The Application of Chaos Equation and Fractal Dimension To Interpret Rice Types Using Hyperspectral Data

• Main Authors: I-Chen Yang, 楊宜蓁
• Other Authors: 李瑞陽
• Format: Others
• Language: zh-TW
• Published: 2016
• Online Access: http://ndltd.ncl.edu.tw/handle/41412943713006732033

碩士 === 逢甲大學 === 土地管理學系 === 104 === Chaos can be explained as the seemingly lack of rules, but it will appear in a certain regularity. In other words, regularities are in the irregular behaviors. The process of each crop growth is a steady time series. However, the spectral reflectance values of the crop are influenced by the crop growth status (such as moisture, nutrition, pest ...etc.), soil, sun illumination as well as atmospheric and climatic conditions. These phenomena have resulted in the spectral reflectance curves of the same crops showing a tiny and irregular deviation. But, as phenology presenting a certain regularity, is quite accord with chaos. The current classification methods are applicable for specific crops and at specific locations and times. However, these methods lack generalizability. Recent studies found that both traditional and machine learning classifiers were difficult to distinguish different crops such as garlic and scallion. Thus, in nonlinear systems under certain conditions, the feasibility of applying chaos theory to remote-sensing crop classification is worthy of exploration. This research utilizes the spectral reflectance values of rice to interpret different crop types by employing chaos theory in Taichung County, Taiwan. Different growth periods of the rice phenology were selected to measure rice spectral reflectance. Two types of analysis, spectral sequence and time series, are used in the study. The chaotic equation and the hyperspectral data collected by the handheld spectra radiometer are used to create chaotic graphs. These graphs are then applied to produce different rice fractal dimension. The MATLAB software is used to develop the models. The fractal theory is used to calculate the fractal dimension of chaotic graphics. The rice fractal dimensions were used to build their fractal dimension ranges. The results demonstrate that the period of planting - growing is easier to be differentiated from other periods by employing the chaos theory and the fractal theory. The near-infrared bands are the best spectrum bands for the rice, including (766.5- 814.9) nm, (844.1-863.7) nm, (903.1-932.9) nm and （764.9-797.1）nm、（805.1-853.9）nm、（870.2-878.4）nm range. For the satellite images, the results indicate that the Red Edge (705 ~ 745 nm) and NIR1 (770 ~ 895 nm) are easier to be identified compared with the other spectrum bands using the fractal dimension. This study explores whether a single crop using the chaos theory and the fractal theory can produce unique graphs. These unique graphs can be used to establish the rice sample models for the purpose of classification.