Synthetic aperture radar (SAR) ship detection based on deep learning has been widely applied in recent years. However, two main obstacles are hindering SAR detection. First, the identification of ships a port is seriously disrupted by presence onshore buildings. It difficult for existing algorithms to effectively distinguish targets from such complex background. Additionally, it appears more complicated accurately locate densely arranged ships. Second, images exist at variety scales due...

Ship detection plays an important role in synthetic aperture radar (SAR) image interpretation. However, there are still some difficulties SAR ship detection. First, ships often have a large aspect ratio and arbitrary directionality images. Traditional algorithms can cause the area to be redundant, which makes it difficult accurately locate target complex scenes. Second, ports densely arranged, effective identification of arranged is complicated. Finally, images exist at variety scales due...

Previous state-of-the-art real-time object detectors have been reported on GPUs which are extremely expensive for processing massive data and in resource-restricted scenarios. Therefore, high efficiency CPU-only devices urgently-needed industry. The floating-point operations (FLOPs <sup xmlns:mml="http://www.w3.org/1998/Math/MathML" xmlns:xlink="http://www.w3.org/1999/xlink">1</sup> ) of networks not strictly proportional to the running speed CPU devices, inspires design an exactly "fast"...

Atmospheric effects are among the primary error sources affecting accuracy of interferometric synthetic aperture radar (InSAR). The topography-dependent atmospheric effect is particularly noteworthy in reservoir areas for landslide monitoring utilizing InSAR, which must be effectively corrected to complete InSAR high-accuracy measurement. This paper proposed a correction method based on Multi-Layer Perceptron (MLP) neural network model combined with topography and spatial data information....

In object detection tasks, the of small size objects is very difficult since these targets are always tightly grouped and interfered by background information. order to solve this problem, we propose a novel network architecture based on YOLOv3 new feature fusion mechanism. We added multi-scale convolution kernels differential receptive fields into extract semantic features using an Inception-like architecture. also optimize weights selecting appropriate channel number ratios. Our model...

Vehicle detection in traffic surveillance videos is a special subtask object detection, where desired objects are vehicles moving on the road while background still within sequence. The disparity of speed each frame, i.e. and static, consistent with vehicle semantic to some extent, thus motions can be extracted enhance appearance foreground. In this paper, we propose motion prior embedded parallel architecture for aiming at illuminating suppressing false positives background. We further...

Motivated by the body structure and movement of sea turtles, this paper proposes a novel mechanical design an innovative turtle-inspired robot system for both efficient fast swimming high spatial maneuverability. The robotic turtle is driven two variable stiffness fore hydrofoils to produce thrust soft hind control direction. Considering long-time observation in wide ocean, centre gravity adjustment mechanism buoyancy are developed keep glide lower pow manner. modular layout overall also...

Abstract The large reservoirs in the southwestern Alpine Canyon region are characterized by long reservoir banks and complex geological structures. problem of finding deformation zone quickly efficiently is urgent needs to be resolved. In this study, taking area 110 km upstream Baihetan dam site as study area, applicability various interferometric synthetic aperture radar (InSAR) techniques was summarized, small baseline subset (SBAS-InSAR) method used carry out large-scale disaster risk...

Deep learning (DL) systems are typically used to accelerate training DL jobs. Training models requires feeding mass input data. It takes a long time transfer data from the storage nodes compute nodes. However, computational resources of GPUs idle during transmission period, which results in waste computing resources. In systems, large number short-term jobs queuing longer than their own execution times. Meanwhile, many multi-GPU suffering long-queuing due not enough free GPU. To best our...

On 22 December 2018, volcano Anak Krakatau, located in Indonesia, erupted and experienced a major lateral collapse. The triggered tsunami killed at least 437 people by the 13-m-high tide. Traditional optical imagery plays great role monitoring volcanic activities, but it is susceptible to cloud fog interference has low temporal resolution. Synthetic aperture radar (SAR) can monitor activities high resolution, immune influence of clouds. In this paper, we propose an automatic method...

Deep neural networks (DNNs) have achieved state-of-the-art performance in a number of domains but suffer intensive complexity. Network quantization can effectively reduce computation and memory costs without changing network structure, facilitating the deployment DNNs on mobile devices. While existing methods obtain good performance, low-bit time-consuming training or access to full dataset is still challenging problem. In this paper, we develop novel method named Compressorbased non-uniform...

To study the radar characteristics of tiltrotor aircraft when considering rotor rotation and tilting actions, a dynamic calculation method (DCM) based on physical optics theory diffraction is presented. The results show that cross section single periodic it rotates, while increasing speed can shorten this period. At fixed tilt angle, overall cabin plus still exhibits various at different azimuths rotor. Increasing angle better improve electromagnetic scattering level rotor, but easily makes...

The ground-based Synthetic Aperture Radar (GB-SAR) technique can be applied to the safety monitoring and early warning of geo-hazards, especially for displacement various types landslide masses. One key techniques processing GB-SAR data is phase unwrapping, which dramatically affected by atmospheric humidity, pressure, sampling interval, etc. In high mountains valleys where environmental change drastic, vulnerable incoherence both spatially temporally. Therefore, an improved unwrapping...

The radar signals of UAVs in urban low altitude airspace has a signal-to-noise ratio (SNR) and strong multipath interference, which is not conducive to the detection task. On other hand, effective exploitation energy signal can also improve SNR. Based on traditional focusing method, this paper proposes an method based multi- parameter estimation. By using scene information specular reflection model, quickly search focus target range-Doppler-angle data, by non-coherent integration....

Segmented Gamma Scanner (SGS) is a commonly used nondestructive testing (Non-Destructive Assay NDA) method. SGS uses radial rotation, axial segmentation, segmented scanning of the non-uniform sample uniform treatment, making it possible to accurately measure radioactivity on each segment and currently widely in field nuclear material management. This paper introduces self-designed measuring device this perform large number experiments different measurement objects laboratory. It also studied...

With the development of deep learning, computer vision has made great progress. Computer training based on learning requires a large number data sets. However, manual obtaining relevant sets is costly, and some special samples are not easy to obtain. Therefore, in order solve lack sets, virtual platform designed. The able render three-dimensional scene simulation by using Unity 3D, automatically generates visual through corresponding script file. Through interactive interface platform, users...