Neuronal soma segmentation plays a crucial role in neuroscience applications.However,the fine structure,such as boundaries,small-volume neuronal somata and fibers,are commonly present in cell images,which pose a chall...Neuronal soma segmentation plays a crucial role in neuroscience applications.However,the fine structure,such as boundaries,small-volume neuronal somata and fibers,are commonly present in cell images,which pose a challenge for accurate segmentation.In this paper,we propose a 3D semantic segmentation network for neuronal soma segmentation to address this issue.Using an encoding-decoding structure,we introduce a Multi-Scale feature extraction and Adaptive Weighting fusion module(MSAW)after each encoding block.The MSAW module can not only emphasize the fine structures via an upsampling strategy,but also provide pixel-wise weights to measure the importance of the multi-scale features.Additionally,a dynamic convolution instead of normal convolution is employed to better adapt the network to input data with different distributions.The proposed MSAW-based semantic segmentation network(MSAW-Net)was evaluated on three neuronal soma images from mouse brain and one neuronal soma image from macaque brain,demonstrating the efficiency of the proposed method.It achieved an F1 score of 91.8%on Fezf2-2A-CreER dataset,97.1%on LSL-H2B-GFP dataset,82.8%on Thy1-EGFP-Mline dataset,and 86.9%on macaque dataset,achieving improvements over the 3D U-Net model by 3.1%,3.3%,3.9%,and 2.3%,respectively.展开更多
Moving object segmentation (MOS) is one of the essential functions of the vision system of all robots,including medical robots. Deep learning-based MOS methods, especially deep end-to-end MOS methods, are actively inv...Moving object segmentation (MOS) is one of the essential functions of the vision system of all robots,including medical robots. Deep learning-based MOS methods, especially deep end-to-end MOS methods, are actively investigated in this field. Foreground segmentation networks (FgSegNets) are representative deep end-to-endMOS methods proposed recently. This study explores a new mechanism to improve the spatial feature learningcapability of FgSegNets with relatively few brought parameters. Specifically, we propose an enhanced attention(EA) module, a parallel connection of an attention module and a lightweight enhancement module, with sequentialattention and residual attention as special cases. We also propose integrating EA with FgSegNet_v2 by taking thelightweight convolutional block attention module as the attention module and plugging EA module after the twoMaxpooling layers of the encoder. The derived new model is named FgSegNet_v2 EA. The ablation study verifiesthe effectiveness of the proposed EA module and integration strategy. The results on the CDnet2014 dataset,which depicts human activities and vehicles captured in different scenes, show that FgSegNet_v2 EA outperformsFgSegNet_v2 by 0.08% and 14.5% under the settings of scene dependent evaluation and scene independent evaluation, respectively, which indicates the positive effect of EA on improving spatial feature learning capability ofFgSegNet_v2.展开更多
To improve the accuracy and efficiency of coal-rock interface recognition,this study proposes a model built on the real-time detection algorithm,you only look once(YOLO),and the lightweight bilateral segmentation netw...To improve the accuracy and efficiency of coal-rock interface recognition,this study proposes a model built on the real-time detection algorithm,you only look once(YOLO),and the lightweight bilateral segmentation network.Simultaneously,the regional similarity transformation function and dragonfly algorithm are introduced to enhance the quality of coal-rock images.The comparison with three other models demonstrates the superior edge inference performance of the proposed model,achieving a mean Average Precision(mAP)of 90.2 at the Intersection over Union(IoU)threshold of 0.50(mAP50)and 81.4 across a range of IoU thresholds from 0.50 to 0.95(mAP[50,95]).Furthermore,to maintain high accuracy and real-time recognition capabilities,the proposed model is optimized using the open visual inference and neural network optimization toolkit,resulting in a 144.97%increase in the mean frames per second.Experimental results on four actual coal faces confirm the efficacy of the proposed model,showing a better balance between accuracy and efficiency in coal-rock image recognition,which supports further advancements in coal mining intelligence.展开更多
Synchrophasor measurements are essential to realtime situational awareness of the smart grid but vulnerable to cyber-attacks during the process of transmission and invocation.To ensure data security and mitigate the i...Synchrophasor measurements are essential to realtime situational awareness of the smart grid but vulnerable to cyber-attacks during the process of transmission and invocation.To ensure data security and mitigate the impact of spoofed synchrophasor measurements,this work proposes a novel object detection method using a Weight-based One-dimensional Convolutional Segmentation Network(WOCSN)with the ability of attack behavior identification and time localization.In WOCSN,automatic data feature extraction can be achieved by onedimensional convolution from the input signal,thereby reducing the impact of handcrafted features.A weight loss function is designed to distribute the contribution for normal and attack signals.Then,attack time is located via the proposed binary method based on pixel segmentation.Furthermore,the actual synchrophasor data collected from four locations are used for the performance evaluation of the WOCSN.Finally,combined with designed evaluation metrics,the time localization ability of WOCSN is validated in the scenarios of composite attacks with different spoofed intensities and time-sensitivities.展开更多
Chinese named entity recognition(CNER)has received widespread attention as an important task of Chinese information extraction.Most previous research has focused on individually studying flat CNER,overlapped CNER,or d...Chinese named entity recognition(CNER)has received widespread attention as an important task of Chinese information extraction.Most previous research has focused on individually studying flat CNER,overlapped CNER,or discontinuous CNER.However,a unified CNER is often needed in real-world scenarios.Recent studies have shown that grid tagging-based methods based on character-pair relationship classification hold great potential for achieving unified NER.Nevertheless,how to enrich Chinese character-pair grid representations and capture deeper dependencies between character pairs to improve entity recognition performance remains an unresolved challenge.In this study,we enhance the character-pair grid representation by incorporating both local and global information.Significantly,we introduce a new approach by considering the character-pair grid representation matrix as a specialized image,converting the classification of character-pair relationships into a pixel-level semantic segmentation task.We devise a U-shaped network to extract multi-scale and deeper semantic information from the grid image,allowing for a more comprehensive understanding of associative features between character pairs.This approach leads to improved accuracy in predicting their relationships,ultimately enhancing entity recognition performance.We conducted experiments on two public CNER datasets in the biomedical domain,namely CMeEE-V2 and Diakg.The results demonstrate the effectiveness of our approach,which achieves F1-score improvements of 7.29 percentage points and 1.64 percentage points compared to the current state-of-the-art(SOTA)models,respectively.展开更多
Human vision depends heavily on retinal tissue.The loss of eyesight may result from infections of the retinal tissues that are treated slowly or do not work at all.Additionally,the diagnosis is susceptible to inaccura...Human vision depends heavily on retinal tissue.The loss of eyesight may result from infections of the retinal tissues that are treated slowly or do not work at all.Additionally,the diagnosis is susceptible to inaccuracies when a large dataset is involved.Therefore,a fully automated transfer learning approach for diagnosing diabetic retinopathy(DR)is suggested to minimize human intervention while maintaining high classification accuracy.To address this issue,we proposed a transfer learning-based trilateral attention network(TaNet)for the classification.To boost the visual quality of the DR pictures,a contrast constrained adaptive histogram equalization approach is applied.The pre-processed pictures are then segmented using a bilateral segmentation network(BiSeNet).The BiSeNet segmented the optic disc and blood vessels individually.After the completion of segmentation,the features are extracted.Feature extraction is based on the wavelet scattering transformation approach.The results of many trials were evaluated against the Messidor-2,EYEPACS,and APTOS 2019 datasets.The proposed model was created using a refined pre-trained technique and transfer learning methodology.Finally,the suggested framework was tested using efficiency assessment methods,and the classification rate was recorded as having above 98%sensitivity,specificity,precision,and accuracy.The proposed approach yields greater performance and shows enhancement towards the existing approach.展开更多
The latest iteration of PaveVision3D Ultra can obtain true 1 mm resolution 3D data at full- lane coverage in all 3 directions at highway speed up to 60 mph. This paper introduces the PaveVision3D Ultra technology for ...The latest iteration of PaveVision3D Ultra can obtain true 1 mm resolution 3D data at full- lane coverage in all 3 directions at highway speed up to 60 mph. This paper introduces the PaveVision3D Ultra technology for rapid network level pavement survey on approximately 1280 center miles of Oklahoma interstate highways. With sophisticated automated distress analyzer （ADA） software interface, the collected 1 mm 3D data provide Oklahoma Department of Transportation （ODOT） with comprehensive solutions for automated eval- uation of pavement surface including longitudinal profile for roughness, transverse profile for rutting, predicted hydroplaning speed for safety analysis, and cracking and various surface defects for distresses. The pruned exact linear time （PELT） method, an optimal partitioning algorithm, is implemented to identify change points and dynamically deter- mine homogeneous segments so as to assist ODOT effectively using the available 1 mm 3D pavement surface condition data for decision-making. The application of 1 mm 3D laser imaging technology for network survey is unprecedented. This innovative technology allows highway agencies to access its options in using the 1 mm 3D system for its design and management purposes, particularly to meet the data needs for pavement management system （PMS）, pavement ME design and highway performance monitoring system （HPMS）.展开更多
基金supported by the STI2030-Major-Projects(No.2021ZD0200104)the National Natural Science Foundations of China under Grant 61771437.
文摘Neuronal soma segmentation plays a crucial role in neuroscience applications.However,the fine structure,such as boundaries,small-volume neuronal somata and fibers,are commonly present in cell images,which pose a challenge for accurate segmentation.In this paper,we propose a 3D semantic segmentation network for neuronal soma segmentation to address this issue.Using an encoding-decoding structure,we introduce a Multi-Scale feature extraction and Adaptive Weighting fusion module(MSAW)after each encoding block.The MSAW module can not only emphasize the fine structures via an upsampling strategy,but also provide pixel-wise weights to measure the importance of the multi-scale features.Additionally,a dynamic convolution instead of normal convolution is employed to better adapt the network to input data with different distributions.The proposed MSAW-based semantic segmentation network(MSAW-Net)was evaluated on three neuronal soma images from mouse brain and one neuronal soma image from macaque brain,demonstrating the efficiency of the proposed method.It achieved an F1 score of 91.8%on Fezf2-2A-CreER dataset,97.1%on LSL-H2B-GFP dataset,82.8%on Thy1-EGFP-Mline dataset,and 86.9%on macaque dataset,achieving improvements over the 3D U-Net model by 3.1%,3.3%,3.9%,and 2.3%,respectively.
基金the National Natural Science Foundation of China(No.61702323)。
文摘Moving object segmentation (MOS) is one of the essential functions of the vision system of all robots,including medical robots. Deep learning-based MOS methods, especially deep end-to-end MOS methods, are actively investigated in this field. Foreground segmentation networks (FgSegNets) are representative deep end-to-endMOS methods proposed recently. This study explores a new mechanism to improve the spatial feature learningcapability of FgSegNets with relatively few brought parameters. Specifically, we propose an enhanced attention(EA) module, a parallel connection of an attention module and a lightweight enhancement module, with sequentialattention and residual attention as special cases. We also propose integrating EA with FgSegNet_v2 by taking thelightweight convolutional block attention module as the attention module and plugging EA module after the twoMaxpooling layers of the encoder. The derived new model is named FgSegNet_v2 EA. The ablation study verifiesthe effectiveness of the proposed EA module and integration strategy. The results on the CDnet2014 dataset,which depicts human activities and vehicles captured in different scenes, show that FgSegNet_v2 EA outperformsFgSegNet_v2 by 0.08% and 14.5% under the settings of scene dependent evaluation and scene independent evaluation, respectively, which indicates the positive effect of EA on improving spatial feature learning capability ofFgSegNet_v2.
基金funded by the National Natural Science Foundation of China(Grant Nos.U21A20153 and 52074258)the Key Research and Development Project of Hubei Province,China(Grant No.2021BCA133)+1 种基金the Outstanding Youth Fund Program of the Natural Science Foundation of Hubei Province,China(Grant No.2022CFA084)the Wuhan Knowledge Innovation Supporting project(Grant No.2022010801010162).
文摘To improve the accuracy and efficiency of coal-rock interface recognition,this study proposes a model built on the real-time detection algorithm,you only look once(YOLO),and the lightweight bilateral segmentation network.Simultaneously,the regional similarity transformation function and dragonfly algorithm are introduced to enhance the quality of coal-rock images.The comparison with three other models demonstrates the superior edge inference performance of the proposed model,achieving a mean Average Precision(mAP)of 90.2 at the Intersection over Union(IoU)threshold of 0.50(mAP50)and 81.4 across a range of IoU thresholds from 0.50 to 0.95(mAP[50,95]).Furthermore,to maintain high accuracy and real-time recognition capabilities,the proposed model is optimized using the open visual inference and neural network optimization toolkit,resulting in a 144.97%increase in the mean frames per second.Experimental results on four actual coal faces confirm the efficacy of the proposed model,showing a better balance between accuracy and efficiency in coal-rock image recognition,which supports further advancements in coal mining intelligence.
基金This work is supported in part by the CURENT Industry Partnership Program,in part by the Engineering Research Center Program of the National Science Foundation,DOE under NSF Award Number EEC-1041877in part by the National Natural Science Foundation of China under award number 52177078in part with the project funded by China Postdoctoral Science Foundation under award number BX20220102.
文摘Synchrophasor measurements are essential to realtime situational awareness of the smart grid but vulnerable to cyber-attacks during the process of transmission and invocation.To ensure data security and mitigate the impact of spoofed synchrophasor measurements,this work proposes a novel object detection method using a Weight-based One-dimensional Convolutional Segmentation Network(WOCSN)with the ability of attack behavior identification and time localization.In WOCSN,automatic data feature extraction can be achieved by onedimensional convolution from the input signal,thereby reducing the impact of handcrafted features.A weight loss function is designed to distribute the contribution for normal and attack signals.Then,attack time is located via the proposed binary method based on pixel segmentation.Furthermore,the actual synchrophasor data collected from four locations are used for the performance evaluation of the WOCSN.Finally,combined with designed evaluation metrics,the time localization ability of WOCSN is validated in the scenarios of composite attacks with different spoofed intensities and time-sensitivities.
基金supported by Yunnan Provincial Major Science and Technology Special Plan Projects(Grant Nos.202202AD080003,202202AE090008,202202AD080004,202302AD080003)National Natural Science Foundation of China(Grant Nos.U21B2027,62266027,62266028,62266025)Yunnan Province Young and Middle-Aged Academic and Technical Leaders Reserve Talent Program(Grant No.202305AC160063).
文摘Chinese named entity recognition(CNER)has received widespread attention as an important task of Chinese information extraction.Most previous research has focused on individually studying flat CNER,overlapped CNER,or discontinuous CNER.However,a unified CNER is often needed in real-world scenarios.Recent studies have shown that grid tagging-based methods based on character-pair relationship classification hold great potential for achieving unified NER.Nevertheless,how to enrich Chinese character-pair grid representations and capture deeper dependencies between character pairs to improve entity recognition performance remains an unresolved challenge.In this study,we enhance the character-pair grid representation by incorporating both local and global information.Significantly,we introduce a new approach by considering the character-pair grid representation matrix as a specialized image,converting the classification of character-pair relationships into a pixel-level semantic segmentation task.We devise a U-shaped network to extract multi-scale and deeper semantic information from the grid image,allowing for a more comprehensive understanding of associative features between character pairs.This approach leads to improved accuracy in predicting their relationships,ultimately enhancing entity recognition performance.We conducted experiments on two public CNER datasets in the biomedical domain,namely CMeEE-V2 and Diakg.The results demonstrate the effectiveness of our approach,which achieves F1-score improvements of 7.29 percentage points and 1.64 percentage points compared to the current state-of-the-art(SOTA)models,respectively.
文摘Human vision depends heavily on retinal tissue.The loss of eyesight may result from infections of the retinal tissues that are treated slowly or do not work at all.Additionally,the diagnosis is susceptible to inaccuracies when a large dataset is involved.Therefore,a fully automated transfer learning approach for diagnosing diabetic retinopathy(DR)is suggested to minimize human intervention while maintaining high classification accuracy.To address this issue,we proposed a transfer learning-based trilateral attention network(TaNet)for the classification.To boost the visual quality of the DR pictures,a contrast constrained adaptive histogram equalization approach is applied.The pre-processed pictures are then segmented using a bilateral segmentation network(BiSeNet).The BiSeNet segmented the optic disc and blood vessels individually.After the completion of segmentation,the features are extracted.Feature extraction is based on the wavelet scattering transformation approach.The results of many trials were evaluated against the Messidor-2,EYEPACS,and APTOS 2019 datasets.The proposed model was created using a refined pre-trained technique and transfer learning methodology.Finally,the suggested framework was tested using efficiency assessment methods,and the classification rate was recorded as having above 98%sensitivity,specificity,precision,and accuracy.The proposed approach yields greater performance and shows enhancement towards the existing approach.
基金partially sponsored by the Oklahoma Department of Transportation(ODOT)
文摘The latest iteration of PaveVision3D Ultra can obtain true 1 mm resolution 3D data at full- lane coverage in all 3 directions at highway speed up to 60 mph. This paper introduces the PaveVision3D Ultra technology for rapid network level pavement survey on approximately 1280 center miles of Oklahoma interstate highways. With sophisticated automated distress analyzer （ADA） software interface, the collected 1 mm 3D data provide Oklahoma Department of Transportation （ODOT） with comprehensive solutions for automated eval- uation of pavement surface including longitudinal profile for roughness, transverse profile for rutting, predicted hydroplaning speed for safety analysis, and cracking and various surface defects for distresses. The pruned exact linear time （PELT） method, an optimal partitioning algorithm, is implemented to identify change points and dynamically deter- mine homogeneous segments so as to assist ODOT effectively using the available 1 mm 3D pavement surface condition data for decision-making. The application of 1 mm 3D laser imaging technology for network survey is unprecedented. This innovative technology allows highway agencies to access its options in using the 1 mm 3D system for its design and management purposes, particularly to meet the data needs for pavement management system （PMS）, pavement ME design and highway performance monitoring system （HPMS）.
