2017 Fiscal Year Annual Research Report
ブリルアン光相関領域リフレクトメトリの性能向上と温度と歪の同時・分離・分布測定
| Project/Area Number |
16J05910
|
|---|
| Research Institution | The University of Tokyo |
|---|
Principal Investigator |
YAO YUGUO 東京大学, 先端科学技術研究センター, 特別研究員(PD)
|
|---|
| Project Period (FY) |
2016-04-22 – 2018-03-31
|
| Keywords | Brillouin scattering / signal processing / machine learning / BOCDA/BOCDR |
|---|
| Outline of Annual Research Achievements |
We propose the signal processing based on machine learning in Brillouin optical correlation domain analysis/ reflectometry (BOCDA/R) for the first time. The implementation of the neural network method is described. However, other machine learning methods are also thought to be adaptive in this signal process, such as the support vector machine.
Machine learning is a fashionable and also promising method that has been applied in many fields, such bio-imaging. Different from the conventional signal processing methods, machine learning infers a reasonable model from massive data that are used to train the model, and shows great power at handling the information when the physical law is no clear or when the law is difficult to achieve from induction.
By introducing the machine learning, the performance of BOCDA/R is expected to be more robust. Also, the speed of signal processing in BOCDA/R is expected to increase without deteriorating the measurement accuracy, if a good model is trained by the training data.
The future work will focus on the collection of the big data in the real experiment, and the debugging of the algorithm. By introducing the machine learning into the BOCDA/R, it is expected that the system performances will be more reliable and precise, and the repeatability issue which has been bothering the researchers will be conquered.
|
| Research Progress Status |
29年度が最終年度であるため、記入しない。
|
| Strategy for Future Research Activity |
29年度が最終年度であるため、記入しない。
|
Research Products
(1 results)
