distributed multi-unmanned aerial vehicle cluster cooperative scheduling system and method thereof | |
| CN111445026A (en) | Deep neural network multi-path reasoning acceleration method for edge intelligent application | |
| CN112767682A (en) | Multi-scale traffic flow prediction method based on graph convolution neural network | |
| CN113709779B (en) | Cellular network fault diagnosis method | |
| CN112270355A (en) | Active safety prediction method based on big data technology and SAE-GRU | |
| CN115240024A (en) | Method and system for segmenting extraterrestrial pictures by combining self-supervised learning and semi-supervised learning | |
| CN114943832A (en) | Road surface crack image lightweight segmentation method based on multi-path convolution feature fusion network and differential temperature knowledge distillation | |
| CN114299286A (en) | Road scene semantic segmentation method based on category grouping in abnormal weather | |
| CN114265913A (en) | Space-time prediction algorithm based on federal learning on industrial Internet of things edge equipment | |
| CN117237559B (en) | Digital twin city-oriented three-dimensional model data intelligent analysis method and system | |
| CN111325259A (en) | Remote sensing image classification method based on deep learning and binary coding | |
| CN117116051B (en) | Intelligent traffic management system and method based on artificial intelligence | |
| CN114331837A (en) | Method for processing and storing panoramic monitoring image of protection system of extra-high voltage converter station | |
| Odetola et al. | A scalable multilabel classification to deploy deep learning architectures for edge devices | |
| CN111178275A (en) | Fire detection method based on convolutional neural network | |
| CN115861664A (en) | Feature matching method and system based on local feature fusion and self-attention mechanism | |
| CN112966748B (en) | Polarized SAR image classification method based on edge perception double-branch FCN | |
| CN115240271A (en) | Video behavior identification method and system based on space-time modeling | |
| CN113486716B (en) | Airport scene target segmentation method and system thereof | |
| CN117522091B (en) | Intelligent scheduling system and method for emergency repair of post-earthquake power equipment | |
| CN116524199B (en) | Image rain removing method and device based on PReNet progressive network |
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 |