A STUDY ON THE ENSEMBLE FORECAST REAL-TIME CORRECTION METHOD

Using real-time correction technology for typhoons, this paper discusses real-time correction for forecasting the track of four typhoons during 2009 and 2010 in Japan, Beijing, Guangzhou, and Shanghai. It was determined that the short-time forecast effect was better than the original objective mode. By selecting four types of integration schemes after multiple mode path integration for those four objective modes, the forecast effect of the multi-mode path integration is better, on average, than any single model. Moreover, multi-mode ensemble forecasting has obvious advantages during the initial 36 h.

Restoring force correction based on online discrete tangent stiffness estimation method for real-time hybrid simulation

In real-time hybrid simulation（RTHS）, it is difficult if not impossible to completely erase the error in restoring force due to actuator response delay using existing displacement-based compensation methods. This paper proposes a new force correction method based on online discrete tangent stiffness estimation（online DTSE） to provide accurate online estimation of the instantaneous stiffness of the physical substructure. Following the discrete curve parameter recognition theory, the online DTSE method estimates the instantaneous stiffness mainly through adaptively building a fuzzy segment with the latest measurements, constructing several strict bounding lines of the segment and calculating the slope of the strict bounding lines, which significantly improves the calculation efficiency and accuracy for the instantaneous stiffness estimation. The results of both computational simulation and real-time hybrid simulation show that：（1） the online DTSE method has high calculation efficiency, of which the relatively short computation time will not interrupt RTHS; and（2） the online DTSE method provides better estimation for the instantaneous stiffness, compared with other existing estimation methods. Due to the quick and accurate estimation of instantaneous stiffness, the online DTSE method therefore provides a promising technique to correct restoring forces in RTHS.

Sensor placement of long-term health monitoring for large bridges based on the real-time correction of finite element model

The process of optimized placement of long-term health monitoring sensors for large bridges generally begins with finite element models, but there will arise great discrepancies between theoretically-calculated results and actual measurements.Therefore, rectified finite element models need to be rectified by virtue of model rectifying technology. Firstly, the result of construction monitoring and finished state load test is used to real-time modification of finite element model. Subsequently, an accurate finite element model is established. Secondly, the optimizing the layout of sensor with following orthogonality guarantees orthogonal property and linear independence for the measured data. Lastly, the effectiveness and feasibility of method in the paper is tested by real-time modifying finite element model and optimizing the layout of sensor for Nujiang Bridge.

Analyzing the Impact of Scene Transitions on Indoor Camera Localization through Scene Change Detection in Real-Time

Real-time indoor camera localization is a significant problem in indoor robot navigation and surveillance systems.The scene can change during the image sequence and plays a vital role in the localization performance of robotic applications in terms of accuracy and speed.This research proposed a real-time indoor camera localization system based on a recurrent neural network that detects scene change during the image sequence.An annotated image dataset trains the proposed system and predicts the camera pose in real-time.The system mainly improved the localization performance of indoor cameras by more accurately predicting the camera pose.It also recognizes the scene changes during the sequence and evaluates the effects of these changes.This system achieved high accuracy and real-time performance.The scene change detection process was performed using visual rhythm and the proposed recurrent deep architecture,which performed camera pose prediction and scene change impact evaluation.Overall,this study proposed a novel real-time localization system for indoor cameras that detects scene changes and shows how they affect localization performance.

Real-time measurement of high rotational projectile axial acceleration based on 2-axis acceleration sensor

The real-time measurement principle of high rotational projectile＇s angular velocity based on 2-axis acceleration sensor and the axial acceleration measurement error caused by the installation error are discussed.The 2-axis acceleration sensor is applied to measure the high rotational projectile＇s angular velocity and the measurement value of axial acceleration,the axial acceleration of the high rotational projectile equals the measurement value of axial acceleration subtracting the centrifugal acceleration component,so that the high-accuracy real-time measurement of axial acceleration is realized.The memory test has confirmed the strike tally of the theoretical analysis and the test result.The measurement technique can satisfy the high-accuracy measurement of the high rotational projectile axial acceleration in the self-determination course correction fuze projectile.

Performance analysis of real-time and post-mission kinematic precise point positioning in marine environments

This article focuses on the performance analysis of both real-time and post-mission kinematic precise point positioning(PPP)in challenging marine environments.For this purpose,a real dynamic experiment lasting 6 h was carried out on a lake dam in?orum City of Turkey.While the kinematic test was continuing,the real-time PPP coordinates were obtained for each measurement epoch with a commercial real-time PPP(RT-PPP)service,namely the Trimble Center Point RTX.Then the post-mission PPP(PM-PPP)coordinates were calculated by using Multi-GNSS data and the Multi-GNSS Experiment(MGEX)precise products.The kinematic RT-PPP and PM-PPP results showed that the PPP coordinates were consistent with the relative solution at centimetre and decimetre level in horizontal and height components,respectively.This study implies that PPP technique is a powerful tool for highly accurate positioning in both real-time and post-mission modes,even for dynamic applications in harsh environments.

Improved Multiple Feature-electrochemical Thermal Coupling Modeling of Lithium-ion Batteries at Low-temperature with Real-time Coefficient Correction

Monitoring various internal parameters plays a core role in ensuring the safety of lithium-ion batteries in power supply applications.It also influences the sustainability effect and online state of charge prediction.An improved multiple feature-electrochemical thermal coupling modeling method is proposed considering low-temperature performance degradation for the complete characteristic expression of multi-dimensional information.This is to obtain the parameter influence mechanism with a multi-variable coupling relationship.An optimized decoupled deviation strategy is constructed for accurate state of charge prediction with real-time correction of time-varying current and temperature effects.The innovative decoupling method is combined with the functional relationships of state of charge and open-circuit voltage to capture energy management ef-fectively.Then,an adaptive equivalent-prediction model is constructed using the state-space equation and iterative feedback correction,making the proposed model adaptive to fractional calculation.The maximum state of charge estimation errors of the proposed method are 4.57% and 0.223% under the Beijing bus dynamic stress test and dynamic stress test conditions,respectively.The improved multiple feature-electrochemical thermal coupling modeling realizes the effective correction of the current and temperature variations with noise influencing coefficient,and provides an efficient state of charge prediction method adaptive to complex conditions.

改进YOLOv8n-Pose的形变QR码校正与识别

针对QR码图像检测过程中因环境复杂性和拍摄角度变化等因素导致QR码读取率低的问题,本文提出一种基于改进YOLOv8n-Pose的形变QR码校正与识别算法.首先,在主干网络引入高效通道注意力机制(efficient channel attention,ECA)模块,该模块通过不降维的方式实现跨通道交互,有效提升网络的特征提取能力和检测精度.其次,采用Slim-neck架构重构颈部网络,减少模型的复杂度,提高对不同尺度QR码的检测能力.最后,通过模型检测到的QR码角点,结合逆透视变换对QR码进行校正,并使用ZBar算法进行读取.实验结果表明,在公开的QR码数据集上,改进的算法相比原算法,mAP50和mAP50-95分别提升1.6%和1.1%,模型参数量和模型计算量分别降低6.5%和9.5%,在CPU和GPU上检测速度分别提升0.3 f/s和0.7 f/s,达到14.2 f/s和59.6 f/s,能够高效地满足QR码角点检测需求.此外,在自制的形变QR码数据集上,基于改进YOLOv8n-Pose的QR码识别方法相比单独使用ZBar算法的QR码识别方法,QR码读取率提高23.66%,达到87.41%.该方法仅需拍摄一张照片就可识别所有货物的信息,能够有效提高货物管理的效率.

一种带高光处理的无缝纹理映射方法

纹理映射作为三维重建的重要步骤,直接关系到生成模型的视觉效果。传统的纹理映射方法通常使用简单的混合方法来获取表面纹理。通过将模型表面投影到每个纹理图像上来获取相应的纹理区域,然后将它们混合在一起以获取表面纹理。然而,由于相机位姿的不准确性,纹理映射结果存在明显的模糊和鬼影等问题。此外,在高光环境下获取的纹理图像很容易包含高光区域,最终导致纹理颜色的丢失,降低纹理的真实性。为了解决这些问题,提出了一种无缝纹理映射方法,可以有效消除高光反射。通过衡量纹理图像的质量,为每个模型表面选择最佳的纹理图像,并利用色度一致性约束优化的相机位姿以消除明显的纹理错位。针对高光问题,提出了一个高光处理模块,使用多视图图像信息并基于双色反射模型定位和处理高光纹理。最后,对纹理图进行纹理颜色一致性调整,以处理纹理之间的颜色差异。实验结果表明,与现有方法相比,提出的算法可以有效消除高光反射的影响,从而获得更好的纹理映射结果。

人体上肢运动姿态多点视觉自动跟踪仿真

由于视觉传感器本身的噪声以及运动模糊等原因,使得采集的运动上肢姿态数据存在噪声和误差,导致多点视觉自动跟踪精度较低。为此,提出人体运动上肢姿态多点视觉自动跟踪方法仿真。布设多维传感器对姿态多点采集,建立传感器加速度误差模型,获取误差修正目标函数,利用蚁群算法(Ant Colony Optimization, ACO)寻优,修正传感器采集过程中的误差;采用卡尔曼滤波算法(Kalman Filtering Algorithm, Kalman),实现人体运动上肢姿态的多点视觉自动跟踪。实验结果表明,所提方法能够在保障跟踪稳定性的用时,提高姿态跟踪的精度和效率,跟踪耗时仅为10ms左右。

基于神经辐射场的视觉位姿分段估计方法

为提升观测载体的视觉位姿估计精度和效率,提出了一种基于神经辐射场的“粗-精”两阶段位姿估计方法。首先基于通道注意力机制构建初始位姿估计网络Rnet,用以估计大失准角情况下的初始位姿;然后设计了基于可信区域约束和兴趣点采样策略的深度纠正模块,通过优化NeRF模型深度,为位姿精确估计提供更高精度先验信息;最后构建PoseNeRF网络,基于SE(3)流形的梯度下降法最小化NeRF像素颜色、深度推理值与真值之间的残差,实现位姿精确估计。在OVIT实测数据集实验中,采用深度监督后,位置精度提升27.8%,姿态精度提升12.6%;与TD-NeRF相比,所提方法位姿收敛效率提升约6倍;相较于iNeRF、NeRF--,所提方法受初始姿态误差影响最小,验证了所提方法具有更高的位姿估计精度、效率和鲁棒性。

基于特征点位置约束的三维激光移动测量位姿改正方法

准确的导航定位信息是激光移动测量系统获取高精度点云的基础,但导航定位系统在短时段失效情况下会导致观测点位置信息不准确,从而无法获得准确的点云数据。为此,提出一种基于特征点位置约束的激光移动测量位姿改正方法,利用导航定位系统失效时段前后测量得到的少量特征点位置信息,构建基于不同时刻特征点配准的位姿改正模型,将导航定位系统失效时刻的观测点位置和姿态作为待估参数进行最小二乘解算,实现位姿参数恢复和点云重构。结果表明,采用基于特征点位置约束的位姿改正方法,导航定位系统短时段失效情况下的点云坐标中误差从米级改正到了分米级,对于提升组合导航系统稳定性和点云移动测量精度有重要的参考价值。

耳温计检定系统机械臂位姿自动修正方法研究

针对耳温计自动检定系统中机械定位随机偏移的问题,提出一种基于机器视觉的机械臂位姿自动修正算法。在恒温槽基准位建立机械臂运动的基准参考坐标系,在此坐标系下设计机械臂作业路径动作;系统作业过程中,通过定位恒温槽上的特征图案生成变换矩阵,产生新参考坐标系;机械臂来到新参考坐标系下拍照,反推变换矩阵偏差值并对其迭代更新;通过对机械臂运动的参考坐标系进行修正,完成对机械臂三维作业的位姿修正。实验结果表明,所提算法对恒温槽上3个平面的定位精度达到0.5 mm以内,为相似场景下的视觉引导提供了一种新思路。

Real-Time Facial Pose Estimation and Tracking by Coarse-to-Fine Iterative Optimization

We present a novel and efficient method for real-time multiple facial poses estimation and tracking in a single frame or video.First,we combine two standard convolutional neural network models for face detection and mean shape learning to generate initial estimations of alignment and pose.Then,we design a bi-objective optimization strategy to iteratively refine the obtained estimations.This strategy achieves faster speed and more accurate outputs.Finally,we further apply algebraic filtering processing,including Gaussian filter for background removal and extended Kalman filter for target prediction,to maintain real-time tracking superiority.Only general RGB photos or videos are required,which are captured by a commodity monocular camera without any priori or label.We demonstrate the advantages of our approach by comparing it with the most recent work in terms of performance and accuracy.

RRVPE:A Robust and Real-Time Visual-Inertial-GNSS Pose Estimator for Aerial Robot Navigation

Self-localization and orientation estimation are the essential capabilities for mobile robot navigation.In this article,a robust and real-time visual-inertial-GNSS(Global Navigation Satellite System)tightly coupled pose estimation(RRVPE)method for aerial robot navigation is presented.The aerial robot carries a front-facing stereo camera for self-localization and an RGB-D camera to generate 3D voxel map.Ulteriorly,a GNSS receiver is used to continuously provide pseudorange,Doppler frequency shift and universal time coordinated(UTC)pulse signals to the pose estimator.The proposed system leverages the Kanade Lucas algorithm to track Shi-Tomasi features in each video frame,and the local factor graph solution process is bounded in a circumscribed container,which can immensely abandon the computational complexity in nonlinear optimization procedure.The proposed robot pose estimator can achieve camera-rate(30 Hz)performance on the aerial robot companion computer.We thoroughly experimented the RRVPE system in both simulated and practical circumstances,and the results demonstrate dramatic advantages over the state-of-the-art robot pose estimators.

Feature-based RGB-D camera pose optimization for real-time 3D reconstruction

In this paper we present a novel featurebased RGB-D camera pose optimization algorithm for real-time 3D reconstruction systems. During camera pose estimation, current methods in online systems suffer from fast-scanned RGB-D data, or generate inaccurate relative transformations between consecutive frames. Our approach improves current methods by utilizing matched features across all frames and is robust for RGB-D data with large shifts in consecutive frames. We directly estimate camera pose for each frame by efficiently solving a quadratic minimization problem to maximize the consistency of3 D points in global space across frames corresponding to matched feature points. We have implemented our method within two state-of-the-art online 3D reconstruction platforms. Experimental results testify that our method is efficient and reliable in estimating camera poses for RGB-D data with large shifts.

任意位姿圆柱曲面透视投影失真的图像校正

机器视觉是检测圆柱物体表面缺陷的重要手段,正对圆柱物体拍摄图像的透视投影失真易于校正但拍摄条件十分苛刻,倾斜状态是圆柱物体视觉检测的常态。针对任意位姿圆柱物体表面透视投影导致的失真变形问题,提出一种基于圆柱曲面位姿估计的图像校正方法。该方法在提取圆柱图像两侧边线的基础上,利用圆柱半径和系统成像参数估计圆柱曲面与相机之间的位姿关系,通过建立的圆柱曲面位姿变步长优化迭代算法,实现圆柱曲面位姿精确计算。在此基础上,构建圆柱曲面细分网格点与原始图像像素坐标点的对应关系,然后将曲面展开为平面,建立展开的细分网格点与校正图像像素坐标点的对应关系,从而创建校正图像像素坐标点与原始图像像素坐标点的映射关系,对原始图像进行重采样得到校正图像。实验结果表明,圆柱曲面位姿估计的平均距离误差为0.3 mm、平均角度误差为0.60°;表面贴有棋盘格圆柱物体的图像相邻角点的平均距离标准差从校正前的12.2 pixels降至校正后的0.8 pixels;校正后的实际场景图像能有效识别出圆柱曲面上的文字,圆柱曲面上缺陷的测量误差不超过0.1 mm。校正后的圆柱物体图像消除了圆柱曲面倾斜投影失真和“近大远小”透视变形,验证了提出方法的有效性。

基于视觉位姿矫正的机载激光实景三维重构

城市测绘中无人机搭载激光雷达悬停采集数据时受机身振动影响,容易产生运动畸变,导致融合建模效果不佳。针对该问题,本文提出了一种基于视觉位姿校正的机载激光实景三维重构方法。首先利用激光雷达扫描过程中双目相机获取的视频信息,处理得到激光雷达位姿变化的轨迹以矫正激光点云的位姿;然后将矫正后的激光点云投影到单目相机坐标系下,通过共线方程获得色彩信息,实现信息融合生成三维真彩点云;最后针对视觉位姿矫正前后的真彩点云,建立直线度、平面度、垂直度、融合赋色4个特征维度的评价标准。试验结果表明,视觉位姿矫正后的真彩点云相较于矫正前,直线度最高提升77.3%,平面度最高提升54.5%,垂直度最高矫正54.53°,色彩附着明显更准确。

A Real-Time Multi-Stage Architecture for Pose Estimation of Zebrafish Head with Convolutional Neural Networks

In order to conduct optical neurophysiology experiments on a freely swimming zebrafish,it is essential to quantify the zebrafish head to determine exact lighting positions.To efficiently quantify a zebrafish head's behaviors with limited resources,we propose a real-time multi-stage architecture based on convolutional neural networks for pose estimation of the zebrafish head on CPUs.Each stage is implemented with a small neural network.Specifically,a light-weight object detector named Micro-YOLO is used to detect a coarse region of the zebrafish head in the first stage.In the second stage,a tiny bounding box refinement network is devised to produce a high-quality bounding box around the zebrafish head.Finally,a small pose estimation network named tiny-hourglass is designed to detect keypoints in the zebrafish head.The experimental results show that using Micro-YOLO combined with RegressNet to predict the zebrafish head region is not only more accurate but also much faster than Faster R-CNN which is the representative of two-stage detectors.Compared with DeepLabCut,a state-of-the-art method to estimate poses for user-defined body parts,our multi-stage architecture can achieve a higher accuracy,and runs 19x faster than it on CPUs.

基于仿射变换的多姿态人脸矫正和识别

为了解决随着人脸姿态的变化,人脸识别率迅速下降的问题,提出了利用仿射变换和成像原理相结合对待识别的多姿态人脸图像进行姿态调整,将其调整为近似于正面人脸的方法,该方法能够有效地将45°范围内的多姿态人脸图像调整为正面人脸图像。同时,使用改进的SURF(加速鲁棒特征算法)算法对校正后的人脸图像进行识别。在FERET等人脸库及拍摄的人脸图像上进行实验,实验结果表明该方法能够在一定程度上克服姿态变化带来的影响,使平均识别率最高可提高7.0%左右。