基于核自适应滤波的无线传感网络定位算法研究

针对动态室内环境的变化及时变的接收信号强度(Received signal strength,RSS)对定位精度的影响,提出了一类基于核自适应滤波算法的农业无线传感器网络室内定位方法。核自适应滤波算法具体包括量化核最小均方(Quantized kernel least mean square,QKLMS)算法及固定预算(Fixed-budget,FB)核递推最小二乘(Kernel recursive least-squares,KRLS)算法。QKLMS算法基于一种简单在线矢量量化方法替代稀疏化,抑制核自适应滤波中径向基函数结构的增长。FB-KRLS算法是一种固定内存预算的在线学习方法,与以往的"滑窗"技术不同,每次时间更新时并不"修剪"最旧的数据,而是旨在"修剪"最无用的数据,从而抑制核矩阵的不断增长。通过构建RSS指纹信息与物理位置之间的非线性映射关系,核自适应滤波算法实现WSN的室内定位,将所提出的算法应用于仿真与物理环境下的不同实例中,在同等条件下,还与其他核学习算法、极限学习机(Extreme learning machine,ELM)等定位算法进行比较。仿真实验中2种算法在3种情形下的平均定位误差分别为0.746、0.443 m,物理实验中2种算法在2种情形下的平均定位误差分别为0.547、0.282 m。实验结果表明,所提出的核自适应滤波算法均能提高定位精度,其在线学习能力使得所提出的定位算法能自适应环境动态的变化。     

基于红外光声光谱的农作物秸秆导热系数定量分析

选用我国华北地区具有代表性的小麦、玉米、水稻秸秆样品,对比研究了偏最小二乘(PLSR)和高斯核支持向量机(RBF-SVR)分别构建单一和混合种类秸秆全波段定量分析模型的效果,探讨了红外光声光谱耦合化学计量学方法构建我国主要粮食作物秸秆导热系数定量分析模型的可行性。研究发现,小麦秸秆和水稻秸秆导热系数RBF-SVR非线性模型,以及玉米秸秆、混合种类秸秆的PLSR线性模型效果较优。进一步应用蚁群算法与上述最优建模方法相结合,构建了更加优化的小麦秸秆、玉米秸秆、水稻秸秆和混合秸秆导热系数模型,验证决定系数(R_p~2)分别为0.77、0.83、0.96和0.79,验证均方差(RMSEP)分别为0.007 8、0.015、0.005 9、0.014 W/(m·K),验证相对分析误差(RPD)分别为2.81、2.41、7.39和2.15。研究结果表明,红外光声光谱技术结合先进适用的化学计量学方法可实现我国主要粮食作物秸秆导热系数的快速定量分析,但混合秸秆模型预测精度仍需进一步提升。     

NCT对小麦衰老生理的影响

以新合成的NCT配制成不同浓度的水溶液，对灌浆初期的大田小麦进行叶面喷施，以喷清水作对照，并以与NCT相同浓度梯度的NAA水溶液作对比处理，测定了喷后不同时期小麦旗叶的相对电导率、叶绿素含量、超氧化物歧化酶（SOD）活力及丙二醛（MDA）含量。结果表明：NCT能延缓灌浆期SOD活力，使叶绿素含量降低，控制相对电导率和MDA含量的升高，并使小麦穗粒重和千粒重增加，其效果明显优于NAA。     

基于Harris算子的籽粒尖端识别方法

通过籽粒尖端形态特征分析,提出了一种基于Harris角点检测的籽粒尖端识别方法。Harris算子在一个局部区域中检测角特征,作为籽粒最明显的角结构,尖端在Harris角点检测中具有最大的响应值,使其对尖端不明显的籽粒也有良好的响应。通过对玉米、南瓜和西葫芦等具有尖端特征的750粒籽粒的测试结果表明,提出算法的尖端检测综合准确率为95.6%。     

基于核密度估计的京津冀地区耕地破碎化分析

为了深入分析京津冀地区耕地破碎化和人工干扰程度,以空间网格为基本单元,采用核密度函数法对耕地进行密度计算,继而采用自然断点法进行密度分区,采用景观格局指数法,对耕地的空间分布特征、破碎化程度进行定量分析,最后采用耕地破碎度指数和人工干扰指数对比分析提出耕地整治的方向。结果表明:耕地核密度变化范围为0~95.08点/km2,85%以上的耕地集中分布在区域中部和东南部的中密度区到高密度区,耕地面积随着密度降低而减少;耕地斑块密度指数、斑块边界密度指数、面积加权分维数和耕地斑块聚集度在不同密度区的变化情况均反映出破碎化程度随着耕地密度减小而增大;耕地破碎度指数和人工干扰度指数均随着耕地密度降低而升高,人工干扰是该区域耕地破碎化程度加剧的主要因素。采用核密度估计法能合理地反映耕地聚集程度和破碎化程度,可为京津冀地区协同发展、耕地整治和土地利用优化布局提供理论支撑。     

视觉导引AGV鲁棒特征识别与精确路径跟踪研究

针对AGV多分支路径与工位点标识的可靠识别以及导引路径的精确跟踪问题,提出了一种基于双视野窗口的鲁棒特征识别与精确路径跟踪方法。采用整幅视野范围作为模式识别窗口,在该窗口采用基于核主成分分析(KPCA)和BP神经网络的识别方法,将路径特征通过核函数映射到高维空间进行PCA降维,再利用BP神经网络识别降维后的样本矩阵。同时提出一种导引扫描窗口设置方法,该窗口范围取决于摄像机竖直视角以及摄像机安装倾斜角,在导引扫描窗口内将导引路径简化为直线模型并用最小二乘法拟合,针对拟合直线计算导引所需的路径偏差。实验结果表明,KPCA-BP方法显著提高了路径特征识别的实时性和鲁棒性,6类路径特征的平均特征识别正确率为99.5%;导引扫描窗口有效减小了导引路径直线拟合的计算误差,直线路径跟踪误差小于3 mm,曲线路径跟踪误差小于30 mm。     

糙米发芽前含水率提升工艺优化

糙米发芽前的吸水过程是导致籽粒裂纹的根本原因,制约着发芽糙米品质和口感。为降低发芽前糙米裂纹增率,探究了完整吸湿区间内各含水率水平糙米的最优吸湿速率。将糙米初始含水率至发芽含水率的完整区间分为若干子区间,在各区间内以不同加湿速率加湿至该区间目标含水率。探究各区间内裂纹增率的变化规律,建立裂纹增率与加湿速率变化规律的数学模型,以低裂纹增率为目标确定最优加湿速率。在此基础上,得出完整区间内以低裂纹增率及高效率为目标的加湿速率数学模型并试验验证。与前期分段加湿工艺相比,本优化工艺可降低发芽前糙米和发芽糙米裂纹增率(41.48±0.15)%和(43.67±0.26)%,糙米发芽率和γ-氨基丁酸含量增加(6.92±0.25)%和(25.03±0.18)%,为高品质发芽糙米的生产方法提供参考。     

Influence of long-term fertilization with farmyard manure on soil organic matter: Characteristics of particle-size fractions

Summary The influence of more than 100 years of fertilization with farmyard manure on soil organic matter in comparison to unfertilized soil was studied in particle-size fractions using elemental (C and N) analyses and pyrolysis-field ionization mass spectrometry. Distinct differences in C and N concentrations and distribution and in the quality of organic matter between the size fractions and the fertilization treatments were observed. Clay-associated C and N were relatively higher in the unfertilized treatment, whereas the application of farmyard manure preferentially increased soil organic matter associated with the fine and medium silt fractions. Pyrolysis-field ionization mass spectrometry of soil fractions <20 m showed increasing values for lignin monomers and dimers and fatty acids with larger equivalent diameters, whereas the proportion of N compounds, mono- and polysaccharides and phenolics decreased in the larger size fractions. Sand fractions were particularly rich in lignin fragments, mono- and polysaccharides, and alkanes/alkenes. These relationships seemed to be independent of management practices. In the same size fractions of the different treatments, however, a higher relative abundance of N-compounds, mono- and polysaccharides, phenolics, lignin monomers, and alkanes/alkenes was observed in the unfertilized variant. Lignin dimers and fatty acids were more abundant in the farmyard manure treatment. Both trends together imply that soil enrichment in organic matter due to the application of farmyard manure largely reflects an increase in lignin building blocks and partly reflects an increase in lipids such as fatty acids in the silt fractions. Therefore these constituents are of particular importance in assessing the positive effects of farmyard manure on soil fertility.     

豫北蔬菜保护地土壤磷素形态及其空间分布特性研究

采用蒋柏藩、顾益初无机P分级方法研究了豫北褐土多年棚龄蔬菜保护地土壤P素形态及其空间分布特性。结果表明,蔬菜保护地0 ~ 20cm土层全P、无机P、有机P、Olsen-P的含量分别为:1385.6 ~ 2896.5、1097.1 ~ 2365.7、270.0 ~ 606.9、109.8 ~ 302.4 mg/kg,Ca2-P、Ca8-P、Al-P、Fe-P、O-P、Ca10-P分别占无机P的百分比平均为:12.5 %、37.2 %、10.8 %、5.8 %、13.3 %、20.5 %,Olsen-P占全P的百分比高达4 % ~ 15 %,平均为10.6 %;土壤各形态P素主要积累在0 ~ 20 cm土层,随着深度的增加土壤全P、有机P、Olsen-P、各形态无机P均减少。     

Mass selection of wheat for grain filling without transient photosynthesis

Summary Post-anthesis chemical desiccation of wheat (Triticum aestivum L.) plants in the field eliminates transtent photosynthesis by killing all green tissues, thus revealing the plant's capacity for grain filling from stored stem reserves, as the case is for post-anthesis stress such as drought or leaf diseases. This study was conducted to investigate whether mass selection for large kernels under chemical desiccation would lead to the improve ment of grain filling in the absence of transient photosynthesis.Six crosses of common spring wheat were subjected to three cycles of mass selection from F₂ through F₁ when selection was performed for large kernels by sieving grains from plants that were erther chemically desiccated after anthesis, or not (controls). The resulting 36 bulks (six crosses by three selection cycles by two selection environments) were compared with their respective F₂ base populations, when tested with and without chemical desiccation.Selection for large kernels under potential conditions (without chemical desiccation) did not improve kernel weight under potnetial conditions, evidently because these materials were lacking in genetic variation for kernel weight under potential conditions. In four of the crosses, 3rd cycle selection for large kernels under potential conditions decreased kernel weight under chemical desiccation. On the other hand, selection for large kernels under chemical desiccation was effective in improving kernel weight and test weight under chemical desiccation, depending on the cross and the selection cycle, with no genetic shift in mean days to heading or mean plant height. Selection for large kernels under chemical desiccation was also effective in some cases in increasing kernel weight under potential conditions. The results are interpreted to show that selection under potential conditions and under chemical desiccation operate on two different sources for grain filling, namely transient photosynthesis and stem reserve utilization, respectively. In order to expose genetic variability for stem reserve utilization to selection pressure, transient photosynthesis must be eliminated, as done by chemical desiccation in this study.