In this paper,we propose an end-to-end cross-layer gated attention network(CLGA-Net)to directly restore fog-free images.Compared with the previous dehazing network,the dehazing model presented in this paper uses the s...In this paper,we propose an end-to-end cross-layer gated attention network(CLGA-Net)to directly restore fog-free images.Compared with the previous dehazing network,the dehazing model presented in this paper uses the smooth cavity convolution and local residual module as the feature extractor,combined with the channel attention mechanism,to better extract the restored features.A large amount of experimental data proves that the defogging model proposed in this paper is superior to previous defogging technologies in terms of structure similarity index(SSIM),peak signal to noise ratio(PSNR)and subjective visual quality.In order to improve the efficiency of decoding and encoding,we also describe a fusion residualmodule and conduct ablation experiments,which prove that the fusion residual is suitable for the dehazing problem.Therefore,we use fusion residual as a fixed module for encoding and decoding.In addition,we found that the traditional defogging model based on the U-net network may cause some information losses in space.We have achieved effective maintenance of low-level feature information through the cross-layer gating structure that better takes into account global and subtle features.We also present the application of our CLGA-Net in challenging scenarios where the best results in both quantity and quality can be obtained.Experimental results indicate that the present cross-layer gating module can be widely used in the same type of network.展开更多
Radar quantitative precipitation estimation(QPE)is a key and challenging task for many designs and applications with meteorological purposes.Since the Z-R relation between radar and rain has a number of parameters on ...Radar quantitative precipitation estimation(QPE)is a key and challenging task for many designs and applications with meteorological purposes.Since the Z-R relation between radar and rain has a number of parameters on different areas,and the rainfall varies with seasons,the traditional methods are incapable of achieving high spatial and temporal resolution and thus difficult to obtain a refined rainfall estimation.This paper proposes a radar quantitative precipitation estimation algorithm based on the spatiotemporal network model(ST-QPE),which designs a convolutional time-series network QPE-Net8 and a multi-scale feature fusion time-series network QPE-Net22 to address these limitations.We report on our investigation into contrast reversal experiments with radar echo and rainfall data collected by the Hunan Meteorological Observatory.Experimental results are verified and analyzed by using statistical and meteorological methods,and show that the ST-QPE model can inverse the rainfall information corresponding to the radar echo at a given moment,which provides practical guidance for accurate short-range precipitation nowcasting to prevent and mitigate disasters efficiently.展开更多
We propose an end-to-end dehazing model based on deep learning(CNN network)and uses the dehazing model re-proposed by AOD-Net based on the atmospheric scattering model for dehazing.Compare to the previously proposed d...We propose an end-to-end dehazing model based on deep learning(CNN network)and uses the dehazing model re-proposed by AOD-Net based on the atmospheric scattering model for dehazing.Compare to the previously proposed dehazing network,the dehazing model proposed in this paper make use of the FPN network structure in the field of target detection,and uses five feature maps of different sizes to better obtain features of different proportions and different sub-regions.A large amount of experimental data proves that the dehazing model proposed in this paper is superior to previous dehazing technologies in terms of PSNR,SSIM,and subjective visual quality.In addition,it achieved a good performance in speed by using EfficientNet B0 as a feature extractor.We find that only using high-level semantic features can not effectively obtain all the information in the image.The FPN structure used in this paper can effectively integrate the high-level semantics and the low-level semantics,and can better take into account the global and local features.The five feature maps with different sizes are not simply weighted and fused.In order to keep all their information,we put them all together and get the final features through decode layers.At the same time,we have done a comparative experiment between ResNet with FPN and EfficientNet with BiFPN.It is proved that EfficientNet with BiFPN can obtain image features more efficiently.Therefore,EfficientNet with BiFPN is chosen as our network feature extraction.展开更多
This paper proposes a new model of facility location problem referred to as k-product uncapacitated facility location problem with multi-type clients. The k-product uncapacitated facility location problem with multi- ...This paper proposes a new model of facility location problem referred to as k-product uncapacitated facility location problem with multi-type clients. The k-product uncapacitated facility location problem with multi- type clients consists of two set of sites, one is the set of demand points where clients are located and the other is the set of sites where facilities of unlimited capacities can be set up to serve the clients. Each facility can provide only one kind of products. Each client needs to be served by a set of facilities depending on which products it needs. Each facility can be set up only for one of the k products with a non-negative fixed cost determined by the product it is designated to provide. There is also a nonnegative cost of shipping goods between each pair of locations. The problem is to determine the set of facilities to be set up and to find an assignment of each client to a set of facilities so that the sum of the setup costs and the shipping costs is minimized. Under the assumption that the setting costs is zero and the shipping costs are in facilities centered metric space, it is shown that the problem with two kinds of clients is NP-complete. Furthermore a heuristic algorithm with worst case performance ratio not more than 2-1/k is presented for any integer k.展开更多
基金This work is supported by theKey Research and Development Program of Hunan Province(No.2019SK2161)the Key Research and Development Program of Hunan Province(No.2016SK2017).
文摘In this paper,we propose an end-to-end cross-layer gated attention network(CLGA-Net)to directly restore fog-free images.Compared with the previous dehazing network,the dehazing model presented in this paper uses the smooth cavity convolution and local residual module as the feature extractor,combined with the channel attention mechanism,to better extract the restored features.A large amount of experimental data proves that the defogging model proposed in this paper is superior to previous defogging technologies in terms of structure similarity index(SSIM),peak signal to noise ratio(PSNR)and subjective visual quality.In order to improve the efficiency of decoding and encoding,we also describe a fusion residualmodule and conduct ablation experiments,which prove that the fusion residual is suitable for the dehazing problem.Therefore,we use fusion residual as a fixed module for encoding and decoding.In addition,we found that the traditional defogging model based on the U-net network may cause some information losses in space.We have achieved effective maintenance of low-level feature information through the cross-layer gating structure that better takes into account global and subtle features.We also present the application of our CLGA-Net in challenging scenarios where the best results in both quantity and quality can be obtained.Experimental results indicate that the present cross-layer gating module can be widely used in the same type of network.
基金This work is supported by the Key Research and Development Program of Hunan Province(No.2019SK2161)the Key Research and Development Program of Hunan Province(No.2016SK2017).
文摘Radar quantitative precipitation estimation(QPE)is a key and challenging task for many designs and applications with meteorological purposes.Since the Z-R relation between radar and rain has a number of parameters on different areas,and the rainfall varies with seasons,the traditional methods are incapable of achieving high spatial and temporal resolution and thus difficult to obtain a refined rainfall estimation.This paper proposes a radar quantitative precipitation estimation algorithm based on the spatiotemporal network model(ST-QPE),which designs a convolutional time-series network QPE-Net8 and a multi-scale feature fusion time-series network QPE-Net22 to address these limitations.We report on our investigation into contrast reversal experiments with radar echo and rainfall data collected by the Hunan Meteorological Observatory.Experimental results are verified and analyzed by using statistical and meteorological methods,and show that the ST-QPE model can inverse the rainfall information corresponding to the radar echo at a given moment,which provides practical guidance for accurate short-range precipitation nowcasting to prevent and mitigate disasters efficiently.
基金the Key Research and Development Program of Hunan Province(No.2019SK2161)the Key Research and Development Program of Hunan Province(No.2016SK2017).
文摘We propose an end-to-end dehazing model based on deep learning(CNN network)and uses the dehazing model re-proposed by AOD-Net based on the atmospheric scattering model for dehazing.Compare to the previously proposed dehazing network,the dehazing model proposed in this paper make use of the FPN network structure in the field of target detection,and uses five feature maps of different sizes to better obtain features of different proportions and different sub-regions.A large amount of experimental data proves that the dehazing model proposed in this paper is superior to previous dehazing technologies in terms of PSNR,SSIM,and subjective visual quality.In addition,it achieved a good performance in speed by using EfficientNet B0 as a feature extractor.We find that only using high-level semantic features can not effectively obtain all the information in the image.The FPN structure used in this paper can effectively integrate the high-level semantics and the low-level semantics,and can better take into account the global and local features.The five feature maps with different sizes are not simply weighted and fused.In order to keep all their information,we put them all together and get the final features through decode layers.At the same time,we have done a comparative experiment between ResNet with FPN and EfficientNet with BiFPN.It is proved that EfficientNet with BiFPN can obtain image features more efficiently.Therefore,EfficientNet with BiFPN is chosen as our network feature extraction.
文摘This paper proposes a new model of facility location problem referred to as k-product uncapacitated facility location problem with multi-type clients. The k-product uncapacitated facility location problem with multi- type clients consists of two set of sites, one is the set of demand points where clients are located and the other is the set of sites where facilities of unlimited capacities can be set up to serve the clients. Each facility can provide only one kind of products. Each client needs to be served by a set of facilities depending on which products it needs. Each facility can be set up only for one of the k products with a non-negative fixed cost determined by the product it is designated to provide. There is also a nonnegative cost of shipping goods between each pair of locations. The problem is to determine the set of facilities to be set up and to find an assignment of each client to a set of facilities so that the sum of the setup costs and the shipping costs is minimized. Under the assumption that the setting costs is zero and the shipping costs are in facilities centered metric space, it is shown that the problem with two kinds of clients is NP-complete. Furthermore a heuristic algorithm with worst case performance ratio not more than 2-1/k is presented for any integer k.