研究目的
To demonstrate identification of object wedge angle and direction using machine learning algorithms based on received beam intensity profiles.
研究成果
CNN-based deep learning method has high accuracy and can stay stable when detecting the wedge angle and the direction. CNN performs better than the other traditional ML methods, which can achieve 100% accuracy, even under extreme slight change of the object. Compared with other works, our method can significantly reduce the hardware implementation complexity, and it only requires a single-shot measurement.
研究不足
The study does not discuss the scalability of the method to larger datasets or more complex objects.
1：Experimental Design and Method Selection:
The study uses a CNN-based deep learning method for object parameter identification, such as wedge angle and direction.
2：Sample Selection and Data Sources:
A large dataset containing numerous collected images with 14 wedge angles and 32 directions is generated.
3：List of Experimental Equipment and Materials:
Includes a Gaussian beam as the probe, SLM to emulate objects with various features, and CCD to capture images.
4：Experimental Procedures and Operational Workflow:
Different ML techniques are trained to identify the image collected by the CCD.
5：Data Analysis Methods:
CNN includes multiple convolutional layers, pooling layers, and fully-connected layers. Linear regression and soft-max algorithms are applied.
独家科研数据包，助您复现前沿成果，加速创新突破
获取完整内容
