研究目的
Extracting physical properties of nematic and cholesteric liquid crystals directly from their textures images using a combination of complexity measures and machine learning techniques.
研究成果
The proposed method accurately predicts physical properties of liquid crystals from texture images, with high precision in regression and classification tasks. It is simple, fast, and scalable, showing potential for broader applications in materials science and imaging-based characterization.
研究不足
The approach relies on specific complexity measures and may not capture all aspects of texture complexity; it requires known physical properties for training, and accuracy can be affected by noise or overlapping features in the complexity-entropy plane. The method is demonstrated on specific liquid crystal types and may need adaptation for other materials or more complex experimental setups.
1：Experimental Design and Method Selection:
The approach involves calculating permutation entropy (H) and statistical complexity (C) from liquid crystal textures, which are used as features in supervised learning tasks (regression and classification) using the k-nearest neighbors algorithm.
2：Sample Selection and Data Sources:
Data includes simulated nematic textures from Monte Carlo simulations, experimental nematic textures from E7 liquid crystal samples, and simulated cholesteric textures from continuum elastic theory.
3：List of Experimental Equipment and Materials:
Polarized optical microscope, temperature controller, E7 liquid crystal samples, rectangular capillaries, computational tools for simulations and image processing.
4：Experimental Procedures and Operational Workflow:
For experimental textures, samples are heated at controlled rates, and images are captured at intervals. For simulations, textures are generated using specific models (e.g., Lebwohl-Lasher potential for nematics, Landau-de Gennes theory for cholesterics). Images are processed to grayscale, and H and C are computed.
5：Data Analysis Methods:
k-nearest neighbors algorithm is used for regression (predicting order parameter and temperature) and classification (predicting pitch length), with cross-validation to optimize parameters and assess accuracy.
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
