Non-maximum suppression (NMS) is widely adopted as a post-processing step in the state-of-the-art object detection pipelines to merge nearby detections around one object. However, its performance affected by objects that are highly overlapped with each other, and localization accuracy depends solely on highest scored detection. To tackle this, an accurate NMS method proposed this letter, which gradually merges iterative way. In iteration, grouped harder threshold regress for new proposal,...

Benefits from the powerful local modeling capability of deep convolutional networks (CNNs), remote sensing image change detection (CD) has made significant progress. In recent years, rise Transformers further driven improvements in global feature extraction for bitemporal images. Some prior efforts have tried to integrate CNN and Transformer, but they suffer limitation inefficiently aggregating features contextual information. Besides, struggle refine boundaries exhibit inferior performance...

Deformable-part-based model (DPM) has shown great success in object detection recent years. However, its performance will degrade on partially occluded objects and is even worse largely real remote sensing applications. To address this problem, a novel partial configuration (PCM) developed paper. Compared to conventional single-layer DPMs, an extra layer, which composed of configurations defined according possible occlusion patterns, introduced PCM block the transmission impact. During...

We present a novel end-to-end partially supervised deep learning approach for video anomaly detection and localization using only normal samples. The insight that motivates this study is the samples can be associated with at least one Gaussian component of Mixture Model (GMM), while anomalies either do not belong to any component. method based on Variational Autoencoder, which learn feature representations as trained learning. A Fully Convolutional Network (FCN) does contain fully-connected...

A flexible new technique is proposed to calibrate the geometric model of line scan cameras. In this technique, camera rigidly coupled a calibrated frame establish pair stereo The linear displacements and rotation angles between two cameras are fixed but unknown. This only requires observe specially designed planar pattern shown at few (at least two) different orientations. At each orientation, obtained including array image image. Radial distortion modeled. calibration scheme includes...

Hyperspectral anomaly detection plays an important role in the field of remote sensing. It provides a way to distinguish interested targets from background without any prior knowledge. The majority pixels hyperspectral dataset belong background, and they can be well represented by several endmembers, so has low-rank property. Anomalous usually account for tiny part dataset, are considered have sparse Recently, matrix decomposition (LRaSMD) technique drawn great attention as method solving...

Numerous hyperspectral anomaly detection (AD) methods suffer from complex background compositions and subpixel objects due to their inadequate Gaussian-distributed representations for nonhomogeneous backgrounds or low discrimination between anomalies the background. To alleviate these issues, a novel AD weighting strategy based on tensor decomposition cluster is proposed in this letter. Equipped with simple but effective as postprocess, performances of generic can be significantly boosted....

With the complexity and refinement of industrial systems, fast fault diagnosis is crucial to ensuring stable operation equipment. The main limitation current methods lack real-time performance in resource-constrained embedded systems. Rapid online detection can help deal with equipment failures time prevent damage. Inspired by ideas compressed sensing (CS) deep extreme learning machines (DELM), a data-driven general method proposed for diagnosis. contains two modules: data sampling module...

Partial configuration model (PCM) is an occluded object detection method in high-resolution remote sensing images (HR-RSIs) based on the deformable part-based (DPM). However, it needs extra category predefinition, considerable partlevel annotation, and repeated multimodel training. In this letter, automatic fast PCM generation proposed a novel part sharing mechanism. We propose to share parts from one trained DPM (tDPM) among different models of partial configurations (PCs) address above...

Partially occluded object detection (POOD) has been an important task for both civil and military applications that use high-resolution remote sensing images (HR-RSIs). This topic is very challenging due to the limited evidence detection. Recent partial configuration model (PCM) based methods deal with occlusion yet suffer from problems of massive manual annotation, separate parameter learning, low training efficiency. To tackle this, a unified PCM framework (UniPCM) proposed in this paper....

The frequent illegal use of drones poses a serious threat to public security and property. Counter-drones are crucial tools. prerequisite for an effective counter-drone is detect accurately. With the rapid advancements in computer vision, vision-based drone detection methods have emerged as hot topic research. However, current reviews less focused on algorithmic summarization analysis. For this reason, survey aims comprehensively review latest complex environments, with goal providing more...

Automatic matching of multimodal remote sensing images remains a vital yet challenging task, particularly for and computer vision applications. Most traditional methods mainly focus on key point detection description the original image, thus ignoring deep semantic feature information such as road features, with result that method can not effectively resist nonlinear grayscale distortion, has low efficiency poor accuracy. Motivated by this, this paper proposes novel automatic named LURF via...

The pose of a rotational symmetry target in wideband radar is the angle between its axis and line sight. It required parameter for absolute attitude measurement using bistatic measuring system. In this letter, 3-D electromagnetic model (3-D EM model)-based method proposed estimating from fully polarimetric measurements. established offline based on geometric structure. Scattering features at different poses can be accurately predicted by model. algorithm, synthetic high-resolution range...

Rapid fault diagnosis of electromagnetic launch and recovery (EMLR) Systems is great significance. It an important research direction to diagnose the system by mining event data collected during operation a certain type arresting gear controller. While have characteristics high sampling rate, multidimensional coupling, nonperiodic transient, which belongs complex large-scale time series, existing steady-state series methods are not applicable. In order solve this problem, article proposes...

Timely and accurate detection of the initiation expansion crack is great significance for improving safe operation civil infrastructures. Image-based visual surface inspection has been an indispensable way long-time infrastructure monitoring. However, existing methods generally suffer from interference complex background, leading to obvious performance drops. To tackle this, improved encoder-decoder architecture based on SegNet proposed in this paper, namely crack-SegNet. The encoder network...

Convolutional neural network (CNN)-based object detection for optical remote sensing images has achieved higher accuracy compared with traditional methods handcrafted features. However, the deep and large CNNs make it hard to be deployed in real-time scenarios limited computation, storage, power bandwidth resources, example, data processing onboard airborne, satellites unmanned aerial vehicles search rescue. Therefore, this paper we present a high-performance approach images. Based on widely...

Anomaly detection (AD) is an important research topic in the hyperspectral remote sensing field. However, owing to complex background distributions and interference of clutter noise practical situations, AD problem far from being addressed satisfactorily. In this paper, a novel method by joining statistical model representation theory for predominant proposed. It mainly consists two parts. A Mahalanobis distance-based anomaly characterization criterion first designed acquire initial result....