深度学习支持下的宅基地复垦项目真实性智能审查技术研究与应用

全域综合整治背景下的宅基地复垦项目审查是保证复垦项目合法性合规性的有效手段,针对当前宅基地复垦项目审查效率低的问题,本文提出了基于遥感影像的农村宅基地复垦项目立项真实性审查方法。首先以重庆市农村宅基地复垦项目为例,建立宅基地复垦项目立项真实性审查模型,利用改进的U-Net模型准确提取建筑物,然后与复垦红线叠加实现复垦项目真实性的智能审查。试验结果表明,该方法能够快速准确地审查复垦项目真实性,对优化土地结构、促进土地流转、实现乡村振兴具有重要意义,可为全域综合整治背景下宅基地复垦项目真实性审查提供技术参考。     

用于盐湖提锂的聚酰胺复合纳滤膜制备及其性能研究

纳滤膜技术以其环境友好、装置集成度高和智能化控制等优势,在卤水锂资源绿色开发领域具有良好的应用前景。实验通过添加聚乙烯醇对纳滤膜功能层进行改性,考察了聚乙烯醇相对含量对纳滤膜性能的影响。结果表明,聚乙烯醇参与界面聚合反应,纳滤膜功能层O/N比增加,聚乙烯醇分子链间氢键作用降低纳滤膜表面粗糙度。羟基官能团的引入有效改善纳滤膜材料亲水效果,膜面羧基密度随聚乙烯醇相对添加量的增加逐渐增加。聚乙烯醇改性纳滤膜截留效果优异,纳滤膜对镁离子的截留率高于95%,锂离子截留率低于-44%,不可逆通量衰减指数低于15%。聚乙烯醇改性纳滤膜材料具有较好的镁锂分离选择效果及抗污染能力,为用于盐湖卤水镁锂分离国产化纳滤膜材料的开发提供新思路。     

基于ALBERT模型的园林植物知识实体与关系抽取方法

园林植物知识图谱可为顾及区域适应性、观赏性和生态性等因子的绿化树种的选型提供知识支持。植物描述文本的实体识别及关系抽取是知识图谱构建的关键环节。针对植物领域未有公开的标注数据集,本文阐述了园林植物数据集的构建流程,定义了园林植物的概念体系结构,完成了园林植物语料库的构建。针对现有Word2vec、ELMo和BERT等语言模型存在无法解决多义词、融合上下文能力差、运行速度慢等缺点,提出了嵌入ALBERT（A Lite BERT）预训练语言模型的实体识别和关系抽取模型。ALBERT预训练的动态词向量能够有效地表示文本特征,将其分别输入到BiGRU-CRF命名实体识别模型和BiGRU-Attention关系抽取模型中进行训练,进一步提升实体识别和关系抽取的效果。在园林植物语料库上进行方法的有效性验证,结果表明ALBERT-BiGRU-CRF命名实体识别模型的F₁值为0.9517,ALBERT-BiGRU-Attention关系抽取模型的F₁值为0.9161,相较于经典的语言模型（如Word2vec、ELMo和BERT等）性能有较为显著的提升。因此基于ALBERT模型的实体与关系抽取任务能有效提高识别分类效果,可将其应用于植物描述文本的实体关系抽取任务中,为园林植物知识图谱自动构建提供方法。     

基于残差神经网络改进的密集人群计数方法

为避免密集人群踩踏事件发生,从监控图像中准确获取密集人群人数信息非常重要。针对密集人群计数难度大、人群目标小、场景尺度变化大等特点,本文提出一种新型神经网络结构VGG-ResNeXt。本网络使用VGG-16的前10层作粗粒度特征提取器,使用改进的残差神经网络作为细粒度特征提取器。利用改进的残差神经网络“多通道,共激活”的特点,使得单列式人群计数神经网络获得了多列式人群计数网络的优点（即从小目标、多尺度的密集人群图像中提取更多人群特征）,同时避免了多列式人群计数网络训练难度大、结构冗余等缺点。实验结果表明本模型在UCF-CC-50数据集、ShangHaiTech B数据集和UCF-QNRF数据集中取得了最高精度,MAE指标分别优于其他同期模型7.5%、18.8%和2.4%,证明了本模型的在计数精度方面的有效性。本研究成果可以有效帮助城市管理,有效缓解公安疏导压力,保障人民生命财产安全。     

基于深度学习的滑坡灾害易发性分析

滑坡灾害成因机理复杂、影响因素众多,深度学习作为当前人工智能领域的热点,能够更好地模拟滑坡灾害的形成并准确预测潜在的斜坡。为了挖掘深度学习在滑坡易发性的应用潜能,本文构建了一维、二维和三维的滑坡数据表达形式,并提出3种基于卷积神经网络模型（Convolutional Neural Networks, CNN）的滑坡易发性分析处理框架：基于CNN分类器、基于CNN与逻辑回归的融合和基于CNN集成,最后以江西省铅山县为研究对象进行验证,结果表明：所有基于CNN的易发性模型都能够获得准确且可靠的滑坡易发性分析结果。其中,基于二维数据的CNN模型在所有单分类器中预测精度最高,为78.95%。此外,二维CNN特征提取能够显著提升逻辑回归的预测精度,其准确率提升7.9%。最后,异质集成策略能够大幅度提升基于CNN分类器的滑坡预测精度,其准确率提升4.35%~8.78%。     

从硫酸锂粗矿中分离富集锂的实验研究

针对从西藏查波措盐湖卤水中获得的硫酸锂粗矿,开展从该矿物中分离提取锂的研究。根据该矿物的组成,设计了去除与锂离子共存的镁、硫酸根等主要离子的工艺路线,通过除镁、硫酸根、钙等过程和蒸发浓缩、结晶分离等步骤,得到了杂质离子含量较低的富锂溶液。实验对除镁步骤的加水量、加料方式、反应时间、Ca O加入量等反应关键控制因素进行了考察,确定了较优的分离条件。     

人影高炮作业数据采集器的移动电源设计

高炮作业是人工影响天气的日常工作,其作业数据的采集与收集尤为重要。高炮作业数据采集器是实现作业数据的采集与存储,一般采用锂电池供电。针对锂电池容量低而无法满足采集器的长时间使用,常在高炮作业的关键时刻电量不足影响高炮作业顺利进行的问题,设计了一种人影高炮作业数据采集器的专用移动电源。该专用移动电源由充放电保护电路、电芯、电芯容量指示电路、DC-DC升压电路、及CN3703充电管理电路等部分组成。经过最终测试,该设计符合实际使用要求、成本低、性能稳定,具有较强的工程应用价值。     

Molecularly imprinted polymers for selective extraction of crystal violet from natural seawater coupled with high-performance liquid chromatographic determination

Molecularly imprinted polymers(MIPs) were prepared by the bulk polymerization using crystal violet as the template molecule, and the methacrylic acid and ethylene glycol dimetheacrylate as functional monomer and cross-linker, respectively. Systematic investigations of synthetic conditions were conducted. The surface morphology and recognition mechanism of the obtained polymers were studied using scanning electron microscope and spectrophotometric analysis. MIPs showed high affinity to template molecule and were successfully applied as special solid-phase extraction sorbent for selective extraction of crystal violet from natural seawater. An off-line molecularly imprinted solid-phase extraction(MISPE) method followed by high-performance liquid chromatography with diodearray detection for the analysis of crystal violet was also established. MISPE columns have good recoveries for crystal violet standard solutions and good linearity was obtained over the concentration range of 0-200 μg L-1(R2 0.99). Finally, two natural seawater samples were investigated. The recoveries of spiked seawater on the MISPE columns were from 44.47% to 62.34%, the relative standard deviation(n=3) being in the range of 2.89%-5.96%.     

基于航摄影像的滇池捞鱼河湿地植物类型提取

为了获取滇池湿地的植物类型分布信息,以捞鱼河湿地为示范区,采用分辨率为0.15 m的航摄影像为影像数据源,利用人机交互和面向对象相结合的影像解译方法,实现捞鱼河湿地植物类型分布信息的提取.在植物类型边界提取过程中,发现过高的影像分辨率(≤0.3m)不利于成片具有空间异质性植物类型边界的正常提取,容易出现同一类型过度分割情况.通过邻域降分辨率重采样技术,将影像分辨率降至0.6m时,能有效解决过度分割问题.当分辨率降至更低时,虽然能解决过度分割问题,但会导致地物边界出现不同程度的锯齿而与地物边界不符的后果.研究所应用的基于Feature Extaction模块的植物类型边界提取方法,能有效减少人机交互方法中人工勾绘边界的工作量,在保证分类精度的同时减少人力投入,对航摄影像在大范围湿地植物类型的提取中具有重要借鉴意义.     

Organic geochemistry of marine source rocks and pyrobitumen-containing reservoir rocks of the Sichuan Basin and neighbouring areas,SW China

Three bitumen fractions were obtained and systematically analysed for the terpane and sterane composition from 30 Paleozoic source rocks and 64 bitumen-containing reservoir rocks within the Upper Sinian, Lower Cambrian, Lower Silurian, Middle Carboniferous, Upper Permian and Lower Triassic strata in the Sichuan Basin and neighbouring areas, China. These bitumen fractions include extractable oils (bitumen I), oil-bearing fluid inclusions and/or closely associated components with the kerogen or pyrobitumen/mineral matrix, released during kerogen or pyrobitumen isolation and demineralization (bitumen II), and bound compounds within the kerogen or pyrobitumen released by confined pyrolysis (bitumen III). In addition, atomic H/C and O/C ratios and carbon isotopic compositions of kerogen and pyrobitumen from some of the samples were measured. Geochemical results and geological information suggest that: (1) in the Central Sichuan Basin, hydrocarbon gases in reservoirs within the fourth section of the Upper Sinian Dengying Formation were derived from both the Lower Cambrian and Upper Sinian source rocks; and (2) in the Eastern Sichuan Basin, hydrocarbon gases in Middle Carboniferous Huanglong Formation reservoirs were mainly derived from Lower Silurian source rocks, while those in Upper Permian and Lower Triassic reservoirs were mainly derived from both Upper Permian and Lower Silurian marine source rocks. For both the source and reservoir rocks, bitumen III fractions generally show relatively lower maturity near the peak oil generation stage, while the other two bitumen fractions show very high maturities based on terpane and sterane distributions. Tricyclic terpanes evolved from the distribution pattern C₂₀ < C₂₁ < C₂₃, through C₂₀ < C₂₁ > C₂₃, finally to C₂₀ > C₂₁ > C₂₃ during severe thermal stress. The concentration of C₃₀ diahopane in bitumen III (the bound components released from confined pyrolysis) is substantially lower than in the other two bitumen fractions for four terrigenous Upper Permian source rocks, demonstrating that this compound originated from free hopanoid precursors, rather than hopanoids bound to the kerogen.