Relations between damage symptoms and nutritional status of Norway spruce stands (Picea abies Karst.) in southwestern Germany

Since the mid-1970s new types of forest damage are observed in the Federal Republic of Germany. Typical damage symptoms in coniferous tree species are needle yellowing, reddening and premature loss of older foliage. Needle analysis carried out in damaged Norway spruce stands indicated that these phenomena, most of all needle yellowing, are influenced by the nutritional status of the trees. For the investigated sites Mg but also other elements, particularly P, S, K, Ca and Zn may play a causal role. As N is the most abundant element in the plant its form of uptake is important for a balanced nutrition. On the contrary to NO₃-N high NH₄-N uptake reduces the uptake of cations, particularly Mg and Ca, eventually inducing foliar discoloration symptoms on specific sites. Fertilizer experiments indicated that Mg fertilization is an appropriate tool to mitigate forest damages associated with Mg deficiency for a sustained time period. The application of N in the form of NH₄-N may impede Mg uptake. This mechanism may cause severe damage in forest areas receiving continuously high NH₄-N deposition.     

林产化学加工学科的现况与发展前景

本文了介绍了并论述了林产化学加工学科所含的五个专业方向，木材化学加工工程，树木提取物化学工程，林产生物化工，木材保护与改性，林产化工装备与控制的研究内容，工业现状，存在问题与发展前景。     

基于SURF与Hough森林的人脸检测研究

为实现复杂场景中的人脸检测与定位,提出了一种基于快速鲁棒特征SURF与Hough森林的人脸检测算法。采用SURF局部特征构建Hough森林分类器,每个叶子节点存储类别信息与特征点到达目标中心的偏移量,在图像局部外观与Hough投票之间建立映射关系,生成有监督的判别式的码本,获得可靠的概率Hough投票,以此预测目标中心位置,提高了检测精度。与此同时,采用SURF局部特征提取图像兴趣点有助于减小计算量、加快检测速度。实验证明了所提算法的有效性。     

Two‐step consecutive reaction model and kinetic parameters relevant to the decomposition of Chinese forest fuels

Decomposition of some Chinese forest fuels has been studied by means of nonisothermal thermogravimetric analysis in oxidative and inert atmosphere at low heating rates. A comparison between thermogravimetric curves obtained in air and nitrogen shows that the existence of oxygen enhances the decomposition rate and changes the mechanisms of thermal degradation. After the water evaporation, two well‐defined decomposition stages have been observed in thermogravimetric curves obtained in air, which correspond to oxidative degradation of main components and oxidation of char formed. A kinetic model named “two‐step consecutive reaction model” is developed to describe the thermal degradation process of these materials in air, and there is a good agreement between the experimental and calculated thermogravimetric and derivative thermogravimetric curves. © 2006 Wiley Periodicals, Inc. J Appl Polym Sci 102: 571–576, 2006     

The effect of nitrogenous fertilizers on the nitrification of forest soils

The effect of fertilization with urea and ammonium nitrate on nitrogen mineralization was studied in a series of laboratory incubation experiments. The samples (humus layer) were collected from field experiments with different applications of fertilizers during a period of 11–14 years. The nitrogen fertilization influenced some chemical properties of the humus layer such as pH, N-Kjeldahl content and the content of inorganic nitrogen, especially in the samples from North Sweden. Nitrate nitrogen was formed in humus samples classified as non-nitrifying in several cases, usually after urea fertilization.     

森林生态税收体系研究现状

由于森林生态系统具有正的外部性,对其生态效益进行补偿就成为必须要考虑的问题。森林生态税作为一种为森林生态效益补偿筹集资金的手段,在发达国家已经有着不错的实施效果,而在我国基本上还处于空白。本文回顾了国内外对于森林生态税收体系的研究进展,比较了国内外在这一领域研究的异同,提出了在我国进行森林生态税研究的一些建议和看法。     

生物质能利用技术的研究及发展

我国的生物质能资源主要是农业废弃物、禽畜类便和林业废弃物。生物质能的利用方式有:直接燃烧、产生沼气等可燃气、发电、转化为液体燃料和加工成高密度的固体燃料。文中对几种利用方式进行了讨论。     

面向非平衡数据集的金融欺诈账户检测研究

针对非平衡金融数据集,提出一种银行欺诈账户检测框架iForest-SMOTE。基于账户的动态交易特点,从统计、时序、监督信息维度抽取账户交易行为特征。针对过采样技术ADASYN在金融账户数据集中存在的跨区域样本合成问题,提出一种基于iForest算法的数据集均衡预处理策略,通过iForest算法对数据进行混合采样,在去除多数类噪声数据的同时降低分类器对少数类的学习难度。在此基础上,设计随机森林分类器实现金融欺诈账户检测。在真实金融账户交易数据集上进行实验,结果表明,与ADASYN、SMOTE等采样技术相比,iForest-SMOTE在召回率和准确率方面具有明显优势,F-value值至少能够提升2.13个百分点。     

基于三维场景和图像边缘特征的浅浮雕建模

针对数字浅浮雕设计,通过对已有的浅浮雕建模方法进行分析,提出一种结合深度场边缘特征自动生成浅浮雕的方法。从三维场景中获取深度图,再将深度图分割成块,利用局部图像结构的优点,用随机结构化森林来学习精确且计算有效的边缘检测器,得到的边缘图再进行处理得到平滑且细节增强的浅浮雕轮廓;视边缘连续图为概率场,提出根据概率场对深度场的梯度域进行自适应的非线性压缩;通过求解泊松方程来恢复浅浮雕的高度场。实验结果表明,该方法对于大部分三维场景能在合理时间内自动生成效果良好的浅浮雕模型。     

环状扫描的强深度森林

深度森林（Deep Forest,DF）,由于此模型超参数少,且参数设置没有过多的要求,训练方便,鲁棒性高,因此在处理大型数据时比神经网络算法更加具有优势。但是,传统的深度森林中,多粒度扫描忽略了边缘数据携带的隐含信息,无法充分地获得各个特征子集,进而会对以后的级联部分产生影响。而且,级联部分每次得到的新特征有限,影响了模型的表征学习能力。针对以上问题,提出一种环状强深度森林（Circular Strong Deep Forest,CSDF）,其通过环状扫描过程,一定程度上得到更充分的特征子集,且强级联森林通过特征选择提高了模型的表征学习能力。经过在不同数据集上的测试,结果表明,CSDF的性能更加优越,尤其是高维数据上更为明显。