Object detection in aerial images is an active yet challenging task computer vision because of the bird’s-eye view perspective, highly complex backgrounds, and variant appearances objects. Especially when detecting densely packed objects images, methods relying on horizontal proposals for common object often introduce mismatches between Region Interests (RoIs) This leads to misalignment final classification confidence localization accuracy. In this paper, we propose a RoI Transformer address...

Object detection in aerial images is an active yet challenging task computer vision because of the birdview perspective, highly complex backgrounds, and variant appearances objects. Especially when detecting densely packed objects images, methods relying on horizontal proposals for common object often introduce mismatches between Region Interests (RoIs) This leads to misalignment final classification confidence localization accuracy. Although rotated anchors have been used tackle this...

This paper develops several new strategies for remote sensing image classification postprocessing (CPP) and conducts a systematic study in this area. CPP is defined as refinement of the labeling classified order to enhance its original accuracy. The current mainstream methods (preprocessing) extract additional spatial features complement spectral information using responses alone. On other hand, however, methods, providing solution improve accuracy by refining initial result, have not...

The rapidly increasing world population and human activities accelerate the crisis of limited freshwater resources. Water quality must be monitored for sustainability Unmanned aerial vehicle (UAV)-borne hyperspectral data can capture fine features water bodies, which have been widely used monitoring quality. In this study, nine machine learning algorithms are systematically evaluated inversion parameters including chlorophyll-a (Chl-a) suspended solids (SS) with UAV-borne data. comparing...

Hyperspectral imagery has been widely used in precision agriculture due to its rich spectral characteristics. With the rapid development of remote sensing technology, airborne hyperspectral shows detailed spatial information and temporal flexibility, which open a new way accurate agricultural monitoring. To extract crop types from images, we propose fine classification method based on multi-feature fusion deep learning. In this research, morphological profiles, GLCM texture endmember...

Soil particle size fractions (PSFs) are important properties for understanding the physical and chemical processes in soil systems. Knowledge about distribution of PSFs is critical sustainable management. Although log-ratio transformations have been widely applied to prediction, statistical original data transformed given by different, resulting biased estimates PSFs. Therefore, multivariate random forest (MRF) was utilized simultaneous prediction PSFs, as it able capture dependencies...

Supervised classification is the commonly used method for extracting ground information from images. However, supervised classification, selection and labelling of training samples an expensive time-consuming task. Recently, automatic indexes have achieved satisfactory results indicating different land-cover classes, which makes it possible to develop instead manual interpretation. In this paper, we propose a high-resolution image classification. way, initial candidate can be provided by...

Unmanned aerial vehicle (UAV)-borne hyperspectral imagery has been applied in precision agriculture, owing to its high spatial and spectral resolution. Specifically, the resolution is conducive revealing textural characteristics of crops, while can depict detailed differences among crops. In this study, we explored potential extended attribute profiles (EAP) modeling spectral-spatial UAV-borne for precise crop classification. two dimensionality reduction approaches, namely, principle...

Many significant applications need land cover information of remote sensing images that are acquired from different areas and times, such as change detection disaster monitoring. However, it is difficult to find a generic classification scheme for due the spectral shift caused by diverse acquisition condition. In this paper, we develop novel method can deal with large-scale data captured widely distributed times. Additionally, establish dataset consisting 150 Gaofen-2 imageries support model...

The total arsenic (TAs) content in the soil is commonly used as an important indicator for evaluating pollution. However, traditional methods investigating TAs concentration over a large area are always labor-intensive and costly. As rapid convenient technique, unmanned aerial vehicle (UAV) equipped with hyperspectral camera offers promising way estimating distribution of TAs. In this study, we utilized UAV-borne data Daye city China mining suspected contaminated to establish deep model...

Prior information about classes plays an important role in the high-resolution image classification. Produced by volunteers with GPS tracking practice and local knowledge, crowdsourced OpenStreetMap (OSM) data have shown potential as a time-saving cost-effective way to provide prior for In this letter, we develop remote-sensing classification method using OSM information. objects of interest except roads are extracted construct training set To decrease misleading errors redundancy OSM,...

The spatial information has been proved to be effective in improving the performance of spectral-based classification. However, it is difficult describe different image scenes by using monofeature owing complexity geospatial scenes. In this letter, a novel framework developed combine multiple spectral and features based on Markov random field (MRF). Specifically, pixels an are separated into reliable unreliable ones according decision multifeature classifications. labels can conveniently...

Land cover classification of urban areas is critical for understanding the environment. High-resolution remotely sensed imagery provides abundant, detailed spatial information classification. In meantime, OpenStreetMap (OSM) data, as typical crowd-sourced geographical information, have been an emerging data source obtaining information. this context, a land method that fuses high-resolution and OSM proposed. Training samples were generated by integrating multiple indexes. which contain class...

Malware classification is a critical task in cybersecurity, as it offers insights into the threats that malware poses to victim device and helps design of countermeasures. For real-time classification, due large amount potential present network, there challenge achieving high accuracy while maintaining low inference latency. We first introduce two self-attention transformer-based classifiers, SeqConvAttn ImgConvAttn, replace currently predominant CNN-based classifiers. then devise...

Aerial image scene classification is a fundamental problem for understanding high-resolution remote sensing images and has become an active research task in the field of due to its important role wide range applications. However, limitations existing datasets classification, such as small scale low-diversity, severely hamper potential usage new generation deep convolutional neural networks (CNNs). Although huge efforts have been made building large-scale very recently, e.g., Image Dataset...

As an effective approach to obtaining agricultural information, the remote sensing technique has been applied in classification of crop types. The unmanned aerial vehicle (UAV)-manned hyperspectral sensors provide imagery with high spatial and spectral resolutions. Moreover, detailed as well abundant properties UAV-manned imagery, opens a new avenue fine crops. In this manuscript, multiscale superpixel-based approaches are proposed for identification crops imagery. Specifically, superpixel...