基于全连接神经网络的自然光谱复现方法研究

为了能够精准和快速地复现可见光波段自然光谱,实现全光谱照明,使用具有强大非线性拟合能力的神经网络完成光谱匹配。首先,利用全连接神经网络的自学习能力,在采用基于修正高斯分布拟合函数的多个单色LED合成光谱模型生成训练和测试数据集的基础上,构建充分体现合成光谱和各单色LED光强系数比例关系的神经网络模型,即该模型能够针对输入光谱得到对应的单色LED光强比例系数,进而实现光谱复现。其次,针对标准太阳光谱和实测得到的不同时刻、天气的自然光谱进行光谱复现,并与采用基于遗传算法的光谱复现方法得到的结果进行对比。结果表明,基于全连接神经网络的自然光谱复现方法能够以小于5%的误差实现标准太阳光谱和实测太阳光谱的匹配,证明可使用同一个训练模型在不改变单色LED数量和种类的条件下得到多种与目标光谱高度吻合的不同类型光谱。对比其他光谱匹配算法,所提方法的拟合速度提高了数倍,还具有稳定性高、调控灵活、操作简便等优势。

A Comparative Study on the Quality Components Between Self Pollinated Seeds and Naturally Pollinated Seeds in Brassica napus L. in Sichuan Ecological Region

[ Objective] This study was to reveal the differences in crude fat and glucosinolates between self pollinated seeds and naturally pollinated seeds in Brassica napus in sichuan ecological region.. [ Method] Near-infrared spectroscopy method （NIRS） was employed to measure the quality components in self pollinated seeds and naturally pollinated seeds of 861 shares of Brassica napus from Sichuan ecological region. And correlation analysis and regression analysis were conducted based on the experimental data via SPSS （statistics package for social science）. [ Result] The contents of crude fat in the self pollinated seeds were commonly a higher than that in the naturally pollinated seeds at 0.01 significant level; while the contents of glucosinolates in the self pollinated seeds and the naturally pollinated seeds were insignificantly different. Both the correlation relationship and linear regression for the crude fat between the self pollinated seeds and naturally pollinated seeds reached the significant level. The regression equations for the contents of crude fat（y1 ） and glucosinolates（ y2 ） in the naturally pollinated seeds and of crude fat（ x1 ） and glucosinolates（ x2 ） in self pollinated seeds were respectively determined to be y1 = 16.844 ＋0.614x1 and y2 = -0.620 ＋ 1.017 x2. [ Conclusion] In Brassica napus breeding, crude fat in naturally pollinated seeds should be emphatically taken into account, meanwhile concurrently considering that in self pollinated seeds; while glucosinolates in both the self pollinated seeds and the naturally pollinated seeds must be simultaneously concerned.

Ghezeljeh nanoclay as a new natural adsorbent for the removal of copper and mercury ions: Equilibrium, kinetics and thermodynamics studies

Heavy metal determination was carried out by applying the solid phase extraction （SPE） method in batch mode followed by atomic absorption spectroscopy （AAS） and inductively coupled plasma atomic emission spectrosco py （ICP-AES） from aqueous solutions using Ghezeljeh montmorillonite nanoclay as a new natural adsorbent. The Ghezeljeh clay is characterized by using Fourier Transform Infrared （FT-IR） Spectroscopy, Scanning Electron Mi- croscopy-Energy Dispersive Spectrometry （SEM-EDS） and X-ray Diffractometry （XRD） and X-ray Fluorescence （XRF）. The results of XRD and FT-IR of nanoclay confirm that montmorillonite is the dominant mineral phase. Based on SEM images of Ghezeljeh clay, it can be seen that the distance between the plates is Nano. The effects of varying parameters such as initial concentration of metal ions, pH and type of buffer solutions, amount of ad- sorbent, contact time, and temperature on the adsorption process were examined. The effect of various interfer- ing ions was studied. The adsorption data correlated with Freundlich, Langmuir, Dubinin-Radushkevich （D-R）, and Temkin isotherms. The Langmuir and Freundlich isotherms showed the best fit to the equilibrium data for Hg（II）, but the equilibrium nature of Cu（ll） adsorption has been described by the Langmuir isotherm. The kinetic data were described with pseudo-first-order, pseudo-second-order and double-exponential models, The adsorp- tion process follows a pseudo-second-order reaction scheme, Calculation of AGσ, △Hσ and ASσ showed that tilenature of Hg（II） ion sorption onto the Ghezeljeh nanoclay was endothermic and was favored at higher temper- attire, and the nature of Cu（II） ion sorption was exothermic and was favored at lower temperature,