伪造文本检测是保证社交安全的重要条件。融合多模特征注意力机制的伪造文本检测,实现多类型伪造文本的检测,并提高其精度。针对目前数据集的不足,根据不同的文本生成模型GPT-2、Grover、LSTM等制作了对应的多伪造类型数据集。融合GLTR...伪造文本检测是保证社交安全的重要条件。融合多模特征注意力机制的伪造文本检测,实现多类型伪造文本的检测,并提高其精度。针对目前数据集的不足,根据不同的文本生成模型GPT-2、Grover、LSTM等制作了对应的多伪造类型数据集。融合GLTR、Grover和LP(Language and Physical)3个多模特征,使用注意力机制将该3个模型输出进行融合。设计时空特征融合网络充分提取时序信息和局部空间信息。最后添加分类层完成伪造文本的分类。实验结果表明:融合多模特征和时空特征融合网络可提取更多的伪造信息,同时充分融合了时序和局部空间信息,提升了伪造文本的表征能力,且泛化能力也优于以往网络。展开更多
Wood transportation begins from the landing or the forest road where the processed wood assortments are placed, they are loaded on trucks and then transported to the wood-working industry and/or the consumption center...Wood transportation begins from the landing or the forest road where the processed wood assortments are placed, they are loaded on trucks and then transported to the wood-working industry and/or the consumption centers. Wood transportation in Greece is carried out, almost exclusively, by trucks. The aim of this study is to analyze the evolution of timber trucking in Greece through the analysis of wood transportation waybills. A total of 1,471 waybills have been collected and analyzed that cover the period 1981-2008 from nine of the most productive and representative forest districts in Greece, in terms of timber production. The final dataset included information on the type and technical characteristics of truck types, the transported wood quantities and haulage distances. Furthermore, data on the wood transportation cost for the year 2010 have been collected and analyzed. According to the results, the number of two-axle trucks that dominated in the 1980s is reducing rapidly, mostly to the benefit of larger three-axle and four-axle trucks. More than three quarters （77%） of wood transportation is conducted in the period June-October. Mean vehicle load per vehicle type differs between conifer and broadleaved species from 7.1% up to 23.9%. Transportation of stacked wood with semitrailers is conducted with overloaded trucks, which could entail serious problems for public safety and the condition of road network. New design standards will enhance serviceability of the wood transportation routes and environmental protection.展开更多
Recent developments of 30 m global land characterization datasets （e.g., land cover, vegetation continues field） represent the finest spatial resolution inputs for global scale studies. Here, we present results from...Recent developments of 30 m global land characterization datasets （e.g., land cover, vegetation continues field） represent the finest spatial resolution inputs for global scale studies. Here, we present results from further improvement to land cover map- ping and impact analysis of spatial resolution on area estimation for different land cover types. We proposed a set of methods to aggregate two existing 30 m resolution circa 2010 global land cover maps, namely FROM-GLC （Finer Resolution Observa- tion and Monitoring-Global Land Cover） and FROM-GLC-seg （Segmentation）, with two coarser resolution global maps on development, i.e., Nighttime Light Impervious Surface Area （NL-ISA） and MODIS urban extent （MODIS-urban）, to produce an improved 30 m global land cover map-FROM-GLC-agg （Aggregation）. It was pos-processed using additional coarse res- olution datasets （i.e., MCD12Q1, GlobCover2009, MOD44W etc.） to reduce land cover type confusion. Around 98.9% pixels remain 30 m resolution after some post-processing to this dataset. Based on this map, majority aggregation and proportion ag- gregation approaches were employed to create a multi-resolution hierarchy （i.e., 250 m, 500 m, 1 km, 5 km, 10 km, 25 km, 50 km, 100 km） of land cover maps to meet requirements for different resolutions from different applications. Through accuracy assessment, we found that the best overall accuracies for the post-processed base map （at 30 m） and the three maps subse- quently aggregated at 250 m, 500 m, 1 km resolutions are 69.50%, 76.65%, 74.65%, and 73.47%, respectively. Our analysis of area-estimation biases for different land cover types at different resolutions suggests that maps at coarser than 5 km resolution contain at least 5% area estimation error for most land cover types. Proportion layers, which contain precise information on land cover percentage, are suggested for use when coarser resolution land cover data are required.展开更多
Land cover is one of the most basic input elements of land surface and climate models. Currently, the direct and indirect effects of land cover data on climate and climate change are receiving increasing attentions. I...Land cover is one of the most basic input elements of land surface and climate models. Currently, the direct and indirect effects of land cover data on climate and climate change are receiving increasing attentions. In this study, a high resolution(30 m) global land cover dataset(Globe Land30) produced by Chinese scientists was, for the first time, used in the Beijing Climate Center Climate System Model(BCC_CSM) to assess the influences of land cover dataset on land surface and climate simulations. A two-step strategy was designed to use the Globe Land30 data in the model. First, the Globe Land30 data were merged with other satellite remote sensing and climate datasets to regenerate plant functional type(PFT) data fitted for the BCC_CSM. Second, the up-scaling based on an area-weighted approach was used to aggregate the fine-resolution Globe Land30 land cover type and area percentage with the coarser model grid resolutions globally. The Globe Land30-based and the BCC_CSM-based land cover data had generally consistent spatial distribution features, but there were some differences between them. The simulation results of the different land cover type dataset change experiments showed that effects of the new PFT data were larger than those of the new glaciers and water bodies(lakes and wetlands). The maximum value was attained when dataset of all land cover types were changed. The positive bias of precipitation in the mid-high latitude of the northern hemisphere and the negative bias in the Amazon, as well as the negative bias of air temperature in part of the southern hemisphere, were reduced when the Globe Land30-based data were used in the BCC_CSM atmosphere model. The results suggest that the Globe Land30 data are suitable for use in the BCC_CSM component models and can improve the performance of the land and atmosphere simulations.展开更多
文摘伪造文本检测是保证社交安全的重要条件。融合多模特征注意力机制的伪造文本检测,实现多类型伪造文本的检测,并提高其精度。针对目前数据集的不足,根据不同的文本生成模型GPT-2、Grover、LSTM等制作了对应的多伪造类型数据集。融合GLTR、Grover和LP(Language and Physical)3个多模特征,使用注意力机制将该3个模型输出进行融合。设计时空特征融合网络充分提取时序信息和局部空间信息。最后添加分类层完成伪造文本的分类。实验结果表明:融合多模特征和时空特征融合网络可提取更多的伪造信息,同时充分融合了时序和局部空间信息,提升了伪造文本的表征能力,且泛化能力也优于以往网络。
文摘Wood transportation begins from the landing or the forest road where the processed wood assortments are placed, they are loaded on trucks and then transported to the wood-working industry and/or the consumption centers. Wood transportation in Greece is carried out, almost exclusively, by trucks. The aim of this study is to analyze the evolution of timber trucking in Greece through the analysis of wood transportation waybills. A total of 1,471 waybills have been collected and analyzed that cover the period 1981-2008 from nine of the most productive and representative forest districts in Greece, in terms of timber production. The final dataset included information on the type and technical characteristics of truck types, the transported wood quantities and haulage distances. Furthermore, data on the wood transportation cost for the year 2010 have been collected and analyzed. According to the results, the number of two-axle trucks that dominated in the 1980s is reducing rapidly, mostly to the benefit of larger three-axle and four-axle trucks. More than three quarters （77%） of wood transportation is conducted in the period June-October. Mean vehicle load per vehicle type differs between conifer and broadleaved species from 7.1% up to 23.9%. Transportation of stacked wood with semitrailers is conducted with overloaded trucks, which could entail serious problems for public safety and the condition of road network. New design standards will enhance serviceability of the wood transportation routes and environmental protection.
基金supported by the National High-tech R&D Program of China(Grant No.2009AA12200101)the National Natural Science Foundation of China(Grant No.41301445)+1 种基金an Open Fund from the State Key Laboratory of Remote Sensing Science(Grant No.OFSLRSS201202)a research grant from Tsinghua University(Grant No.2012Z02287)
文摘Recent developments of 30 m global land characterization datasets （e.g., land cover, vegetation continues field） represent the finest spatial resolution inputs for global scale studies. Here, we present results from further improvement to land cover map- ping and impact analysis of spatial resolution on area estimation for different land cover types. We proposed a set of methods to aggregate two existing 30 m resolution circa 2010 global land cover maps, namely FROM-GLC （Finer Resolution Observa- tion and Monitoring-Global Land Cover） and FROM-GLC-seg （Segmentation）, with two coarser resolution global maps on development, i.e., Nighttime Light Impervious Surface Area （NL-ISA） and MODIS urban extent （MODIS-urban）, to produce an improved 30 m global land cover map-FROM-GLC-agg （Aggregation）. It was pos-processed using additional coarse res- olution datasets （i.e., MCD12Q1, GlobCover2009, MOD44W etc.） to reduce land cover type confusion. Around 98.9% pixels remain 30 m resolution after some post-processing to this dataset. Based on this map, majority aggregation and proportion ag- gregation approaches were employed to create a multi-resolution hierarchy （i.e., 250 m, 500 m, 1 km, 5 km, 10 km, 25 km, 50 km, 100 km） of land cover maps to meet requirements for different resolutions from different applications. Through accuracy assessment, we found that the best overall accuracies for the post-processed base map （at 30 m） and the three maps subse- quently aggregated at 250 m, 500 m, 1 km resolutions are 69.50%, 76.65%, 74.65%, and 73.47%, respectively. Our analysis of area-estimation biases for different land cover types at different resolutions suggests that maps at coarser than 5 km resolution contain at least 5% area estimation error for most land cover types. Proportion layers, which contain precise information on land cover percentage, are suggested for use when coarser resolution land cover data are required.
基金supported by the National High Technology Research and Development Program of China (Grant No. 2009AA122005)the Public Welfare Meteorology Research Project of China (Grant Nos. 201506023, 201306048)the National Natural Science Foundation of China (Grant Nos. 41275076, 40905046)
文摘Land cover is one of the most basic input elements of land surface and climate models. Currently, the direct and indirect effects of land cover data on climate and climate change are receiving increasing attentions. In this study, a high resolution(30 m) global land cover dataset(Globe Land30) produced by Chinese scientists was, for the first time, used in the Beijing Climate Center Climate System Model(BCC_CSM) to assess the influences of land cover dataset on land surface and climate simulations. A two-step strategy was designed to use the Globe Land30 data in the model. First, the Globe Land30 data were merged with other satellite remote sensing and climate datasets to regenerate plant functional type(PFT) data fitted for the BCC_CSM. Second, the up-scaling based on an area-weighted approach was used to aggregate the fine-resolution Globe Land30 land cover type and area percentage with the coarser model grid resolutions globally. The Globe Land30-based and the BCC_CSM-based land cover data had generally consistent spatial distribution features, but there were some differences between them. The simulation results of the different land cover type dataset change experiments showed that effects of the new PFT data were larger than those of the new glaciers and water bodies(lakes and wetlands). The maximum value was attained when dataset of all land cover types were changed. The positive bias of precipitation in the mid-high latitude of the northern hemisphere and the negative bias in the Amazon, as well as the negative bias of air temperature in part of the southern hemisphere, were reduced when the Globe Land30-based data were used in the BCC_CSM atmosphere model. The results suggest that the Globe Land30 data are suitable for use in the BCC_CSM component models and can improve the performance of the land and atmosphere simulations.
