CN109143851B - Method for recognizing multi-mark fault deep learning and intelligently expressing result thereof - Google Patents
Method for recognizing multi-mark fault deep learning and intelligently expressing result thereof Download PDFInfo
- Publication number
- CN109143851B CN109143851B CN201810760970.XA CN201810760970A CN109143851B CN 109143851 B CN109143851 B CN 109143851B CN 201810760970 A CN201810760970 A CN 201810760970A CN 109143851 B CN109143851 B CN 109143851B
- Authority
- CN
- China
- Prior art keywords
- fault
- network
- deep
- manufacturing process
- label
- 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
Images
Classifications
-
- G—PHYSICS
- G05—CONTROLLING; REGULATING
- G05B—CONTROL OR REGULATING SYSTEMS IN GENERAL; FUNCTIONAL ELEMENTS OF SUCH SYSTEMS; MONITORING OR TESTING ARRANGEMENTS FOR SUCH SYSTEMS OR ELEMENTS
- G05B13/00—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion
- G05B13/02—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric
- G05B13/04—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric involving the use of models or simulators
- G05B13/042—Adaptive control systems, i.e. systems automatically adjusting themselves to have a performance which is optimum according to some preassigned criterion electric involving the use of models or simulators in which a parameter or coefficient is automatically adjusted to optimise the performance
Landscapes
- Engineering & Computer Science (AREA)
- Health & Medical Sciences (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Evolutionary Computation (AREA)
- Medical Informatics (AREA)
- Software Systems (AREA)
- Physics & Mathematics (AREA)
- General Physics & Mathematics (AREA)
- Automation & Control Theory (AREA)
- Image Analysis (AREA)
Abstract
The invention discloses a method for recognizing multi-label fault deep learning and intelligently expressing a result thereof, which relates to the field of intelligent manufacturing process control.
Description
Technical Field
The invention relates to the field of intelligent manufacturing and machine learning, in particular to a method for recognizing multi-mark fault deep learning and intelligently expressing a result of the multi-mark fault deep learning.
Background
Deep learning and recognition are a popular field in machine learning, and are successfully applied to computer vision, natural language processing, intelligent search and the like at present, deep fusion and integration theories of information technology, intelligent technology and equipment manufacturing process technology are less involved, many higher schools and scientific research institutions in China are engaged in research and application development work in the aspect, advanced ideas such as parallelism, agility, networking, reconfiguration and the like are added in the aspect of theoretical research, and object-oriented, component, agent and the like are adopted in the aspect of system design to obtain a lot of beneficial results.
Disclosure of Invention
The invention effectively solves the problems through the establishment and training of a deep recognition network, the deep learning of multi-mark faults and the intelligent expression of recognition results.
The invention provides a method for recognizing multi-mark fault deep learning and intelligently expressing a result thereof, which comprises the following steps of:
A. acquiring mass data generated in the manufacturing process as input data of a deep recognition network;
B. establishing a deep recognition network of the fault in the manufacturing process, collecting mass data generated in the manufacturing process as input data of the deep recognition network, and training the deep recognition network;
C. performing multi-mark deep learning of manufacturing process faults;
D. data monitored in the manufacturing process are collected, and a fault identification result of the manufacturing process is intelligently expressed.
Further, the step B comprises the following steps:
b1, conducting guiding learning of a supervision mode on network characteristics, establishing an information interaction mechanism between a deep sparse network and a SoftmBx classification model, highlighting discrimination characteristics extracted by the deep sparse network, restraining random characteristics extracted by the deep sparse network, and integrating the deep sparse network and the classification model under a frame of a deep recognition network;
b2, correcting a Relu activation function, relaxing the limit of the Relu activation function on zero response of negative errors, and suppressing the gradient disappearance phenomenon caused by traditional activation functions such as Sigmoid and TBnh by the relaxed Relu activation function by using the characteristics of single-side suppression, wide excitation boundary and the like of the Relu activation function, so that under the condition of keeping the advantages of the Relu function, fault identification information contained in the negative errors is fully utilized, and a depth identification network based on the corrected Relu activation function is further established;
b3, adaptively determining a network sparse term coefficient and a network node number through iterative solution, and realizing automatic training of a deep recognition network;
b4, randomly resetting the network node output during each iteration, converting the deep recognition network training into integrated training of a multi-network structure, effectively avoiding the over-fitting phenomenon, then improving the training strategy of Dropout, avoiding hardware resource waste caused by the output of the calculation node through a mode of randomly freezing the network node, training the deep recognition network quickly and reasonably, and ensuring the timeliness of fault recognition in the manufacturing process.
Further, the step C includes the steps of:
c1, constructing a manufacturing process fault multi-marking system;
c2, constructing a multi-label deep learning framework of the fault in the manufacturing process, and training a deep recognition network based on the multi-label fault;
c3, learning jointly with the multi-label fault depth recognition network to obtain joint probability distribution among the multi-label faults, reasoning missing labels, and solving the problem that some labels are missing in the multi-label fault depth recognition network training;
and C4, establishing a multi-label fault depth identification network of multi-source information fusion, thereby improving the diagnosis rate of the multi-label fault in the manufacturing process.
Further, the manufacturing process fault multi-label system in the step C1 specifically includes: the method has the advantages that data of the position, type and degree of faults in the manufacturing process are collected, manufacturing process fault marks of all subsystems of a factory are respectively established, the running state of the manufacturing process is marked efficiently and comprehensively, and the problem that the running state of the manufacturing process cannot be completely described by a single mark is solved.
Further, the constructing a multi-label deep learning framework in the step C2 specifically includes: and expanding a topological structure of a depth recognition network output layer, respectively establishing recognition tasks of each marking system on the network output layer, and defining a loss function of each recognition task by using cross entropy.
Further, the specific implementation manner of the multi-marker fault depth recognition network for establishing multi-source information fusion in step C4 is as follows: and expanding the input layer structure of the multi-mark depth recognition network, performing cross migration learning on multi-source signals, and obtaining the cooperative feature transformation between the multi-source signals and the shared fault features.
Further, the step D includes the steps of:
d1, splitting the weight of the single-layer network training into a plurality of combinations of basis functions;
d2, visualizing the deep network parameters;
d3, obtaining an expression mode of the fault information from shallow to deep;
d4, expressing the fault identification result based on the deep network, forming a health situation map of the manufacturing process, and showing the health state of the manufacturing process.
Further, the specific implementation manner of visualization of the deep-layer network parameters in step D2 is as follows: and searching the maximum input mode for activating the neuron node of the current layer, and taking the node as the linear weighted combination of the filter of the previous layer, thereby realizing the visualization of the deep network parameters.
Further, the specific implementation manner of obtaining the expression mode of the fault information from shallow to deep in the step D3 is as follows: the characteristics extracted by the layer-by-layer visual network are searched for the characteristic space similarity characteristic and the change trend between the current layer and the previous layer, the internal structure of the monitoring big data is visually reflected, and the understanding of the essence of the monitoring big data is enhanced.
The invention has the beneficial effects that: according to the method, the timeliness of fault recognition in the manufacturing process is effectively improved through the establishment and training of the deep recognition network, the diagnosis rate of multi-marker faults in the manufacturing process is effectively improved through the deep learning of the multi-marker faults, and the health state of the manufacturing process is displayed from multiple aspects through the intelligent expression of recognition results.
Drawings
The invention is further illustrated with reference to the following figures and examples.
FIG. 1 is a flow chart of a method of the present invention for recognition of multi-marker fault deep learning and intelligent expression of its results;
FIG. 2 is a flow chart of method step B of the present invention for recognition of multi-marker fault deep learning and intelligent expression of its results;
FIG. 3 is a flowchart of method step C of the present invention for recognition of multi-marker fault deep learning and intelligent expression of its results;
FIG. 4 is a flowchart of method step D of the present invention for recognition of multi-marker fault deep learning and intelligent expression of its results.
Detailed Description
Referring to fig. 1-4, the present invention provides an embodiment of a method for identifying deep learning of multi-marker faults and intelligently expressing the results thereof, comprising the following steps:
A. acquiring mass data generated in the manufacturing process as input data of a deep recognition network;
B. establishing a deep recognition network of the fault in the manufacturing process, collecting mass data generated in the manufacturing process as input data of the deep recognition network, and training the deep recognition network;
C. performing multi-mark deep learning of manufacturing process faults;
D. data monitored in the manufacturing process are collected, and a fault identification result of the manufacturing process is intelligently expressed.
Further, the step B comprises the following steps:
b1, conducting guiding learning of a supervision mode on network characteristics, establishing an information interaction mechanism between a deep sparse network and a SoftmBx classification model, highlighting discrimination characteristics extracted by the deep sparse network, restraining random characteristics extracted by the deep sparse network, and integrating the deep sparse network and the classification model under a frame of a deep recognition network;
b2, correcting a Relu activation function, relaxing the limit of the Relu activation function on zero response of negative errors, and suppressing the gradient disappearance phenomenon caused by traditional activation functions such as Sigmoid and TBnh by the relaxed Relu activation function by using the characteristics of single-side suppression, wide excitation boundary and the like of the Relu activation function, so that under the condition of keeping the advantages of the Relu function, fault identification information contained in the negative errors is fully utilized, and a depth identification network based on the corrected Relu activation function is further established;
b3, adaptively determining a network sparse term coefficient and a network node number through iterative solution, and realizing automatic training of a deep recognition network;
b4, randomly resetting the network node output during each iteration, converting the deep recognition network training into integrated training of a multi-network structure, effectively avoiding the over-fitting phenomenon, then improving the training strategy of Dropout, avoiding hardware resource waste caused by the output of the calculation node through a mode of randomly freezing the network node, training the deep recognition network quickly and reasonably, and ensuring the timeliness of fault recognition in the manufacturing process.
Further, the step C includes the steps of:
c1, constructing a manufacturing process fault multi-marking system;
c2, constructing a multi-label deep learning framework of the fault in the manufacturing process, and training a deep recognition network based on the multi-label fault;
c3, learning jointly with the multi-label fault depth recognition network to obtain joint probability distribution among the multi-label faults, reasoning missing labels, and solving the problem that some labels are missing in the multi-label fault depth recognition network training;
and C4, establishing a multi-label fault depth identification network of multi-source information fusion, thereby improving the diagnosis rate of the multi-label fault in the manufacturing process.
Further, the manufacturing process fault multi-label system in the step C1 specifically includes: the method has the advantages that data of the position, type and degree of faults in the manufacturing process are collected, manufacturing process fault marks of all subsystems of a factory are respectively established, the running state of the manufacturing process is marked efficiently and comprehensively, and the problem that the running state of the manufacturing process cannot be completely described by a single mark is solved.
Further, the constructing a multi-label deep learning framework in the step C2 specifically includes: and expanding a topological structure of a depth recognition network output layer, respectively establishing recognition tasks of each marking system on the network output layer, and defining a loss function of each recognition task by using cross entropy.
Further, the specific implementation manner of the multi-marker fault depth recognition network for establishing multi-source information fusion in step C4 is as follows: and expanding the input layer structure of the multi-mark depth recognition network, performing cross migration learning on multi-source signals, and obtaining the cooperative feature transformation between the multi-source signals and the shared fault features.
Further, the step D includes the steps of:
d1, splitting the weight of the single-layer network training into a plurality of combinations of basis functions;
d2, visualizing the deep network parameters;
d3, obtaining an expression mode of the fault information from shallow to deep;
d4, expressing the fault identification result based on the deep network, forming a health situation map of the manufacturing process, and showing the health state of the manufacturing process.
Further, the specific implementation manner of visualization of the deep-layer network parameters in step D2 is as follows: and searching the maximum input mode for activating the neuron node of the current layer, and taking the node as the linear weighted combination of the filter of the previous layer, thereby realizing the visualization of the deep network parameters.
Further, the specific implementation manner of obtaining the expression mode of the fault information from shallow to deep in the step D3 is as follows: the characteristics extracted by the layer-by-layer visual network are searched for the characteristic space similarity characteristic and the change trend between the current layer and the previous layer, the internal structure of the monitoring big data is visually reflected, and the understanding of the essence of the monitoring big data is enhanced.
The above description is only a preferred embodiment of the present invention, and the present invention is not limited to the above embodiment, and the present invention shall fall within the protection scope of the present invention as long as the technical effects of the present invention are achieved by the same means.
Claims (7)
1. A method for recognizing multi-mark fault deep learning and intelligently expressing the result thereof is characterized by comprising the following steps:
A. collecting data generated in the manufacturing process as input data of a depth recognition network;
B. establishing a deep recognition network of the fault in the manufacturing process, and training the deep recognition network;
C. performing multi-mark deep learning of the fault in the manufacturing process, and establishing a multi-mark fault deep recognition network;
D. collecting data monitored in the manufacturing process, and intelligently expressing a fault identification result of the manufacturing process;
wherein the step B comprises the following steps:
b1, performing guiding learning of a supervision mode on the network characteristics;
b2, relaxing the limit of the Relu activation function to zero negative error response, and restraining the gradient disappearance phenomenon caused by the Sigmoid activation function and the TBnh activation function through the relaxed Relu activation function;
b3, adaptively determining a network sparse term coefficient and a network node number through iterative solution;
b4, randomly clearing the output of the network node during each iteration, and randomly freezing the network node;
wherein the step C comprises the following steps:
c1, constructing a manufacturing process fault multi-marking system;
c2, constructing a multi-label deep learning framework of the fault in the manufacturing process, and training a deep recognition network based on the multi-label fault;
c3, performing joint learning with the multi-label fault depth recognition network to obtain joint probability distribution among the multi-label faults and reasoning missing labels;
and C4, establishing a multi-mark fault depth recognition network of multi-source information fusion.
2. The method for identifying the deep learning of multi-label fault and intelligently expressing the result thereof according to claim 1, wherein the manufacturing process fault multi-label system in the step C1 specifically comprises: and collecting data of the position, type and degree of the fault in the manufacturing process, and respectively establishing a manufacturing process fault mark of each subsystem of the factory.
3. The method for identifying the multi-label fault deep learning and intelligently expressing the result thereof according to claim 1, wherein the step C2 of constructing the multi-label fault deep learning framework specifically comprises: and expanding a topological structure of a depth recognition network output layer, respectively establishing recognition tasks of each marking system on the network output layer, and defining a loss function of each recognition task by using cross entropy.
4. The method for multi-label fault deep learning identification and intelligent expression of the result thereof according to claim 1, wherein the specific implementation manner of establishing the multi-label fault deep learning network with multi-source information fusion in the step C4 is as follows: and expanding the input layer structure of the multi-mark depth recognition network, performing cross migration learning on multi-source signals, and obtaining the cooperative feature transformation between the multi-source signals and the shared fault features.
5. The method for recognizing the deep learning of multi-label fault and intelligently expressing the result thereof according to claim 1, wherein the step D comprises the following steps:
d1, splitting the weight of the single-layer network training into a plurality of combinations of basis functions;
d2, visualizing the deep network parameters;
d3, obtaining an expression mode of the fault information from shallow to deep;
d4, expressing the fault identification result based on the deep network, and forming a health situation map of the manufacturing process.
6. The method for identifying the deep learning of multi-label fault and intelligently expressing the result thereof according to claim 5, wherein the specific visualization manner of the deep network parameters in the step D2 is as follows: and finding the maximum input mode for activating the neuron node of the current layer, and taking the node as the linear weighted combination of the filter of the previous layer.
7. The method for multi-label fault deep learning identification and intelligent expression of the result thereof according to claim 5, wherein the specific implementation manner of obtaining the fault information from shallow to deep expression mode in step D3 is as follows: and (3) visualizing the network extracted features layer by layer, and searching the similar characteristics and the variation trend of the feature space between the current layer and the previous layer.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810760970.XA CN109143851B (en) | 2018-07-11 | 2018-07-11 | Method for recognizing multi-mark fault deep learning and intelligently expressing result thereof |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810760970.XA CN109143851B (en) | 2018-07-11 | 2018-07-11 | Method for recognizing multi-mark fault deep learning and intelligently expressing result thereof |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN109143851A CN109143851A (en) | 2019-01-04 |
| CN109143851B true CN109143851B (en) | 2021-06-01 |
Family
ID=64800209
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810760970.XA Active CN109143851B (en) | 2018-07-11 | 2018-07-11 | Method for recognizing multi-mark fault deep learning and intelligently expressing result thereof |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109143851B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN111537945B (en) * | 2020-06-28 | 2021-05-11 | 南方电网科学研究院有限责任公司 | Intelligent ammeter fault diagnosis method and equipment based on federal learning |
| CN111722043B (en) * | 2020-06-29 | 2021-09-14 | 南方电网科学研究院有限责任公司 | Power equipment fault detection method, device and system |
Citations (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102637019A (en) * | 2011-02-10 | 2012-08-15 | 武汉科技大学 | Intelligent integrated fault diagnosis method and device in industrial production process |
| WO2017112591A1 (en) * | 2015-12-20 | 2017-06-29 | Prophecy Sensors, Llc | Machine fault detection based on a combination of sound capture and on spot feedback |
| CN107133569A (en) * | 2017-04-06 | 2017-09-05 | 同济大学 | The many granularity mask methods of monitor video based on extensive Multi-label learning |
| CN107430679A (en) * | 2015-01-29 | 2017-12-01 | 尤尼伐控股有限公司 | Computer vision system |
| JP2018037078A (en) * | 2016-08-26 | 2018-03-08 | 伊達 博 | Power generation monitoring system with failure detection function |
| CN107797537A (en) * | 2017-11-10 | 2018-03-13 | 上海第二工业大学 | A kind of prognostic and health management method applied to automatic production line |
| WO2018063539A1 (en) * | 2016-09-30 | 2018-04-05 | Sony Interactive Entertainment Inc. | Method and system for classifying virtual reality (vr) content based on modeled discomfort of a user |
| CN108038847A (en) * | 2017-12-05 | 2018-05-15 | 国网内蒙古东部电力有限公司 | Transformer inspection digital image recognition and fault detection system based on deep learning |
| CN108052954A (en) * | 2017-11-01 | 2018-05-18 | 佛山科学技术学院 | The method for diagnosing faults of sample space based on multistage high dimensional feature |
| CN108088495A (en) * | 2017-11-01 | 2018-05-29 | 佛山科学技术学院 | The hybrid system Fault Locating Method of multi-sensor monitoring data-driven |
| CN108126987A (en) * | 2017-05-04 | 2018-06-08 | 汉威广园(广州)机械设备有限公司 | A kind of high-speed rod-rolling mill remote online monitoring and intelligent diagnosis system and method |
| CN108182445A (en) * | 2017-12-13 | 2018-06-19 | 东北大学 | Procedure fault recognition methods based on big data intelligence core independent component analysis |
| CN108197648A (en) * | 2017-12-28 | 2018-06-22 | 华中科技大学 | A kind of Fault Diagnosis Method of Hydro-generating Unit and system based on LSTM deep learning models |
-
2018
- 2018-07-11 CN CN201810760970.XA patent/CN109143851B/en active Active
Patent Citations (13)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102637019A (en) * | 2011-02-10 | 2012-08-15 | 武汉科技大学 | Intelligent integrated fault diagnosis method and device in industrial production process |
| CN107430679A (en) * | 2015-01-29 | 2017-12-01 | 尤尼伐控股有限公司 | Computer vision system |
| WO2017112591A1 (en) * | 2015-12-20 | 2017-06-29 | Prophecy Sensors, Llc | Machine fault detection based on a combination of sound capture and on spot feedback |
| JP2018037078A (en) * | 2016-08-26 | 2018-03-08 | 伊達 博 | Power generation monitoring system with failure detection function |
| WO2018063539A1 (en) * | 2016-09-30 | 2018-04-05 | Sony Interactive Entertainment Inc. | Method and system for classifying virtual reality (vr) content based on modeled discomfort of a user |
| CN107133569A (en) * | 2017-04-06 | 2017-09-05 | 同济大学 | The many granularity mask methods of monitor video based on extensive Multi-label learning |
| CN108126987A (en) * | 2017-05-04 | 2018-06-08 | 汉威广园(广州)机械设备有限公司 | A kind of high-speed rod-rolling mill remote online monitoring and intelligent diagnosis system and method |
| CN108052954A (en) * | 2017-11-01 | 2018-05-18 | 佛山科学技术学院 | The method for diagnosing faults of sample space based on multistage high dimensional feature |
| CN108088495A (en) * | 2017-11-01 | 2018-05-29 | 佛山科学技术学院 | The hybrid system Fault Locating Method of multi-sensor monitoring data-driven |
| CN107797537A (en) * | 2017-11-10 | 2018-03-13 | 上海第二工业大学 | A kind of prognostic and health management method applied to automatic production line |
| CN108038847A (en) * | 2017-12-05 | 2018-05-15 | 国网内蒙古东部电力有限公司 | Transformer inspection digital image recognition and fault detection system based on deep learning |
| CN108182445A (en) * | 2017-12-13 | 2018-06-19 | 东北大学 | Procedure fault recognition methods based on big data intelligence core independent component analysis |
| CN108197648A (en) * | 2017-12-28 | 2018-06-22 | 华中科技大学 | A kind of Fault Diagnosis Method of Hydro-generating Unit and system based on LSTM deep learning models |
Non-Patent Citations (2)
| Title |
|---|
| Fault Diagnosis Based on Deep Learning;Feiya Lv,等;《AACC》;20161231;第6851-6856页 * |
| 大数据下机械智能故障诊断的机遇与挑战;雷亚国，等;《机械工程学报》;20180331;第98-99页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN109143851A (en) | 2019-01-04 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN107563381B (en) | Multi-feature fusion target detection method based on full convolution network | |
| Brahimi et al. | Deep learning for plant diseases: detection and saliency map visualisation | |
| Liang et al. | A real-time detection framework for on-tree mango based on SSD network | |
| CN109993748B (en) | Three-dimensional grid object segmentation method based on point cloud processing network | |
| CN109614874B (en) | Human behavior recognition method and system based on attention perception and tree skeleton point structure | |
| CN107247952B (en) | Deep supervision-based visual saliency detection method for cyclic convolution neural network | |
| CN111401149B (en) | Lightweight video behavior identification method based on long-short-term time domain modeling algorithm | |
| US20210182615A1 (en) | Alexnet-based insulator self-explosion recognition method | |
| CN114419464B (en) | Construction method of twin network change detection model based on deep learning | |
| CN110993037A (en) | Protein activity prediction device based on multi-view classification model | |
| CN109143851B (en) | Method for recognizing multi-mark fault deep learning and intelligently expressing result thereof | |
| CN116599857B (en) | Digital twin application system suitable for multiple scenes of Internet of things | |
| Oh et al. | Convolutional neural network and 2-D image based fault diagnosis of bearing without retraining | |
| CN107194415A (en) | Peak clustering method based on Laplace centrality | |
| CN112217674A (en) | Alarm root cause identification method based on causal network mining and graph attention network | |
| CN116150509B (en) | Threat information identification method, system, equipment and medium for social media network | |
| CN113420827A (en) | Semantic segmentation network training and image semantic segmentation method, device and equipment | |
| CN114202026A (en) | Multitask model training method and device and multitask processing method and device | |
| Zainudin et al. | A Framework for Chili Fruits Maturity Estimation using Deep Convolutional Neural Network. | |
| CN116977265A (en) | Training method and device for defect detection model, computer equipment and storage medium | |
| CN104835154A (en) | Color image object acquisition method based on random walk | |
| CN114036319A (en) | Power knowledge extraction method, system, device and storage medium | |
| CN115019342A (en) | Endangered animal target detection method based on class relation reasoning | |
| Heng et al. | Anti-vibration hammer detection in UAV image | |
| Li et al. | Region proposal ranking via fusion feature for object detection |
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 |