CN109712183A - Electronic speckle interference intelligent information retrieval method based on deep learning - Google Patents
Electronic speckle interference intelligent information retrieval method based on deep learning Download PDFInfo
- Publication number
- CN109712183A CN109712183A CN201811436904.3A CN201811436904A CN109712183A CN 109712183 A CN109712183 A CN 109712183A CN 201811436904 A CN201811436904 A CN 201811436904A CN 109712183 A CN109712183 A CN 109712183A
- Authority
- CN
- China
- Prior art keywords
- network
- training
- image
- test
- deep learning
- 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
- 238000000034 method Methods 0.000 title claims abstract description 58
- 238000013135 deep learning Methods 0.000 title claims abstract description 12
- 238000012549 training Methods 0.000 claims abstract description 41
- 238000012360 testing method Methods 0.000 claims abstract description 21
- 238000013527 convolutional neural network Methods 0.000 claims abstract description 16
- 230000006870 function Effects 0.000 claims abstract description 13
- 238000000605 extraction Methods 0.000 claims abstract description 8
- 230000004913 activation Effects 0.000 claims abstract description 6
- 230000000644 propagated effect Effects 0.000 claims abstract description 6
- 238000013519 translation Methods 0.000 claims abstract description 5
- 230000008569 process Effects 0.000 claims description 23
- 239000000284 extract Substances 0.000 abstract description 7
- 238000013459 approach Methods 0.000 description 4
- 238000007781 pre-processing Methods 0.000 description 4
- 238000002474 experimental method Methods 0.000 description 3
- 230000003287 optical effect Effects 0.000 description 3
- 238000005457 optimization Methods 0.000 description 3
- 238000011160 research Methods 0.000 description 3
- 238000013528 artificial neural network Methods 0.000 description 2
- 238000004422 calculation algorithm Methods 0.000 description 2
- 238000011161 development Methods 0.000 description 2
- 230000018109 developmental process Effects 0.000 description 2
- 238000004458 analytical method Methods 0.000 description 1
- 238000013473 artificial intelligence Methods 0.000 description 1
- 230000006399 behavior Effects 0.000 description 1
- 230000009286 beneficial effect Effects 0.000 description 1
- 230000015572 biosynthetic process Effects 0.000 description 1
- 230000000052 comparative effect Effects 0.000 description 1
- 238000005094 computer simulation Methods 0.000 description 1
- 238000013461 design Methods 0.000 description 1
- 238000001514 detection method Methods 0.000 description 1
- 238000010586 diagram Methods 0.000 description 1
- 238000006073 displacement reaction Methods 0.000 description 1
- 230000000694 effects Effects 0.000 description 1
- 238000001914 filtration Methods 0.000 description 1
- 238000005755 formation reaction Methods 0.000 description 1
- 238000011478 gradient descent method Methods 0.000 description 1
- 238000003709 image segmentation Methods 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 238000005305 interferometry Methods 0.000 description 1
- 238000010801 machine learning Methods 0.000 description 1
- 238000013507 mapping Methods 0.000 description 1
- 238000012986 modification Methods 0.000 description 1
- 230000004048 modification Effects 0.000 description 1
- 230000010363 phase shift Effects 0.000 description 1
- 238000002203 pretreatment Methods 0.000 description 1
- 238000012545 processing Methods 0.000 description 1
- 230000009467 reduction Effects 0.000 description 1
- 238000007616 round robin method Methods 0.000 description 1
- 239000002689 soil Substances 0.000 description 1
- 238000010998 test method Methods 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
- 230000000007 visual effect Effects 0.000 description 1
- XLYOFNOQVPJJNP-UHFFFAOYSA-N water Substances O XLYOFNOQVPJJNP-UHFFFAOYSA-N 0.000 description 1
- 239000002023 wood Substances 0.000 description 1
Abstract
The electronic speckle interference intelligent information retrieval method based on deep learning that the invention discloses a kind of: initialization network parameter provides the size, convolution kernel number, translation stride of convolution kernel；Training image is input to U-net convolutional neural networks；Training image realizes feature extraction by convolution in a network, and feature is transmitted to next layer by weighted sum bias operation, and using activation primitive；The update of weight is carried out come the loss of training of judgement image by loss function；It reaches iteration maximum times or obtains network optimal weights, then deconditioning；Test image is input to trained network；Output data is obtained by propagated forward；It is compared with label value, statistical result；The statistical result for completing all test images, exits network.The present invention passes through using coding-decoder architecture U-net convolutional neural networks training data is based on, and obtains optimal skeleton and extracts model, to realize that intelligent ESPI striped picture skeleton line extracts.
Description
Technical field
The invention belongs to optical image securities and machine learning field, and more specifically, it relates to one kind to be based on depth
The electronic speckle interference intelligent information retrieval method of habit.
Background technique
Electronic speckle pattern interferometry is a kind of modern optical method to grow up on the basis of modern high technology achievement,
It has many advantages, such as the whole audience, non-contact, high-precision and highly sensitive, quickly in real time and can on-line checking.The technology is in mechanical, soil
There is very important status in the detection in the fields such as wood, water conservancy, electric appliance, aerospace, weapon industry and biomedicine.
The key that applying electronic speckle interference technology obtains object deformation displacement information is accurately to extract phase.Currently,
There are three types of main phase extraction methods: phase shift method, Fourier transform and striped skeleton method, in addition, there are also space phases to examine
Survey method, PGC demodulation round-robin method etc..In above-mentioned phase extraction method, the skeleton method based on electronic speckle bar graph is to extract to do
Relate to the most straightforward approach of phase.Traditional skeleton technique needs in order to automatically extract the skeleton of striped to optical interference
It is filtered noise reduction and improves contrast processing, recycle refinement or peak time tracking method etc. then to detect the central point of striped.
In recent years, with scientific and technological progress, artificial intelligence field has obtained development at full speed, deep learning especially convolution mind
The a large amount of concerns and research of domestic and foreign scholars have been obtained through network.Intelligent algorithm based on convolutional neural networks has been applied to perhaps
It is multi-field.Compared to traditional-handwork design feature, convolutional neural networks simplify the process of image preprocessing, in training data
Driving is lower to be described characteristics of image using network end to end.Energy is extracted using the very strong nonlinear characteristic of neural network
Power, this method can learn the information important for sample data automatically from training data, and convolutional neural networks can be layer-by-layer
The feature of image layer from low to high is extracted in training, to have preferable scale invariability and adaptivity.Convolutional neural networks
Started one research boom in industry and academia, for computer vision field bring unprecedented life and
It changes.
Summary of the invention
Purpose of the invention is to overcome the shortcomings in the prior art, it is intended to realize based on convolutional neural networks method
Electronic speckle interference fringe pattern skeleton line drawing simplifies image preprocessing process, so that the skeleton line extracted is more accurate, proposes
A kind of electronic speckle interference intelligent information retrieval method based on deep learning, by using based on coding-decoder architecture
U-net convolutional neural networks training data, the optimization of model parameter is independently carried out by successive ignition, is obtained optimal skeleton and is mentioned
Modulus type, to realize that intelligent ESPI striped picture skeleton line extracts.
The purpose of the present invention is what is be achieved through the following technical solutions.
Electronic speckle interference intelligent information retrieval method based on deep learning of the invention, comprising the following steps:
Step 1: the training process of network
(1) initialization network parameter provides the size, convolution kernel number, translation stride parameter of convolution kernel；
(2) training image is input to U-net convolutional neural networks；
(3) training image realizes feature extraction by convolution in a network, by weighted sum bias operation, and utilizes activation
Feature is transmitted to next layer by function；
(4) update of weight is carried out come the loss of training of judgement image by loss function；
(5) iteration maximum times are reached or obtain network optimal weights, then deconditioning；
Step 2: the test process of network
(1) test image is input to trained network；
(2) output data is obtained by propagated forward；
(3) it is compared with label value, statistical result；
(4) statistical result for completing all test images, exits network.
Compared with prior art, the beneficial effects brought by the technical solution of the present invention are as follows:
There are pre-treatment step complexity, extraction accuracy be not high for electronic speckle interference fringe pattern skeleton line drawing by the present invention
The problem of, the present invention is based on convolutional neural networks to propose a kind of new method of intelligent extraction skeleton line, and this method utilizes convolution
The multiscale analysis characteristic of neural network, depth are excavated the mapping between electronic speckle interference fringe pattern and stripe fixed position and are closed
System, the further feature of image is combined with outside visual information.Compared to traditional skeleton line extracting method, which can
It is significantly reduced the process of image preprocessing, and can accurately extract the electronic speckle interference fringe pattern with much noise
Skeleton line.
Detailed description of the invention
Fig. 1 is network training flow chart；
Fig. 2 is the big variation density of electronic speckle interference fringe pattern that width experiment obtains；
Fig. 3 is the skeleton line that Coupled PDE method obtains；
Fig. 4 is the skeleton line that direction Coupled PDE obtains；
Fig. 5 is the skeleton line that anisotropic approaches obtain；
Fig. 6 is the skeleton line that method proposed by the present invention obtains.
Specific embodiment
The invention will be further described with reference to the accompanying drawing.
The present invention is directed to realize the electronic speckle interference fringe pattern skeleton line drawing based on convolutional neural networks method, simplify
Image preprocessing process, so that the skeleton line extracted is more accurate.The technical solution adopted by the present invention is that by using based on volume
The U-net convolutional neural networks training data of code-decoder architecture, the optimization of model parameter is independently carried out by successive ignition,
It obtains optimal skeleton and extracts model, to realize that intelligent ESPI striped picture skeleton line extracts.
Electronic speckle interference intelligent information retrieval method based on deep learning of the invention, is based on deep learning
The research of ESPI skeleton line extracting method mainly includes two parts content: the training process of network and the test process of network.Instruction
Practice network development process and is divided into forward data transmitting and reversed error propagation two parts.Compared to training process, the test process of network
There is no reversed error propagation process, does not need to calculate loss function, but need to calculate the classification accuracy of network to verify net
The performance of network.The key step of the training process of network and the test process of network is given below:
Step 1: the training process of network
(1) initialization network parameter provides the parameters such as the size, convolution kernel number, translation stride of convolution kernel；
(2) training image is input to U-net convolutional neural networks；
(3) training image realizes feature extraction by convolution in a network, by weighted sum bias operation, and utilizes activation
Feature is transmitted to next layer by function；
(4) update of weight is carried out come the loss of training of judgement image by loss function；
(5) iteration maximum times are reached or obtain network optimal weights, then deconditioning.
Step 2: the test process of network
(1) test image is input to trained network；
(2) output data is obtained by propagated forward；
(3) it is compared with label value, statistical result；
(4) statistical result for completing all test images, exits network.
Next the electronic speckle interference intelligent information retrieval method method proposed by the present invention based on deep learning is applied
Skeleton line drawing is carried out to ESPI bar graph.Specific step is as follows:
Step 1: the training process of network
Step 1: initialization network parameter: selecting the convolution kernel of four kinds of different scales, convolution kernel size difference in the invention
Are as follows: 1 × 1,3 × 3,5 × 5,7 × 7；When propagated forward, the height of the filter of each convolution kernel in input feature vector figure
It is translated on degree and width, the translation step size designed in the present invention is 1；And provide maximum number of iterations.
Step 2: piecemeal being carried out to training image and corresponding skeleton line chart, due to obtaining electronic speckle interference in an experiment
Striped is relatively difficult, so the input picture designed in the present invention is small ESPI image block, which is comprising pixel neighbour
The super-pixel of domain structure information, after other nonlinear operations such as a series of convolution, pond, exporting is corresponding point
The probability of class classification.Assuming that the size of original image is W × H, the size of image block is p × p, and one is chosen in original image
Extracted region images block, wherein abscissa range beOrdinate range isIn above-mentioned zone
Center of the n point as image block is randomly selected, then cuts out the n image blocks having a size of p × p as training sample.Wherein,
The center of selected image block need to ensure in ESPI stripe pattern effective coverage.We choose 1/10 work of training image piecemeal
For verifying collection, remaining sample is as training set.
Step 3: the more sizes of training input convolutional neural networks model.What the present invention selected is based on coding-decoder knot
The U-net network model of structure, the model have been achieved for good effect in medical image segmentation field, which includes
Cataloged procedure and decoding process, the two form U-shaped symmetrical structure.Meanwhile having between coding and the characteristic layer of decoding process generation
There is jump connection structure, this connection type can combine the local detail information of bottom with top layer global characteristics.In the model
Middle there are two class formations: convolutional layer and pond layer.Convolutional layer is responsible for extracting the characteristic information of image, the input picture in first layer
Convolution algorithm, which is carried out, with 4 convolution kernels mentioned in step 1 obtains 4 kinds of characteristic informations.Pass through biasing and activation primitive behaviour later
Make to enter pond layer, present invention selection: ReLu activation primitive.The forward direction transmitting of data is realized by convolutional layer and pond layer.
Step 4: the loss that training image is calculated by loss function realizes that reversed weight updates, and the present invention selects:
Softmax function is as loss function.The cross entropy loss function of Softmax can incessantly be updated weighting parameter
Optimization, and when actually calculating, keep value more stable since Softmax classifier contains normalized operation.
Step 5: reaching iteration maximum times or obtain network optimal weights, then deconditioning.
In the training of convolutional neural networks model, the training of multiple iteration cycles is needed, network can just converge to training
Collection, if continuing training pattern, network will generate different degrees of over-fitting.The present invention is enhanced and is increased by data
The mode of layer living prevents over-fitting.This model optimizes parameter in back-propagation process by gradient descent method, works as error
It is gradually reduced when tending towards stability, network, which can consider, has restrained, and the corresponding label probability figure of output ESPI bar graph obtains net
The optimal weights and offset of network save as the training pattern of the database, or when training pattern reaches preset iteration time
Number, deconditioning.
Step 2: the test process of network, specific embodiment:
Step 1: test image being input to and has trained mature network；
Step 2: output data is obtained by propagated forward；
Step 3: being compared with label value, statistical result；
Step 4: completing the statistical result of all test images, exit network.
The test of the model and training flow chart are shown in attached drawing 1.
For the validity of verification method, experimental results.
The model structure that the present invention tests is built based on TensorFlow and Keras deep learning frame, is emulated
Server hardware configuration used: CPU is Intel Xeon CPU E5-1650, RAM 16G, GPU are NVIDIA Quadro K
2200；Software environment: Ubuntu16.04 system, TensorFlow and Keras deep learning frame.
The experimental data that the present invention uses is 50 electronic speckle interference fringe patterns of computer simulation, and picture size is equal
Are as follows: 512 × 512 pixels.Every bar graph all corresponds to a skeleton line chart.By dicing treatment we available 50000
Sample image, wherein 40000 training samples the most, 10000 are used as test sample.By training and the present invention after test
Method can accurately obtain skeleton line.
For test validity, the mentioned method of this patent and Coupled PDE method, direction Coupled PDE
Method and anisotropy parameter skeleton line extracting method are made comparisons.Fig. 2 is that the big variation density of electronic that width experiment obtains dissipates
Spot interference fringe picture, Fig. 3 are the skeleton lines that Coupled PDE method obtains, and Fig. 4 is that direction Coupled PDE obtains
Skeleton line, Fig. 5 is the skeleton line that anisotropic approaches obtain, and Fig. 6 is the skeleton line that method proposed by the present invention obtains.It is above-mentioned
Three kinds of comparative approach have all done the pretreatment of image filtering before application, but still it can be seen that there are bifurcated, bridging,
Phenomena such as disconnected.
Although function and the course of work of the invention are described above in conjunction with attached drawing, the invention is not limited to
Above-mentioned concrete function and the course of work, the above mentioned embodiment is only schematical, rather than restrictive, ability
The those of ordinary skill in domain under the inspiration of the present invention, is not departing from present inventive concept and scope of the claimed protection situation
Under, many forms can also be made, all of these belong to the protection of the present invention.
It will be appreciated by those skilled in the art that attached drawing is the schematic diagram of a preferred embodiment, the embodiments of the present invention
Serial number is for illustration only, does not represent the advantages or disadvantages of the embodiments.
The foregoing is merely a prefered embodiment of the invention, is not intended to limit the invention, all in the spirit and principles in the present invention
Within, any modification, equivalent replacement, improvement and so on should be included within the scope of the present invention.
Claims (1)
1. a kind of electronic speckle interference intelligent information retrieval method based on deep learning, which comprises the following steps:
Step 1: the training process of network
(1) initialization network parameter provides the size, convolution kernel number, translation stride parameter of convolution kernel；
(2) training image is input to U-net convolutional neural networks；
(3) training image realizes feature extraction by convolution in a network, by weighted sum bias operation, and utilizes activation primitive
Feature is transmitted to next layer；
(4) update of weight is carried out come the loss of training of judgement image by loss function；
(5) iteration maximum times are reached or obtain network optimal weights, then deconditioning；
Step 2: the test process of network
(1) test image is input to trained network；
(2) output data is obtained by propagated forward；
(3) it is compared with label value, statistical result；
(4) statistical result for completing all test images, exits network.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811436904.3A CN109712183A (en) | 2018-11-28 | 2018-11-28 | Electronic speckle interference intelligent information retrieval method based on deep learning |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201811436904.3A CN109712183A (en) | 2018-11-28 | 2018-11-28 | Electronic speckle interference intelligent information retrieval method based on deep learning |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN109712183A true CN109712183A (en) | 2019-05-03 |
Family
ID=66255096
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201811436904.3A Pending CN109712183A (en) | 2018-11-28 | 2018-11-28 | Electronic speckle interference intelligent information retrieval method based on deep learning |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN109712183A (en) |
Cited By (7)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110472637A (en) * | 2019-07-29 | 2019-11-19 | 天津大学 | Deep learning variable density low quality electronic speckle stripe direction extracting method |
| CN111696632A (en) * | 2020-06-22 | 2020-09-22 | 钢铁研究总院 | Full-field quantitative statistical distribution characterization method for gamma' phase microstructure in metal material |
| CN111812647A (en) * | 2020-07-11 | 2020-10-23 | 桂林电子科技大学 | Phase unwrapping method for interferometric synthetic aperture radar |
| CN112033280A (en) * | 2020-09-03 | 2020-12-04 | 合肥工业大学 | Speckle interference phase calculation method combining Fourier transform model and deep learning |
| CN112101362A (en) * | 2020-08-25 | 2020-12-18 | 中国科学院空间应用工程与技术中心 | Semantic segmentation method and system for space science experimental data |
| CN112308863A (en) * | 2020-10-27 | 2021-02-02 | 苏州大学 | Method for segmenting myopic macular degeneration area in retina OCT image based on improved U-shaped network |
| CN113312848A (en) * | 2021-06-10 | 2021-08-27 | 太原理工大学 | Intelligent design method of optical system with adaptive target information extraction algorithm as target |
Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103514604A (en) * | 2013-10-08 | 2014-01-15 | 天津工业大学 | Method for extracting skeleton line of electronic speckle interference fringe image |
| CN104268837A (en) * | 2014-09-26 | 2015-01-07 | 天津工业大学 | Method for extracting phase position information of electronic speckle interference fringe pattern |
| CN106289051A (en) * | 2016-08-02 | 2017-01-04 | 天津大学 | The direction of big change density of electronic speckle interference fringe pattern and density processing method |
| CN106447626A (en) * | 2016-09-07 | 2017-02-22 | 华中科技大学 | Blurred kernel dimension estimation method and system based on deep learning |
| CN108734661A (en) * | 2018-05-25 | 2018-11-02 | 南京信息工程大学 | High-definition picture prediction technique based on image texture information architecture loss function |
-
2018
- 2018-11-28 CN CN201811436904.3A patent/CN109712183A/en active Pending
Patent Citations (5)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN103514604A (en) * | 2013-10-08 | 2014-01-15 | 天津工业大学 | Method for extracting skeleton line of electronic speckle interference fringe image |
| CN104268837A (en) * | 2014-09-26 | 2015-01-07 | 天津工业大学 | Method for extracting phase position information of electronic speckle interference fringe pattern |
| CN106289051A (en) * | 2016-08-02 | 2017-01-04 | 天津大学 | The direction of big change density of electronic speckle interference fringe pattern and density processing method |
| CN106447626A (en) * | 2016-09-07 | 2017-02-22 | 华中科技大学 | Blurred kernel dimension estimation method and system based on deep learning |
| CN108734661A (en) * | 2018-05-25 | 2018-11-02 | 南京信息工程大学 | High-definition picture prediction technique based on image texture information architecture loss function |
Non-Patent Citations (6)
| Title |
|---|
| OLAF RONNEBERGER ET AL.: "U-Net: Convolutional Networks for Biomedical Image Segmentation", pages 234 - 241 * |
| 唐晨 等: "基于变分图像分解的电子散斑干涉信息提取方法", no. 03 * |
| 唐晨 等: "电子散斑干涉条纹处理偏微分方程方法的回顾与展望", no. 02 * |
| 王海涛 等: "基于条纹中心线法的电子散斑干涉条纹信息提取", no. 01 * |
| 金雪峰 等: "电子散斑干涉条纹骨架线提取的新方法", no. 02 * |
| 韩成: "《基于结构光的计算机视觉》", 国防工业出版社, pages: 69 * |
Cited By (12)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN110472637A (en) * | 2019-07-29 | 2019-11-19 | 天津大学 | Deep learning variable density low quality electronic speckle stripe direction extracting method |
| CN111696632A (en) * | 2020-06-22 | 2020-09-22 | 钢铁研究总院 | Full-field quantitative statistical distribution characterization method for gamma' phase microstructure in metal material |
| CN111696632B (en) * | 2020-06-22 | 2023-10-10 | 钢铁研究总院有限公司 | Method for characterizing full-view-field quantitative statistical distribution of gamma' -phase microstructure in metal material |
| CN111812647A (en) * | 2020-07-11 | 2020-10-23 | 桂林电子科技大学 | Phase unwrapping method for interferometric synthetic aperture radar |
| CN111812647B (en) * | 2020-07-11 | 2022-06-21 | 桂林电子科技大学 | Phase unwrapping method for interferometric synthetic aperture radar |
| CN112101362A (en) * | 2020-08-25 | 2020-12-18 | 中国科学院空间应用工程与技术中心 | Semantic segmentation method and system for space science experimental data |
| CN112033280A (en) * | 2020-09-03 | 2020-12-04 | 合肥工业大学 | Speckle interference phase calculation method combining Fourier transform model and deep learning |
| CN112033280B (en) * | 2020-09-03 | 2021-09-24 | 合肥工业大学 | Speckle interference phase calculation method combining Fourier transform model and deep learning |
| CN112308863A (en) * | 2020-10-27 | 2021-02-02 | 苏州大学 | Method for segmenting myopic macular degeneration area in retina OCT image based on improved U-shaped network |
| CN112308863B (en) * | 2020-10-27 | 2023-06-06 | 苏州大学 | OCT (optical coherence tomography) image myopic macular lesion segmentation method based on improved U-shaped network |
| CN113312848A (en) * | 2021-06-10 | 2021-08-27 | 太原理工大学 | Intelligent design method of optical system with adaptive target information extraction algorithm as target |
| CN113312848B (en) * | 2021-06-10 | 2022-10-04 | 太原理工大学 | Intelligent design method of optical system with adaptive target information extraction algorithm as target |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN109712183A (en) | Electronic speckle interference intelligent information retrieval method based on deep learning | |
| CN110472483A (en) | A kind of method and device of the small sample semantic feature enhancing towards SAR image | |
| CN108229444A (en) | A kind of pedestrian's recognition methods again based on whole and local depth characteristic fusion | |
| CN108960086A (en) | Based on the multi-pose human body target tracking method for generating confrontation network positive sample enhancing | |
| Li et al. | Ship motion prediction using dynamic seasonal RvSVR with phase space reconstruction and the chaos adaptive efficient FOA | |
| CN110120020A (en) | A kind of SAR image denoising method based on multiple dimensioned empty residual error attention network | |
| CN108399248A (en) | A kind of time series data prediction technique, device and equipment | |
| CN108509910A (en) | Deep learning gesture identification method based on fmcw radar signal | |
| CN109086700A (en) | Radar range profile's target identification method based on depth convolutional neural networks | |
| CN105844627B (en) | A kind of sea-surface target image background suppressing method based on convolutional neural networks | |
| CN110033473A (en) | Motion target tracking method based on template matching and depth sorting network | |
| CN108182450A (en) | A kind of airborne Ground Penetrating Radar target identification method based on depth convolutional network | |
| CN107529650A (en) | The structure and closed loop detection method of network model, related device and computer equipment | |
| CN110033440A (en) | Biological cell method of counting based on convolutional neural networks and Fusion Features | |
| CN109522855A (en) | In conjunction with low resolution pedestrian detection method, system and the storage medium of ResNet and SENet | |
| Liu et al. | Coastline extraction method based on convolutional neural networks—A case study of Jiaozhou Bay in Qingdao, China | |
| CN110163213A (en) | Remote sensing image segmentation method based on disparity map and multiple dimensioned depth network model | |
| CN108879732A (en) | Transient stability evaluation in power system method and device | |
| Liu et al. | Image edge recognition of virtual reality scene based on multi-operator dynamic weight detection | |
| CN109490861A (en) | A kind of blast-furnace line extracting method | |
| Chen et al. | A wavelet based deep learning method for underwater image super resolution reconstruction | |
| CN115310515A (en) | Fault-labeled seismic data sample set amplification method based on generation countermeasure network | |
| CN108230313A (en) | Based on the adaptively selected SAR image object detection method with discrimination model of component | |
| Majewski et al. | Developing particle image velocimetry software based on a deep neural network | |
| CN117079132A (en) | Remote sensing image target detection method based on Gaussian distance loss |
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 | ||
| WD01 | Invention patent application deemed withdrawn after publication |
Application publication date: 20190503 |