Extracting roads from high-resolution remote sensing images (HRSIs) is vital in a wide variety of applications, such as autonomous driving, path planning, and road navigation. Due to the long thin shape well shades induced by vegetation buildings, small-sized are more difficult discern. In order improve reliability accuracy extraction when multiple sizes coexist an HRSI, enhanced deep neural network model termed Dual-Decoder-U-Net (DDU-Net) proposed this paper. Motivated U-Net model, small...

Landslides are one of the most serious natural hazards along Sichuan-Tibet transportation corridor, which crosses complicated region in world terms topography and geology. Landslide susceptibility mapping (LSM) is high demand for risk assessment disaster reduction this mountainous region. A new model, namely Convolutional-Squeeze Excitation-long short-term memory network (Conv-SE-LSTM), proposed to map landslide corridor. Compared with conventional deep learning models, Conv-SE-LSTM...

Landslide is one of the most dangerous and frequently occurred natural disasters. The semantic segmentation technique efficient for wide area landslide identification from high-resolution remote sensing images (HRSIs). However, considerable challenges exist because effects sediments, vegetation, human activities over long periods time make visually blurred old landslides very challenging to detect based upon HRSIs. Moreover, terrain features like slopes, aspect altitude variations cannot be...

The early identification of potential landslide hazards is great practical significance for disaster warning and prevention. study used different machine learning methods to identify active landslides along a 15 km buffer zone on both sides Jinsha River (Panzhihua-Huize section), China. morphology texture features were characterized with InSAR deformation monitoring data high-resolution optical remote sensing data, combined 17 influencing factors. In the area, 83 accumulation areas 54...

Lakes are an important component of global water resources. In order to achieve accurate lake extractions on a large scale, this study takes the Tibetan Plateau as area and proposes Automated Lake Extraction Workflow (ALEW) based Google Earth Engine (GEE) deep learning in response problems low identification accuracy efficiency complex situations. It involves pre-processing massive images creating database examples extraction Plateau. A lightweight convolutional neural network named...

Recent developments in hyperspectral satellites have dramatically promoted the wide application of large-scale quantitative remote sensing. As an essential part preprocessing, cloud detection is great significance for subsequent analysis. For Gaofen-5 (GF-5) data producers, daily hundreds scenes a challenging task. Traditional methods cannot meet strict demands production, especially GF-5 satellites, which massive volumes. Deep learning technology, however, able to perform efficiently...

The geological characteristics of old landslides can provide crucial information for the task landslide protection. However, detecting from high-resolution remote sensing images (HRSIs) is great challenges due to their partially or strongly transformed morphology over a long time and thus limited difference with surroundings. Additionally, small-sized datasets restrict in-depth learning. To address these challenges, this paper proposes new iterative classification semantic segmentation...

Automatic landslide detection based on very high spatial resolution remote sensing images is crucial for disaster prevention and mitigation applications. With the rapid development of deep-learning techniques, state-of-the-art semantic segmentation methods fully convolutional network (FCNN) have achieved outstanding performance in task. However, most existing articles only utilize visual features. Even if advanced FCNN models are applied, there still a certain amount falsely detected miss...

Few-shot hyperspectral classification is a challenging problem that involves obtaining effective spatial–spectral features in an unsupervised or semi-supervised manner. In recent years, as result of the development computer vision, deep learning techniques have demonstrated their superiority tackling problems unmixing (HU) and classification. this paper, we present new pipeline for few-shot classification, where endmember abundance maps obtained by HU are treated latent A cube-based...

Early discovery and monitoring of the active deformation areas potential landslides are important for geohazard risk prevention. The objective study is to propose a one-step strategy automatically mapping from Sentinel-1 SAR dataset. First, we built generalized convolutional neural network (CNN) based on activity topographic characteristics. Second, conducted comparative analysis performance various multi-channel combiners detecting landslides. Third, verified transferability pretrained CNN...

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...

Landslides are common hazardous geological events, and accurate efficient landslide identification methods important for hazard assessment post-disaster response to disasters. Deep learning (DL) based on remote sensing data currently widely used in tasks. The recently proposed segment anything model (SAM) has shown strong generalization capabilities zero-shot semantic segmentation. Nevertheless, SAM heavily relies user-provided prompts, performs poorly identifying landslides images. In this...

矿物识别是高光谱遥感技术优势之一,已在地质矿产领域取得了显著应用效果。随着光谱分辨率的不断提高,高光谱遥感矿物识别逐渐从识别矿物种类向矿物亚类、矿物成分等精细信息识别发展,且随着应用实践的不断深入,对矿物精细信息的需求也越来越大。而光谱分辨率和矿物识别方法是制约高光谱矿物精细识别的主要因素。高分五号(GF-5)超高的光谱分辨率为矿物精细识别提供了可能。首先在分析总结已有高光谱矿物识别方法优缺性的基础上,提出了综合光谱特征增强匹配度和特征参量的矿物识别方法;其次,选取甘肃柳园和美国Cuprite两个研究较多的地区为研究对象,基于GF-5卫星数据开展了矿物精细识别,在完成矿物种类、亚类识别的基础上,进一步对绢云母成分信息进行了反演;最后,结合上述地区已有机载高光谱数据及填图结果开展对比分析。结果表明:GF-5矿物识别信息分布与机载HyMap、AVIRIS一致性很好,相较机载数据GF-5矿物识别平均正确率优于90%,说明本研究提出的矿物识别方法能够满足GF-5矿物精细识别,可为后续业务化应用提供技术支撑,同时认为超高的光谱分辨率使得GF-5在矿物成分信息识别上更具优势。

A hyperspectral image (HSI) contains hundreds of spectral bands, which provide detailed information, thus offering an inherent advantage in classification. The successful launch the Gaofen-5 and ZY-1 02D satellites has promoted need for large-scale geological applications, such as mineral lithological mapping (LM). In recent years, following success computer vision, deep learning methods have shown their solving problem However, combination HSI to solve is insufficient. We propose a new 3D...

Landslides are one of the most destructive natural disasters in world, posing a serious threat to human life and safety. The development foundation models has provided new research paradigm for large-scale landslide detection. Segment Anything Model (SAM) garnered widespread attention field image segmentation. However, our experiment found that SAM performed poorly task We propose TransLandSeg, which is transfer learning approach semantic segmentation based on vision model (VFM)....

With the advancement of artificial intelligence, deep learning has become instrumental in land cover classification. While there been a notable emphasis on refining model structures to improve classification accuracy, it is imperative also emphasize pivotal role data-driven optimization techniques. This paper presents an in-depth investigation into optimizing multi-class using high-resolution multispectral images from Worldview3. We explore various strategies, including refined sampling data...

The Guanting Reservoir supplied drinking water to Beijing until 1997, following which the quality of reservoir deteriorated. chlorophyll-a concentration (Cchl-a) is an important indicator eutrophication. Therefore, changes in Cchl-a should be monitored and analysed. For more than 30 years, monitoring inland waterbodies has only been possible using Landsat 5 Thematic Mapper (TM), 7 Enhanced (ETM), 8 Operational Land Imager (OLI). However, there are data consistency problems these sensors. To...

Abstract. With the rapid development of artificial intelligence, significant progress has been made in land cover classification using deep learning methods. However, existing research, most studies focus more on improving accuracy by optimizing model structure and less mining value data itself. In this paper, experiments remote sensing multi-class were conducted based Worldview3 data, strategies to improve proposed terms sampling methods, band combination, loss function, optimization....

Abstract. Geometric correction, a pivotal step in the preprocessing of airborne remote sensing imagery, is critical for ensuring accuracy subsequent quantitative analyses. Achieving precise and efficient geometric correction hyperspectral data remains significant challenge field. This study presents new method system-level fine-scale uncontrolled images utilizing DEM data, which integrates forward inverse transformation algorithms. Furthermore, an optimized workflow proposed to facilitate...

Relic landslide, formed over a long period, possess the potential for reactivation, making them hazardous geological phenomenon. While reliable relic landslide detection benefits effective monitoring and prevention of disaster, semantic segmentation using high-resolution remote sensing images landslides faces many challenges, including object visual blur problem, due to changes appearance caused by prolonged natural evolution human activities, small-sized dataset difficulty in recognizing...