论文部分内容阅读
基于生物组织的粒状模型以及光谱对米氏(Mie)散射体形态的灵敏性,构建了类上皮组织模型的偏振后向散射光谱反演模型。针对反演模型的多参量、多极值、非线性,涉及到非常复杂的三角函数计算,模型空间极小的特点,采用每代保留最优个体的浮点遗传算法对类上皮组织模型的偏振后向散射光谱反演。对编码策略、适值调整策略及选择策略进行了讨论。研究结果表明:经过70代反演迭代,每个参量的相对误差趋于稳定,最小的达到0.02%左右,最大的不超过3%。采用基于实数编码的遗传算法能从偏振散射光谱中同时反演获得表层粒子的形态参量,具有全局收敛性和良好的反演精度与抗噪声能力。
Based on the granular model of biological tissue and the sensitivity of the spectrum to Mie scatterers, a polarization backscattered spectral inversion model of the epithelial tissue model was constructed. The multi-parameters, multi-polarity, and nonlinearity of the inversion model involve very complex trigonometric functions with minimal model space. The floating-point genetic algorithm that preserves the optimal individuals in each generation is used to polarize the epithelium-like tissue model. Backscattering spectra inversion. The coding strategy, fitness adjustment strategy and selection strategy were discussed. The results show that after 70 iterations, the relative error of each parameter tends to be stable, with the smallest reaching 0.02% and the largest not exceeding 3%. The genetic algorithm based on real coding can simultaneously obtain the morphological parameters of the surface particles from the polarization scattering spectrum, with global convergence and good inversion accuracy and anti-noise ability.