CN117972364B - Low-altitude airplane real-time identification method combining seismograph with deep learning - Google Patents
Low-altitude airplane real-time identification method combining seismograph with deep learning Download PDFInfo
- Publication number
- CN117972364B CN117972364B CN202410377473.7A CN202410377473A CN117972364B CN 117972364 B CN117972364 B CN 117972364B CN 202410377473 A CN202410377473 A CN 202410377473A CN 117972364 B CN117972364 B CN 117972364B
- Authority
- CN
- China
- Prior art keywords
- time
- aircraft
- real
- short
- low
- 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 27
- 238000013135 deep learning Methods 0.000 title claims abstract description 15
- 238000012549 training Methods 0.000 claims abstract description 25
- 238000012360 testing method Methods 0.000 claims abstract description 18
- 238000012545 processing Methods 0.000 claims abstract description 13
- 238000001914 filtration Methods 0.000 claims abstract description 12
- 238000004364 calculation method Methods 0.000 claims abstract description 9
- 238000012952 Resampling Methods 0.000 claims abstract description 8
- 230000009466 transformation Effects 0.000 claims abstract description 3
- 230000006870 function Effects 0.000 claims description 19
- 238000001228 spectrum Methods 0.000 claims description 8
- 230000004913 activation Effects 0.000 claims description 5
- 230000007246 mechanism Effects 0.000 claims description 5
- 238000010606 normalization Methods 0.000 claims description 5
- 230000003993 interaction Effects 0.000 claims description 4
- 238000011176 pooling Methods 0.000 claims description 3
- 238000005070 sampling Methods 0.000 claims description 3
- ORILYTVJVMAKLC-UHFFFAOYSA-N Adamantane Natural products C1C(C2)CC3CC1CC2C3 ORILYTVJVMAKLC-UHFFFAOYSA-N 0.000 claims 1
- 238000013507 mapping Methods 0.000 claims 1
- 238000001514 detection method Methods 0.000 abstract description 17
- 230000000694 effects Effects 0.000 abstract description 7
- 230000006872 improvement Effects 0.000 description 4
- 230000008569 process Effects 0.000 description 3
- 230000009467 reduction Effects 0.000 description 3
- 238000010586 diagram Methods 0.000 description 2
- 238000012544 monitoring process Methods 0.000 description 2
- 238000013528 artificial neural network Methods 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000008901 benefit Effects 0.000 description 1
- 238000011161 development Methods 0.000 description 1
- 230000004069 differentiation Effects 0.000 description 1
- 230000010339 dilation Effects 0.000 description 1
- 230000007613 environmental effect Effects 0.000 description 1
- 230000007274 generation of a signal involved in cell-cell signaling Effects 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 239000013307 optical fiber Substances 0.000 description 1
- 238000005457 optimization Methods 0.000 description 1
- 230000008447 perception Effects 0.000 description 1
- 239000000047 product Substances 0.000 description 1
- 238000011897 real-time detection Methods 0.000 description 1
- 230000004044 response Effects 0.000 description 1
- 230000035945 sensitivity Effects 0.000 description 1
- 239000013589 supplement Substances 0.000 description 1
- 230000001131 transforming effect Effects 0.000 description 1
- 239000013585 weight reducing agent Substances 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
- Y02T90/00—Enabling technologies or technologies with a potential or indirect contribution to GHG emissions mitigation
Landscapes
- Measurement Of Mechanical Vibrations Or Ultrasonic Waves (AREA)
Abstract
The invention discloses a low-altitude aircraft real-time identification method combining seismograph with deep learning, relates to the technical field of data processing and identification, and solves the problem that the existing low-altitude aircraft detection method is easily influenced by external environment, so that the detection effect is poor; the method comprises the steps of collecting original vibration signal data through seismometers arranged near an airport railway; sequentially carrying out filtering, short-time Fourier transformation, resampling and standardization on original vibration signal data to obtain a data set, wherein the data set comprises a training set and a testing set; training and testing the improved MobileNetV-Small network based on the data set as an aircraft signal recognition model; carrying out real-time identification on the low-altitude airplane signals by using the acquired airplane signal identification model; according to the invention, the lightweight network MobilenetV-Small training in deep learning is used as an identification model, so that on the premise of high accuracy, the calculation resource is saved, and the method is suitable for real-time identification of low-altitude airplane signals.
Description
Technical Field
The invention relates to the technical field of data processing and recognition, in particular to a low-altitude aircraft real-time recognition method combining seismometers with deep learning.
Background
Seismometers are a common instrument used in the field of geophysical prospecting for detecting and recording seismic waves and their propagation paths. As a sensor with high sensitivity and a wide frequency band, a seismograph can record seismic data in real time, and accurately measure the amplitude, frequency and propagation velocity of seismic waves.
Aircraft, one of the main vehicles, is naturally an important item in traffic monitoring. The radar is an effective flight track monitoring technical means at present, and the radar system can accurately determine the information such as the position, the speed, the direction and the like of the aircraft through the reflection and the scattering of electromagnetic waves and objects generated by the aircraft. Meanwhile, the radar also has high time resolution. It can be scanned and measured at a relatively high rate to provide positional information of the aircraft in real time. This accurate real-time detection capability is important for aviation traffic management, aircraft tracking, and quick response in emergency situations. Radar observations still suffer from some unavoidable drawbacks. First, when the aircraft flight altitude is low, the cross-sectional area presented to the radar is insufficient, making it difficult for radar technology to detect low-altitude aircraft. Secondly, the radar itself is huge and cannot move, and is easy to find and lacks flexibility; thirdly, radar detection has a terrain shielding blind area under the condition of complex terrain, and the whole area cannot be completely covered; fourth, radar detection is easily affected by weather conditions, and accuracy is not high in extreme weather.
Aiming at the problem, students at home and abroad put forward different low-altitude airplane detection methods based on different sensors to be used as the supplement of radar. Such as infrared detection, image detection, acoustic detection. In addition, in the military field, with the development of stealth technology of low-altitude sudden attack aircrafts, the accuracy of infrared detection is also reduced. The image detection is easily influenced by environmental weather and light sources, detection work cannot be carried out at night, and all-weather perception of an airplane is difficult to realize. The detection distance of the acoustic wave detection is small, and most of the detection distances cannot be estimated to be more than 150 m. And secondly, the acoustic wave signals are vulnerable to strong pollution of natural noise such as wind mines and human activity noise.
Disclosure of Invention
In order to solve the problems in the prior art, the invention provides a low-altitude airplane real-time identification method combining seismometers with deep learning, which solves the problem that the detection effect is poor due to the fact that the existing low-altitude airplane detection method is easily influenced by external environment.
A low-altitude airplane real-time identification method combining seismograph with deep learning comprises the following steps:
step 1: collecting original vibration signal data through a laid seismograph;
Step 2: sequentially carrying out filtering, short-time Fourier transformation, resampling and standardization on original vibration signal data to obtain a data set, wherein the data set comprises a training set and a testing set;
Step 3: training and testing improvements based on datasets The network is used as an aircraft signal recognition model;
step 4: and (3) carrying out real-time identification on the low-altitude airplane signals by using the airplane signal identification model obtained in the step (3).
Preferably, the step 1 includes: and recording vibration signal data through the arranged earthquake station array, simultaneously manually recording the passing time of the airplane, selecting the vibration signal data according to the passing time of the airplane, wherein the selected vibration signal data cover airplane vibration signals of different types and non-airplane noise vibration signals of different working conditions.
Preferably, the step 2 includes:
Step 2.1: performing high-pass filtering on the original vibration signal data collected in the step 1;
step 2.2: processing the vibration signal data after filtering into a short-time Fourier spectrogram by using short-time Fourier transform;
step 2.3: resampling the short-time Fourier spectrogram by adopting a regional interpolation method;
Step 2.4: normalizing the resampled short-time Fourier spectrogram;
step 2.5: and extracting part of the short-time Fourier spectrogram as a test set according to the time, and taking the rest part as a training set, wherein the proportion of the short-time Fourier spectrogram in the first half period in the test set is smaller than that in the second half period.
Preferably, the step 2.2 includes: and carrying out sliding window sampling on the vibration signal data subjected to filtering processing by using a window with fixed duration and a sliding window to generate a series of short-time Fourier spectrograms, and classifying and marking the spectrograms by combining the manually recorded aircraft passing time.
Preferably, the normalizing the short-time Fourier spectrum in the step 2.4 includesThe channels perform normalized calculations:
wherein the method comprises the steps of Represents the mean value/>Represents standard deviation,/>For the value before normalization,/>Is a normalized value.
Preferably, the improvement in said step 3The mechanism of attention of the network is/>Module, activation function is/>Function, said/>After the modules are subjected to global pooling with dimension reduction and unchanged channel dimension in space, capturing local cross-channel interaction through the relation between each channel and adjacent channels, ensuring efficiency and further improving training effect, wherein the channel is a channel with the same dimension as the channel with the same dimension, and the channel is a channel with the same dimension as the channel with the same dimension, wherein the channel with the same dimension is the same as the channel with the same dimension, and the channel with the same dimension is the channel with the same dimension as the channel with the same dimensionThe function is used after the linear layer to map the activated result to a probability distribution.
Preferably, the training and testing improvement in step 3The loss function used in the network is a binary cross entropy loss function, and the specific calculation formula is as follows:
Is a true label,/> For predicting the result, learning rate is employed/>And (5) adjusting by an optimizer.
Preferably, the step 4 includes that in the real-time processing, the vibration signal data recorded by the seismometer is processed in the same data processing steps as those in the steps 1 and 2, and then the aircraft signal recognition model obtained through training in the step 3 is used for judging whether the aircraft is detected, if so, the corresponding time of the short-time fourier spectrum is recorded as the time of the aircraft passing.
The beneficial effects of the invention include:
1. the seismograph has portability, high precision and strong anti-interference capability, and can accurately measure low-altitude aircraft vibration signals under complex environments and severe weather conditions.
2. The aircraft vibration signal does not have a significant degree of differentiation in waveform characteristics, and the present invention utilizes short-time fourier transform techniques to convert the vibration signal from the time-energy domain to the time-frequency domain. The short-time Fourier spectrogram of the converted aircraft vibration signal has obvious characteristics.
3. The invention utilizes a lightweight network in deep learningTraining is used as an identification model, on the premise of high accuracy, the calculation resources are saved, and the method is suitable for real-time identification of low-altitude airplane signals.
Drawings
FIG. 1 is a flow chart of an embodiment of a method for implementing real-time low-altitude aircraft identification using deep learning based on seismometers in accordance with the present invention.
Fig. 2 is a short-time fourier transform spectrum diagram of vibration signal generation in the second step according to example 1.
Fig. 3 is a schematic structural diagram of the aircraft signal recognition model used in the third step of embodiment 1.
Detailed Description
For the purpose of making the objects, technical solutions and advantages of the embodiments of the present application more apparent, the technical solutions of the embodiments of the present application will be clearly and completely described below with reference to the accompanying drawings in the embodiments of the present application, and it is apparent that the described embodiments are only some embodiments of the present application, not all embodiments. Thus, the following detailed description of the embodiments of the application, as presented in the figures, is not intended to limit the scope of the application, as claimed, but is merely representative of selected embodiments of the application. All other embodiments, which can be made by a person skilled in the art without making any inventive effort, are intended to be within the scope of the present application.
Example 1
A specific embodiment of the present invention will be described in detail with reference to fig. 1;
Step 1: by an array of seismic stations arranged at various locations of an airport, railway, etc. at 1000 Is recorded with the vibration signal data. The vibration signals recorded in a plurality of periods of time are collected, meanwhile, the passing time of an airplane and a train is manually recorded, and vibration signal data are selected according to the passing time of the airplane and the train, wherein the vibration signals of different types of airplane and the noise vibration signals of non-airplane when pedestrians pass under different working conditions, such as trains and pedestrians, are covered.
Step 2: the flow of making the aircraft signal data set by using the vibration signal data selected in the step 1 is shown in fig. 1, and specifically comprises the following steps:
(1) Third-order Butterworth 1 on collected vibration signal data High pass filtering removes baseline wander caused by low frequency signals.
(2) Because the aircraft signal is difficult to identify through waveform characteristics, the embodiment utilizes the short-time Fourier transform) Techniques to convert vibration signals from the time-energy domain to the time-frequency domain.
And carrying out sliding window sampling on the vibration signal data after filtering by using a window of forty seconds and a sliding window of five seconds to generate a series of short-time Fourier spectrograms, wherein the short-time Fourier spectrograms are shown in figure 2, and then classifying and marking the spectrograms according to the manually recorded aircraft passing time. Wherein the short-time fourier transform provides frequency domain information while preserving the characteristics of the signal over time, so that the time at which the aircraft signal occurs can be determined from the short-time fourier spectrum.
The short-time fourier transform is calculated as follows:
1. A time window function of length 0.256 seconds is selected and is point-wise multiplied with each local period of the signal.
2. Discrete fourier transforming the product signal）。
3. Moving the time window in an overlapping proportion of 8%, repeating steps 1 and 2 until the whole window signal is covered.
The specific formula is as follows:
Wherein: representing the time domain discrete points of the signal,/> As a window function,/>For the length of the window it is,For frequency index,/>Is a time index.
(3) According to the embodiment, the short-time Fourier spectrogram of the first 4/5 time and the short-time Fourier spectrogram of the last 1/5 time are extracted and put into the test set, the rest of data are used as the training set, and in general, the training set accounts for 70% and the test set accounts for 30%, so that the data with the front time and the rear time are reused for testing the data of the training model in the follow-up training model, and the prediction result can more objectively reflect the model prediction effect.
(4) Resampling the dataset image: and the short-time Fourier spectrogram is resampled by adopting a region interpolation method, so that the pixel size is 256×256, the calculation cost is reduced, the training of a subsequent model and the real-time processing of an airplane signal are facilitated, and the resampled spectrogram is standardized, so that the robustness of the model of the subsequent training is ensured. Wherein:
The resampling process is as follows: the image is divided into different regions using an area relationship, each region having an average pixel value, and the pixel value of each region is then copied in the new image according to the target size.
The spectrogram normalization process includes, for each pixel pointThe channels were normalized as follows:
wherein the method comprises the steps of Represents the mean value/>Representing standard deviation.
In particular, in the present embodiment, for each pixel pointThe channel, minus the mean (0.5,0.5,0.5), divided by the standard deviation (0.5,0.5,0.5), scales the pixel value range between [ -1, 1 ].
Step 3: the embodiment uses the improved typeThe model is shown in FIG. 3 as a base model,/>Is a lightweight neural network architecture commonly used for real-time image classification and object detection tasks on mobile devices and embedded systems. In order to achieve the goal of light weight,Many optimizations are made to the model architecture. First,/>Depth separable convolution is employed to divide the standard volume into two steps: depth convolution and point-by-point convolution. The depth convolution performs a convolution operation on the channel, while the point-by-point convolution performs a convolution operation in space. Doing so can significantly reduce the number of parameters and the amount of computation, thereby reducing the complexity of the model. Then/>The bottleneck structure is adopted, and the standard convolution is replaced by 1x1 convolution, so that the number of input and output channels is reduced, and the calculated amount is reduced. In addition, in the case of the optical fiber,An inverse residual structure is used, placing the nonlinear activation function after the bottleneck layer, rather than in the previous residual structure. The structure can reduce the calculation amount while increasing the depth of the model and improve the accuracy of the model.
In this embodiment, in order to further enhance the model training effect and ensure the weight reduction of the model, use is made of(Effective attention mechanism Module) instead of/>/>, In the model(Channel attention mechanism) module. Specifically, given an input feature,/>The module first pools globally for each channel individually, then uses two nonlinear fully-connected layers and one/>The function generates channel weights. Two/>The layers are intended to capture nonlinear cross-channel interactions and control model complexity by dimension reduction. And/>After the modules are subjected to global pooling with dimension reduction and unchanged channel dimension in space, local cross-channel interaction is captured through the relation between each channel and adjacent channels, so that the training effect is further improved while the efficiency is ensured. After the improvement, the output result is adjusted to 2 through one full-connection layer, and the output result is used after the linear layerThe function maps the activated result to a probability distribution. The network configuration of the model is specifically shown in table 1.
First row: the input size is 224x224 and the number of channels is 3. The operation used is a 3x3 convolution @3X 3), the number of output channels is 16. The stride is 2.
Second row: the input size becomes 112x112 through the previous step, and the number of channels becomes 16. The operation of the method is that the bottleneck module is%3X 3), the expansion size is 16, and the number of output channels is 16. Here, a stride of 2 is used, and the activation function type is RE (ReLU).
Third row: the input size becomes 56x56, leaving 16 channels unchanged. The operation of the application is still a bottleneck module3X 3), the expansion size was 72, and the number of output channels increased to 24. The stride is 2.
Fourth and fifth rows: the input size is still 28x28, the number of channels is 24. All the bottleneck modules are used3X 3), but the expansion sizes are different (88 and 96), the number of output channels remains unchanged. The steps are 1 and 2, respectively.
From the sixth row up to the last row of the table, each layer contains similar information: input size, operation type, dilation size, number of output channels, attention mechanism, activation function, stride.
Last two rows of'1x1, />"Means a convolution operation using a 1x1 size,/>Indicating no batch normalization (/ >)). Therein, "/>"Represents the final output class number, usually set at the back of the network, to accommodate different class tasks.
Table 1 improved generationNetwork configuration
In the model training process, the loss function used is a binary cross entropy loss function to measure the difference between the model prediction result and the real label. The specific calculation formula is as follows:
By using The optimizer adjusts the learning rate.
Specifically, the test results of the trained aircraft signal recognition model on the test set are shown in table 2:
table 2 model test results
Step 4: in real-time processing, the vibration signal data recorded by the seismometer are subjected to the same data processing steps as in the steps 1 and 2, short-time Fourier spectrum is generated by high-pass filtering in sequence, whether the aircraft is detected or not is judged by the aircraft signal recognition model obtained through the training in the step 3 after resampling and standardization processing, and if the aircraft is judged to pass, the time of the corresponding short-time Fourier spectrum is recorded as the time of the aircraft.
The above examples merely illustrate specific embodiments of the application, which are described in more detail and are not to be construed as limiting the scope of the application. It should be noted that it is possible for a person skilled in the art to make several variants and modifications without departing from the technical idea of the application, which fall within the scope of protection of the application.
Claims (6)
1. A low-altitude airplane real-time identification method combining seismograph with deep learning is characterized by comprising the following steps:
step 1: collecting original vibration signal data through a laid seismograph;
Step 2: sequentially carrying out filtering, short-time Fourier transformation, resampling and standardization on original vibration signal data to obtain a data set, wherein the data set comprises a training set and a testing set;
the step 2 comprises the following steps:
Step 2.1: performing high-pass filtering on the original vibration signal data collected in the step 1;
step 2.2: processing the vibration signal data after filtering into a short-time Fourier spectrogram by using short-time Fourier transform;
step 2.3: resampling the short-time Fourier spectrogram by adopting a regional interpolation method;
Step 2.4: normalizing the resampled short-time Fourier spectrogram;
Step 2.5: extracting a part of short-time Fourier spectrogram as a test set according to time, and taking the rest part as a training set, wherein the proportion of the short-time Fourier spectrogram in the first half period in the test set is smaller than that in the second half period;
Step 3: training and testing the improved MobileNetV-Smal network based on the data set as an aircraft signal recognition model;
The attention mechanism of the improved Mobi leNetV-Smal network in the step 3 is an ECA module, the activation function is a Softmax function, after the ECA module is subjected to global pooling with unchanged channel dimension and dimension in space, local cross-channel interaction is captured through the relation between each channel and adjacent channels, and the Softmax function is used for mapping the activated result into probability distribution after a linear layer;
step 4: and (3) carrying out real-time identification on the low-altitude airplane signals by using the airplane signal identification model obtained in the step (3).
2. The method for real-time identification of low-altitude aircraft by combining seismograph with deep learning according to claim 1, wherein the step 1 includes: and recording vibration signal data through the arranged earthquake station array, simultaneously manually recording the passing time of the airplane, selecting the vibration signal data according to the passing time of the airplane, wherein the selected vibration signal data cover airplane vibration signals of different types and non-airplane noise vibration signals of different working conditions.
3. The method for real-time identification of low-altitude aircraft by combining seismograph with deep learning according to claim 1, wherein the step 2.2 includes: and carrying out sliding window sampling on the vibration signal data subjected to filtering processing by using a window with fixed duration and a sliding window to generate a series of short-time Fourier spectrograms, and classifying and marking the spectrograms by combining the manually recorded aircraft passing time.
4. The method for real-time recognition of a low-altitude aircraft by combining seismograph with deep learning according to claim 1, wherein the step 2.4 of normalizing the short-time fourier spectrum map includes performing normalized calculation on RGB channels of each pixel point:
new_value＝(old_value-Mean)/Standard_Deviation
Where Mean represents Mean, standard_Deviation represents Standard Deviation, old_value is the value before normalization, new_value is the value after normalization.
5. The method for identifying the low-altitude aircraft by combining the seismograph with the deep learning in real time according to claim 1, wherein a loss function used in training and testing the improved MobileNetV-Smal network in the step 3 is a binary cross entropy loss function, and a specific calculation formula is as follows:
loss＝-(y_true*log(y_pred)+(1-y_true)*log(1-y_pred))
y_true is a real label, y_pred is a prediction result, and the learning rate is adjusted by an Adam optimizer.
6. The method for recognizing the low-altitude aircraft by combining the seismograph with the deep learning according to claim 1, wherein the step 4 is characterized in that in the real-time processing, vibration signal data recorded by the seismograph is processed in the same manner as the steps 1 and 2, and then whether the aircraft is detected is judged by the aircraft signal recognition model obtained by training in the step 3, and if the aircraft is judged to pass, the time of the corresponding short-time Fourier spectrum is recorded as the time of the aircraft passing.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410377473.7A CN117972364B (en) | 2024-03-29 | 2024-03-29 | Low-altitude airplane real-time identification method combining seismograph with deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202410377473.7A CN117972364B (en) | 2024-03-29 | 2024-03-29 | Low-altitude airplane real-time identification method combining seismograph with deep learning |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN117972364A CN117972364A (en) | 2024-05-03 |
| CN117972364B true CN117972364B (en) | 2024-06-18 |
Family
ID=90851836
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202410377473.7A Active CN117972364B (en) | 2024-03-29 | 2024-03-29 | Low-altitude airplane real-time identification method combining seismograph with deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN117972364B (en) |
Family Cites Families (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20150085604A1 (en) * | 2013-09-20 | 2015-03-26 | Ubiterra Corporation | Apparatus, system, and method for real-time seismic data acquisition management |
| CN112560913B (en) * | 2020-12-04 | 2023-11-03 | 东北大学 | Industrial equipment fault signal positioning and identifying method based on microphone array |
| CN114993444B (en) * | 2022-05-09 | 2023-03-17 | 武汉理工大学 | Airport runway safety monitoring method and system based on grating sensor array |
| CN115355166A (en) * | 2022-08-30 | 2022-11-18 | 杭州展德软件技术有限公司 | Air compressor fault diagnosis method and system based on short-time Fourier transform |
-
2024
- 2024-03-29 CN CN202410377473.7A patent/CN117972364B/en active Active
Non-Patent Citations (2)
| Title |
|---|
| Automatic recognition and classification of microseismic waveforms based on computer vision;Jiaming Li等;《Tunnelling and Underground Space Technology incorporating Trenchless Technology Research》;20211223;第121卷;第1-14页 * |
| 基于声震融合的低空涡扇飞机探测识别方法研究;郑晶涵;《中国优秀硕士学位论文全文数据库 工程科技Ⅱ辑》;20220115(第1期);第C031-600页 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN117972364A (en) | 2024-05-03 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Stork et al. | Application of machine learning to microseismic event detection in distributed acoustic sensing data | |
| US4906940A (en) | Process and apparatus for the automatic detection and extraction of features in images and displays | |
| US5373486A (en) | Seismic event classification system | |
| Sueur et al. | Global estimation of animal diversity using automatic acoustic sensors | |
| Jin et al. | Vehicle classification based on seismic signatures using convolutional neural network | |
| CN112735473B (en) | Method and system for identifying unmanned aerial vehicle based on voice | |
| CN114266892A (en) | Pavement disease identification method and system for multi-source data deep learning | |
| CN114966899B (en) | Regional visibility prediction method based on multi-source multi-element remote sensing technology cooperation | |
| CN111983676A (en) | Earthquake monitoring method and device based on deep learning | |
| van den Ende et al. | Deep deconvolution for traffic analysis with distributed acoustic sensing data | |
| Meyer et al. | Systematic identification of external influences in multi-year microseismic recordings using convolutional neural networks | |
| Martínez et al. | Advanced signal recognition methods applied to seismo-volcanic events from Planchon Peteroa Volcanic Complex: Deep Neural Network classifier | |
| CN115826042A (en) | Edge-cloud-combined distributed seismic data processing method and device | |
| CN113640768B (en) | Low-resolution radar target identification method based on wavelet transformation | |
| CN109409285B (en) | Remote sensing video target detection method based on overlapped slices | |
| CN114417908A (en) | Multi-mode fusion-based unmanned aerial vehicle detection system and method | |
| CN117972364B (en) | Low-altitude airplane real-time identification method combining seismograph with deep learning | |
| CN109815773A (en) | A kind of low slow small aircraft detection method of view-based access control model | |
| CN115331096B (en) | Mining subsidence area identification method, system, storage medium and electronic equipment | |
| Dorian et al. | Bi-class classification of humpback whale sound units against complex background noise with Deep Convolution Neural Network | |
| CN116910690A (en) | Target classification system based on data fusion | |
| CN113553912A (en) | Unmanned aerial vehicle image processing method based on RGB distribution | |
| CN113343930A (en) | Unmanned aerial vehicle image processing method based on Gaussian denoising | |
| KR102672778B1 (en) | Method and apparatus for identifying aircraft | |
| Nugroho et al. | Seismic Data Quality Analysis Based on Image Recognition Using Convolutional Neural 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 |