红树林共生真菌Paecilomyces sp· Tree 1-7代谢产物的研究

用硅胶柱色谱及重结晶等方法,并根据理化性质和光谱数据从一株红树林共生真菌Paecilomycessp.Tree 1-7中分离鉴定了四个化合物,分别为secalon ic ac id A(1),tenellic ac id A(2),大黄素(3),大黄酚(4)。并首次对化合物1和2进行了抗人的肝癌细胞hepG2的活性测试。结果显示,化合物1有很强的抗肿瘤活性,IC50=2.0μg/m l,化合物2的抗肿瘤活性IC50=62.1μg/m l。

一种基于分类-回归决策树的烤烟产区识别模型

为建立不同产区及风格烤烟烟叶与化学指标之间的关系,对2005~2009年湖南省、河南省、福建省和云南省等4省份共计1040份烟叶样品的21种化学成分进行检测,并利用Mining Tree模型进行分类-回归决策树（C＆RT analysis）数据分析。结果表明：（1）从21个烟叶化学成分中,共筛选出14个特征化学指标,其中钾和硝酸根为4省烟叶共有特征化学指标;（2）湖南、河南、福建和云南4省份识别概率最高的特征化学指标分段组合（Segment）分别为：还原糖（≤24.93%）、钾（〉1.98%）、p H值（≤5.37）（p=0.658）;钾（≤1.98%）、灰分（〉11.03%）、水溶性灰份碱度（≤0.49）、硝酸根（≤0.06%）（p=0.776）;还原糖（〉24.93%）、钾（〉1.98%）、硝酸根（≤0.06%）、蛋白质（〉5.01%）（p=0.914）;钾（≤1.98%）、灰分（≤11.03%）、水溶性总糖（〉28.94%）、硫酸根（≤1.43%）（p=0.957）。分类-回归决策树方法在建立烤烟化学成分识别模型中具有重要的应用价值。

基于改进FPN与SVM的树障检测方法

针对目前无人机搭载传感器的树障检测方法无法实现自动检测的问题,提出一种基于改进的FPN(特征金字塔网络)与SVM(支持向量机)的树障检测算法。在传统的FPN基础上,进行自下而上的反向侧边连接并融合,采用ResNet 50(深度残差网络)和改进的FPN作为特征提取网络得到特征向量,并将其输入基于遗传算法的SVM中进行二分类,进而判断所检测图像中是否存在树障隐患。实验结果表明,本算法用于树障检测的准确率达到93.4%,处理图像的平均速度达到每秒11张,漏检率和误检率较低,具有较强的泛化能力。

基于自适应全局定位算法的带钢表面缺陷检测

针对热轧带钢表面缺陷检测存在的智能化水平低、检测精度低和检测速度慢等问题,提出了一种基于自适应全局定位网络(Adaptive global localization network,AGLNet)的深度学习缺陷检测算法.首先,引入一种残差网络(Residual network,ResNet)与特征金字塔网络(Feature pyramid network,FPN)集成的特征提取结构,减少缺陷语义信息在层级传递间的消失;其次,提出基于TPE(Tree-structure Parzen estimation)的自适应树型候选框提取网络(Adaptive treestructure region proposal extraction network,AT-RPN),无需先验知识的积累,避免了人为调参的训练模式;最后,引入全局定位回归算法,以全局定位的模式在复杂的缺陷检测中实现缺陷更精确定位.本文实现一种快速、准确、更智能化、更适用于实际应用的热轧带钢表面缺陷的算法.实验结果表明,AGLNet在NEU-DET热轧带钢表面缺陷数据集上的检测速度保持在11.8帧/s,平均精度达到79.90%,优于目前其他深度学习带钢表面缺陷检测算法.另外,该算法还具备较强的泛化能力.

基于卡尔曼滤波的红外图像增强算法

针对红外图像中的非均匀性噪声的去除问题,提出基于卡尔曼滤波的红外图像去噪及增强算法.在Bayes-ian-MAP框架下分析卡尔曼滤波器对去噪问题的适用性.由于成像电路内部温度上升和参数的细微变化,每个像元的固定模式噪声(FPN)在帧间缓慢变化.基于此点,建立暗帧的噪声模型.将卡尔曼滤波器作用于红外暗帧序列,估计出暗帧中每个像元的FPN水平.引入噪声影响因子(NIF)来评估FPN噪声对像元输出信号的影响.根据NIF自适应地选取每个像元的FPN噪声权重.实际带噪图像减去加权FPN噪声,即得到增强图像.将该算法应用于实拍红外图像,用平均灰度梯度(GMG)评估算法的性能.在目标区域,GMG下降了5.1%,说明算法在去噪的同时很好地保留了目标的边缘.而在平滑区域,GMG下降了85.5%.结果表明,该算法在去除非均匀性噪声,提高图像的对比度方面,取得较好的效果.

一种基于频繁模式树的正负关联规则挖掘算法

当前关联规则挖掘主要着眼于正关联规则,如A→B的关联规则的挖掘,这种单一的只对正关联规则的挖掘方式存在严重的弊端,他掩盖了数据之间存在的隐含负关联规则,进而无法得出一些正关联规则中某些项目间相互制约的负关联关系。在关联规则概念和性质的基础上提出了基于频繁模式树的拓展式的正、负项目的关联规则挖掘算法,通过对数据库的遍历形成前缀链表,不仅挖掘包含所有正项目的关联规则,而且还能够挖掘出所有包含负项目的关联规则,不会造成负关联规则的淹没。并对算法的效率和可行性进行分析,该算法在描述关联规则项目间的相互独立程度上比已有的单一挖掘负项目的关联规则算法更具优势。

Tracking leaf area index and coefficient of light extinction over the harvesting cycle of black wattle

The amount of photosynthetic radiation inter- cepted by a crop is a function of the incident solar radiation on the plants, the leaf area index （LAI）, and the light extinction coefficient （k）. We quantified LAI and k in stands of black wattle （Acacia mearnsii De Wild.） over a 7-year growth cycle at two locations in the state of Rio Grande do Sul, Brazil. Our study was conducted in commercial stands in agroecological regions with high densities of black wattle plantations. LAI was calculated as the ratio between the leaf area of a tree and its planting space, and k was derived from Beer＇s law. LAI depends on the planting site and stand age. Between the two sites, the LAI was similar over time, the amount of variation differed. Values of k depended only on stand age, with the highest average observed for stands up to 5 years old. The trend of k during the plantation cycle was inversely proportional to LAI and was correlated with LAI, leaf area, leaf dry mass, canopy volume, height, branches dry mass, total dry mass, and crown diameter.

Dynamics of Soil N<span>H</span>⁺<sub style="margin-left:-10px;">4</sub>-N and N<span>O</span>^-<sub style="margin-left:-4px;">3</sub>-N in Rubber Plantation in Wenchang, South China

In this study, the dynamics of soil available nitrogen in rubber plantations was studied to provide a basis for the sustainable management of soil nutrients in rubber plantations. Soil samples were collected from 0 - 20 cm and 21 - 40 cm depth of rubber plantations in Wenchang area of Hainan Province to determine the soil ammonium nitrogen and nitrate nitrogen in March, June, September and December. The results showed that the total available nitrogen in the rubber plantation soil was mainly composed of nitrate nitrogen (>57%), and this nitrate nitrogen was always the absolute composition advantage in the total available nitrogen composition, and this composition advantage did not change with the change of soil depth. Nitrate and ammonium nitrogen in rubber plantation soil showed obvious seasonal variation characteristics in 0 - 20 cm soil layer and 21 - 40 cm soil depth. With the seasonal time change, the contents of soil nitrate and ammonium nitrogen increased first and then decreased. Summer and autumn were the most active periods of soil nitrate and ammonium nitrogen. There was a significant positive correlation between soil nitrate nitrogen content and ammonium nitrogen content (r = 0.6532**). Based on the correlation between soil ammonium nitrogen content and nitrate nitrogen content, the content of ammonium nitrogen in soil can be estimated according to the amount of soil nitrate nitrogen.

卡培他滨联合奥沙利铂治疗晚期或转移性结直肠癌的成本-效果分析

目的:对卡培他滨联合奥沙利铂(Cape OX方案)和5-氟尿嘧啶/亚叶酸钙联合奥沙利铂(m FOLFOX6方案)治疗晚期或转移性结直肠癌进行经济学评价。方法:基于Cape OX方案和m FOLFOX6方案治疗晚期或转移性结直肠癌疗效评价的Meta分析结果,各方案的终点结局为用药后不同疾病状态。以直接医疗成本进行成本计算,用Tree Age Pro软件构建决策树,进行成本-效果分析及敏感性分析。结果:与m FOLFOX6组比较,Cape OX组的成本-效果比值大,影响因素分析及单因素敏感性分析表明成本-效果分析的结果在有些情况下不太稳定,如采用国内品牌卡培他滨计算成本则Cape OX组的成本-效果比值小。结论:Cape OX组和m FOLFOX6组疗效相当,而成本-效果分析表明完成相应治疗周期后,采用国内品牌卡培他滨的Cape OX组更具有经济学优势。应用该结论时要注意相关概率和成本选择对结果稳定性的影响。