Now object detection based on deep learning tries different strategies.It uses fewer data training networks to achieve the effect of large dataset training.However,the existing methods usually do not achieve the balan...Now object detection based on deep learning tries different strategies.It uses fewer data training networks to achieve the effect of large dataset training.However,the existing methods usually do not achieve the balance between network parameters and training data.It makes the information provided by a small amount of picture data insufficient to optimize model parameters,resulting in unsatisfactory detection results.To improve the accuracy of few shot object detection,this paper proposes a network based on the transformer and high-resolution feature extraction(THR).High-resolution feature extractionmaintains the resolution representation of the image.Channels and spatial attention are used to make the network focus on features that are more useful to the object.In addition,the recently popular transformer is used to fuse the features of the existing object.This compensates for the previous network failure by making full use of existing object features.Experiments on the Pascal VOC and MS-COCO datasets prove that the THR network has achieved better results than previous mainstream few shot object detection.展开更多
The separation and detection of particles in suspension are essential for a wide spectrum of applications including medical diagnostics.In this field,microfluidic deterministic lateral displacement(DLD)holds a promise...The separation and detection of particles in suspension are essential for a wide spectrum of applications including medical diagnostics.In this field,microfluidic deterministic lateral displacement(DLD)holds a promise due to the ability of continuous separation of particles by size,shape,deformability,and electrical properties with high resolution.DLD is a passive microfluidic separation technique that has been widely implemented for various bioparticle separations from blood cells to exosomes.DLD techniques have been previously reviewed in 2014.Since then,the field has matured as several physics of DLD have been updated,new phenomena have been discovered,and various designs have been presented to achieve a higher separation performance and throughput.Furthermore,some recent progress has shown new clinical applications and ability to use the DLD arrays as a platform for biomolecules detection.This review provides a thorough discussion on the recent progress in DLD with the topics based on the fundamental studies on DLD models and applications for particle separation and detection.Furthermore,current challenges and potential solutions of DLD are also discussed.We believe that a comprehensive understanding on DLD techniques could significantly contribute toward the advancements in the field for various applications.In particular,the rapid,low-cost,and high-throughput particle separation and detection with DLD have a tremendous impact for point-of-care diagnostics.展开更多
This study presents a new idea of using only mid-span displacement measurement for damage detection of simply supported beams. Equivalent element concept is introduced at the beginning. In order to relate the damage d...This study presents a new idea of using only mid-span displacement measurement for damage detection of simply supported beams. Equivalent element concept is introduced at the beginning. In order to relate the damage detectability by means of mid-span displacement measurement with the damage-caused local stiffness change, a novel index termed as symmetrical mid-span displacement difference index (SMDDI) is proposed. The proposed method based on SMDDI is sensitive to tiny damage and comparatively small quantities of measurements are required during the application process. Another significant attraction of this method is putting aside the knowledge a-priori of the intact state. An example using simulated data has been conducted to examine the suitability of this method and assess its comparative advantages over the previous modal method.展开更多
High-resolution approaches such as multiple signal classification and estimation of signal parameters via rotational invariance techniques(ESPRIT) are currently employed widely in multibeam echo-sounder(MBES)syste...High-resolution approaches such as multiple signal classification and estimation of signal parameters via rotational invariance techniques(ESPRIT) are currently employed widely in multibeam echo-sounder(MBES)systems for sea floor bathymetry,where a uniform line array is also required.However,due to the requirements in terms of the system coverage/resolution and installation space constraints,an MBES system usually employs a receiving array with a special shape,which means that high-resolution algorithms cannot be applied directly.In addition,the short-term stationary echo signals make it difficult to estimate the covariance matrix required by the high-resolution approaches,which further increases the complexity when applying the high-resolution algorithms in the MBES systems.The ESPRIT with multiple-angle subarray beamforming is employed to reduce the requirements in terms of the signal-to-noise ratio,number of snapshots,and computational effort.The simulations show that the new processing method can provide better fine-structure resolution.Then a highresolution bottom detection(HRBD) algorithm is developed by combining the new processing method with virtual array transformation.The application of the HRBD algorithm to a U-shaped array is also discuss.The computer simulations and experimental data processing results verify the effectiveness of the proposed algorithm.展开更多
A large number of debris flow disasters(called Seismic debris flows) would occur after an earthquake, which can cause a great amount of damage. UAV low-altitude remote sensing technology has become a means of quickly ...A large number of debris flow disasters(called Seismic debris flows) would occur after an earthquake, which can cause a great amount of damage. UAV low-altitude remote sensing technology has become a means of quickly obtaining disaster information as it has the advantage of convenience and timeliness, but the spectral information of the image is so scarce, making it difficult to accurately detect the information of earthquake debris flow disasters. Based on the above problems, a seismic debris flow detection method based on transfer learning(TL) mechanism is proposed. On the basis of the constructed seismic debris flow disaster database, the features acquired from the training of the convolutional neural network(CNN) are transferred to the disaster information detection of the seismic debris flow. The automatic detection of earthquake debris flow disaster information is then completed, and the results of object-oriented seismic debris flow disaster information detection are compared and analyzed with the detection results supported by transfer learning.展开更多
The method proposed in this paper is based on the fact that the damage in different types of structural members has distinctive influence on the structural stiffness. The intrinsic mechanical property of the structure...The method proposed in this paper is based on the fact that the damage in different types of structural members has distinctive influence on the structural stiffness. The intrinsic mechanical property of the structure is tapped and fully utilized for damage detection. The simplified model of the flexibility of frames treats the individual storeys as springs in series and the frame as an equivalent column. It fully considers the main deformation of all beams and columns in the frame. The deformation property of the simplified model accorded well with that of the actual frame model. The obtained increment of lateral displacement change (IOLDC) at the storey level was found to be very sensitive to the local damage in the frame. A damage detection method is pro- posed using the IOLDCs as the damage identification parameters. Numerical examples demonstrate the potential applicability of this method.展开更多
Multichannel high-resolution and wide-swath(HRWS)imaging is an advanced digital beamforming technique for future synthetic aperture radar(SAR)systems.However,radio frequency interference(RFI)is a critical concern for ...Multichannel high-resolution and wide-swath(HRWS)imaging is an advanced digital beamforming technique for future synthetic aperture radar(SAR)systems.However,radio frequency interference(RFI)is a critical concern for HRWS SAR missions,which distorts measure-ments and produces image artifacts.In this paper,the spatial cross-correlation coefficients of multichannel HRWS SAR signals are investigated for RFI detection.It is found when the two channels are correlated,RFI-polluted areas present lower coherence values than non-polluted areas in the same scenarios,which makes previous methods fail.Further,this paper studies the case of two fully decorrelated channels to maximize the coherence difference among RFI and target echoes,and RFI detection is realized by exploiting the anomaly value of coherence.Experimental results of real air-borne multichannel SAR data demonstrate that the RFI can be detected successfully.展开更多
A novel magnetic grating based on calibration was proposed.Two tracks,look-up track and index track,were used to realize absolute output.Magnets of look-up track were magnetized according to N-S-N-S,and magnetic field...A novel magnetic grating based on calibration was proposed.Two tracks,look-up track and index track,were used to realize absolute output.Magnets of look-up track were magnetized according to N-S-N-S,and magnetic field was sensed by 6 linear Hall sensors.Three signals whose phase shift is 120° were obtained through difference,and the offset of magnetic head in a signal period could be obtained by look-up table;Magnets of index track were magnetized according to pseudorandom binary sequence.Hall sensors were used to get the absolute offset of the signal period to which the magnetic head is belonged.The magnetic grating was calibrated using a higher resolution optical grating:output of optical grating and signals from magnetic grating were sampled at the same time and transmitted to computer,the relation between them could be got and stored in MCU for looking-up.The displacement was got according to Hall signals while in working state.A magnetic grating prototype was made,and it could realize absolute detecting in 2048 mm and the resolution could achieve to 0.001 mm.Its structure is simple,cost is very low and it is suitable for mass production.展开更多
Currently, most researches use signals, such as the coil current or voltage of solenoid, to identify parameters; typically, parameter identification method based on variation rate of coil current is applied for positi...Currently, most researches use signals, such as the coil current or voltage of solenoid, to identify parameters; typically, parameter identification method based on variation rate of coil current is applied for position estimation. The problem exists in these researches that the detected signals are prone to interference and difficult to obtain. This paper proposes a new method for detecting the core position by using flux characteristic quantity, which adds a new group of secondary winding to the coil of the ordinary switching electromagnet. On the basis of electromagnetic coupling theory analysis and simulation research of the magnetic field regarding the primary and secondary winding coils, and in accordance with the fact that under PWM control mode varying core position and operating current of windings produce different characteristic of flux increment of the secondary winding. The flux increment of the electromagnet winding can be obtained by conducting time domain integration for the induced voltage signal of the extracted secondary winding, and the core position from the two-dimensional fitting curve of the operating winding current and flux-linkage characteristic quantity of solenoid are calculated. The detecting and testing system of solenoid core position is developed based on the theoretical research. The testing results show that the flux characteristic quantity of switching electromagnet magnetic circuit is able to effectively show the core position and thus to accomplish the non-displacement transducer detection of the said core position of the switching electromagnet. This paper proposes a new method for detecting the core position by using flux characteristic quantity, which provides a new theory and method for switch solenoid to control the proportional valve.展开更多
The value of the high-resolution data lies in the high-precision information discovery.The fine-detailed landform element extraction is thus the basis of high-fidelity application of the high-resolution digital elevat...The value of the high-resolution data lies in the high-precision information discovery.The fine-detailed landform element extraction is thus the basis of high-fidelity application of the high-resolution digital elevation models(DEMs).However,the results of landform element extraction generated by classical methods might be ungrounded on high-resolution DEMs.This paper presents our research on using the aspect to reinforce the applicability and robustness of the classical approaches in landform element extraction.First,according to the research of pattern recognition,we assume that aspect-enhanced landform representation is robust to rotation,scaling and affine variance.To testify the role of aspect,we respectively integrated the aspect into three classical approaches:mean curvaturebased fuzzy classification,elevation-based feature descriptor,and object-based segmentation.In the experiment,based on four types of high-resolution DEMs(1 m,2 m,4 m and 8 m),we compare each classical approaches and their corresponding aspect-enhanced approaches based on extracting the rims of two craters having different landforms,and the ridgelines and valleylines of a region covered by few vegetables and man-made buildings.In comparison to the results generated by curvature-based fuzzy classification,the aspect enhanced curvature-based fuzzy classification can effectively filter a number of noises outperforms the curvature-based one.Otherwise,the aspect-enhanced feature descriptor can detect more landform elements than the elevation-based feature descriptor.Moreover,the aspect-based segmentation can detect the main structure of landform,while the boundaries segmented by classical approaches are messing and meaningless.The systematic experiments on meter-level resolution DEMs proved that the aspect in topography could effectively to improve the classical method-system,including fuzzy-based classification,feature descriptors-based detection and object-based segmentation.The value of aspect is significantly great to be worthy of attentions in landform representation.展开更多
Forest is the largest carbon reservoir and carbon absorber on earth.Thus,mapping forest cover change accurately is of great significance to achieving the global carbon neutrality goal.Accurate forest change informatio...Forest is the largest carbon reservoir and carbon absorber on earth.Thus,mapping forest cover change accurately is of great significance to achieving the global carbon neutrality goal.Accurate forest change information could be acquired by deep learning methods using high-resolution remote sensing images.However,deforestation detection based on deep learning on a large-scale region with high-resolution images required huge computational resources.Therefore,there was an urgent need for a fast and accurate deforestation detection model.In this study,we proposed an interesting but effective re-parameterization deforestation detection model,named RepDDNet.Unlike other existing models designed for deforestation detection,the main feature of RepDDNet was its decoupling feature,which means that it allowed the multi-branch structure in the training stages to be converted into a plain structure in the inference stage,thus the computation efficiency can be significantly improved in the inference stage while maintaining the accuracy unchanged.A large-scale experiment was carried out in Ankang city with 2-meter high-resolution remote sensing images(the total area of it was over 20,000 square kilometers),and the result indicated that the model computation efficiency could be improved by nearly 30%compared with the model without re-parameterization.Additionally,compared with other lightweight models,RepDDNet also displayed a trade-off between accuracy and computation efficiency.展开更多
受GNSS硬件设备、通讯链路以及观测环境等因素影响,GNSS位移监测数据往往包含粗差,无法反映真实的变形特征。针对该问题,本文提出将稳健随机分割森林(robust random cut forest,RRCF)算法应用于GNSS位移监测数据粗差实时检测。仿真数据...受GNSS硬件设备、通讯链路以及观测环境等因素影响,GNSS位移监测数据往往包含粗差,无法反映真实的变形特征。针对该问题,本文提出将稳健随机分割森林(robust random cut forest,RRCF)算法应用于GNSS位移监测数据粗差实时检测。仿真数据处理结果表明,RRCF算法粗差实时检测的准确率、精确率与召回率分别优于95%、98%、96%。地质灾害位移监测数据处理结果表明,GNSS位移监测数据发生异常突变时,RRCF方法检测结果与实际异常值情况吻合且误判率较低。总体而言,RRCF算法对GNSS位移监测数据异常实时检测的准确率和可用性均较好。展开更多
基金the National Natural Science Foundation of China under grant 62172059 and 62072055Hunan Provincial Natural Science Foundations of China under Grant 2020JJ4626+2 种基金Scientific Research Fund of Hunan Provincial Education Department of China under Grant 19B004“Double First-class”International Cooperation and Development Scientific Research Project of Changsha University of Science and Technology under Grant 2018IC25the Young Teacher Growth Plan Project of Changsha University of Science and Technology under Grant 2019QJCZ076.
文摘Now object detection based on deep learning tries different strategies.It uses fewer data training networks to achieve the effect of large dataset training.However,the existing methods usually do not achieve the balance between network parameters and training data.It makes the information provided by a small amount of picture data insufficient to optimize model parameters,resulting in unsatisfactory detection results.To improve the accuracy of few shot object detection,this paper proposes a network based on the transformer and high-resolution feature extraction(THR).High-resolution feature extractionmaintains the resolution representation of the image.Channels and spatial attention are used to make the network focus on features that are more useful to the object.In addition,the recently popular transformer is used to fuse the features of the existing object.This compensates for the previous network failure by making full use of existing object features.Experiments on the Pascal VOC and MS-COCO datasets prove that the THR network has achieved better results than previous mainstream few shot object detection.
基金the scholarship from NUS Graduate School for integrative science and engineering and funding support from Ministry of Education Academic Research Fund,Singapore(AcRF:R-397-000-270-114,R-397-000-183-112).
文摘The separation and detection of particles in suspension are essential for a wide spectrum of applications including medical diagnostics.In this field,microfluidic deterministic lateral displacement(DLD)holds a promise due to the ability of continuous separation of particles by size,shape,deformability,and electrical properties with high resolution.DLD is a passive microfluidic separation technique that has been widely implemented for various bioparticle separations from blood cells to exosomes.DLD techniques have been previously reviewed in 2014.Since then,the field has matured as several physics of DLD have been updated,new phenomena have been discovered,and various designs have been presented to achieve a higher separation performance and throughput.Furthermore,some recent progress has shown new clinical applications and ability to use the DLD arrays as a platform for biomolecules detection.This review provides a thorough discussion on the recent progress in DLD with the topics based on the fundamental studies on DLD models and applications for particle separation and detection.Furthermore,current challenges and potential solutions of DLD are also discussed.We believe that a comprehensive understanding on DLD techniques could significantly contribute toward the advancements in the field for various applications.In particular,the rapid,low-cost,and high-throughput particle separation and detection with DLD have a tremendous impact for point-of-care diagnostics.
基金National Key Basic Research and Development Program of China (973) (No. 2002CB412709)
文摘This study presents a new idea of using only mid-span displacement measurement for damage detection of simply supported beams. Equivalent element concept is introduced at the beginning. In order to relate the damage detectability by means of mid-span displacement measurement with the damage-caused local stiffness change, a novel index termed as symmetrical mid-span displacement difference index (SMDDI) is proposed. The proposed method based on SMDDI is sensitive to tiny damage and comparatively small quantities of measurements are required during the application process. Another significant attraction of this method is putting aside the knowledge a-priori of the intact state. An example using simulated data has been conducted to examine the suitability of this method and assess its comparative advantages over the previous modal method.
基金The National Natural Science Foundation of China under contract No.41706066the National Key R&D Program of China under contract No.2016YFC1400200the China-ASEAN Maritime Cooperation Fund
文摘High-resolution approaches such as multiple signal classification and estimation of signal parameters via rotational invariance techniques(ESPRIT) are currently employed widely in multibeam echo-sounder(MBES)systems for sea floor bathymetry,where a uniform line array is also required.However,due to the requirements in terms of the system coverage/resolution and installation space constraints,an MBES system usually employs a receiving array with a special shape,which means that high-resolution algorithms cannot be applied directly.In addition,the short-term stationary echo signals make it difficult to estimate the covariance matrix required by the high-resolution approaches,which further increases the complexity when applying the high-resolution algorithms in the MBES systems.The ESPRIT with multiple-angle subarray beamforming is employed to reduce the requirements in terms of the signal-to-noise ratio,number of snapshots,and computational effort.The simulations show that the new processing method can provide better fine-structure resolution.Then a highresolution bottom detection(HRBD) algorithm is developed by combining the new processing method with virtual array transformation.The application of the HRBD algorithm to a U-shaped array is also discuss.The computer simulations and experimental data processing results verify the effectiveness of the proposed algorithm.
基金supported by the National Natural Science Foundation of China(41701499)the Sichuan Science and Technology Program(2018GZ0265)the Geomatics Technology and Application Key Laboratory of Qinghai Province(QHDX-2018-07)
文摘A large number of debris flow disasters(called Seismic debris flows) would occur after an earthquake, which can cause a great amount of damage. UAV low-altitude remote sensing technology has become a means of quickly obtaining disaster information as it has the advantage of convenience and timeliness, but the spectral information of the image is so scarce, making it difficult to accurately detect the information of earthquake debris flow disasters. Based on the above problems, a seismic debris flow detection method based on transfer learning(TL) mechanism is proposed. On the basis of the constructed seismic debris flow disaster database, the features acquired from the training of the convolutional neural network(CNN) are transferred to the disaster information detection of the seismic debris flow. The automatic detection of earthquake debris flow disaster information is then completed, and the results of object-oriented seismic debris flow disaster information detection are compared and analyzed with the detection results supported by transfer learning.
文摘The method proposed in this paper is based on the fact that the damage in different types of structural members has distinctive influence on the structural stiffness. The intrinsic mechanical property of the structure is tapped and fully utilized for damage detection. The simplified model of the flexibility of frames treats the individual storeys as springs in series and the frame as an equivalent column. It fully considers the main deformation of all beams and columns in the frame. The deformation property of the simplified model accorded well with that of the actual frame model. The obtained increment of lateral displacement change (IOLDC) at the storey level was found to be very sensitive to the local damage in the frame. A damage detection method is pro- posed using the IOLDCs as the damage identification parameters. Numerical examples demonstrate the potential applicability of this method.
基金supported by the National Natural Foundation of China(Nos.41001282,40871205,and 62271408)partly by Shanghai Aerospace Science and Technology Innovation Fund(No.SAST2021-044)。
文摘Multichannel high-resolution and wide-swath(HRWS)imaging is an advanced digital beamforming technique for future synthetic aperture radar(SAR)systems.However,radio frequency interference(RFI)is a critical concern for HRWS SAR missions,which distorts measure-ments and produces image artifacts.In this paper,the spatial cross-correlation coefficients of multichannel HRWS SAR signals are investigated for RFI detection.It is found when the two channels are correlated,RFI-polluted areas present lower coherence values than non-polluted areas in the same scenarios,which makes previous methods fail.Further,this paper studies the case of two fully decorrelated channels to maximize the coherence difference among RFI and target echoes,and RFI detection is realized by exploiting the anomaly value of coherence.Experimental results of real air-borne multichannel SAR data demonstrate that the RFI can be detected successfully.
基金Sponsored by the Scientific Research Foundation of Heilongjiang Province for Returned Chinese Scholars(Grant No. 2006212)
文摘A novel magnetic grating based on calibration was proposed.Two tracks,look-up track and index track,were used to realize absolute output.Magnets of look-up track were magnetized according to N-S-N-S,and magnetic field was sensed by 6 linear Hall sensors.Three signals whose phase shift is 120° were obtained through difference,and the offset of magnetic head in a signal period could be obtained by look-up table;Magnets of index track were magnetized according to pseudorandom binary sequence.Hall sensors were used to get the absolute offset of the signal period to which the magnetic head is belonged.The magnetic grating was calibrated using a higher resolution optical grating:output of optical grating and signals from magnetic grating were sampled at the same time and transmitted to computer,the relation between them could be got and stored in MCU for looking-up.The displacement was got according to Hall signals while in working state.A magnetic grating prototype was made,and it could realize absolute detecting in 2048 mm and the resolution could achieve to 0.001 mm.Its structure is simple,cost is very low and it is suitable for mass production.
基金supported by National Natural Science Foundation of China(Grant No.51175362)
文摘Currently, most researches use signals, such as the coil current or voltage of solenoid, to identify parameters; typically, parameter identification method based on variation rate of coil current is applied for position estimation. The problem exists in these researches that the detected signals are prone to interference and difficult to obtain. This paper proposes a new method for detecting the core position by using flux characteristic quantity, which adds a new group of secondary winding to the coil of the ordinary switching electromagnet. On the basis of electromagnetic coupling theory analysis and simulation research of the magnetic field regarding the primary and secondary winding coils, and in accordance with the fact that under PWM control mode varying core position and operating current of windings produce different characteristic of flux increment of the secondary winding. The flux increment of the electromagnet winding can be obtained by conducting time domain integration for the induced voltage signal of the extracted secondary winding, and the core position from the two-dimensional fitting curve of the operating winding current and flux-linkage characteristic quantity of solenoid are calculated. The detecting and testing system of solenoid core position is developed based on the theoretical research. The testing results show that the flux characteristic quantity of switching electromagnet magnetic circuit is able to effectively show the core position and thus to accomplish the non-displacement transducer detection of the said core position of the switching electromagnet. This paper proposes a new method for detecting the core position by using flux characteristic quantity, which provides a new theory and method for switch solenoid to control the proportional valve.
基金Supported by the National Natural Science Foundation of China (No.50378041) and the Specialized Research Fund for the Doctoral Program of Higher Education (No.2003487016).
基金Under the auspices of Priority Academic Program Development of Jiangsu Higher Education Institutions(No.140119001)Science&Technology Department of Liaoning Province(No.20180550831)。
文摘The value of the high-resolution data lies in the high-precision information discovery.The fine-detailed landform element extraction is thus the basis of high-fidelity application of the high-resolution digital elevation models(DEMs).However,the results of landform element extraction generated by classical methods might be ungrounded on high-resolution DEMs.This paper presents our research on using the aspect to reinforce the applicability and robustness of the classical approaches in landform element extraction.First,according to the research of pattern recognition,we assume that aspect-enhanced landform representation is robust to rotation,scaling and affine variance.To testify the role of aspect,we respectively integrated the aspect into three classical approaches:mean curvaturebased fuzzy classification,elevation-based feature descriptor,and object-based segmentation.In the experiment,based on four types of high-resolution DEMs(1 m,2 m,4 m and 8 m),we compare each classical approaches and their corresponding aspect-enhanced approaches based on extracting the rims of two craters having different landforms,and the ridgelines and valleylines of a region covered by few vegetables and man-made buildings.In comparison to the results generated by curvature-based fuzzy classification,the aspect enhanced curvature-based fuzzy classification can effectively filter a number of noises outperforms the curvature-based one.Otherwise,the aspect-enhanced feature descriptor can detect more landform elements than the elevation-based feature descriptor.Moreover,the aspect-based segmentation can detect the main structure of landform,while the boundaries segmented by classical approaches are messing and meaningless.The systematic experiments on meter-level resolution DEMs proved that the aspect in topography could effectively to improve the classical method-system,including fuzzy-based classification,feature descriptors-based detection and object-based segmentation.The value of aspect is significantly great to be worthy of attentions in landform representation.
基金supported by the Shenzhen Science and Technology Innovation Project(No.ZDSYS20210623091808026)supported in part by the National Natural Science Foundation of China(General Program,No.42071351)+1 种基金the National Key Research and Development Program of China(No.2020YFA0608501)the Chongqing Science and Technology Bureau technology innovation and application development special(cstc2021jscx-gksb0116).
文摘Forest is the largest carbon reservoir and carbon absorber on earth.Thus,mapping forest cover change accurately is of great significance to achieving the global carbon neutrality goal.Accurate forest change information could be acquired by deep learning methods using high-resolution remote sensing images.However,deforestation detection based on deep learning on a large-scale region with high-resolution images required huge computational resources.Therefore,there was an urgent need for a fast and accurate deforestation detection model.In this study,we proposed an interesting but effective re-parameterization deforestation detection model,named RepDDNet.Unlike other existing models designed for deforestation detection,the main feature of RepDDNet was its decoupling feature,which means that it allowed the multi-branch structure in the training stages to be converted into a plain structure in the inference stage,thus the computation efficiency can be significantly improved in the inference stage while maintaining the accuracy unchanged.A large-scale experiment was carried out in Ankang city with 2-meter high-resolution remote sensing images(the total area of it was over 20,000 square kilometers),and the result indicated that the model computation efficiency could be improved by nearly 30%compared with the model without re-parameterization.Additionally,compared with other lightweight models,RepDDNet also displayed a trade-off between accuracy and computation efficiency.
文摘受GNSS硬件设备、通讯链路以及观测环境等因素影响,GNSS位移监测数据往往包含粗差,无法反映真实的变形特征。针对该问题,本文提出将稳健随机分割森林(robust random cut forest,RRCF)算法应用于GNSS位移监测数据粗差实时检测。仿真数据处理结果表明,RRCF算法粗差实时检测的准确率、精确率与召回率分别优于95%、98%、96%。地质灾害位移监测数据处理结果表明,GNSS位移监测数据发生异常突变时,RRCF方法检测结果与实际异常值情况吻合且误判率较低。总体而言,RRCF算法对GNSS位移监测数据异常实时检测的准确率和可用性均较好。