Landslide disasters are extremely destructive. Accurate identification of landslides plays an important role in disaster assessment, loss control and post-disaster reconstruction. This paper proposes a semantic segmen...
Landslide disasters are extremely destructive. Accurate identification of landslides plays an important role in disaster assessment, loss control and post-disaster reconstruction. This paper proposes a semantic segmentation landslide identification method based on improved U-Net. The deep convolution neural network and jump connection method is used for end-to-end semantic segmentation to achieve deep feature extraction and fusion of different receptive fields, thus enriching feature information. SENet modules are adopted to enhance the ability of the model to extract important features, so as to further improve the accuracy of model recognition. Extensive experiments show that our improved U-Net achieves better performance than the original algorithm on our landslide datasets. The results of Iou are improved by 4.12% which demonstrates our work is of great significance for the research of landslide area identification. Finally, the model is deployed to the web and applied to the geological hazard intelligent monitoring system to realize the landslide identification task.
Deep learning is currently the mainstream method for ceramic defect detection, and it requires a large number of defect samples to train the network. However, collecting these defect samples is very time-consuming and...
详细信息
Deep learning is currently the mainstream method for ceramic defect detection, and it requires a large number of defect samples to train the network. However, collecting these defect samples is very time-consuming and deep learning suffers from few-shot learning problems. In this study, a StyleGAN3-based data augmentation method for ceramic defect detection was proposed which can generate ceramic defect samples and thus reduce the data collection work. Experiments show that our method uses less training time, has a more stable training process, and can improve the accuracy of the detection network.
This paper focuses on the bounded tracking control of general linear multi-agent systems(MASs), considering the effects of inevitable communication time-delays, measurement noises, and uncertain disturbances in practi...
详细信息
This paper focuses on the bounded tracking control of general linear multi-agent systems(MASs), considering the effects of inevitable communication time-delays, measurement noises, and uncertain disturbances in practical applications. Firstly, the bounded tracking control problem of uncertain MASs under multiplicative noises is transformed into the boundedness problem of stochastic differential delay equations. Then, the upper bound of agent tracking is calculated by means of linear variation, variation of constants formula and stochastic analysis theory, and sufficient conditions are given for the system to achieve the bounded tracking control.
Landslide displacement prediction is an important and indispensable part of landslide monitoring and warning. The change of the displacement is always considered being related to inducing factors, which are aimed at i...
Landslide displacement prediction is an important and indispensable part of landslide monitoring and warning. The change of the displacement is always considered being related to inducing factors, which are aimed at improving accuracy of the predicted model. However, the seasonal characteristic of the displacement, which has not been carefully analyzed, reveals the law of inducing factors. In order to gain a deeper understanding of characteristics, the Baijiabao landslide is taken as an example. The variational mode decomposition (VMD) method, which can extract effective information well, is introduced to decompose the displacement. Introducing the seasonal parameters, the seasonal autoregressive integrated moving average (SARIMA) model is established to predict the displacement subseries. Finally, accumulative displacement prediction values are obtained by superimposing the predicted subseries. With higher accuracy and lower error, the VMD-SARIMA model proves a better option in application compared with VMD-ARIMA, SARIMA and ARIMA models.
This paper addresses the problem of state estimation for Markov jump genetic oscillator networks with time-varying delays based on hidden Markov model. Two non-identical types of time-varying delays, that is, the inte...
This paper addresses the problem of state estimation for Markov jump genetic oscillator networks with time-varying delays based on hidden Markov model. Two non-identical types of time-varying delays, that is, the intercellular coupling delay, and the regulatory delay are considered in consideration in genetic oscillator networks. Then a state estimator is designed by solving a set of linear matrix inequalities that can be solved with existing software. Finally, The effectiveness of state estimation approach can then be demonstrated through a numerical example.
In this paper, a template matching and trend feature analysis-based data pre-processing method for seismic wave detection is proposed with two stages. In the first stage, it involves extracting the rock physical param...
In this paper, a template matching and trend feature analysis-based data pre-processing method for seismic wave detection is proposed with two stages. In the first stage, it involves extracting the rock physical parameters from seismic wave detection results using OCR (Optical Character Recognition) method, and extracting the original rock physical parameters from the raw rock property table using keyword matching method. Using the rock physical parameters as a template, a template matching approach is employed to eliminate abnormal values from the original rock physical parameters. In the next stage, a technique is proposed to extract trend features of rock physical parameters for conducting advanced geological forecasting, which considered the expertise of experts in interpreting seismic wave detection data. Finally, the effectiveness of the proposed method is verified by the compared simulation results.
The irradiance-power curve is an important basis for examining the operating status of photovoltaic power stations. In the actual operation process, sensor failure, abnormal communication and equipment damage will bri...
详细信息
The irradiance-power curve is an important basis for examining the operating status of photovoltaic power stations. In the actual operation process, sensor failure, abnormal communication and equipment damage will bring a large number of abnormal values to the output data of photovoltaic power plants. It will have a significant impact on a variety of applications based on photovoltaic output data. This paper analyzes the typical outliers on the irradiance-power curve and proposes a photovoltaic output data cleaning method based on fuzzy clustering algorithm and quartile algorithm. By comparing with the quartile method, it is proved that this method can effectively identify abnormal data when there are a large number of outliers in the photovoltaic output data.
Single cell RNA sequencing (scRNA-seq) technology can study gene expression in single cell resolution and solve cell heterogeneity that cannot be solved by the traditional RNA sequencing (Bulk RNA-seq) technology. It ...
详细信息
High precision modeling in industrial systems is difficult and costly. Model-free intelligentcontrol methods, represented by reinforcement learning, have been applied in industrial systems broadly. The hard evaluated...
详细信息
High precision modeling in industrial systems is difficult and costly. Model-free intelligentcontrol methods, represented by reinforcement learning, have been applied in industrial systems broadly. The hard evaluated of production states and the low value density of processing data causes sparse rewards, which lead to an insufficient performance of reinforcement learning. To overcome the difficulty of reinforcement learning in sparse reward scenes, a reinforcement learning method with reward shaping and hybrid exploration is proposed. By perfecting the rewards distribution in the state space of environment, the reward shaping can make the state-value estimation of reinforcement learning more accurate. By improving the rewards distribution in time dimension, the hybrid exploration can make the iteration of reinforcement learning more efficient and more stable. Finally, the effectiveness of the proposed method is verified by simulations.
Leaks in natural gas pipelines can cause very serious safety accidents, and timely detection and remedial action can greatly reduce the losses. In recent years, pipeline leak detection has received extensive studies. ...
Leaks in natural gas pipelines can cause very serious safety accidents, and timely detection and remedial action can greatly reduce the losses. In recent years, pipeline leak detection has received extensive studies. Most methods use pressure sensors or acoustic sensors to detect pipelines, but there are certain limitations on the usage scenarios and detection time delays. On this basis, this paper selects maglev vibration detector to detect the vibration signal of pipelines. The difficulty lies in that, sudden changes in vibration signals due to external disturbances, may lead to false alarms. Therefore, this paper proposes a pipeline leak detection method using Multivariate Gaussian Distribution based Kullback-Leibler Divergence (MGD-KLD) and on-delay timer to reduce false alarms during the detection process. In this paper, by constructing a simulated pipeline platform for leak experiments and applying the above method to process the experimental data, the false alarm rate of pipeline leak detection can be effectively reduced.
暂无评论