Stream networks are considered important units in many environmental decision making processes. The extraction of streams using digital elevation models (DEMs) presents many advantages. However it is very sensitive to...Stream networks are considered important units in many environmental decision making processes. The extraction of streams using digital elevation models (DEMs) presents many advantages. However it is very sensitive to the uncertainty of the elevation datasets used. The main aim of this paper is to implement geostatistical simulations and assess the propagated uncertainty and map the error of location streams. First, point sampled elevations are used to fit a variogram model. Next two hundred DEM realizations are generated using conditional sequential Gaussian simulation;the stream network map is extracted for each of these realizations, and the collection of stream networks is analyzed to quantify the error propagation. At each grid cell, the probability of the occurrence of a stream and the propagated error are estimated. The more probable stream network are delineated and compared with the digital stream network derived from topographic map. The method is illustrated using a small dataset (8742 sampled elevations) for Anaguid Saharan platform. All computations are run in two free softwares: R and SAGA. R is used to fit variogram and to run sequential Gaussian simulation. SAGA is used to extract streams via RSAGA library.展开更多
Micro-expressions(ME)recognition is a complex task that requires advanced techniques to extract informative features fromfacial expressions.Numerous deep neural networks(DNNs)with convolutional structures have been pr...Micro-expressions(ME)recognition is a complex task that requires advanced techniques to extract informative features fromfacial expressions.Numerous deep neural networks(DNNs)with convolutional structures have been proposed.However,unlike DNNs,shallow convolutional neural networks often outperform deeper models in mitigating overfitting,particularly with small datasets.Still,many of these methods rely on a single feature for recognition,resulting in an insufficient ability to extract highly effective features.To address this limitation,in this paper,an Improved Dual-stream Shallow Convolutional Neural Network based on an Extreme Gradient Boosting Algorithm(IDSSCNN-XgBoost)is introduced for ME Recognition.The proposed method utilizes a dual-stream architecture where motion vectors(temporal features)are extracted using Optical Flow TV-L1 and amplify subtle changes(spatial features)via EulerianVideoMagnification(EVM).These features are processed by IDSSCNN,with an attention mechanism applied to refine the extracted effective features.The outputs are then fused,concatenated,and classified using the XgBoost algorithm.This comprehensive approach significantly improves recognition accuracy by leveraging the strengths of both temporal and spatial information,supported by the robust classification power of XgBoost.The proposed method is evaluated on three publicly available ME databases named Chinese Academy of Sciences Micro-expression Database(CASMEII),Spontaneous Micro-Expression Database(SMICHS),and Spontaneous Actions and Micro-Movements(SAMM).Experimental results indicate that the proposed model can achieve outstanding results compared to recent models.The accuracy results are 79.01%,69.22%,and 68.99%on CASMEII,SMIC-HS,and SAMM,and the F1-score are 75.47%,68.91%,and 63.84%,respectively.The proposed method has the advantage of operational efficiency and less computational time.展开更多
In this paper, we propose a multi-source multi-path video streaming system for supporting high quality concurrent video-on-demand (VoD) services over wireless mesh networks (WMNs), and leverage forward error correctio...In this paper, we propose a multi-source multi-path video streaming system for supporting high quality concurrent video-on-demand (VoD) services over wireless mesh networks (WMNs), and leverage forward error correction to enhance the error resilience of the system. By taking wireless interference into consideration, we present a more realistic networking model to capture the characteristics of WMNs and then design a route selection scheme using a joint rate/interference-distortion optimiza- tion framework to help the system optimally select concurrent streaming paths. We mathematically formulate such a route selec- tion problem, and solve it heuristically using genetic algorithm. Simulation results demonstrate the effectiveness of our proposed scheme.展开更多
In peer-assisted streaming network, service provider needs to pre-pushing resources to some users for caching. This scheme can increase the resource availability. In traditional strategy, in order to guarantee the use...In peer-assisted streaming network, service provider needs to pre-pushing resources to some users for caching. This scheme can increase the resource availability. In traditional strategy, in order to guarantee the user＇s quality of service （QoS）, the pre-pushing action can only start at the time point that the certain user has not used the computer for more than 20 minutes. If the user comes back, the pre-pushing action will be stopped. We claim that this is not an efficient scheme. In this paper, we propose a novel pre-pushing scheme to improve the pre-pushing efficiency, while meeting the user＇s QoS requirements. The basic idea is using the user＇s available bandwidth as much as possible in the condition of meeting the user＇s QoS requirements. Then we design an available bandwidth calculating strategy. The numerical simulation demonstrates that our proposed scheme outperforms the traditional one.展开更多
With the continual growth of the variety and complexity of network crime means, the traditional packet feature matching cannot detect all kinds of intrusion behaviors completely. It is urgent to reassemble network str...With the continual growth of the variety and complexity of network crime means, the traditional packet feature matching cannot detect all kinds of intrusion behaviors completely. It is urgent to reassemble network stream to perform packet processing at a semantic level above the network layer. This paper presents an efficient TCP stream reassembly mechanism for real-time processing of high-speed network traffic. By analyzing the characteristics of network stream in high-speed network and TCP connection establishment process, several polices for designing the reassembly mechanism are built. Then, the reassembly implementation is elaborated in accordance with the policies. Finally, the reassembly mechanism is compared with the traditional reassembly mechanism by the network traffic captured in a typical gigabit gateway. Experiment results illustrate that the reassembly mechanism is efficient and can satisfy the real-time property requirement of traffic analysis system in high-speed network.展开更多
Over the past years, we have witnessed an explosive growth in the use of multimedia applications such as audio and video streaming with mobile and static devices. Multimedia streaming applications need new approaches ...Over the past years, we have witnessed an explosive growth in the use of multimedia applications such as audio and video streaming with mobile and static devices. Multimedia streaming applications need new approaches to multimedia transmissions to meet the growing volume demand and quality expectations of multimedia traffic. This paper studies network coding which is a promising paradigm that has the potential to improve the performance of networks for multimedia streaming applications in terms of packet delivery ratio (PDR), latency and jitter. This paper examines several network coding protocols for ad hoc wireless mesh networks and compares their performance on multimedia streaming applications with optimized broadcast protocols, e.g., BCast, Simplified Multicast Forwarding (SMF), and Partial Dominant Pruning (PDP). The results show that the performance increases significantly with the Random Linear Network Coding (RLNC) scheme.展开更多
The support for multiple video streams in an ad-hoc wireless network requires appropriate routing and rate allocation measures ascertaining the set of links for transmitting each stream and the encoding rate of the vi...The support for multiple video streams in an ad-hoc wireless network requires appropriate routing and rate allocation measures ascertaining the set of links for transmitting each stream and the encoding rate of the video to be delivered over the chosen links. The routing and rate allocation procedures impact the sustained quality of each video stream measured as the mean squared error (MSE) distortion at the receiver, and the overall network congestion in terms of queuing delay per link. We study the trade-off between these two competing objectives in a convex optimization formulation, and discuss both centralized and dis- tributed solutions for joint routing and rate allocation for multiple streams. For each stream, the optimal allocated rate strikes a balance between the selfish motive of minimizing video distortion and the global good of minimizing network congestions, while the routes are chosen over the least-congested links in the network. In addition to detailed analysis, network simulation results using ns-2 are presented for studying the optimal choice of parameters and to confirm the effectiveness of the proposed measures.展开更多
With the rapid growth of the Internet in recent years, the ability to analyze and identify its users has become increasingly important. Authorship analysis provides a means to glean information about the author of a d...With the rapid growth of the Internet in recent years, the ability to analyze and identify its users has become increasingly important. Authorship analysis provides a means to glean information about the author of a document originating from the internet or elsewhere, including but not limited to the author’s gender. There are well-known linguistic differences between the writing of men and women, and these differences can be effectively used to predict the gender of a document’s author. Capitalizing on these linguistic nuances, this study uses a set of stylometric features and a set of word count features to facilitate automatic gender discrimination on emails from the popular Enron email dataset. These features are used in conjunction with the Modified Balanced Winnow Neural Network proposed by Carvalho and Cohen, an improvement on the original Balanced Winnow created by Littlestone. Experiments with the Modified Balanced Winnow show that it is effectively able to discriminate gender using both stylometric and word count features, with the word count features providing superior results.展开更多
As the Internet offers increased connectivity between human beings, it has fallen prey to malicious users who exploit its resources to gain illegal access to critical information. In an effort to protect computer netw...As the Internet offers increased connectivity between human beings, it has fallen prey to malicious users who exploit its resources to gain illegal access to critical information. In an effort to protect computer networks from external attacks, two common types of Intrusion Detection Systems (IDSs) are often deployed. The first type is signature-based IDSs which can detect intrusions efficiently by scanning network packets and comparing them with human-generated signatures describing previously-observed attacks. The second type is anomaly-based IDSs able to detect new attacks through modeling normal network traffic without the need for a human expert. Despite this advantage, anomaly-based IDSs are limited by a high false-alarm rate and difficulty detecting network attacks attempting to blend in with normal traffic. In this study, we propose a StreamPreDeCon anomaly-based IDS. StreamPreDeCon is an extension of the preference subspace clustering algorithm PreDeCon designed to resolve some of the challenges associated with anomalous packet detection. Using network packets extracted from the first week of the DARPA '99 intrusion detection evaluation dataset combined with Generic Http, Shellcode and CLET attacks, our IDS achieved 94.4% sensitivity and 0.726% false positives in a best case scenario. To measure the overall effectiveness of the IDS, the average sensitivity and false positive rates were calculated for both the maximum sensitivity and the minimum false positive rate. With the maximum sensitivity, the IDS had 80% sensitivity and 9% false positives on average. The IDS also averaged 63% sensitivity with a 0.4% false positive rate when the minimal number of false positives is needed. These rates are an improvement on results found in a previous study as the sensitivity rate in general increased while the false positive rate decreased.展开更多
Dynamic adaptation of multimedia content is seen as an important feature of next generation networks and pervasive systems enabling terminals and applications to adapt to changes in e.g. context, access network, and a...Dynamic adaptation of multimedia content is seen as an important feature of next generation networks and pervasive systems enabling terminals and applications to adapt to changes in e.g. context, access network, and available Quality-of-Service(QoS) due to mobility of users, devices or sessions. We present the architecture of a multimedia stream adaptation service which enables communication between terminals having heterogeneous hardware and software capabilities and served by heterogeneous networks. The service runs on special content adaptation nodes which can be placed at any location within the network. The flexible structure of our architecture allows using a variety of different adaptation engines. A generic transcoding engine is used to change the codec of streams. An MPEG-21 Digital Item Adaptation (DIA) based transformation engine allows adjusting the data rate of scalable media streams. An intelligent decision-taking engine implements adaptive flow control which takes into account current network QoS parameters and congestion information. Measurements demonstrate the quality gains achieved through adaptive congestion control mechanisms under conditions typical for a heterogeneous network.展开更多
The rising popularity of online social networks (OSNs), such as Twitter, Facebook, MySpace, and LinkedIn, in recent years has sparked great interest in sentiment analysis on their data. While many methods exist for id...The rising popularity of online social networks (OSNs), such as Twitter, Facebook, MySpace, and LinkedIn, in recent years has sparked great interest in sentiment analysis on their data. While many methods exist for identifying sentiment in OSNs such as communication pattern mining and classification based on emoticon and parts of speech, the majority of them utilize a suboptimal batch mode learning approach when analyzing a large amount of real time data. As an alternative we present a stream algorithm using Modified Balanced Winnow for sentiment analysis on OSNs. Tested on three real-world network datasets, the performance of our sentiment predictions is close to that of batch learning with the ability to detect important features dynamically for sentiment analysis in data streams. These top features reveal key words important to the analysis of sentiment.展开更多
多特征模态融合时存在噪声的叠加,而为减小模态间的差异采用的级联方式的结构也未充分利用模态间的特征信息,因此设计一种跨模态双流交替交互网络(DAINet)方法。首先,构建双流交替增强(DAE)模块,以交互双分支形式融合模态特征,并通过学...多特征模态融合时存在噪声的叠加,而为减小模态间的差异采用的级联方式的结构也未充分利用模态间的特征信息,因此设计一种跨模态双流交替交互网络(DAINet)方法。首先,构建双流交替增强(DAE)模块,以交互双分支形式融合模态特征,并通过学习模态数据的映射关系,以红外-可见光-红外(IR-VIS-IR)和可见光-红外-可见光(VIS-IR-VIS)的双向反馈调节实现模态间噪声的交叉抑制;然后,构建跨模态特征交互(CMFI)模块,并引入残差结构将红外-可见光模态内以及模态间的低层特征和高层特征进行有效融合,从而减小模态间的差异并充分利用模态间的特征信息;最后,在自建红外-可见光多模态台风数据集及RGB-NIR多模态公开场景数据集上进行实验,以验证DAE模块和CMFI模块的有效性。实验结果表明,与简单级联融合方法相比,所提的基于DAINet的特征融合方法在自建台风数据集上的红外模态和可见光模态上的总体分类精度分别提高了6.61和3.93个百分点,G-mean值分别提高了6.24和2.48个百分点,表明所提方法在类别不均衡分类任务上的通用性;所提方法在RGB-NIR数据集上的2种测试模态下的总体分类精度分别提高了13.47和13.90个百分点。同时,所提方法在2个数据集上分别与IFCNN(general Image Fusion framework based on Convolutional Neural Network)和DenseFuse方法进行对比的实验结果表明,所提方法在自建台风数据集上的2种测试模态下的总体分类精度分别提高了9.82、6.02和17.38、1.68个百分点。展开更多
基金This work was supported by the Natural Science Foundation of Jiangsu Province(Grant No.BK20181310)the National Natural Science Foundation of China(Grant No.52079039).
文摘Stream networks are considered important units in many environmental decision making processes. The extraction of streams using digital elevation models (DEMs) presents many advantages. However it is very sensitive to the uncertainty of the elevation datasets used. The main aim of this paper is to implement geostatistical simulations and assess the propagated uncertainty and map the error of location streams. First, point sampled elevations are used to fit a variogram model. Next two hundred DEM realizations are generated using conditional sequential Gaussian simulation;the stream network map is extracted for each of these realizations, and the collection of stream networks is analyzed to quantify the error propagation. At each grid cell, the probability of the occurrence of a stream and the propagated error are estimated. The more probable stream network are delineated and compared with the digital stream network derived from topographic map. The method is illustrated using a small dataset (8742 sampled elevations) for Anaguid Saharan platform. All computations are run in two free softwares: R and SAGA. R is used to fit variogram and to run sequential Gaussian simulation. SAGA is used to extract streams via RSAGA library.
基金supported by the Key Research and Development Program of Jiangsu Province under Grant BE2022059-3,CTBC Bank through the Industry-Academia Cooperation Project,as well as by the Ministry of Science and Technology of Taiwan through Grants MOST-108-2218-E-002-055,MOST-109-2223-E-009-002-MY3,MOST-109-2218-E-009-025,and MOST431109-2218-E-002-015.
文摘Micro-expressions(ME)recognition is a complex task that requires advanced techniques to extract informative features fromfacial expressions.Numerous deep neural networks(DNNs)with convolutional structures have been proposed.However,unlike DNNs,shallow convolutional neural networks often outperform deeper models in mitigating overfitting,particularly with small datasets.Still,many of these methods rely on a single feature for recognition,resulting in an insufficient ability to extract highly effective features.To address this limitation,in this paper,an Improved Dual-stream Shallow Convolutional Neural Network based on an Extreme Gradient Boosting Algorithm(IDSSCNN-XgBoost)is introduced for ME Recognition.The proposed method utilizes a dual-stream architecture where motion vectors(temporal features)are extracted using Optical Flow TV-L1 and amplify subtle changes(spatial features)via EulerianVideoMagnification(EVM).These features are processed by IDSSCNN,with an attention mechanism applied to refine the extracted effective features.The outputs are then fused,concatenated,and classified using the XgBoost algorithm.This comprehensive approach significantly improves recognition accuracy by leveraging the strengths of both temporal and spatial information,supported by the robust classification power of XgBoost.The proposed method is evaluated on three publicly available ME databases named Chinese Academy of Sciences Micro-expression Database(CASMEII),Spontaneous Micro-Expression Database(SMICHS),and Spontaneous Actions and Micro-Movements(SAMM).Experimental results indicate that the proposed model can achieve outstanding results compared to recent models.The accuracy results are 79.01%,69.22%,and 68.99%on CASMEII,SMIC-HS,and SAMM,and the F1-score are 75.47%,68.91%,and 63.84%,respectively.The proposed method has the advantage of operational efficiency and less computational time.
文摘In this paper, we propose a multi-source multi-path video streaming system for supporting high quality concurrent video-on-demand (VoD) services over wireless mesh networks (WMNs), and leverage forward error correction to enhance the error resilience of the system. By taking wireless interference into consideration, we present a more realistic networking model to capture the characteristics of WMNs and then design a route selection scheme using a joint rate/interference-distortion optimiza- tion framework to help the system optimally select concurrent streaming paths. We mathematically formulate such a route selec- tion problem, and solve it heuristically using genetic algorithm. Simulation results demonstrate the effectiveness of our proposed scheme.
基金the National Natural Science Foundation of China under Grant No.61202079,61308018，the China Postdoctoral Science Foundation under Grant No.2014T70031，the Fundamental Research Funds for the Central Universities under Grant No.FRF-TP-14-046A2
文摘In peer-assisted streaming network, service provider needs to pre-pushing resources to some users for caching. This scheme can increase the resource availability. In traditional strategy, in order to guarantee the user＇s quality of service （QoS）, the pre-pushing action can only start at the time point that the certain user has not used the computer for more than 20 minutes. If the user comes back, the pre-pushing action will be stopped. We claim that this is not an efficient scheme. In this paper, we propose a novel pre-pushing scheme to improve the pre-pushing efficiency, while meeting the user＇s QoS requirements. The basic idea is using the user＇s available bandwidth as much as possible in the condition of meeting the user＇s QoS requirements. Then we design an available bandwidth calculating strategy. The numerical simulation demonstrates that our proposed scheme outperforms the traditional one.
基金National High-Tech Research and Development Program of China (863 Program) (No.2007AA01Z309)
文摘With the continual growth of the variety and complexity of network crime means, the traditional packet feature matching cannot detect all kinds of intrusion behaviors completely. It is urgent to reassemble network stream to perform packet processing at a semantic level above the network layer. This paper presents an efficient TCP stream reassembly mechanism for real-time processing of high-speed network traffic. By analyzing the characteristics of network stream in high-speed network and TCP connection establishment process, several polices for designing the reassembly mechanism are built. Then, the reassembly implementation is elaborated in accordance with the policies. Finally, the reassembly mechanism is compared with the traditional reassembly mechanism by the network traffic captured in a typical gigabit gateway. Experiment results illustrate that the reassembly mechanism is efficient and can satisfy the real-time property requirement of traffic analysis system in high-speed network.
文摘Over the past years, we have witnessed an explosive growth in the use of multimedia applications such as audio and video streaming with mobile and static devices. Multimedia streaming applications need new approaches to multimedia transmissions to meet the growing volume demand and quality expectations of multimedia traffic. This paper studies network coding which is a promising paradigm that has the potential to improve the performance of networks for multimedia streaming applications in terms of packet delivery ratio (PDR), latency and jitter. This paper examines several network coding protocols for ad hoc wireless mesh networks and compares their performance on multimedia streaming applications with optimized broadcast protocols, e.g., BCast, Simplified Multicast Forwarding (SMF), and Partial Dominant Pruning (PDP). The results show that the performance increases significantly with the Random Linear Network Coding (RLNC) scheme.
基金Project (No. CCR-0325639) partially supported by the National Science Foundation, USA
文摘The support for multiple video streams in an ad-hoc wireless network requires appropriate routing and rate allocation measures ascertaining the set of links for transmitting each stream and the encoding rate of the video to be delivered over the chosen links. The routing and rate allocation procedures impact the sustained quality of each video stream measured as the mean squared error (MSE) distortion at the receiver, and the overall network congestion in terms of queuing delay per link. We study the trade-off between these two competing objectives in a convex optimization formulation, and discuss both centralized and dis- tributed solutions for joint routing and rate allocation for multiple streams. For each stream, the optimal allocated rate strikes a balance between the selfish motive of minimizing video distortion and the global good of minimizing network congestions, while the routes are chosen over the least-congested links in the network. In addition to detailed analysis, network simulation results using ns-2 are presented for studying the optimal choice of parameters and to confirm the effectiveness of the proposed measures.
文摘With the rapid growth of the Internet in recent years, the ability to analyze and identify its users has become increasingly important. Authorship analysis provides a means to glean information about the author of a document originating from the internet or elsewhere, including but not limited to the author’s gender. There are well-known linguistic differences between the writing of men and women, and these differences can be effectively used to predict the gender of a document’s author. Capitalizing on these linguistic nuances, this study uses a set of stylometric features and a set of word count features to facilitate automatic gender discrimination on emails from the popular Enron email dataset. These features are used in conjunction with the Modified Balanced Winnow Neural Network proposed by Carvalho and Cohen, an improvement on the original Balanced Winnow created by Littlestone. Experiments with the Modified Balanced Winnow show that it is effectively able to discriminate gender using both stylometric and word count features, with the word count features providing superior results.
文摘As the Internet offers increased connectivity between human beings, it has fallen prey to malicious users who exploit its resources to gain illegal access to critical information. In an effort to protect computer networks from external attacks, two common types of Intrusion Detection Systems (IDSs) are often deployed. The first type is signature-based IDSs which can detect intrusions efficiently by scanning network packets and comparing them with human-generated signatures describing previously-observed attacks. The second type is anomaly-based IDSs able to detect new attacks through modeling normal network traffic without the need for a human expert. Despite this advantage, anomaly-based IDSs are limited by a high false-alarm rate and difficulty detecting network attacks attempting to blend in with normal traffic. In this study, we propose a StreamPreDeCon anomaly-based IDS. StreamPreDeCon is an extension of the preference subspace clustering algorithm PreDeCon designed to resolve some of the challenges associated with anomalous packet detection. Using network packets extracted from the first week of the DARPA '99 intrusion detection evaluation dataset combined with Generic Http, Shellcode and CLET attacks, our IDS achieved 94.4% sensitivity and 0.726% false positives in a best case scenario. To measure the overall effectiveness of the IDS, the average sensitivity and false positive rates were calculated for both the maximum sensitivity and the minimum false positive rate. With the maximum sensitivity, the IDS had 80% sensitivity and 9% false positives on average. The IDS also averaged 63% sensitivity with a 0.4% false positive rate when the minimal number of false positives is needed. These rates are an improvement on results found in a previous study as the sensitivity rate in general increased while the false positive rate decreased.
基金Project supported by IST FP6 Integrated Project DAIDALOS (No. IST-2002-506997) and the German Research Foundation (DFG) within the AKOM Framework (No. HA2207/2-3)
文摘Dynamic adaptation of multimedia content is seen as an important feature of next generation networks and pervasive systems enabling terminals and applications to adapt to changes in e.g. context, access network, and available Quality-of-Service(QoS) due to mobility of users, devices or sessions. We present the architecture of a multimedia stream adaptation service which enables communication between terminals having heterogeneous hardware and software capabilities and served by heterogeneous networks. The service runs on special content adaptation nodes which can be placed at any location within the network. The flexible structure of our architecture allows using a variety of different adaptation engines. A generic transcoding engine is used to change the codec of streams. An MPEG-21 Digital Item Adaptation (DIA) based transformation engine allows adjusting the data rate of scalable media streams. An intelligent decision-taking engine implements adaptive flow control which takes into account current network QoS parameters and congestion information. Measurements demonstrate the quality gains achieved through adaptive congestion control mechanisms under conditions typical for a heterogeneous network.
文摘The rising popularity of online social networks (OSNs), such as Twitter, Facebook, MySpace, and LinkedIn, in recent years has sparked great interest in sentiment analysis on their data. While many methods exist for identifying sentiment in OSNs such as communication pattern mining and classification based on emoticon and parts of speech, the majority of them utilize a suboptimal batch mode learning approach when analyzing a large amount of real time data. As an alternative we present a stream algorithm using Modified Balanced Winnow for sentiment analysis on OSNs. Tested on three real-world network datasets, the performance of our sentiment predictions is close to that of batch learning with the ability to detect important features dynamically for sentiment analysis in data streams. These top features reveal key words important to the analysis of sentiment.
文摘多特征模态融合时存在噪声的叠加,而为减小模态间的差异采用的级联方式的结构也未充分利用模态间的特征信息,因此设计一种跨模态双流交替交互网络(DAINet)方法。首先,构建双流交替增强(DAE)模块,以交互双分支形式融合模态特征,并通过学习模态数据的映射关系,以红外-可见光-红外(IR-VIS-IR)和可见光-红外-可见光(VIS-IR-VIS)的双向反馈调节实现模态间噪声的交叉抑制;然后,构建跨模态特征交互(CMFI)模块,并引入残差结构将红外-可见光模态内以及模态间的低层特征和高层特征进行有效融合,从而减小模态间的差异并充分利用模态间的特征信息;最后,在自建红外-可见光多模态台风数据集及RGB-NIR多模态公开场景数据集上进行实验,以验证DAE模块和CMFI模块的有效性。实验结果表明,与简单级联融合方法相比,所提的基于DAINet的特征融合方法在自建台风数据集上的红外模态和可见光模态上的总体分类精度分别提高了6.61和3.93个百分点,G-mean值分别提高了6.24和2.48个百分点,表明所提方法在类别不均衡分类任务上的通用性;所提方法在RGB-NIR数据集上的2种测试模态下的总体分类精度分别提高了13.47和13.90个百分点。同时,所提方法在2个数据集上分别与IFCNN(general Image Fusion framework based on Convolutional Neural Network)和DenseFuse方法进行对比的实验结果表明,所提方法在自建台风数据集上的2种测试模态下的总体分类精度分别提高了9.82、6.02和17.38、1.68个百分点。
