EVALUATION OF BARLEY AND MALT QUALITY USING NEAR-INFRARED REFLECTANCE TECHNIQUES

A scanning near-infrared reflectance spectrophotometer was calibrated for the prediction of barley aleurone colour and malt moisture. The malt moisture was predicted on malt ground for the determination of malt extract (coarse grind) making the method suitable for moisture correction in malt extract estimation. Calibrations for the prediction of malt extract and endosperm modification from barley and malt were also attempted. A correlation (r= 0.851 n = 135) was found between malt hot water extract and the percentage of the endosperm estimated as being modified by microscopy following staining with Calcofluor. Probably because of this influence of modification on malt extract, the use of near-infrared reflectance to predict malt extract was most successful at predicting the malt extract values obtained following micro-malting in the absence of the additives, gibberellic acid and potassium bromate.     

基于自组织神经网络的实时统计特征抽取方法

应用人工神经网络讨论统计了特征抽取的原理，提出了相应的抽取方法。该方法具有实时性，容错性，高数据压缩率和良好的特征分辨率。通过计算机仿真和桥梁索力状态的实例计算分析证实了该技术具有广泛的应用价值。     

Chlorosis in sugarcane: Associated soil properties,leaf mineral composition,and crop response to iron and manganese

Attempts were made to find out the nature and cause of a widespread sugarcane chlorosis in a cane growing area. Soils from chlorotic and nonchlorotic fields did not differ markedly in pH, CaCO₃, electrical conductivity, organic C, and soil test P, Zn, Cu, and Mn levels, but the chlorotic field soils had relatively more NH₄OAc-extractable K and less clay and DTPA-extractable Fe. Chlorotic and green leaf blades contained about the same concentration of P, S, Mg, Fe, Zn, Cu, and Mo, but the former leaf blades had more K and less Ca and HCl soluble Fe than the green ones. Green leaves of plants that seemed to have naturally recovered in chlorotic fields were higher in Mn. A foliage spray with 2.5% FeSO₄.7H₂O solution resulted in greening of leaves and a field experiment showed marked crop response to Fe and some response to Mn. The results thus suggest that the chlorosis is due to a lime-induced Fe-deficiency with the possibility of some role of Mn in Fe nutrition in calcareous soil conditions.Deceased 22 September 1988     

Foliar fertilization of high-yielding peanuts during the pod-filling period

A 2-year field experiment was carried out in a sandy soil (Xeropsament-Torripsament) at Nir Yizhaq, Israel, where commercially grown peanut plants were sprayed with different NPKS solutions during the pod-filling period. All foliar sprays were applied in addition to the fertilizer which had been added to the soil before planting. The differences in yield between the treatments were not significant in either year. In the first year, there was a tendency toward increased yields of pods (up to 13%) and of hay (up to 16%) when four foliar fertilizer applications (10 kg N, 1 kg P, 3 kg K and 0.5 kg S/ha/application) were given, at one-week intervals. N was added mainly as urea, P and K as potassium polyphosphate, and S as ammonium sulphate. In the second year, the highest yield was obtained in the control plots and the differences between the treatments were not significant. Utilization of foliar nutrient application seems to be dependent upon availability of these nutrients in the soil.Contribution from the Agricultural Research Organization, The Volcani Center, Bet Dagan, Israel. No. 1976-E. 1087 series.     

An economical synthesis of the major sex attractant of the oak leaf roller—cis-10-tetradecenyl acetate

The major sex attractant of the oak leaf roller (Archips semiferanus Walker),cis-10-tetradecenyl acetate (I), was synthesized in 55–62% overall yield by a highly economical route from inexpensive azelaic acid (II).Authorized as Paper No. 4779 in the Journal Series of the Pennsylvania Agricultural Experiment Station.     

一种基于支持向量机的模糊施肥预测方法研究

提出了一种利用支持向量机(SVM)学习算法提取模糊规则进而实现施肥预测的方法.对于农业施肥中常用的正交实验法,由于其数据均衡分散使得曲线拟合的回归预测方法效果不佳.提出了一种利用SVM学习样本数据,再利用隶属度来提取模糊规则,通过阈值和可信度来控制规则的激活和准确性的预测方法,这一方法不仅避免了回归预测所产生的误差,并且模糊规则更具有实际意义,从而大大提高了知识获取的能力.     

基于数学归纳法抽取循环程序研究

广义上,程序设计过程就是定理证明过程,因而程序综合与机器定理证明关系密切。由于归结原理的反证法本质,不能保证其证明是构造性的,因此不能保证能够抽取循环程序,但是利用数学归纳法可以使用归结原理来抽取循环体,从而抽取循环程序。文中利用数学归纳法进行了循环程序的提取。     

基于卡口上下文和深度置信网络的车辆轨迹预测模型研究

针对车辆轨迹预测中节点序列的时序特性和实际路网中的空间关联性,该文提出一种基于深度置信网络和SoftMax (DBN-SoftMax)轨迹预测方法.首先,考虑到轨迹在节点集合中的强稀疏性和一般特征学习方法对新特征的泛化能力不足,该文利用深度置信网络(DBN)较强的无监督特征学习能力,达到提取轨迹局部空间特性的目的;然后,针对轨迹的时序特性,该文采用逻辑回归的预测思路,用当前轨迹集在路网特征空间中的线性组合来预测轨迹;最后,结合自然语言处理领域中的词嵌入的思想,基于实际轨迹中节点存在的上下文关系,运用节点的向量集表征了节点间的交通时空关系.实验结果表明该模型不仅能够有效地提取轨迹特征,并且在拓扑结构复杂的路网中也能得到较好的预测结果.     

基于非线性重构模型的植物叶片图像集分类方法

提出一种基于非线性重构模型的植物叶片图像集的分类识别方法。该方法首先使用高斯受限玻尔兹曼机(GRBMs)通过非监督预训练来初始化模型的权值;然后针对每一个植物叶片图像集用初始化的模型训练得到一个特定的模型;最后根据测试样本的最小重构误差和测试样本集的最多投票策略来判定测试样本集的类别。该方法通过图像预处理来处理图像,避免了图像在缩放时发生形变,并采用基于k-means的特征提取方法来提取植物叶片图像特征。实验结果表明,该方法能够准确地对植物叶片图像集进行分类识别。