中国物理B ›› 2005, Vol. 14 ›› Issue (10): 2041-2045.doi: 10.1088/1009-1963/14/10/020
赵国良1, 杨春玲2, 王宇野3, 赵东阳3
Yang Chun-Ling (杨春玲)ab, Wang Yu-Ye (王宇野)b, Zhao Dong-Yang (赵东阳)b, Zhao Guo-Liang (赵国良)a
摘要: There exist a considerable variety of factors affecting the spectral emissivity of an object. The authors have designed an improved combined neural network emissivity model, which can identify the continuous spectral emissivity and true temperature of any object only based on the measured brightness temperature data. In order to improve the accuracy of approximate calculations, the local minimum problem in the algorithm must be solved.Therefore, the authors design an optimal algorithm, i.e. a hybrid chaotic optimal algorithm, in which the chaos is used to roughly seek for the parameters involved in the model, and then a second seek for them is performed using the steepest descent. The modelling of emissivity settles the problems in assumptive models in multi-spectral theory.
中图分类号: (Nonlinear dynamics and chaos)