Background Coronary artery calcification is a well-known marker of atherosclerotic plaque burden.High-resolution intravascular optical coherence tomography(OCT)imaging has shown the potential to characterize the detai...Background Coronary artery calcification is a well-known marker of atherosclerotic plaque burden.High-resolution intravascular optical coherence tomography(OCT)imaging has shown the potential to characterize the details of coronary calcification in vivo.In routine clinical practice,it is a time-consuming and laborious task for clinicians to review the over 250 images in a single pullback.Besides,the imbalance label distribution within the entire pullbacks is another problem,which could lead to the failure of the classifier model.Given the success of deep learning methods with other imaging modalities,a thorough understanding of calcified plaque detection using Convolutional Neural Networks(CNNs)within pullbacks for future clinical decision was required.Methods All 33 IVOCT clinical pullbacks of 33 patients were taken from Affiliated Drum Tower Hospital,Nanjing University between December 2017 and December 2018.For ground-truth annotation,three trained experts determined the type of plaque that was present in a B-Scan.The experts assigned the labels'no calcified plaque','calcified plaque'for each OCT image.All experts were provided the all images for labeling.The final label was determined based on consensus between the experts,different opinions on the plaque type were resolved by asking the experts for a repetition of their evaluation.Before the implement of algorithm,all OCT images was resized to a resolution of 300×300,which matched the range used with standard architectures in the natural image domain.In the study,we randomly selected 26 pullbacks for training,the remaining data were testing.While,imbalance label distribution within entire pullbacks was great challenge for various CNNs architecture.In order to resolve the problem,we designed the following experiment.First,we fine-tuned twenty different CNNs architecture,including customize CNN architectures and pretrained CNN architectures.Considering the nature of OCT images,customize CNN architectures were designed that the layers were fewer than 25 layers.Then,three with good performance were selected and further deep fine-tuned to train three different models.The difference of CNNs was mainly in the model architecture,such as depth-based residual networks,width-based inception networks.Finally,the three CNN models were used to majority voting,the predicted labels were from the most voting.Areas under the receiver operating characteristic curve(ROC AUC)were used as the evaluation metric for the imbalance label distribution.Results The imbalance label distribution within pullbacks affected both convergence during the training phase and generalization of a CNN model.Different labels of OCT images could be classified with excellent performance by fine tuning parameters of CNN architectures.Overall,we find that our final result performed best with an accuracy of 90%of'calcified plaque'class,which the numbers were less than'no calcified plaque'class in one pullback.Conclusions The obtained results showed that the method is fast and effective to classify calcific plaques with imbalance label distribution in each pullback.The results suggest that the proposed method could be facilitating our understanding of coronary artery calcification in the process of atherosclerosis andhelping guide complex interventional strategies in coronary arteries with superficial calcification.展开更多
Non-orthogonal multiple access(NOMA), featuring high spectrum efficiency, massive connectivity and low latency, holds immense potential to be a novel multi-access technique in fifth-generation(5G) communication. Succe...Non-orthogonal multiple access(NOMA), featuring high spectrum efficiency, massive connectivity and low latency, holds immense potential to be a novel multi-access technique in fifth-generation(5G) communication. Successive interference cancellation(SIC) is proved to be an effective method to detect the NOMA signal by ordering the power of received signals and then decoding them. However, the error accumulation effect referred to as error propagation is an inevitable problem. In this paper,we propose a convolutional neural networks(CNNs) approach to restore the desired signal impaired by the multiple input multiple output(MIMO) channel. Especially in the uplink NOMA scenario,the proposed method can decode multiple users' information in a cluster instantaneously without any traditional communication signal processing steps. Simulation experiments are conducted in the Rayleigh channel and the results demonstrate that the error performance of the proposed learning system outperforms that of the classic SIC detection. Consequently, deep learning has disruptive potential to replace the conventional signal detection method.展开更多
The fifth-generation (5G) communication requires a highly accurate estimation of the channel state information (CSI)to take advantage of the massive multiple-input multiple-output(MIMO) system. However, traditional ch...The fifth-generation (5G) communication requires a highly accurate estimation of the channel state information (CSI)to take advantage of the massive multiple-input multiple-output(MIMO) system. However, traditional channel estimation methods do not always yield reliable estimates. The methodology of this paper consists of deep residual shrinkage network (DRSN)neural network-based method that is used to solve this problem.Thus, the channel estimation approach, based on DRSN with its learning ability of noise-containing data, is first introduced. Then,the DRSN is used to train the noise reduction process based on the results of the least square (LS) channel estimation while applying the pilot frequency subcarriers, where the initially estimated subcarrier channel matrix is considered as a three-dimensional tensor of the DRSN input. Afterward, a mixed signal to noise ratio (SNR) training data strategy is proposed based on the learning ability of DRSN under different SNRs. Moreover, a joint mixed scenario training strategy is carried out to test the multi scenarios robustness of DRSN. As for the findings, the numerical results indicate that the DRSN method outperforms the spatial-frequency-temporal convolutional neural networks (SF-CNN)with similar computational complexity and achieves better advantages in the full SNR range than the minimum mean squared error (MMSE) estimator with a limited dataset. Moreover, the DRSN approach shows robustness in different propagation environments.展开更多
According to the oversampling imaging characteristics, an infrared small target detection method based on deep learning is proposed. A 7-layer deep convolutional neural network(CNN) is designed to automatically extrac...According to the oversampling imaging characteristics, an infrared small target detection method based on deep learning is proposed. A 7-layer deep convolutional neural network(CNN) is designed to automatically extract small target features and suppress clutters in an end-to-end manner. The input of CNN is an original oversampling image while the output is a cluttersuppressed feature map. The CNN contains only convolution and non-linear operations, and the resolution of the output feature map is the same as that of the input image. The L1-norm loss function is used, and a mass of training data is generated to train the network effectively. Results show that compared with several baseline methods, the proposed method improves the signal clutter ratio gain and background suppression factor by 3–4 orders of magnitude, and has more powerful target detection performance.展开更多
The accurate and efficient prediction of explosive detonation properties has important engineering significance for weapon design.Traditional methods for predicting detonation performance include empirical formulas,eq...The accurate and efficient prediction of explosive detonation properties has important engineering significance for weapon design.Traditional methods for predicting detonation performance include empirical formulas,equations of state,and quantum chemical calculation methods.In recent years,with the development of computer performance and deep learning methods,researchers have begun to apply deep learning methods to the prediction of explosive detonation performance.The deep learning method has the advantage of simple and rapid prediction of explosive detonation properties.However,some problems remain in the study of detonation properties based on deep learning.For example,there are few studies on the prediction of mixed explosives,on the prediction of the parameters of the equation of state of explosives,and on the application of explosive properties to predict the formulation of explosives.Based on an artificial neural network model and a one-dimensional convolutional neural network model,three improved deep learning models were established in this work with the aim of solving these problems.The training data for these models,called the detonation parameters prediction model,JWL equation of state(EOS)prediction model,and inverse prediction model,was obtained through the KHT thermochemical code.After training,the model was tested for overfitting using the validation-set test.Through the model-accuracy test,the prediction accuracy of the model for real explosive formulations was tested by comparing the predicted value with the reference value.The results show that the model errors were within 10%and 3%for the prediction of detonation pressure and detonation velocity,respectively.The accuracy refers to the prediction of tested explosive formulations which consist of TNT,RDX and HMX.For the prediction of the equation of state for explosives,the correlation coefficient between the prediction and the reference curves was above 0.99.For the prediction of the inverse prediction model,the prediction error of the explosive equation was within 9%.This indicates that the models have utility in engineering.展开更多
The open-circuit fault is one of the most common faults of the automatic ramming drive system(ARDS),and it can be categorized into the open-phase faults of Permanent Magnet Synchronous Motor(PMSM)and the open-circuit ...The open-circuit fault is one of the most common faults of the automatic ramming drive system(ARDS),and it can be categorized into the open-phase faults of Permanent Magnet Synchronous Motor(PMSM)and the open-circuit faults of Voltage Source Inverter(VSI). The stator current serves as a common indicator for detecting open-circuit faults. Due to the identical changes of the stator current between the open-phase faults in the PMSM and failures of double switches within the same leg of the VSI, this paper utilizes the zero-sequence voltage component as an additional diagnostic criterion to differentiate them.Considering the variable conditions and substantial noise of the ARDS, a novel Multi-resolution Network(Mr Net) is proposed, which can extract multi-resolution perceptual information and enhance robustness to the noise. Meanwhile, a feature weighted layer is introduced to allocate higher weights to characteristics situated near the feature frequency. Both simulation and experiment results validate that the proposed fault diagnosis method can diagnose 25 types of open-circuit faults and achieve more than98.28% diagnostic accuracy. In addition, the experiment results also demonstrate that Mr Net has the capability of diagnosing the fault types accurately under the interference of noise signals(Laplace noise and Gaussian noise).展开更多
Dense captioning aims to simultaneously localize and describe regions-of-interest(RoIs)in images in natural language.Specifically,we identify three key problems:1)dense and highly overlapping RoIs,making accurate loca...Dense captioning aims to simultaneously localize and describe regions-of-interest(RoIs)in images in natural language.Specifically,we identify three key problems:1)dense and highly overlapping RoIs,making accurate localization of each target region challenging;2)some visually ambiguous target regions which are hard to recognize each of them just by appearance;3)an extremely deep image representation which is of central importance for visual recognition.To tackle these three challenges,we propose a novel end-to-end dense captioning framework consisting of a joint localization module,a contextual reasoning module and a deep convolutional neural network(CNN).We also evaluate five deep CNN structures to explore the benefits of each.Extensive experiments on visual genome(VG)dataset demonstrate the effectiveness of our approach,which compares favorably with the state-of-the-art methods.展开更多
Individual identification of dairy cows is the prerequisite for automatic analysis and intelligent perception of dairy cows'behavior.At present,individual identification of dairy cows based on deep convolutional n...Individual identification of dairy cows is the prerequisite for automatic analysis and intelligent perception of dairy cows'behavior.At present,individual identification of dairy cows based on deep convolutional neural network had the disadvantages in prolonged training at the additions of new cows samples.Therefore,a cow individual identification framework was proposed based on deep feature extraction and matching,and the individual identification of dairy cows based on this framework could avoid repeated training.Firstly,the trained convolutional neural network model was used as the feature extractor;secondly,the feature extraction was used to extract features and stored the features into the template feature library to complete the enrollment;finally,the identifies of dairy cows were identified.Based on this framework,when new cows joined the herd,enrollment could be completed quickly.In order to evaluate the application performance of this method in closed-set and open-set individual identification of dairy cows,back images of 524 cows were collected,among which the back images of 150 cows were selected as the training data to train feature extractor.The data of the remaining 374 cows were used to generate the template data set and the data to be identified.The experiment results showed that in the closed-set individual identification of dairy cows,the highest identification accuracy of top-1 was 99.73%,the highest identification accuracy from top-2 to top-5 was 100%,and the identification time of a single cow was 0.601 s,this method was verified to be effective.In the open-set individual identification of dairy cows,the recall was 90.38%,and the accuracy was 89.46%.When false accept rate(FAR)=0.05,true accept rate(TAR)=84.07%,this method was verified that the application had certain research value in open-set individual identification of dairy cows,which provided a certain idea for the application of individual identification in the field of intelligent animal husbandry.展开更多
Acoustic source localization(ASL)and sound event detection(SED)are two widely pursued independent research fields.In recent years,in order to achieve a more complete spatial and temporal representation of sound field,...Acoustic source localization(ASL)and sound event detection(SED)are two widely pursued independent research fields.In recent years,in order to achieve a more complete spatial and temporal representation of sound field,sound event localization and detection(SELD)has become a very active research topic.This paper presents a deep learning-based multioverlapping sound event localization and detection algorithm in three-dimensional space.Log-Mel spectrum and generalized cross-correlation spectrum are joined together in channel dimension as input features.These features are classified and regressed in parallel after training by a neural network to obtain sound recognition and localization results respectively.The channel attention mechanism is also introduced in the network to selectively enhance the features containing essential information and suppress the useless features.Finally,a thourough comparison confirms the efficiency and effectiveness of the proposed SELD algorithm.Field experiments show that the proposed algorithm is robust to reverberation and environment and can achieve higher recognition and localization accuracy compared with the baseline method.展开更多
[Objective]Urban floods are occurring more frequently because of global climate change and urbanization.Accordingly,urban rainstorm and flood forecasting has become a priority in urban hydrology research.However,two-d...[Objective]Urban floods are occurring more frequently because of global climate change and urbanization.Accordingly,urban rainstorm and flood forecasting has become a priority in urban hydrology research.However,two-dimensional hydrodynamic models execute calculations slowly,hindering the rapid simulation and forecasting of urban floods.To overcome this limitation and accelerate the speed and improve the accuracy of urban flood simulations and forecasting,numerical simulations and deep learning were combined to develop a more effective urban flood forecasting method.[Methods]Specifically,a cellular automata model was used to simulate the urban flood process and address the need to include a large number of datasets in the deep learning process.Meanwhile,to shorten the time required for urban flood forecasting,a convolutional neural network model was used to establish the mapping relationship between rainfall and inundation depth.[Results]The results show that the relative error of forecasting the maximum inundation depth in flood-prone locations is less than 10%,and the Nash efficiency coefficient of forecasting inundation depth series in flood-prone locations is greater than 0.75.[Conclusion]The result demonstrated that the proposed method could execute highly accurate simulations and quickly produce forecasts,illustrating its superiority as an urban flood forecasting technique.展开更多
Deep learning has achieved excellent results in various tasks in the field of computer vision,especially in fine-grained visual categorization.It aims to distinguish the subordinate categories of the label-level categ...Deep learning has achieved excellent results in various tasks in the field of computer vision,especially in fine-grained visual categorization.It aims to distinguish the subordinate categories of the label-level categories.Due to high intra-class variances and high inter-class similarity,the fine-grained visual categorization is extremely challenging.This paper first briefly introduces and analyzes the related public datasets.After that,some of the latest methods are reviewed.Based on the feature types,the feature processing methods,and the overall structure used in the model,we divide them into three types of methods:methods based on general convolutional neural network(CNN)and strong supervision of parts,methods based on single feature processing,and meth-ods based on multiple feature processing.Most methods of the first type have a relatively simple structure,which is the result of the initial research.The methods of the other two types include models that have special structures and training processes,which are helpful to obtain discriminative features.We conduct a specific analysis on several methods with high accuracy on pub-lic datasets.In addition,we support that the focus of the future research is to solve the demand of existing methods for the large amount of the data and the computing power.In terms of tech-nology,the extraction of the subtle feature information with the burgeoning vision transformer(ViT)network is also an important research direction.展开更多
基金supported in part by the National Natural Science Foundation of China ( NSFC ) ( 11772093)ARC ( FT140101152)
文摘Background Coronary artery calcification is a well-known marker of atherosclerotic plaque burden.High-resolution intravascular optical coherence tomography(OCT)imaging has shown the potential to characterize the details of coronary calcification in vivo.In routine clinical practice,it is a time-consuming and laborious task for clinicians to review the over 250 images in a single pullback.Besides,the imbalance label distribution within the entire pullbacks is another problem,which could lead to the failure of the classifier model.Given the success of deep learning methods with other imaging modalities,a thorough understanding of calcified plaque detection using Convolutional Neural Networks(CNNs)within pullbacks for future clinical decision was required.Methods All 33 IVOCT clinical pullbacks of 33 patients were taken from Affiliated Drum Tower Hospital,Nanjing University between December 2017 and December 2018.For ground-truth annotation,three trained experts determined the type of plaque that was present in a B-Scan.The experts assigned the labels'no calcified plaque','calcified plaque'for each OCT image.All experts were provided the all images for labeling.The final label was determined based on consensus between the experts,different opinions on the plaque type were resolved by asking the experts for a repetition of their evaluation.Before the implement of algorithm,all OCT images was resized to a resolution of 300×300,which matched the range used with standard architectures in the natural image domain.In the study,we randomly selected 26 pullbacks for training,the remaining data were testing.While,imbalance label distribution within entire pullbacks was great challenge for various CNNs architecture.In order to resolve the problem,we designed the following experiment.First,we fine-tuned twenty different CNNs architecture,including customize CNN architectures and pretrained CNN architectures.Considering the nature of OCT images,customize CNN architectures were designed that the layers were fewer than 25 layers.Then,three with good performance were selected and further deep fine-tuned to train three different models.The difference of CNNs was mainly in the model architecture,such as depth-based residual networks,width-based inception networks.Finally,the three CNN models were used to majority voting,the predicted labels were from the most voting.Areas under the receiver operating characteristic curve(ROC AUC)were used as the evaluation metric for the imbalance label distribution.Results The imbalance label distribution within pullbacks affected both convergence during the training phase and generalization of a CNN model.Different labels of OCT images could be classified with excellent performance by fine tuning parameters of CNN architectures.Overall,we find that our final result performed best with an accuracy of 90%of'calcified plaque'class,which the numbers were less than'no calcified plaque'class in one pullback.Conclusions The obtained results showed that the method is fast and effective to classify calcific plaques with imbalance label distribution in each pullback.The results suggest that the proposed method could be facilitating our understanding of coronary artery calcification in the process of atherosclerosis andhelping guide complex interventional strategies in coronary arteries with superficial calcification.
基金supported by the National Natural Science Foundation of China (61471021)。
文摘Non-orthogonal multiple access(NOMA), featuring high spectrum efficiency, massive connectivity and low latency, holds immense potential to be a novel multi-access technique in fifth-generation(5G) communication. Successive interference cancellation(SIC) is proved to be an effective method to detect the NOMA signal by ordering the power of received signals and then decoding them. However, the error accumulation effect referred to as error propagation is an inevitable problem. In this paper,we propose a convolutional neural networks(CNNs) approach to restore the desired signal impaired by the multiple input multiple output(MIMO) channel. Especially in the uplink NOMA scenario,the proposed method can decode multiple users' information in a cluster instantaneously without any traditional communication signal processing steps. Simulation experiments are conducted in the Rayleigh channel and the results demonstrate that the error performance of the proposed learning system outperforms that of the classic SIC detection. Consequently, deep learning has disruptive potential to replace the conventional signal detection method.
基金supported by the National Key Scientific Instrument and Equipment Development Project(61827801).
文摘The fifth-generation (5G) communication requires a highly accurate estimation of the channel state information (CSI)to take advantage of the massive multiple-input multiple-output(MIMO) system. However, traditional channel estimation methods do not always yield reliable estimates. The methodology of this paper consists of deep residual shrinkage network (DRSN)neural network-based method that is used to solve this problem.Thus, the channel estimation approach, based on DRSN with its learning ability of noise-containing data, is first introduced. Then,the DRSN is used to train the noise reduction process based on the results of the least square (LS) channel estimation while applying the pilot frequency subcarriers, where the initially estimated subcarrier channel matrix is considered as a three-dimensional tensor of the DRSN input. Afterward, a mixed signal to noise ratio (SNR) training data strategy is proposed based on the learning ability of DRSN under different SNRs. Moreover, a joint mixed scenario training strategy is carried out to test the multi scenarios robustness of DRSN. As for the findings, the numerical results indicate that the DRSN method outperforms the spatial-frequency-temporal convolutional neural networks (SF-CNN)with similar computational complexity and achieves better advantages in the full SNR range than the minimum mean squared error (MMSE) estimator with a limited dataset. Moreover, the DRSN approach shows robustness in different propagation environments.
基金supported by the National Key Research and Development Program of China(2016YFB0500901)the Natural Science Foundation of Shanghai(18ZR1437200)the Satellite Mapping Technology and Application National Key Laboratory of Geographical Information Bureau(KLSMTA-201709)
文摘According to the oversampling imaging characteristics, an infrared small target detection method based on deep learning is proposed. A 7-layer deep convolutional neural network(CNN) is designed to automatically extract small target features and suppress clutters in an end-to-end manner. The input of CNN is an original oversampling image while the output is a cluttersuppressed feature map. The CNN contains only convolution and non-linear operations, and the resolution of the output feature map is the same as that of the input image. The L1-norm loss function is used, and a mass of training data is generated to train the network effectively. Results show that compared with several baseline methods, the proposed method improves the signal clutter ratio gain and background suppression factor by 3–4 orders of magnitude, and has more powerful target detection performance.
文摘The accurate and efficient prediction of explosive detonation properties has important engineering significance for weapon design.Traditional methods for predicting detonation performance include empirical formulas,equations of state,and quantum chemical calculation methods.In recent years,with the development of computer performance and deep learning methods,researchers have begun to apply deep learning methods to the prediction of explosive detonation performance.The deep learning method has the advantage of simple and rapid prediction of explosive detonation properties.However,some problems remain in the study of detonation properties based on deep learning.For example,there are few studies on the prediction of mixed explosives,on the prediction of the parameters of the equation of state of explosives,and on the application of explosive properties to predict the formulation of explosives.Based on an artificial neural network model and a one-dimensional convolutional neural network model,three improved deep learning models were established in this work with the aim of solving these problems.The training data for these models,called the detonation parameters prediction model,JWL equation of state(EOS)prediction model,and inverse prediction model,was obtained through the KHT thermochemical code.After training,the model was tested for overfitting using the validation-set test.Through the model-accuracy test,the prediction accuracy of the model for real explosive formulations was tested by comparing the predicted value with the reference value.The results show that the model errors were within 10%and 3%for the prediction of detonation pressure and detonation velocity,respectively.The accuracy refers to the prediction of tested explosive formulations which consist of TNT,RDX and HMX.For the prediction of the equation of state for explosives,the correlation coefficient between the prediction and the reference curves was above 0.99.For the prediction of the inverse prediction model,the prediction error of the explosive equation was within 9%.This indicates that the models have utility in engineering.
基金supported by the Natural Science Foundation of Jiangsu Province (Grant Nos. BK20210347)。
文摘The open-circuit fault is one of the most common faults of the automatic ramming drive system(ARDS),and it can be categorized into the open-phase faults of Permanent Magnet Synchronous Motor(PMSM)and the open-circuit faults of Voltage Source Inverter(VSI). The stator current serves as a common indicator for detecting open-circuit faults. Due to the identical changes of the stator current between the open-phase faults in the PMSM and failures of double switches within the same leg of the VSI, this paper utilizes the zero-sequence voltage component as an additional diagnostic criterion to differentiate them.Considering the variable conditions and substantial noise of the ARDS, a novel Multi-resolution Network(Mr Net) is proposed, which can extract multi-resolution perceptual information and enhance robustness to the noise. Meanwhile, a feature weighted layer is introduced to allocate higher weights to characteristics situated near the feature frequency. Both simulation and experiment results validate that the proposed fault diagnosis method can diagnose 25 types of open-circuit faults and achieve more than98.28% diagnostic accuracy. In addition, the experiment results also demonstrate that Mr Net has the capability of diagnosing the fault types accurately under the interference of noise signals(Laplace noise and Gaussian noise).
基金Project(2020A1515010718)supported by the Basic and Applied Basic Research Foundation of Guangdong Province,China。
文摘Dense captioning aims to simultaneously localize and describe regions-of-interest(RoIs)in images in natural language.Specifically,we identify three key problems:1)dense and highly overlapping RoIs,making accurate localization of each target region challenging;2)some visually ambiguous target regions which are hard to recognize each of them just by appearance;3)an extremely deep image representation which is of central importance for visual recognition.To tackle these three challenges,we propose a novel end-to-end dense captioning framework consisting of a joint localization module,a contextual reasoning module and a deep convolutional neural network(CNN).We also evaluate five deep CNN structures to explore the benefits of each.Extensive experiments on visual genome(VG)dataset demonstrate the effectiveness of our approach,which compares favorably with the state-of-the-art methods.
基金Supported by the National Key Research and Development Program of China(2019YFE0125600)China Agriculture Research System(CARS-36)。
文摘Individual identification of dairy cows is the prerequisite for automatic analysis and intelligent perception of dairy cows'behavior.At present,individual identification of dairy cows based on deep convolutional neural network had the disadvantages in prolonged training at the additions of new cows samples.Therefore,a cow individual identification framework was proposed based on deep feature extraction and matching,and the individual identification of dairy cows based on this framework could avoid repeated training.Firstly,the trained convolutional neural network model was used as the feature extractor;secondly,the feature extraction was used to extract features and stored the features into the template feature library to complete the enrollment;finally,the identifies of dairy cows were identified.Based on this framework,when new cows joined the herd,enrollment could be completed quickly.In order to evaluate the application performance of this method in closed-set and open-set individual identification of dairy cows,back images of 524 cows were collected,among which the back images of 150 cows were selected as the training data to train feature extractor.The data of the remaining 374 cows were used to generate the template data set and the data to be identified.The experiment results showed that in the closed-set individual identification of dairy cows,the highest identification accuracy of top-1 was 99.73%,the highest identification accuracy from top-2 to top-5 was 100%,and the identification time of a single cow was 0.601 s,this method was verified to be effective.In the open-set individual identification of dairy cows,the recall was 90.38%,and the accuracy was 89.46%.When false accept rate(FAR)=0.05,true accept rate(TAR)=84.07%,this method was verified that the application had certain research value in open-set individual identification of dairy cows,which provided a certain idea for the application of individual identification in the field of intelligent animal husbandry.
基金supported by the National Natural Science Foundation of China(61877067)the Foundation of Science and Technology on Near-Surface Detection Laboratory(TCGZ2019A002,TCGZ2021C003,6142414200511)the Natural Science Basic Research Program of Shaanxi(2021JZ-19)。
文摘Acoustic source localization(ASL)and sound event detection(SED)are two widely pursued independent research fields.In recent years,in order to achieve a more complete spatial and temporal representation of sound field,sound event localization and detection(SELD)has become a very active research topic.This paper presents a deep learning-based multioverlapping sound event localization and detection algorithm in three-dimensional space.Log-Mel spectrum and generalized cross-correlation spectrum are joined together in channel dimension as input features.These features are classified and regressed in parallel after training by a neural network to obtain sound recognition and localization results respectively.The channel attention mechanism is also introduced in the network to selectively enhance the features containing essential information and suppress the useless features.Finally,a thourough comparison confirms the efficiency and effectiveness of the proposed SELD algorithm.Field experiments show that the proposed algorithm is robust to reverberation and environment and can achieve higher recognition and localization accuracy compared with the baseline method.
文摘[Objective]Urban floods are occurring more frequently because of global climate change and urbanization.Accordingly,urban rainstorm and flood forecasting has become a priority in urban hydrology research.However,two-dimensional hydrodynamic models execute calculations slowly,hindering the rapid simulation and forecasting of urban floods.To overcome this limitation and accelerate the speed and improve the accuracy of urban flood simulations and forecasting,numerical simulations and deep learning were combined to develop a more effective urban flood forecasting method.[Methods]Specifically,a cellular automata model was used to simulate the urban flood process and address the need to include a large number of datasets in the deep learning process.Meanwhile,to shorten the time required for urban flood forecasting,a convolutional neural network model was used to establish the mapping relationship between rainfall and inundation depth.[Results]The results show that the relative error of forecasting the maximum inundation depth in flood-prone locations is less than 10%,and the Nash efficiency coefficient of forecasting inundation depth series in flood-prone locations is greater than 0.75.[Conclusion]The result demonstrated that the proposed method could execute highly accurate simulations and quickly produce forecasts,illustrating its superiority as an urban flood forecasting technique.
基金supported by the National Natural Science Foundation of China(61571453,61806218).
文摘Deep learning has achieved excellent results in various tasks in the field of computer vision,especially in fine-grained visual categorization.It aims to distinguish the subordinate categories of the label-level categories.Due to high intra-class variances and high inter-class similarity,the fine-grained visual categorization is extremely challenging.This paper first briefly introduces and analyzes the related public datasets.After that,some of the latest methods are reviewed.Based on the feature types,the feature processing methods,and the overall structure used in the model,we divide them into three types of methods:methods based on general convolutional neural network(CNN)and strong supervision of parts,methods based on single feature processing,and meth-ods based on multiple feature processing.Most methods of the first type have a relatively simple structure,which is the result of the initial research.The methods of the other two types include models that have special structures and training processes,which are helpful to obtain discriminative features.We conduct a specific analysis on several methods with high accuracy on pub-lic datasets.In addition,we support that the focus of the future research is to solve the demand of existing methods for the large amount of the data and the computing power.In terms of tech-nology,the extraction of the subtle feature information with the burgeoning vision transformer(ViT)network is also an important research direction.
