CN108733909A - A kind of melting area detection method - Google Patents
A kind of melting area detection method Download PDFInfo
- Publication number
- CN108733909A CN108733909A CN201810468869.7A CN201810468869A CN108733909A CN 108733909 A CN108733909 A CN 108733909A CN 201810468869 A CN201810468869 A CN 201810468869A CN 108733909 A CN108733909 A CN 108733909A
- Authority
- CN
- China
- Prior art keywords
- detection method
- melting area
- module
- area according
- time series
- 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
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
Abstract
The invention belongs to polar region change detection fields.The present invention discloses a kind of melting area detection method, using depth network model, melting area is detected based on annual daily microwave radiometer bright temperature data, precision is substantially improved compared with conventional model, and there is the adaptive ability of height to different ice and snow types, good basis can be provided for acquisition of information such as polar region freeze thawing regional change, freezing and thawing cycle variations.
Description
Technical field
The present invention relates to polar region change detection fields, and in particular to a kind of melting area detection method.
Background technology
Ice sheet freeze thawing is to pay close attention to problem in the Changeement of polar region.The bright temperature time series data usually quilt of microwave radiometer
It is divided for the region melted in 1 year to ice sheet and the region that do not melt.Conventional method can generally melt
Change the jump signal for finding timing curve near season, and is judged whether the region occurs to melt by threshold value.And
There may be significant differences for the property of different zones ice and snow, therefore cause thawing signal spatially multifarious so that tradition
Method detection accuracy is extremely limited.
Invention content
To solve the above-mentioned problems, the present invention discloses a kind of melting area detection method, includes the following steps：It inputs micro-
The bright temperature time series data of wave radiation meter is simultaneously standardized；The timing variations signal of different scale is detected by depth network model,
Increase the adaptability and robustness to imperfect time series data, the feature of detection is integrated, finally judges that input sample is
The probability of melt zone；And output is to the recognition result of melt zone and non-melt zone.
In the melting area detection method of the present invention, it is preferable that the bright temperature time series data of microwave radiometer is based on z-
Score methods are standardized.
In the melting area detection method of the present invention, it is preferable that the bright temperature time series data of microwave radiometer is whole year
365 days data.
In the melting area detection method of the present invention, it is preferable that the depth network model includes the first full convolution mould
Block, the first random drop module, the second full convolution module, the second random drop module, the full convolution module of third, global average pond
Change module, the first full articulamentum and the second full articulamentum.
In the melting area detection method of the present invention, it is preferable that the first full convolution module, the second full convolution
Module and the full convolution module of the third are respectively used to the clock signal variation characteristic of detection different scale, include the office of small scale
The global change signal of portion's variable signal and large scale.
The present invention melting area detection method in, it is preferable that the first random drop module and described second with
Machine discard module is used to increase the adaptability and robustness to imperfect time series data, and mitigates over-fitting.
In the melting area detection method of the present invention, it is preferable that the global average pondization and first full articulamentum
It is integrated for the feature to detection.
In the melting area detection method of the present invention, it is preferable that the second full articulamentum is inputted for finally judging
Sample is the probability of melt zone.
The present invention uses the depth network model of full convolutional coding structure and multilayer perceptron structure containing there are three, based on annual every
Day, microwave radiometer bright temperature data detected melting area, and precision is substantially improved compared with conventional model, and to difference
Ice and snow type has the adaptive ability of height, can be provided for acquisition of information such as polar region freeze thawing regional change, freezing and thawing cycle variations good
Good basis.
Description of the drawings
Fig. 1 is the flow chart of melting area detection method.
Fig. 2 is the Organization Chart of depth network model.
Fig. 3 is the verification precision of melting area detection method.
Specific implementation mode
In order to make the purpose , technical scheme and advantage of the present invention be clearer, below in conjunction with the embodiment of the present invention
In attached drawing, technical scheme in the embodiment of the invention is clearly and completely described, it should be understood that described herein
Specific examples are only used to explain the present invention, is not intended to limit the present invention.Described embodiment is only the present invention one
Divide embodiment, instead of all the embodiments.Based on the embodiments of the present invention, those of ordinary skill in the art are not making
The all other embodiment obtained under the premise of creative work, shall fall within the protection scope of the present invention.
Fig. 1 is a kind of flow chart of melting area of the invention detection method.As shown in Figure 1, a kind of ice sheet of the present invention melts
Change area's detection method to include the following steps：
In step sl, it inputs the bright temperature time series data of microwave radiometer and is standardized.Wherein, the bright temperature of microwave radiometer
The data that time series data is annual 365 days.The bright temperature time series data of microwave radiometer is based on standard deviation and standardizes (z-score) method
It is standardized.
In step s 2, by depth network model detect different scale timing variations signal, increase to it is imperfect when
The adaptability and robustness of ordinal number evidence, integrate the feature of detection, finally judge input sample for the probability of melt zone.
Specifically, as shown in Fig. 2, depth network model includes the first full convolution module, the first random drop module, the
Two full convolution modules, the second random drop module, the full convolution module of third, global average pond module, the first full articulamentum and
Second full articulamentum.
Wherein, the first full convolution module, the second full convolution module and the full convolution module of third are respectively used to detect different rulers
The clock signal variation characteristic of degree includes the global change signal of the localized variation signal of small scale and large scale.Three convolution
Module window size is respectively 8,5 and 3, and filter (feature) quantity is respectively 256,128 and 256.
First random drop module and the second random drop module be used to increase to the adaptability of imperfect time series data with
Robustness, and mitigate over-fitting.The loss ratio of first random drop module and the second random drop module is all 20%, is used
In respectively to the feature progress random drop of first and second convolution module output.
The average pondization of the overall situation and first full articulamentum are for integrating the feature of detection.The average pond layer of the overall situation is complete
Office's mean value pond makes each characteristic pattern be converted into an output, is the integration of sequential dimension.First full articulamentum is for being rolled up
Weighting between the feature that product obtains is integrated, and is the integration of characteristic dimension.Second full articulamentum is for finally judging input sample
For the probability of melt zone.
The depth network model of the present invention need to first be based on automatic weather station measured data, when each sample temperature is more than 0
The corresponding date is labeled as melting, and temperature corresponds to the date and is labeled as non-thawing when being less than 0, and marks samples based on these and instructed
Practice, can then input the bright temperature time series data of microwave radiometer and whether extensive area occurred in 1 year thawing and visit
It surveys.
In step s3, it exports to melt zone (region that thawing occurred in 1 year) and non-melt zone (in 1 year
Do not occurred melt region) recognition result.
For the technique effect for the melting area detection method that the present invention is further explained.According to corresponding automatic meteorological
956 sample points of data decimation of standing, wherein 80% is used to train, 20% for verifying.By the training of 60 bouts, use
Random optimization device (Adam), this method obtain 93% overall verification precision, and widely used generalized Gaussian distribution method
Precision is about 70%-80% in detection.The verification precision of the melting area detection method of the present invention is shown in FIG. 3.
The present invention uses the depth network model of full convolutional coding structure and multilayer perceptron structure containing there are three, based on annual every
Day, microwave radiometer bright temperature data detected melting area, and precision is substantially improved compared with conventional model, and to difference
Ice and snow type has the adaptive ability of height, can be provided for acquisition of information such as polar region freeze thawing regional change, freezing and thawing cycle variations good
Good basis.
More than, it is described in detail for the specific implementation mode of the melting area detection method of the present invention, still
The present invention is not limited thereto.The specific implementation mode of each step according to circumstances can be different.In addition, the sequence of part steps can be with
It exchanges, part steps can be omitted.
The above description is merely a specific embodiment, but scope of protection of the present invention is not limited thereto, any
Those familiar with the art in the technical scope disclosed by the present invention, all answer by the change or replacement that can be readily occurred in
It is included within the scope of the present invention.
Claims (8)
1. a kind of melting area detection method, which is characterized in that
Include the following steps：
The bright temperature time series data of input microwave radiometer is simultaneously standardized；
The timing variations signal of different scale is detected by depth network model, increase to the adaptability of imperfect time series data and
Robustness integrates the feature of detection, finally judges input sample for the probability of melt zone；And
Export the recognition result to melt zone and non-melt zone.
2. melting area according to claim 1 detection method, which is characterized in that
The bright temperature time series data of microwave radiometer is standardized based on standard deviation standardized method.
3. melting area according to claim 1 detection method, which is characterized in that
The data that the bright temperature time series data of microwave radiometer is annual 365 days.
4. melting area according to claim 1 detection method, which is characterized in that
The depth network model include the first full convolution module, the first random drop module, the second full convolution module, second with
The full convolution module of machine discard module, third, global average pond module, the first full articulamentum and the second full articulamentum.
5. melting area according to claim 4 detection method, which is characterized in that
It is different that the first full convolution module, the second full convolution module and the full convolution module of the third are respectively used to detection
The clock signal variation characteristic of scale includes the global change signal of the localized variation signal of small scale and large scale.
6. melting area according to claim 4 detection method, which is characterized in that
The first random drop module and the second random drop module are used to increase the adaptation to imperfect time series data
Property and robustness, and mitigate over-fitting.
7. melting area according to claim 4 detection method, which is characterized in that
The global average pondization and first full articulamentum are for integrating the feature of detection.
8. melting area according to claim 4 detection method, which is characterized in that
The second full articulamentum is for finally judging input sample for the probability of melt zone.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810468869.7A CN108733909A (en) | 2018-05-16 | 2018-05-16 | A kind of melting area detection method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN201810468869.7A CN108733909A (en) | 2018-05-16 | 2018-05-16 | A kind of melting area detection method |
Publications (1)
| Publication Number | Publication Date |
|---|---|
| CN108733909A true CN108733909A (en) | 2018-11-02 |
Family
ID=63938369
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN201810468869.7A Pending CN108733909A (en) | 2018-05-16 | 2018-05-16 | A kind of melting area detection method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN108733909A (en) |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN107273828A (en) * | 2017-05-29 | 2017-10-20 | 浙江师范大学 | A kind of guideboard detection method of the full convolutional neural networks based on region |
| US20170344888A1 (en) * | 2016-05-27 | 2017-11-30 | Kabushiki Kaisha Toshiba | Information processing apparatus and information processing method |
| CN107832897A (en) * | 2017-11-30 | 2018-03-23 | 浙江工业大学 | A kind of Stock Price Forecasting method based on deep learning |
| CN107908876A (en) * | 2017-11-16 | 2018-04-13 | 宁波工程学院 | Motor vehicle driven by mixed power operating mode Forecasting Methodology based on multiple dimensioned convolutional neural networks |
-
2018
- 2018-05-16 CN CN201810468869.7A patent/CN108733909A/en active Pending
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US20170344888A1 (en) * | 2016-05-27 | 2017-11-30 | Kabushiki Kaisha Toshiba | Information processing apparatus and information processing method |
| CN107273828A (en) * | 2017-05-29 | 2017-10-20 | 浙江师范大学 | A kind of guideboard detection method of the full convolutional neural networks based on region |
| CN107908876A (en) * | 2017-11-16 | 2018-04-13 | 宁波工程学院 | Motor vehicle driven by mixed power operating mode Forecasting Methodology based on multiple dimensioned convolutional neural networks |
| CN107832897A (en) * | 2017-11-30 | 2018-03-23 | 浙江工业大学 | A kind of Stock Price Forecasting method based on deep learning |
Non-Patent Citations (1)
| Title |
|---|
| 万艳丽 等: "基于深度卷积神经网络的脑部图像疾病类别特征库构建", 《医学信息学杂志》 * |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN110335270A (en) | Transmission line of electricity defect inspection method based on the study of hierarchical regions Fusion Features | |
| CN110223510B (en) | Multi-factor short-term traffic flow prediction method based on neural network LSTM | |
| CN108227041B (en) | Horizontal visibility forecasting method based on site measured data and mode result | |
| CN109376960A (en) | Load Forecasting based on LSTM neural network | |
| CN103678766B (en) | A kind of abnormal Electricity customers detection method based on PSO algorithm | |
| CN104834977B (en) | Traffic alert grade prediction technique based on learning distance metric | |
| CN109272146A (en) | A kind of Forecasting Flood method corrected based on deep learning model and BP neural network | |
| CN109299812A (en) | A kind of Forecasting Flood method based on deep learning model and KNN real time correction | |
| CN104573879A (en) | Photovoltaic power station output predicting method based on optimal similar day set | |
| CN109785629A (en) | A kind of short-term traffic flow forecast method | |
| CN116304962B (en) | Intelligent anomaly monitoring method for water meter metering data | |
| CN104570161B (en) | Typhoon based on the global lattice point forecast data of EC/JMA automates forecasting procedure | |
| CN110378520A (en) | A kind of PM2.5 concentration prediction and method for early warning | |
| CN105069537A (en) | Constructing method of combined air quality forecasting model | |
| CN111814407B (en) | Flood forecasting method based on big data and deep learning | |
| CN106875048A (en) | Emergent Flood Forecasting Method based on success experience pattern | |
| CN102524024A (en) | Crop irrigation system based on computer vision | |
| CN102867221A (en) | Method for dynamically and transversely correcting medium and long-term electric quantity forecasting | |
| CN111210093A (en) | Daily water consumption prediction method based on big data | |
| CN102213593A (en) | Method for rapidly acquiring abnormal land | |
| CN109358381A (en) | A kind of website forecast correction wind method | |
| CN102357934A (en) | Quality monitor soft sensing method based on rubber mixing process | |
| CN113095694A (en) | Method for constructing rainfall sand transportation model suitable for multi-landform type area | |
| CN116384591A (en) | Drought prediction method, system and medium based on big data | |
| CN108733909A (en) | A kind of melting area detection method |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| RJ01 | Rejection of invention patent application after publication | ||
| RJ01 | Rejection of invention patent application after publication |
Application publication date: 20181102 |