Video camouflaged object detection(VCOD)has become a fundamental task in computer vision that has attracted significant attention in recent years.Unlike image camouflaged object detection(ICOD),VCOD not only requires ...Video camouflaged object detection(VCOD)has become a fundamental task in computer vision that has attracted significant attention in recent years.Unlike image camouflaged object detection(ICOD),VCOD not only requires spatial cues but also needs motion cues.Thus,effectively utilizing spatiotemporal information is crucial for generating accurate segmentation results.Current VCOD methods,which typically focus on exploring motion representation,often ineffectively integrate spatial and motion features,leading to poor performance in diverse scenarios.To address these issues,we design a novel spatiotemporal network with an encoder-decoder structure.During the encoding stage,an adjacent space-time memory module(ASTM)is employed to extract high-level temporal features(i.e.,motion cues)from the current frame and its adjacent frames.In the decoding stage,a selective space-time aggregation module is introduced to efficiently integrate spatial and temporal features.Additionally,a multi-feature fusion module is developed to progressively refine the rough prediction by utilizing the information provided by multiple types of features.Furthermore,we incorporate multi-task learning into the proposed network to obtain more accurate predictions.Experimental results show that the proposed method outperforms existing cutting-edge baselines on VCOD benchmarks.展开更多
Convective initiations(CIs)in western Jiangnan,China,were examined using radar data spanning April–September 2018–2021.Our approach combined objective identification and subjective validation to identify,track,and v...Convective initiations(CIs)in western Jiangnan,China,were examined using radar data spanning April–September 2018–2021.Our approach combined objective identification and subjective validation to identify,track,and validate the CIs,thereby producing a highly accurate CI dataset.Using this dataset,we investigated the spatiotemporal variations and environmental conditions associated with CIs and revealed distinct seasonal and diurnal patterns of CI events.Spatially,CIs occur more frequently south of the Nanling Mountains and less frequently in the north.Seasonally,they were most frequent from June to August,and least frequent in April and September,following a unimodal distribution.The CIs exhibited pronounced afternoon convection,particularly from June to August,when most occurred between 11:00 and 19:00 local time(UTC+8 h).Terrain significantly influenced the spatial variation in the CIs.North of the Nanling Mountains,CIs occurred near higher mountains,whereas south of the range,they were concentrated near smaller mountains and along the Guangdong coast.Using K-means clustering,CIs that could develop into Mesoscale Convective Systems were classified into four circulation types:Western Pacific subtropical high(WPSH)control(TypeⅠ),WPSH edge(TypeⅡ),southwest airflow(TypeⅢ),and low trough shear(TypeⅣ).The CIs of TypesⅠandⅡwere primarily attributed to afternoon thermal convection,which occurs under conditions of high moisture and thermal instability.Triggers for these CIs mostly occurred near high-elevation terrain.In contrast,TypesⅢandⅣwere driven primarily by the synergy of abundant moisture conditions and synoptic dynamic factors,such as low-level jets,upper-level troughs,and shear lines.These types exhibited higher frequency in the south,with high-frequency CI trigger zones particularly observed in regions with strong moisture-flux convergence and near complex terrain.展开更多
Coronavirus disease 2019(COVID-19)is continuing to spread globally and still poses a great threat to human health.Since its outbreak,it has had catastrophic effects on human society.A visual method of analyzing COVID-...Coronavirus disease 2019(COVID-19)is continuing to spread globally and still poses a great threat to human health.Since its outbreak,it has had catastrophic effects on human society.A visual method of analyzing COVID-19 case information using spatio-temporal objects with multi-granularity is proposed based on the officially provided case information.This analysis reveals the spread of the epidemic,from the perspective of spatio-temporal objects,to provide references for related research and the formulation of epidemic prevention and control measures.The case information is abstracted,descripted,represented,and analyzed in the form of spatio-temporal objects through the construction of spatio-temporal case objects,multi-level visual expressions,and spatial correlation analysis.The rationality of the method is verified through visualization scenarios of case information statistics for China,Henan cases,and cases related to Shulan.The results show that the proposed method is helpful in the research and judgment of the development trend of the epidemic,the discovery of the transmission law,and the spatial traceability of the cases.It has a good portability and good expansion performance,so it can be used for the visual analysis of case information for other regions and can help users quickly discover the potential knowledge this information contains.展开更多
Spatiotemporal data represent the real-world objects that move in geographic space over time.The enormous numbers of mobile sensors and location tracking devices continuously produce massive amounts of such data.This ...Spatiotemporal data represent the real-world objects that move in geographic space over time.The enormous numbers of mobile sensors and location tracking devices continuously produce massive amounts of such data.This leads to the need for scalable spatiotemporal data management systems.Such systems shall be capable of representing spatiotemporal data in persistent storage and in memory.They shall also provide a range of query processing operators that may scale out in a cloud setting.Currently,very few researches have been conducted to meet this requirement.This paper proposes a Hadoop extension with a spatiotemporal algebra.The algebra consists of moving object types added as Hadoop native types,and operators on top of them.The Hadoop file system has been extended to support parameter passing for files that contain spatiotemporal data,and for operators that can be unary or binary.Both the types and operators are accessible for the MapReduce jobs.Such an extension allows users to write Hadoop programs that can perform spatiotemporal analysis.Certain queries may call more than one operator for different jobs and keep these operators running in parallel.This paper describes the design and implementation of this algebra,and evaluates it using a benchmark that is specific to moving object databases.展开更多
Three-dimensional(3D)visual perception systems hold transformative potential for artificial intelligence by enabling machines to interpret dynamic environments with human-like spatial awareness[1,2].In biological visi...Three-dimensional(3D)visual perception systems hold transformative potential for artificial intelligence by enabling machines to interpret dynamic environments with human-like spatial awareness[1,2].In biological vision,the human brain seamlessly integrates temporal and spatial information through hierarchical processing:retinal signals are decomposed into near/far spatial features by the visual cortex,while neural synapses employ spatiotemporal filtering to prioritize moving objects through attention mechanisms[3,4].This integration allows humans to simultaneously perceive an object’s depth,trajectory,and velocity,critical for tasks like catching a ball or navigating crowded spaces[5].展开更多
With the development of the modern information society, more and more multimedia information is available. So the technology of multimedia processing is becoming the important task for the irrelevant area of scientist...With the development of the modern information society, more and more multimedia information is available. So the technology of multimedia processing is becoming the important task for the irrelevant area of scientist. Among of the multimedia, the visual informarion is more attractive due to its direct, vivid characteristic, but at the same rime the huge amount of video data causes many challenges if the video storage, processing and transmission.展开更多
摘要Video camouflaged object detection(VCOD)has become a fundamental task in computer vision that has attracted significant attention in recent years.Unlike image camouflaged object detection(ICOD),VCOD not only requires spatial cues but also needs motion cues.Thus,effectively utilizing spatiotemporal information is crucial for generating accurate segmentation results.Current VCOD methods,which typically focus on exploring motion representation,often ineffectively integrate spatial and motion features,leading to poor performance in diverse scenarios.To address these issues,we design a novel spatiotemporal network with an encoder-decoder structure.During the encoding stage,an adjacent space-time memory module(ASTM)is employed to extract high-level temporal features(i.e.,motion cues)from the current frame and its adjacent frames.In the decoding stage,a selective space-time aggregation module is introduced to efficiently integrate spatial and temporal features.Additionally,a multi-feature fusion module is developed to progressively refine the rough prediction by utilizing the information provided by multiple types of features.Furthermore,we incorporate multi-task learning into the proposed network to obtain more accurate predictions.Experimental results show that the proposed method outperforms existing cutting-edge baselines on VCOD benchmarks.
基金Major Project of Hunan Provincial Natural Science Foundation(2021JC0009)CMA High-Impact Weather Key Open Laboratory Project(2024-G-10)+1 种基金Hunan Provincial Natural Science Foundation Youth Fund Project(2023JJ40369)Natural Science Foundation of China Meteorological Joint Fund(U2242201)。
摘要Convective initiations(CIs)in western Jiangnan,China,were examined using radar data spanning April–September 2018–2021.Our approach combined objective identification and subjective validation to identify,track,and validate the CIs,thereby producing a highly accurate CI dataset.Using this dataset,we investigated the spatiotemporal variations and environmental conditions associated with CIs and revealed distinct seasonal and diurnal patterns of CI events.Spatially,CIs occur more frequently south of the Nanling Mountains and less frequently in the north.Seasonally,they were most frequent from June to August,and least frequent in April and September,following a unimodal distribution.The CIs exhibited pronounced afternoon convection,particularly from June to August,when most occurred between 11:00 and 19:00 local time(UTC+8 h).Terrain significantly influenced the spatial variation in the CIs.North of the Nanling Mountains,CIs occurred near higher mountains,whereas south of the range,they were concentrated near smaller mountains and along the Guangdong coast.Using K-means clustering,CIs that could develop into Mesoscale Convective Systems were classified into four circulation types:Western Pacific subtropical high(WPSH)control(TypeⅠ),WPSH edge(TypeⅡ),southwest airflow(TypeⅢ),and low trough shear(TypeⅣ).The CIs of TypesⅠandⅡwere primarily attributed to afternoon thermal convection,which occurs under conditions of high moisture and thermal instability.Triggers for these CIs mostly occurred near high-elevation terrain.In contrast,TypesⅢandⅣwere driven primarily by the synergy of abundant moisture conditions and synoptic dynamic factors,such as low-level jets,upper-level troughs,and shear lines.These types exhibited higher frequency in the south,with high-frequency CI trigger zones particularly observed in regions with strong moisture-flux convergence and near complex terrain.
基金National Key Research and Development Program of China,No.2016YFB0502300。
摘要Coronavirus disease 2019(COVID-19)is continuing to spread globally and still poses a great threat to human health.Since its outbreak,it has had catastrophic effects on human society.A visual method of analyzing COVID-19 case information using spatio-temporal objects with multi-granularity is proposed based on the officially provided case information.This analysis reveals the spread of the epidemic,from the perspective of spatio-temporal objects,to provide references for related research and the formulation of epidemic prevention and control measures.The case information is abstracted,descripted,represented,and analyzed in the form of spatio-temporal objects through the construction of spatio-temporal case objects,multi-level visual expressions,and spatial correlation analysis.The rationality of the method is verified through visualization scenarios of case information statistics for China,Henan cases,and cases related to Shulan.The results show that the proposed method is helpful in the research and judgment of the development trend of the epidemic,the discovery of the transmission law,and the spatial traceability of the cases.It has a good portability and good expansion performance,so it can be used for the visual analysis of case information for other regions and can help users quickly discover the potential knowledge this information contains.
摘要Spatiotemporal data represent the real-world objects that move in geographic space over time.The enormous numbers of mobile sensors and location tracking devices continuously produce massive amounts of such data.This leads to the need for scalable spatiotemporal data management systems.Such systems shall be capable of representing spatiotemporal data in persistent storage and in memory.They shall also provide a range of query processing operators that may scale out in a cloud setting.Currently,very few researches have been conducted to meet this requirement.This paper proposes a Hadoop extension with a spatiotemporal algebra.The algebra consists of moving object types added as Hadoop native types,and operators on top of them.The Hadoop file system has been extended to support parameter passing for files that contain spatiotemporal data,and for operators that can be unary or binary.Both the types and operators are accessible for the MapReduce jobs.Such an extension allows users to write Hadoop programs that can perform spatiotemporal analysis.Certain queries may call more than one operator for different jobs and keep these operators running in parallel.This paper describes the design and implementation of this algebra,and evaluates it using a benchmark that is specific to moving object databases.
基金supported by the National Key R&D Program of China(2024YFB3614500)the National Natural Science Foundation of China(52373194,52403301,and U24A20293),and the Haihe Laboratory of Sustainable Chemical Transformations.
摘要Three-dimensional(3D)visual perception systems hold transformative potential for artificial intelligence by enabling machines to interpret dynamic environments with human-like spatial awareness[1,2].In biological vision,the human brain seamlessly integrates temporal and spatial information through hierarchical processing:retinal signals are decomposed into near/far spatial features by the visual cortex,while neural synapses employ spatiotemporal filtering to prioritize moving objects through attention mechanisms[3,4].This integration allows humans to simultaneously perceive an object’s depth,trajectory,and velocity,critical for tasks like catching a ball or navigating crowded spaces[5].
摘要With the development of the modern information society, more and more multimedia information is available. So the technology of multimedia processing is becoming the important task for the irrelevant area of scientist. Among of the multimedia, the visual informarion is more attractive due to its direct, vivid characteristic, but at the same rime the huge amount of video data causes many challenges if the video storage, processing and transmission.