CN110826636A - Aircraft anomaly detection system and anomaly detection method thereof - Google Patents
Aircraft anomaly detection system and anomaly detection method thereof Download PDFInfo
- Publication number
- CN110826636A CN110826636A CN201911098431.5A CN201911098431A CN110826636A CN 110826636 A CN110826636 A CN 110826636A CN 201911098431 A CN201911098431 A CN 201911098431A CN 110826636 A CN110826636 A CN 110826636A
- Authority
- CN
- China
- Prior art keywords
- layer
- fault
- detection
- aircraft
- feature extraction
- 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
- 238000001514 detection method Methods 0.000 title claims abstract description 87
- 238000000605 extraction Methods 0.000 claims abstract description 27
- 238000011176 pooling Methods 0.000 claims abstract description 20
- 230000009466 transformation Effects 0.000 claims abstract description 18
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 8
- 230000004913 activation Effects 0.000 claims abstract description 6
- 238000012549 training Methods 0.000 claims description 29
- 238000000034 method Methods 0.000 claims description 16
- 238000012423 maintenance Methods 0.000 claims description 13
- 238000012795 verification Methods 0.000 claims description 11
- 238000004804 winding Methods 0.000 claims description 9
- 238000012800 visualization Methods 0.000 claims description 8
- 230000002159 abnormal effect Effects 0.000 claims description 3
- 230000005540 biological transmission Effects 0.000 claims description 3
- 238000012545 processing Methods 0.000 claims description 2
- 238000005457 optimization Methods 0.000 claims 2
- 230000005856 abnormality Effects 0.000 description 4
- 230000006870 function Effects 0.000 description 4
- 238000013024 troubleshooting Methods 0.000 description 3
- 238000006243 chemical reaction Methods 0.000 description 2
- 238000007689 inspection Methods 0.000 description 2
- 238000012821 model calculation Methods 0.000 description 2
- 238000004364 calculation method Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 230000008034 disappearance Effects 0.000 description 1
- 239000011159 matrix material Substances 0.000 description 1
- 238000011179 visual inspection Methods 0.000 description 1
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/24—Classification techniques
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/23—Clustering techniques
-
- 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
- G06N3/048—Activation functions
Abstract
The invention relates to an aircraft anomaly detection system, which comprises a feature extraction model and an improved YOLOv3 detector; the feature extraction model comprises three layers, wherein the 1 st layer and the 2 nd layer respectively comprise a convolution operation unit, a maximum pooling unit and a nonlinear transformation unit, and the 3 rd layer only comprises a convolution operation unit; the convolution operation unit of the 1 st layer comprises 32 convolution kernels of 3 x 3, and the convolution operation units of the 2 nd and 3 rd layers comprise 64 convolution kernels of 3 x 3; the maximum pooling units of the 1 st layer and the 2 nd layer both adopt 2 multiplied by 2 pooling windows, and the nonlinear transformation units both adopt ReLU as a nonlinear activation function to carry out nonlinear transformation; the improved YOLOv3 detector generates 3 anchor frames for target detection through a clustering algorithm aiming at small target image samples to obtain fault coordinates, fault types and fault existence probability information. The invention improves the model operation efficiency and the detection accuracy.
Description
Technical Field
The invention belongs to the technical field of aircraft anomaly detection, and relates to an aircraft anomaly detection system.
Background
The system stores fault information of all airports and corresponding maintenance method information in a database, and each airport can quickly obtain foreperson troubleshooting experience by consulting the database through a client, so that the troubleshooting speed is increased; meanwhile, the new troubleshooting experience of the airport can be stored in a database to be shared with other airports; although the system can improve the information timeliness, the aircraft abnormity needs to be subjectively judged by technicians, and the defects that different release personnel have larger difference in the inspection result of the same aircraft, and the missed inspection is very easy to occur only by a visual inspection mode.
Disclosure of Invention
The invention aims to provide an aircraft anomaly detection system with high detection efficiency and good accuracy.
In order to solve the above technical problem, the aircraft anomaly detection system of the present invention includes a feature extraction model and an improved YOLOv3 detector;
the feature extraction model comprises three layers, wherein the 1 st layer and the 2 nd layer respectively comprise a convolution operation unit, a maximum pooling unit and a nonlinear transformation unit, and the 3 rd layer only comprises a convolution operation unit; the convolution operation unit of the 1 st layer comprises 32 convolution kernels of 3 x 3, and the convolution operation units of the 2 nd and 3 rd layers comprise 64 convolution kernels of 3 x 3; the maximum pooling units of the 1 st layer and the 2 nd layer both adopt 2 multiplied by 2 pooling windows, and the nonlinear transformation units both adopt ReLU as a nonlinear activation function to carry out nonlinear transformation; the improved YOLOv3 detector generates 3 anchor frames for target detection through a clustering algorithm aiming at small target image samples to obtain fault coordinates, fault types and fault existence probability information.
Furthermore, the invention also comprises a result visualization module; and the result visualization module is used for visualizing information such as fault coordinates, fault types and fault existence probabilities and finally outputting images containing the information of the fault coordinates, the fault types and the fault existence probabilities.
The parameters in the feature extraction model and the improved YOLOv3 detector are obtained by pre-training and optimizing, and the training and optimizing method comprises the following steps:
a. the method comprises the steps that an intelligent shooting and recording device is used for collecting a video of a winding machine and transmitting the video to a system server in a network transmission mode through a portable terminal;
b. the system server preprocesses the video of the winding machine through LabelImg software, marks a fault detection target on the video image of the winding machine, and generates an image sample;
c. the image sample dataset is divided into 2 parts: training and verifying sets; inputting image sample data of a training set into a feature extraction model to obtain a last layer of feature map, inputting the last layer of feature map into an improved YOLOv3 detector to obtain a detection result, and comparing the detection result with real mark information to obtain error information; training by using a random gradient descent algorithm, repeatedly optimizing parameters of a feature extraction model and an improved YOLOv3 detector, reducing errors, and finally determining model weight parameters to obtain a detection model; verifying the obtained detection model by using a verification set, and if the difference between the detection accuracy of the verification set and the detection accuracy of the training set is more than 5%, reducing the training times until the accuracy of the training set and the detection accuracy of the verification set is more than 90%, and obtaining a trained feature extraction model and an improved YOLOv3 detector; wherein the original parameters in the feature extraction model and the modified YOLOv3 detector are both chosen randomly.
The division ratio of the training set to the verification set is 4: 1.
The method for detecting the aircraft abnormality by using the aircraft abnormality detection system comprises the following steps:
firstly, inputting aircraft image data acquired by a shooting device into a trained feature extraction model, wherein the aircraft image data firstly passes through a layer 1 convolution operation unit, a maximum pooling unit and a nonlinear transformation unit to obtain a first feature map, and then passes through a layer 2 convolution operation unit, a maximum pooling unit and a nonlinear transformation unit to obtain a second feature map; finally, a final layer of feature map is obtained through a layer 3 convolution operation unit;
inputting the last layer of feature map into an improved YOLOv3 detector for fault detection, and outputting a detection result image;
judging whether the detection result image is abnormal or not, if so, representing that the aircraft has a fault, executing the step four, otherwise, representing that the detection result is normal, and executing the step six;
feeding back the detection result to a maintenance knowledge base, wherein a plurality of fault types of the aircraft and corresponding maintenance methods are stored in the maintenance knowledge base;
feeding back the detected fault type and the corresponding maintenance method to the portable terminal;
and sixthly, feeding back the normal information of the detection result to the portable terminal.
Aiming at the characteristic that the number of image samples for aircraft anomaly detection is small, the invention designs the trunk network comprising 3 convolutional layers, 2 maximum pooling layers and 2 nonlinear conversion layers to replace the complex and huge Darknet53 trunk network of the YOLOv3 detection model in the prior art, greatly simplifies the trunk network, reduces the parameter operation cost and the data sample training time, effectively avoids the problem that small target features disappear along with the increase of the number of convolutional layers, and improves the model operation efficiency and the detection accuracy. Aiming at the detection characteristics of small targets, the YOLOv3 detector is improved, only 3 anchor frames are used, and only the last layer of feature map of the backbone network is used for detection, so that the model is greatly simplified, the detection accuracy is improved, and the model calculation amount is simplified.
Drawings
The present invention will be described in further detail with reference to the accompanying drawings and specific embodiments.
Fig. 1 is a block diagram of an aircraft anomaly detection system according to the present invention.
FIG. 2 is a flow chart of an aircraft anomaly detection method.
Detailed Description
As shown in fig. 1, the aircraft anomaly detection system of the present invention includes a feature extraction model, an improved YOLOv3 detector, and a result visualization module.
The feature extraction model comprises three layers, wherein the 1 st layer and the 2 nd layer respectively comprise a convolution operation unit, a maximum pooling unit and a nonlinear transformation unit, and the 3 rd layer only comprises a convolution operation unit; the convolution operation unit of the 1 st layer comprises 32 convolution kernels of 3 x 3, and the convolution operation units of the 2 nd and 3 rd layers comprise 64 convolution kernels of 3 x 3; the maximum pooling units of the 1 st and 2 nd layers adopt 2 x 2 pooling windows, and the nonlinear transformation units adopt ReLU as nonlinear activation functions to carry out nonlinear transformation.
The improved YOLOv3 detector generates 3 small-size anchor frames of a class by a clustering algorithm for target detection aiming at small targets to obtain fault coordinates, fault types and fault existence probability information.
The result visualization module performs visualization processing on information such as fault coordinates, fault types and fault existence probabilities, and finally outputs an image (200 × 200 × 3) containing the information such as the fault coordinates, the fault types and the fault existence probabilities.
The parameters in the feature extraction model and the YOLOv3 detector are obtained by pre-training and optimizing, and the training and optimizing method comprises the following steps:
a. the method comprises the steps that an intelligent shooting and recording device is used for collecting a video of a winding machine and transmitting the video to a system server in a network transmission mode through a portable terminal;
b. the system server preprocesses the video of the winding machine through LabelImg software, marks a fault detection target on each frame image of the video of the winding machine, and generates an image sample;
c. the image sample dataset is divided into 2 parts: training and verifying sets, wherein the proportion of division is 4: 1; inputting image sample data of a training set into a feature extraction model to obtain a last layer of feature map, inputting the last layer of feature map into an improved YOLOv3 detector to obtain a detection result, and comparing the detection result with real mark information to obtain error information; training by using a random gradient descent algorithm (wherein the learning rate is constant to be lr equal to 0.001), repeatedly optimizing parameters of a feature extraction model and an improved YOLOv3 detector, and reducing errors to obtain a detection model; then, the obtained detection model is verified by using a verification set, if the difference between the detection accuracy of the verification set and the detection accuracy of the training set is more than 5%, the training times are reduced until the accuracy of the training set and the detection accuracy of the verification set is more than 90%, and a trained feature extraction model and an improved YOLOv3 detector are obtained; wherein the original parameters in the feature extraction model and the modified YOLOv3 detector are both chosen randomly.
The Darknet53 backbone network used by YOLOv3 in the prior art has large scale, various parameters and various network layers, large data samples are required to be adopted for training, and small target characteristics disappear along with the increase of the network layers; and in the technical field, technicians generally consider that the more the number of network layers, the more accurate the detection result, so when the deviation of the detection result is large, the technicians generally reduce the deviation by increasing the number of network layers to improve the detection accuracy. The inventor of the invention overcomes the technical bias, and in contrast to the technical bias, aiming at the characteristic that the aircraft anomaly detection image samples are few, the main network (namely the feature extraction model) comprising 3 convolutional layers, 2 maximum pooling layers and 2 nonlinear conversion layers is designed to replace the heavy Darknet53 main network of the existing YOLOv3 detection model, so that the main network is greatly simplified, the parameter operation cost and the data sample training time are reduced, the model operation efficiency is improved, the small target features are ensured not to disappear, the detection accuracy can reach 86%, and the detection accuracy is improved by 6% compared with the Darknet53 main network used by the existing YOLOv 3.
In the prior art, the YOLOv3 detector has a great variety of targets, and needs to generate 9 anchor frames for target detection through a clustering algorithm, and extract features in 3 different feature layers for detection respectively, so as to cope with detection targets of different size types. Aiming at the detection characteristics of small targets, the YOLOv3 detector is improved, only 3 frames are used, and only the last layer of feature map of the backbone network is used for detection, so that the model is greatly simplified, the detection accuracy is improved, and the model calculation amount is simplified.
The method for detecting the aircraft abnormality by using the aircraft abnormality detection system comprises the following steps:
firstly, aircraft image data (200 multiplied by 3) collected by a shooting device are input into a trained feature extraction model, the aircraft image data firstly pass through a layer 1 convolution operation unit, a maximum pooling unit and a nonlinear transformation unit to obtain a first feature map (198 multiplied by 32), and then pass through a layer 2 convolution operation unit, a maximum pooling unit and a nonlinear transformation unit to obtain a second feature map (99 multiplied by 32); finally, a final layer of feature map (46 multiplied by 64) is obtained through a layer 3 convolution operation unit;
inputting the feature map of the last layer into a YOLOv3 detector for fault detection, and outputting a detection result image (46 multiplied by 18);
judging whether the detection result image is abnormal or not, if so, representing that the aircraft has a fault, executing the step four, otherwise, representing that the detection result is normal, and executing the step six;
feeding back the detection result to a maintenance knowledge base, wherein a plurality of fault types of the aircraft and corresponding maintenance methods are stored in the maintenance knowledge base;
feeding back the detected fault type and the corresponding maintenance method to the portable terminal;
and sixthly, feeding back the normal information of the detection result to the portable terminal.
The convolution kernel calculation formula in the invention is as follows:
wherein w is convolution kernel weight, x is input image data, b is offset, c is convolution result, k is convolution kernel number, i, j and m, n are position marks of the input image data and the convolution kernel matrix data respectively.
The nonlinear activation function ReLU is as follows:
f(x)=max(0,x) (2)
where x is the input image data. The nonlinear activation function has the advantages of being capable of reducing model parameters and avoiding gradient disappearance and the like.
Claims (5)
1. An aircraft anomaly detection system, characterized by comprising a feature extraction model and a modified YOLOv3 detector;
the feature extraction model comprises three layers, wherein the 1 st layer and the 2 nd layer respectively comprise a convolution operation unit, a maximum pooling unit and a nonlinear transformation unit, and the 3 rd layer only comprises a convolution operation unit; the convolution operation unit of the 1 st layer comprises 32 convolution kernels of 3 x 3, and the convolution operation units of the 2 nd and 3 rd layers comprise 64 convolution kernels of 3 x 3; the maximum pooling units of the 1 st layer and the 2 nd layer both adopt 2 multiplied by 2 pooling windows, and the nonlinear transformation units both adopt ReLU as a nonlinear activation function to carry out nonlinear transformation; the improved YOLOv3 detector generates 3 anchor frames for target detection through a clustering algorithm aiming at small target image samples to obtain fault coordinates, fault types and fault existence probability information.
2. The aircraft anomaly detection system according to claim 1, characterized in that it further comprises a result visualization module; and the result visualization module is used for performing visualization processing on the fault coordinate, the fault type and the fault existence probability information and finally outputting an image containing the fault coordinate, the fault type and the fault existence probability information.
3. The aircraft anomaly detection system according to claim 1, wherein the parameters of the feature extraction model and the improved YOLOv3 detector are obtained by pre-training optimization, and the training optimization method comprises the following steps:
a. the method comprises the steps that an intelligent shooting and recording device is used for collecting a video of a winding machine and transmitting the video to a system server in a network transmission mode through a portable terminal;
b. the system server preprocesses the video of the winding machine through LabelImg software, marks a fault detection target on the video image of the winding machine, and generates an image sample;
c. the image sample dataset is divided into 2 parts: training and verifying sets; inputting image sample data of a training set into a feature extraction model to obtain a last layer of feature map, inputting the last layer of feature map into an improved YOLOv3 detector to obtain a detection result, and comparing the detection result with real mark information to obtain error information; training by using a random gradient descent algorithm, repeatedly optimizing parameters of a feature extraction model and an improved YOLOv3 detector, reducing errors, and finally determining model weight parameters to obtain a detection model; verifying the obtained detection model by using a verification set, and if the difference between the detection accuracy of the verification set and the detection accuracy of the training set is more than 5%, reducing the training times until the accuracy of the training set and the detection accuracy of the verification set is more than 90%, and obtaining a trained feature extraction model and an improved YOLOv3 detector; wherein the original parameters in the feature extraction model and the modified YOLOv3 detector are both chosen randomly.
4. The aircraft anomaly detection system according to claim 3, wherein said training set is divided by a verification set in a ratio of 4: 1.
5. A method of aircraft anomaly detection using the aircraft anomaly detection system of claim 1, comprising:
firstly, inputting aircraft image data acquired by a shooting device into a trained feature extraction model, wherein the aircraft image data firstly passes through a layer 1 convolution operation unit, a maximum pooling unit and a nonlinear transformation unit to obtain a first feature map, and then passes through a layer 2 convolution operation unit, a maximum pooling unit and a nonlinear transformation unit to obtain a second feature map; finally, a final layer of feature map is obtained through a layer 3 convolution operation unit;
inputting the last layer of feature map into an improved YOLOv3 detector for fault detection, and outputting a detection result image;
judging whether the detection result image is abnormal or not, if so, representing that the aircraft has a fault, executing the step four, otherwise, representing that the detection result is normal, and executing the step six;
feeding back the detection result to a maintenance knowledge base, wherein a plurality of fault types of the aircraft and corresponding maintenance methods are stored in the maintenance knowledge base;
feeding back the detected fault type and the corresponding maintenance method to the portable terminal;
and sixthly, feeding back the normal information of the detection result to the portable terminal.
Priority Applications (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201911098431.5A CN110826636A (en) | 2019-11-12 | 2019-11-12 | Aircraft anomaly detection system and anomaly detection method thereof |
Applications Claiming Priority (1)
Application Number | Priority Date | Filing Date | Title |
---|---|---|---|
CN201911098431.5A CN110826636A (en) | 2019-11-12 | 2019-11-12 | Aircraft anomaly detection system and anomaly detection method thereof |
Publications (1)
Publication Number | Publication Date |
---|---|
CN110826636A true CN110826636A (en) | 2020-02-21 |
Family
ID=69554150
Family Applications (1)
Application Number | Title | Priority Date | Filing Date |
---|---|---|---|
CN201911098431.5A Pending CN110826636A (en) | 2019-11-12 | 2019-11-12 | Aircraft anomaly detection system and anomaly detection method thereof |
Country Status (1)
Country | Link |
---|---|
CN (1) | CN110826636A (en) |
Citations (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102760501A (en) * | 2012-07-02 | 2012-10-31 | 华北电力大学 | Method and system for troubleshooting of equipment in nuclear power plant |
WO2015045851A1 (en) * | 2013-09-27 | 2015-04-02 | Ntn株式会社 | Work-assisting system and work-assisting method |
CN106709511A (en) * | 2016-12-08 | 2017-05-24 | 华中师范大学 | Urban rail transit panoramic monitoring video fault detection method based on depth learning |
CN107145908A (en) * | 2017-05-08 | 2017-09-08 | 江南大学 | A kind of small target detecting method based on R FCN |
JP2017199263A (en) * | 2016-04-28 | 2017-11-02 | 三菱重工業株式会社 | Failure information acquisition system and failure information acquisition method |
CN108375974A (en) * | 2018-05-21 | 2018-08-07 | 上海星融汽车科技有限公司 | It is a kind of to be used to remotely record, analysis, diagnose, the method and system of maintenance vehicle failure |
CN109255044A (en) * | 2018-08-31 | 2019-01-22 | 江苏大学 | A kind of image intelligent mask method based on YOLOv3 deep learning network |
CN110175658A (en) * | 2019-06-26 | 2019-08-27 | 浙江大学 | A kind of distress in concrete recognition methods based on YOLOv3 deep learning |
CN110222769A (en) * | 2019-06-06 | 2019-09-10 | 大连理工大学 | A kind of Further aim detection method based on YOLOV3-tiny |
JP2020039490A (en) * | 2018-09-07 | 2020-03-19 | キヤノン株式会社 | Image processing device, medical diagnostic system, learned model, image processing method, and program |
US20220188577A1 (en) * | 2020-12-16 | 2022-06-16 | Fractal Analytics Private Limited | System and method for identifying object information in image or video data |
-
2019
- 2019-11-12 CN CN201911098431.5A patent/CN110826636A/en active Pending
Patent Citations (11)
Publication number | Priority date | Publication date | Assignee | Title |
---|---|---|---|---|
CN102760501A (en) * | 2012-07-02 | 2012-10-31 | 华北电力大学 | Method and system for troubleshooting of equipment in nuclear power plant |
WO2015045851A1 (en) * | 2013-09-27 | 2015-04-02 | Ntn株式会社 | Work-assisting system and work-assisting method |
JP2017199263A (en) * | 2016-04-28 | 2017-11-02 | 三菱重工業株式会社 | Failure information acquisition system and failure information acquisition method |
CN106709511A (en) * | 2016-12-08 | 2017-05-24 | 华中师范大学 | Urban rail transit panoramic monitoring video fault detection method based on depth learning |
CN107145908A (en) * | 2017-05-08 | 2017-09-08 | 江南大学 | A kind of small target detecting method based on R FCN |
CN108375974A (en) * | 2018-05-21 | 2018-08-07 | 上海星融汽车科技有限公司 | It is a kind of to be used to remotely record, analysis, diagnose, the method and system of maintenance vehicle failure |
CN109255044A (en) * | 2018-08-31 | 2019-01-22 | 江苏大学 | A kind of image intelligent mask method based on YOLOv3 deep learning network |
JP2020039490A (en) * | 2018-09-07 | 2020-03-19 | キヤノン株式会社 | Image processing device, medical diagnostic system, learned model, image processing method, and program |
CN110222769A (en) * | 2019-06-06 | 2019-09-10 | 大连理工大学 | A kind of Further aim detection method based on YOLOV3-tiny |
CN110175658A (en) * | 2019-06-26 | 2019-08-27 | 浙江大学 | A kind of distress in concrete recognition methods based on YOLOv3 deep learning |
US20220188577A1 (en) * | 2020-12-16 | 2022-06-16 | Fractal Analytics Private Limited | System and method for identifying object information in image or video data |
Non-Patent Citations (2)
Title |
---|
方青云,王兆魁: "基于改进YOLOv3网络的遥感目标快速检测方法", 《上海航天》 * |
王思元,王俊杰: "基于改进YOLOv3算法的高密度人群目标实时检测方法研究", 《安全与环境工程》 * |
Similar Documents
Publication | Publication Date | Title |
---|---|---|
CN107808139B (en) | Real-time monitoring threat analysis method and system based on deep learning | |
CN109359520B (en) | Crowd counting method, system, computer readable storage medium and server | |
CN114445706A (en) | Power transmission line target detection and identification method based on feature fusion | |
CN112801146B (en) | Target detection method and system | |
CN111507370A (en) | Method and device for obtaining sample image of inspection label in automatic labeling image | |
CN112448836A (en) | Fault root cause determination method, device, server and computer readable medium | |
CN109858352A (en) | A kind of method for diagnosing faults based on compressed sensing and the multiple dimensioned network of improvement | |
CN111382740B (en) | Text picture analysis method, text picture analysis device, computer equipment and storage medium | |
CN112613569B (en) | Image recognition method, training method and device for image classification model | |
CN109800712A (en) | A kind of vehicle detection method of counting and equipment based on depth convolutional neural networks | |
CN114926747A (en) | Remote sensing image directional target detection method based on multi-feature aggregation and interaction | |
CN111522951A (en) | Sensitive data identification and classification technical method based on image identification | |
CN116229052B (en) | Method for detecting state change of substation equipment based on twin network | |
CN110365639A (en) | A kind of malicious traffic stream detection method and system based on depth residual error network | |
CN110113327A (en) | A kind of method and device detecting DGA domain name | |
CN114511784A (en) | Environment monitoring and early warning method, device, equipment and storage medium | |
Liu et al. | Slippage fault diagnosis of dampers for transmission lines based on faster R-CNN and distance constraint | |
Kato et al. | Parsing line chart images using linear programming | |
CN115346068A (en) | Automatic generation method for bolt loss fault image of railway freight train | |
CN114694130A (en) | Method and device for detecting telegraph poles and pole numbers along railway based on deep learning | |
CN111862036B (en) | Tool abnormal placement detection method and system based on double-index measurement learning | |
CN116503399B (en) | Insulator pollution flashover detection method based on YOLO-AFPS | |
CN113093695A (en) | Data-driven SDN controller fault diagnosis system | |
CN110826636A (en) | Aircraft anomaly detection system and anomaly detection method thereof | |
CN116993681A (en) | Substation inspection defect detection method and system |
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 | ||
RJ01 | Rejection of invention patent application after publication | ||
RJ01 | Rejection of invention patent application after publication |
Application publication date: 20200221 |