Convolutional Neural Networks (CNNs), such as U-Net, have shown competitive performance in the automatic extraction of buildings from Very High-Resolution (VHR) aerial images. However, due to unstable multi-scale context aggregation, insufficient combination multi-level features and lack consideration semantic boundary, most existing CNNs produce incomplete segmentation for large-scale result predictions with huge uncertainty at building boundaries. This paper presents a novel network...

Vehicle targets in unmanned aerial vehicle (UAV) images are generally small, so a significant amount of detailed information on may be lost after neural computing, which leads to the poor performances existing recognition algorithms. Based convolutional networks that utilize YOLOv3 algorithm, this article focuses development quick automatic detection method for UAV images. First, dataset target is constructed. Then, novel framework proposed according following characteristics: The image...

This paper considers a reconnaissance task allocation problem for multiple unmanned aerial vehicles (UAVs) in 3D urban environments. In this paper, we present an extended heterogeneous targets model which introduced cuboid environment to improve the fidelity of model. A method is designed each type target, and mission described as target multi-traveling salesman solving complex optimization problems with constraints.To address these problems, multi-group symbiotic organisms search algorithms...

Abstract Development of shape memory polymer materials with integrated self‐healing ability, property, and outstanding mechanical properties is a challenge. Herein, isophorone diisocyanate, polytetramethylene ether glycol, dimethylglyoxime, glycerol have been used to preparation polyurethane by reacting at 80°C for 6 h. Then, graphene oxide (GO) was added the reaction keep 4 h obtain polyurethane/GO composite properties. Scanning electron microscopy shows that GO sheets were dispersed...

As an inevitable phenomenon in most optical remote-sensing images, the effect of shadows is prominent urban scenes. Shadow detection critical for exploiting and recovering distorted information. Unfortunately, general, automatic shadow methods aerial images cannot achieve satisfactory performance due to limitation feature patterns lack consideration non-local contextual To address this challenging problem, global-spatial-context-attention (GSCA) module was developed self-adaptively aggregate...

Remote sensing image change detection method plays a great role in land cover research, disaster assessment, medical diagnosis, video surveillance, and other fields, so it has attracted wide attention. Based on small sample dataset from SZTAKI AirChange Benchmark Set, order to solve the problem that deep learning network needs large number of samples, this work first uses nongenerative augmentation generative based convolutional adversarial networks, then, constructs remote model an improved...

Image registration is an important basis of image processing, which great significance in mosaicking, target recognition, and change detection. Aiming at the automatic problem multi-angle optical images for ground scenes, a method combining point features line to register proposed. Firstly, LSD (Line Segment Detector) algorithm used extract images. The obtained segments whose length are less than given threshold eliminated by visual significant algorithm. Then, affine transform model...

In agricultural remote sensing monitoring, the quality of optical image data acquisition is often affected by climate, and acquired satellite imagery results usually contain cloud information, which can lead to a lack ground information.Unlike thick clouds, semi-transparent nature thin clouds prevents from completely obscuring scene.In order remove in cultivated land restore real information as much possible, we proposed removal method spatial fusion self-attention generative adversarial...

Cloud is a serious problem that affects the quality of remote-sensing (RS) images Existing cloud removal techniques suffer from notable limitations, such as being specific to certain data types, conditions, and spatial extents, well requiring auxiliary data, which hampers their generalizability flexibility. To address issue, we propose maximum-value compositing approach by generating masks. We acquired 432 daily MOD09GA L2 MODIS imageries covering vast region with persistent cover various...

Since technologies in image fusion, splicing, and target recognition have developed rapidly, as the basis of many applications, performance registration directly affects subsequent work. In this work, for rich features satellite-borne optical imagery such panchromatic multispectral images, Harris corner algorithm is combined with scale invariant feature transform (SIFT) operator point extraction. Our rough matching strategy uses K-D (K-Dimensional) tree BBF (Best Bin First) method,...

Since the vehicle targets in unmanned aerial (UAV) images are generally small, existing neural network calculations will cause detailed information to be seriously lost, which leads poor effects of recognition algorithms. Here we improved one most representative algorithms field target refer characteristics UAV image datasets. There two main tasks: first, improve Faster R-CNN structure so that it can supplement small targets; second, verify has higher accuracy, model discrimination and...

With the rapid development of unmanned aerial vehicle (UAV) technology, UAV remote sensing images are increasing sharply. However, due to limitation perspective sensing, obtained from different viewpoints a same scene need be stitched together for further applications. Therefore, an automatic registration method based on deep residual features is proposed in this work. It needs no additional training and does not depend image features, such as points, lines shapes, or specific contents. This...

Image registration plays a vital role in the mosaic process of multiple UAV (Unmanned Aerial Vehicle) images acquired from different spatial positions same scene. Aimed at problem that many fast methods cannot provide both high speed and accuracy simultaneously for visible light images, this work proposes novel framework based on popular baseline algorithm, ORB—the Oriented FAST (Features Accelerated Segment Test) Rotated BRIEF (Binary Robust Independent Elemental Features) algorithm. First,...

A two-dimensional finite element model was used to analyze the thermal and mechanical behavior during solidification of strand in a continuous bloom casting mold. The coupled heat transfer deformation were analyzed simulate formation air gap between mold strand. investigate influence taper on temperature stress distributions results show that mainly forms around corner, causing hotter thinner solidifying shell this region. partially compensates for shrinkage reduces transfer. compresses...

The semantic segmentation of laser point clouds is critical for many applications aerial clouds. However, most the existing deep learning networks do not make full use cloud data information. PointNet++ was chosen as baseline network, and a deep-residual enhanced encoding method multi-feature information proposed in this work. Firstly, more efficient network structure to enhance geometric constructed, called GEO–PointNet layer. Then, novel feature aggregation, named SEP–PointNet, introduced...

In obtaining Digital Elevation Model (DEM), most methods of acquiring the tie points are generated automatically by software and then manually screened, which is time-consuming labor-intensive, accuracy cannot be guaranteed. Therefore, this paper proposes an automatic stereo matching method combining Speeded Up Robust Features (SURF) Rational Function (RFM) to reconstruct 3D model remote sensing generalized image pairs. There two main tasks: first, apply SURF algorithm images, screen at same...

In the procedure of hyperspectral data dimensionality reduction (DR), intrinsic (ID) high-dimensional is normally obtained through linear analysis methods. This article applies a kind unsupervised learning method, manifold to for and gives manifold-learning-based algorithm data. The experiments use ISOMAP, LLE, LE LTSA algorithms estimate simulated real data, get two-dimension figures discuss advantages disadvantages these in analysis.

In this paper, a novel condition trend prediction method named WHMAR for electronic systems is presented, which based on weighed Hidden Markov model (HMM) and autoregressive model(AR). The basic idea constructing AR cells as the output of HMM, leads to segmentation time series into different models. hidden state sequence chain chosen predicted firstly by means method. second step, results are computed output. tested complex chaotic typical equipment's BIT states, experiment promising.