Recently, the leading performance of human pose estimation is dominated by top-down methods. Being a fundamental component in training and inference, data processing has not been systematically considered community, to best our knowledge. In this paper, we focus on problem find that devil estimator biased processing. Specifically, investigating standard state-of-the-art approaches mainly including transformation encoding-decoding, results obtained common flipping strategy are unaligned with...

Abstract Rail surface defects (RSDs) are a major problem that reduces operation safety. Unfortunately, the existing RSD detection systems have very limited accuracy. Current image processing methods not tailored for railway track and many fully convolutional networks (FCN)‐based suffer from blurry rail edges (RE). This paper proposes new boundary guidance network (RBGNet) salient RS detection. First, novel architecture is proposed to utilize complementarity between RE accurately identify...

Abstract In the United States, to ensure railroad safety and keep its efficient operation, regular track inspections on component defects are required by Federal Railroad Administration (FRA). Various types of inspection equipment applied, such as ground penetrating radar, laser, LiDAR, but they usually very expensive require extensive training rich experience operate. To date, still rely heavily manual which low‐efficiency, subjective, not accurate desired, especially for missing broken...

Leveraging line features to improve localization accuracy of point-based visual-inertial SLAM (VINS) is gaining interest as they provide additional constraints on scene structure. However, real-time performance when incorporating in VINS has not been addressed. This paper presents PL-VINS, a optimization-based monocular method with point and features, developed based the state-of-the-art VINS-Mono \cite{vins}. We observe that current works use LSD \cite{lsd} algorithm extract features;...

In this paper, we address the problem of monocular depth estimation when only a limited number training image-depth pairs are available. To achieve high regression accuracy, state-of-the-art methods rely on CNNs trained with large pairs, which prohibitively costly or even infeasible to acquire. Aiming break curse such expensive data collections, propose semi-supervised adversarial learning framework that utilizes small in conjunction easily-available images performance. particular, use one...

Rail surface defects have negative impacts on riding comfort and track safety, could even lead to accidents. Based the safety database (2020) of Federal Railroad Administration (FRA), rail been among main factors causing derailments. During past decades, there many efforts detect such defects. However, applications earlier methods are limited by high requirements specialized equipment personnel training. To date, defect inspection is still a very labor-intensive time-consuming process, which...

In the realm of maritime target detection, infrared imaging technology has become predominant modality. Detecting small ships on sea surface is crucial for national defense and security. However, challenge detecting targets persists, especially in complex scenes surface. As a response to this challenge, we propose MAPC-Net, an enhanced algorithm based existing network. Unlike conventional approaches, our method focuses addressing intricacies sparse pixel occupancy ships. MAPC-Net...

The automatic monitoring and detection of maritime targets hold paramount significance in safeguarding national sovereignty, ensuring rights, advancing development. Among the principal means surveillance, infrared (IR) small ship technology stands out. However, due to their minimal pixel occupancy lack discernible color texture information, IR ships have persistently posed a formidable challenge realm target detection. Additionally, intricate backgrounds often exacerbate issue by inducing...

In this paper, we report a real-time gesture driven interactive system with multimodal feedback for performing arts, especially dance. The consists of two major parts., recognition engine and engine. provides the performer's based on 3D marker coordinates from marker-based motion capture system. According to results, produces associated visual audio performer. This is simple implement robust errors in data. Satisfactory dance performances have been successfully created presented using reported

The real-time detection and recognition of pitaya fruit is an important prerequisite for automatic picking. We combined with the current deep learning method good accuracy to realize identification fruit. Firstly, we collected a large number pictures labeling, completed production data sets Pitaya Then use YOLOV3, YOLOV3-tiny MobileNet-YOLO network models train. After training, test performance trained model on set. experimental results show that improved has better speed than YOLOV3 model,...

In the United States, highway-railroad grade crossings are easily congested, which not only causes significant traffic delays to travelers but also brings potential threats first responders for emergencies. Unfortunately, very limited research efforts have been dedicated developing practical systems that can assess conditions at overcrowded crossings. The main challenge in evaluating congestion is different instance classes (i.e., vehicle, train, and pedestrian) need be accurately detected,...

Government investment is one of the main sources financing for large-scale infrastructure development. These government-invested construction projects are often characterized by large investments, long periods, and relatively high risks. In this study, process was divided into three stages, namely pre-decision, construction, post-evaluation. Research methods literature review expert interviews were adopted to determine factors that influenced risk projects. A corresponding evaluation index...

Parameter estimation for fractional-order chaotic systems is an important issue in control and synchronization could be essentially formulated as a multidimensional optimization problem. A novel algorithm called quantum parallel particle swarm (QPPSO) proposed to solve the parameter systems. The characteristic of computing used QPPSO. This increases calculation each generation exponentially. behavior particles space restrained by evolution equation, which consists current rotation angle,...

This paper presents a robust computational framework for monocular 3D tracking of human movement. The main innovation the proposed is to explore underlying data structures body silhouette and pose spaces by constructing low-dimensional silhouettes poses manifolds, establishing intermanifold mappings, performing in such manifolds using particle filter. In addition, novel vectorized descriptor introduced achieve low-dimensional, noise-resilient representation. articulated motion tracker...

In this paper, a robust 3D dance posture recognition system using two cameras is proposed. A pair of wide-baseline video with approximately orthogonal looking directions used to reduce pose ambiguities. Silhouettes extracted from these views are represented Gaussian mixture models (GMM) and as features for recognition. Relevance vector machine (RVM) deployed The proposed trained synthesized silhouettes created animation software motion capture data. experimental results on synthetic real...

Control and synchronization of fractional-order chaotic systems have attracted wide attention due to their numerous potential applications. To get suitable control method parameters for systems, the stability analysis time-varying should be discussed in first place. Therefore, this paper analyzes presents a theorem system with order 0<α<1. This is sufficient condition which can discriminate conveniently. Feedback controllers are designed Lü system. The simulation results demonstrate...

Texture is a key component for human visual perception and plays an important role in image-related applications. This paper combines perceptual texture features Gabor wavelet image classification. Three new which are proved to be accordance with introduced. These include directionality, contrast coarseness domain. We test our proposed method using the Brodatz database, experimental results show scheme has produced promising results.