沙漠化过程土壤种子库特征的研究

采用野外定点取样与室内萌发实验相结合的方法 ,研究了 4种不同类型退化沙地土壤种子库特征 ,主要结论是 :(1)土壤种子库密度随沙漠化程度的增加而下降 ,不同沙漠化发展阶段下降速率不同 ,从固定到半固定沙地是种子库密度衰减最快的时期 ;(2 )土壤种子库植物种数从固定到半流动沙地变化很小 ,半流动到流动沙地衰减速度明显加快 ,是种子库植物种数衰减最快的时期 ;(3) 4种类型退化沙地土壤种子库的种类组成以 1、2年生草本植物为主 ,所占比例 79%～ 88% ,多年生草本和灌木所占比例很小 ;(4 )土壤种子库物种多样性指数并非随着沙漠化程度的增加而下降 ,其中半流动沙地的物种多样性指数最小 ;(5 ) 4种类型退化沙地土壤种子库组成上的相似性系数变化范围为 0 .82 9～ 1.0 0 0 ,表现出较小的空间异质性。     

Analysis of the association among three viruses infecting hop in Australia

Associations among Hop latent virus (HpLV), Hop mosaic virus (HpMV), and Apple mosaic virus (ApMV) were assessed in five hop cultivars at four commercial hop-growing regions in Victoria and Tasmania, Australia. The presence or absence of each virus was confirmed by double-antibody sandwich enzyme-linked immunosorbent assay (DAS-ELISA). Spatial patterns of virus-infected plants were characterized using the Spatial Analysis by Distance IndicEs ( sadie ) system of pattern analysis. The association among viruses (occurrence and covariation) was assessed using the Jaccard similarity index, Spearman's rank correlation coefficient, and sadie . The spatial pattern of plants infected by HpLV and HpMV ranged from random to highly aggregated depending upon the cultivar infected and the mean disease incidence. The spatial pattern of plants infected by ApMV was aggregated in six of the seven plots where ApMV was present. A strong positive association between HpLV and HpMV was found in all cultivars at all locations. This association may be the result of the viruses sharing a common aphid vector species, the presence of one virus enhancing the ability of the aphid vector to acquire the other virus either through transencapsidation or influences on virus titre, or mixed infections within source plants. Significant associations, positive or negative, were found less frequently between HpLV and ApMV, and HpMV and ApMV.     

基于FCN的无人机玉米遥感图像垄中心线提取

为解决农业机器人在玉米田行间行走的全局路径规划问题,该研究提出一种基于全卷积神经网络(Fully ConvolutionalNetworks,FCN)的无人机玉米遥感图像垄中心线提取方法。基于无人机获取的高精度可见光遥感图像,设计了针对农田垄中心线提取的数据集标注方法,采用滑动窗口法进行图像分块,利用深度学习语义分割网络FCN对垄中心线附近7～17像素宽度范围的垄线区域进行提取,模型在测试田块上精确率达66.1%～83.4%,召回率达51.1%～73.9%,调和平均值为57.6%～78.4%;对拼接后的图像使用影像分割投影法提取中心线,探究了垄线区域宽度对垄中心线提取精度的影响,训练采用9像素的垄区域宽度,可得到垄中心线在77 mm左右偏差范围准确率为91.2%,在31.5 mm左右偏差范围内为61.5%。结果表明,基于FCN对无人机玉米遥感图像进行处理,可得到整片田地的垄中心线栅格地图,方便农业机器人进行全局路径规划。     

基于红外热成像的夜间农田实时语义分割

农田环境实时语义分割是构成智能农机的视觉环境感知的重要环节，夜间农田语义分割可以使智能农机在夜间通过视觉感知农田环境进行全天候作业，而夜间无光环境下，可见光摄像头成像效果较差，将造成语义分割精度的下降。为保证夜间农田环境下红外图像语义分割的精度与实时性，该研究提出了一种适用于红外图像的红外实时双边语义分割网络（Infrared Real-time Bilateral Semantic Segmentation Network，IR-BiSeNet），根据红外图像分辨率低，细节模糊的特点该网络在实时双边语义分割网络（Bilateral Semantic Segmentation Net，BiSeNet）结构基础上进行改进，在其空间路径上，进一步融合红外图像低层特征，在该网络构架中的注意力提升模块、特征融合模块上使用全局最大池化层替换全局平均池化层以保留红外图像纹理细节信息。为验证提出方法的有效性，通过在夜间使用红外热成像采集的农田数据集上进行试验，数据集分割目标包括田地、行人、植物、障碍物、背景。经试验验证，提出方法在夜间农田红外数据集上达到了85.1%的平均交并比（Mean Intersection over Union，MIoU），同时达到40帧/s的处理速度，满足对夜间农田的实时语义分割。     

医学图像精细化分割方法研究

指出了深度学习方法在医学图像分割中取得较大的进展,但医学图像处理的复杂性使得全自动分割方法难以取得较好的分割效果。在卷积网络分割的基础上,结合应用场景使用适当的后处理手段来提升图像的分割效果是一种比较有临床意义的研究方法。主要研究了基于概率图的全连接条件随机场模型和基于用户交互的深度编辑网络,并分析总结了这两种方法的实现原理和各自优势,并对未来的研究工作进行了一些展望。     

ON STUDY DIGGED FIELD MOVEMENT''S LAW BY EXPERIMENTAL METHOD OF LASER HOLOGRAPHIC MOIRE INTERFERMETRY

The high-precision displacement measuring technology of optic-measuring mechanics is, for the first time,successfully applied on the similar model of the model land-layer with a new-type variable modulus of elasticity. The angle of movement and the distributed displacement of upper and lower rock-layers are measured. The technology is a combination of laser's technology and the modern chemical sticker-it is called the sticking moire interferometry. Compared with the measured values, the result is very satisfactory.     

Reasoning Method Based on Motorcycle Design

According to the needs of setting up a professional system based on motorcycle design,the CBR(case based reasoning) and RBR(rule based reasoning) methods are introduced.After discussing the difference between the two kinds of reasoning,a method of integrating them into the motorcycle design flow is put forward to realize the design reusable technology and improve the design efficiency,by using the CBR in the step of selecting the similar motorcycle and the RBR in project's modification according to correlative rules.     

Knowledge Representation for Rotative Machine Fault Diagnosis Based on Semantic Net

Aiming at knowledge representation problem of expert system for rotary machinery fault diagnosis, semantic net knowledge representation is discussed. Semantic net to represent expert knowledge of rotary machinery fault diagnosis is explored. Its realizing method is afforded in computer language, in addition, demostrating model is also developed in Visual C++. The result shows good validity of the knowledge representation.     

Weighting Coefficient Optimization Unification of Lexicographic Method in Fuzzy Multi -objective Optimization Design

The Lexicographic method has subjective and empirical character,so a weighting coefficient is introduced by considering the importance and proportion of each individual object,the preferable result of every individual object can be confirmed by comparing the degree of similarity between the perfect one and the ecumenic one and the Fuzzy optimization will be perfect.The effectiveness of these methods is illustrated by a fuzzy optimal design of the three-bar truss.     

跌水坎流体动力性噪声模拟试验研

以流体声学相似原理为依据，对三种跌水坎形式进行噪声模拟试验。利用古典声学理论对背景噪声进行分析与修正，评估降噪的效果，确定跌水坎的修改形式。并对修改方案进行不同比尺的试验，对声强进行曲线拟合，绘制趋势线，推求原型噪声的大小，为工程建设提供参考依据。