Detection of cigarettes and other tobacco products by giant African pouched rats (Cricetomys gambianus)

If the illicit tobacco trade were eliminated, governments could gain at least $31.3 billion a year, and more than 164,000 premature deaths a year could be avoided after 2030 (Joossens, Merriman, Ross, and Raw, 2009). Dogs have been used successfully in tobacco control programs, and there is a good chance that rats could also play an important role. In the present experiment, giant African pouched rats were trained to respond to filters that had been stored together with cigarettes (i.e., soaked) and to not respond to filters that had been soaked with noncigarette items. Generalization to untrained types of tobacco was then tested. The sensitivity of 4 rats trained on filters soaked with 1 of 7 types of cigarettes ranged from 86% to 100% (mean, 95%). There was very little evidence of generalization when the rats were tested on tobacco leaves and snuff but good evidence of generalization when the rats were tested on cigarettes that had been soaked with strong-smelling additives. These findings suggest that rats may be a valuable asset in the global effort to control illicit cigarette trade.     

基于聚焦滤波的林业专题地图制图综合方法

森林资源规划设计调查成果总是以连续的面状图斑呈现。当大比例尺二类调查成果图缩编为更小比例尺地图时,常常需要制图综合。如何在制图综合时既要简化多边形边界,又要保持各类型林相面积和位置的相对不变,是林业专题制图的一个难题。经大量实验发现,将实施制图综合的连续面状图斑要素从矢量格式转换为栅格格式,经Focal滤波器以Majority方式滤波处理后,再将滤波后结果从栅格数据转换为矢量数据,可较好地实现森林资源专题连续面状要素信息的制图综合。以云南省某地一个1万 hm²的实验区为例,使用该方法将1∶2.5万的森林资源二类调查成果图缩编为1∶25万。结果表明,缩编后,多边形边界简化,目视效果较好,各类型林相面积变化绝对值平均为0.2%,位置精确度平均值为94.68%。此方法已应用于云南森林资源状况图集相关专题图等的生产实践。     

地图缩编中小图斑综合问题研究

随着城市的发展,人们对各种数字地图中各种比例尺的需求越来越大.对土地利用数据的综合成为了多比例尺地图生产的一种方式.该研究分析了在地图缩编过程针对小图斑综合存在的问题,提出自己的小图斑综合方案与评价方案:通过建立小图斑制图综合的规则,确定最小上图面积,利用专家操作经验设定合理综合范围形成数学模型使小图斑的选取更客观、更具科学性.     

基于Transformer的强泛化苹果叶片病害识别模型

模型泛化能力是病害识别模型多场景应用的关键,该研究针对不同环境下的苹果叶片病害数据,提出一种可以提取多类型特征的强泛化苹果叶片病害识别模型CaTNet。该模型采用双分支结构,首先设计了一种卷积神经网络分支,负责提取苹果叶片图像的局部特征,其次构建了具有挤压和扩充功能的视觉Transformer分支,该分支能够提取苹果叶片图像的全局特征,最后将两种特征进行融合,使Transformer分支可以学习局部特征,使卷积神经网络分支学习全局特征。与多种卷积神经网络模型和Transformer模型相比,该模型具有更好的泛化能力,仅需学习实验室环境叶片数据,即可在自然环境数据下达到80%的识别精度,相较卷积神经网络EfficientNetV2提升7.21个百分点,相较Transformer网络PVT提升26.63个百分点,能够有效提升对不同环境数据的识别精度,解决了深度学习模型训练成本高,泛化能力弱的问题。     

A computerized soil information system for arable land in Japan

A computerized soil information system was compared to manual punch card methods for storage and retrieval of Japanese soil survey data. In spite of the high cost, a soil information system using computers is the most appropriate system for the future, because by this method more than three factors can be analyzed simultaneously, various kinds of soil map data can be stored, retrieved and updated and many kinds of interpretative maps can be constructed automatically. There are large amounts of valuable soil survey data available in Japan which previously were used only for drawing soil maps. These should be utilized to develop a computerized soil information system. At the end of this report, essential processes and the possible structure of a Japanese Soil Information System were discussed.     

工程堆积体坡面砾石分布及含量概化

为室内试验中更好地模拟工程堆积体,利用图像分析法和经典统计学方法对6大典型水蚀区的368座工程堆积体中粒径D1cm的坡面砾石的分布和含量进行概化分析,得到工程堆积体下垫面上中下坡位砾石含量的概化区间。结果表明:工程堆积体坡面各粒径砾石含量大多服从正态分布,且山区和平原丘陵区之间的砾石总含量有显著性差异。工程堆积体坡面砾石重力分选作用明显,且随着砾石粒径的增加,分选程度增大。在室内进行工程堆积体下垫面设计时,山区堆积体的砾石总含量宜设置在50%~60%,其中,细粒砾石含量宜设置在20%~30%,中粒砾石含量宜设置在10%~15%,粗粒砾石和巨粒砾石含量宜设置为10%,堆积体上中下坡位坡面砾石含量配比宜设置为1∶1.3∶1.5;平原丘陵区的砾石总含量宜设置在15%~20%,其中细粒砾石、中粒砾石、粗粒砾石、巨粒砾石含量分别宜设置为5%~10%,3%~7%,2%~5%,2%~3%,堆积体上中下坡位坡面砾石含量配比宜设置为1∶1.5∶2.5。     

基于规则的水资源模拟配置模型

针对区域性水资源系统结构复杂、要素繁多的系统特性,给出了适用于多水源、多工程、多用户的大型水资源系统的系统概化规则和模型计算原则。介绍了反映各种水源及工程供水特点的建模思路,并在配置规则的基础上建立了水资源模拟模型,该模型在河北省水资源合理配置中进行了实际应用,配置结果合理可靠。     

A geospatial and temporal framework for modeling gaseous N and other N losses from forest soils and basins,with application to the Turkey Lakes Watershed Project,in Ontario,Canada

This article quantifies pre- to post-harvest gaseous N emissions and other N losses from forest soils and basins geospatially and temporally via digital elevation and hydrological modeling, using daily rain, snow and air temperature records, annual atmospheric N deposition rates, and basin-specific soil and forest specifications as input. The approach relates gaseous N losses from soils to soil temperature and water-filled pore space (WFPS) as affected by the depth-to-water (DTW) below the soil surface. The approach is applied to the Turkey Lakes Watershed Project (TLW) in Ontario, 60 km north of Sault St. Marie, where basin-wide N losses due to denitrification would mostly be restricted to the wetland portions of the basin. Basin-wide N losses via denitrification and stream export (mineral N and dissolved organic N) were empirically related to upland N mineralization and soil leaching as controlling processes. The calibrated model calculations, set to conform to the field-monitored N concentrations in TLW streams, suggest that the harvest-induced nitrification and denitrification pulses would be strongest near the end of the first post-harvest year, dropping to background levels within about 4–5 years later. The article concludes with assessing basin-specific denitrification efficiencies in relation to atmospheric N deposition and basin-to-basin wetland coverage.     

对农业技术自组织推广能力提升问题的理论探讨

农业技术自组织推广能力包括推广的资金自组织筹运、人员素质自组织提升、信息自组织收送、风险自组织处理、组织自组织缔结等能力。现代农业建设的推进、政府他组织推广思路的转变和农业持久发展动力的获得,对自组织推广能力的提升提出了客观要求;而推广体系的耗散结构特征及其内部组分间竞争与协同的存在,为其形成提供了可能性。应从耗散结构特征的强化及内部组分间竞争与协同的促成等,提升农业技术自组织推广能力。     

土地利用数据综合中图斑多边形决策推理

针对土地利用数据综合中图斑的化简处理,将智能体技术引入到图斑的综合处理过程中,建立了顾及多种条件的图斑重要性判断的决策机制,将可操作的几何、语义控制指标融入到推理中,实现了从图斑重要性判断到尺度变换算子选择的自动化过程,最后通过实际生产进行检验,结果显示该过程能够大幅度地提高生产效率,并能够很好地维护土地利用数据的面积平衡性以及全覆盖、无重叠、无缝隙的特征.该研究可为多比例尺国土资源数据库提供技术支持.