Landslides pose a greater potential risk to the Sichuan-Tibet Transportation Project, and extensive landslide inventory mapping are essential prevent control geological hazards along Corridor (STTC). Recently proposed detection methods mainly focused on new landslides with high vegetation. In addition, there still challenges in automatic of old using optical images. this paper, two methods, namely mask region-based convolutional neural networks (Mask R-CNN) transfer learning Mask R-CNN...

Space debris detection is important in space situation awareness and asset protection. In this article, we propose a method to detect using feature learning of candidate regions. The acquired optical image sequences are first processed remove hot pixels flicker noise, the nonuniform background information removed by proposed one dimensional mean iteration method. Then, regions (FLCR) extract debris. precisely extracted, then classified trained deep network. model large number simulated with...

Hyperspectral remote sensing image (HSI) classification is very useful in different applications, and recently, deep learning has been applied for HSI successfully. However, the number of training samples usually limited, causing difficulty use models. We propose a wide Fourier network to learn features efficiently by using pruned extracted frequency domain. It composed multiple layers extract hierarchical layer-by-layer efficiently. Each layer includes large transforms domain from local...

Landslide inventory mapping (LIM) is a key prerequisite for landslide susceptibility evaluation and disaster mitigation. It aims to record the location, size, extent of landslides in each map scale. Machine learning algorithms, such as support vector machine (SVM) random forest (RF), have been increasingly applied detection using remote sensing images recent decades. However, their limitations impeded wide application. Furthermore, despite widespread use deep algorithms sensing, LIM, are...

We present a high-accuracy, low false-alarm rate, and computational-cost methodology for removing stars noise detecting space debris with signal-to-noise ratio (SNR) in optical image sequences. First, time-index filtering bright star intensity enhancement are implemented to remove effectively. Then, multistage quasi-hypothesis-testing method is proposed detect the pieces of continuous discontinuous trajectories. For this purpose, defined generated. Experimental results show that can...

As a successful application of machine learning in remote sensing and natural language processing, image captioning remote-sensing images has been promoted developed. Remote are large width, complex features, contain abundant information. It is difficult task to extract available visual features based domain knowledge behind sufficiently utilize extracted feature for generation sufficiently. In order overcome this difficulty, we propose novel model "encoder-decoder" framework, termed with...

Old landslides in the Loess Plateau, Northwest China usually occurred over a relatively long period, and their sizes are smaller compared to old alpine valley areas of Sichuan, Yunnan, Southeast Tibet. These landslide may have been changed either partially or greatly, they covered with vegetation similar surrounding environment. Therefore, it is great challenge detect them using high-resolution remote sensing images only orthophoto view. This paper proposes optimal-view multi-view strategic...

Landslide inventory mapping (LIM) is an important prerequisite for disaster emergency rescue and landslide sensitivity analysis. It has been proven that convolutional neural networks have better performance LIM than traditional machine learning methods such as support vector machines (SVM), random forests (RF). However, the accuracy of existing based only on optical images low due to complex background. Moreover, multi-scale features landslides are not considered in network methods....

Land cover (LC) information plays an important role in different geoscience applications such as land resources and ecological environment monitoring. Enhancing the automation degree of LC classification updating at a fine scale by remote sensing has become key problem, capability data acquisition is constantly being improved terms spatial temporal resolution. However, present methods generating are relatively inefficient, manually selecting training samples among multitemporal observations,...

Precise vegetation maps of mountainous areas are great significance to grasp the situation an ecological environment and forest resources. In this paper, while multi-source geospatial data can generally be quickly obtained at present, realize effective mapping in when samples difficult collect due their perilous terrain inaccessible deep forest, we propose a novel intelligent method sample collection for machine-learning (ML)-based mapping. First, employ geo-objects (i.e., polygons) from...

Recently, deep learning has been successfully and widely used in hyperspectral image (HSI) classification. Considering the difficulty of acquiring HSIs, there are usually a small number pixels as training instances. Therefore, it is hard to fully use advantages networks; for example, very layers with large parameters lead overfitting. This paper proposed dynamic wide neural network (DWDNN) HSI classification, which includes multiple efficient sliding window subsampling (EWSWS) networks can...

Fast space debris detection is valuable to safety and vehicle operation. In this paper, we proposed a fast method with grid-based learning. The image divided into 14×14 grids, then the neural network (FGBNN) used locate position of in grids. experiments show that, test set, model can reach accuracy 98.83% recall 95.99% SNR 2 debris, it process 430 images per second given experiment settings prediction process. It proved that has excellent performance speed.

Deep learning networks have achieved great success in many areas, such as large-scale image processing. They usually need large computing resources and time process easy hard samples inefficiently the same way. Another undesirable problem is that network generally needs to be retrained learn new incoming data. Efforts been made reduce realize incremental by adjusting architectures, scalable effort classifiers, multi-grained cascade forest (gcForest), conditional deep (CDL), tree CNN,...

A full-sky autonomous star identification algorithm aimed at solving the "lost-in-space" problem is presented in this paper. It mainly consists of two steps: an initial match step and a reliability evaluation step. Oriented singular value feature matching adopted to search for corresponding candidates stars detected match. After obtaining stars' results, method applied estimate from voting acquiring final unique image. Experiments show that our more robust position noise magnitude than...

Recently, deep learning methods, for example, convolutional neural networks (CNNs), have achieved high performance in hyperspectral image (HSI) classification. The limited training samples of HSI images make it hard to use methods with many layers and a large number kernels as scale imagery tasks, CNN-based usually need long time. In this paper, we present wide sliding window subsampling network (WSWS Net) It is based on transform windows (WSWS). can be extended the direction learn both...

An algorithm was proposed to detect the space debris in optical astronomical images. maximum projection induced reduce data amount three dimensional space. when detected, it tracked by particle filter thus precise location information can be got at any time. simulation results show that method perfectly meet requirement of detection with a high accuracy (rms error less than 1 pixel) and probability, low false alarm rejection.

As an important error in star centroid location estimation, the systematic greatly restricts accuracy of three-axis attitude supplied by a sensor. In this paper, analytical study about behavior center mass (CoM) estimation method under different Gaussian widths starlight energy distribution is presented means frequency field analysis and numerical simulations. Subsequently, optimized extreme learning machine (ELM) based on bat algorithm (BA) adopted to predict actual position then compensate...

With the rapid development of research on machine learning models, especially deep learning, more and endeavors have been made designing new models with properties such as fast training good convergence, incremental to overcome catastrophic forgetting. In this paper, we propose a scalable wide neural network (SWNN), composed multiple multi-channel RBF networks (MWRBF). The MWRBF focuses different regions data nonlinear transformations can be performed Gaussian kernels. number MWRBFs for...

Star tracker has been widely used as a precise and reliable device for the attitude measuring of spacecraft. The accuracy star location will affect identification finally measurement. This paper proposed novel method to locate position with phase transfer function (PTF). numerical expressions are deduced diffraction model point in 1-D given directly 2-D. Then calculation is performed better than 2.1% pixels (SNR=20) 3×3 window airy disk, which higher traditional centroid method. Different...

The pre-processing method of star image, extracting and high accuracy location are studied to improve the reduce amount calculation during processing. background noise CCD (Charge Coupled Device) image analyzed, is performed: correct light dead point, estimate remove phenomenon non-uniformity with local mean iteration method. improved cross projection proposed extract stars in reduced because scanning marking process not needed compared traditional connected domain And stretching range...