为解决荧光油膜灰度与厚度现标定方法(微小高度装置法)存在采集工况复杂、标定周期较长等缺陷,提出了一种新的近效测量方法,该方法只需极少量标定数据,便可达到与现标定方法相近的效果.近效方法利用了 Elman动态神经网络对小样本数据进行量程扩展并引入一维插值算法使扩展后的数据项平滑,对解算出的三维荧光油膜厚度数据采用二维插值算法进行二次平滑以求得完整的厚度分布图.经模拟试验结果显示,通过该近效方法进行厚度测量最终可以清晰、准确、定量地显示荧光油膜厚度分布,与目前广泛使用的微小高度装置法相比效果相近,在荧光油膜汇集处(较厚区域)误差最大不超过±2.5 pm,在平滑适中及较薄区域误差不超过±2 μm,达到荧光油膜厚度工程测量标准,为飞行器全局摩阻测量提供了一种新标定思路,具有一定的实际工程应用意义.

针对基于先验的传统光流法存在前提条件苛刻的问题,提出使用基于深度学习的光流法进行荧光油膜全局速度测量.采用数值仿真试验对基于先验的改进HS光流法和基于深度学习的FlowNet2光流法进行对比,结果显示:在不外加干扰时,改进HS光流法和FlowNet2光流法的平均端点误差分别为0.458 7像素/s和0.381 7像素/s;在亮度变化、噪声干扰或不同的演化时间下,FlowNet2光流法的平均端点误差均明显低于改进HS光流法,平均端点误差差值最大可达5.19像素/s;风洞试验进一步证明,FlowNet2光流法能够获得正确、清晰、定量的荧光油膜全局速度场,较改进HS光流法鲁棒性更高,对风洞工程应用具有一定的参考价值.

传统的液体粘度测量方法如旋转法、落球法和毛细管法等需将传感器或装置浸入液体内部实现测量,然而在某些场景下这些方法无法适用,因此发展了一种基于超声横波反射的液体粘度测量方法,建立超声横波在固/液界面的反射模型,利用入射波、反射波的幅值与相位求解声波反射系数,再根据反射系数与液体声阻抗关系得到液体的粘度系数.引入固液声阻抗比值,构建了反射系数幅值与声阻抗比值的关系模型,定量分析了声阻抗比值与测量范围和灵敏度之间的关系,并给出固体匹配材料的选择依据.仿真实验结果表明,针对铝、花岗岩、石墨、PMMA和聚四氟乙烯五种声阻抗不同的固体材料,所提方法能有效筛选出合适的匹配材料,为实际测量提供参考.

Traditional optical flow algorithm based on prior is sensitive to large displacement, in order to solve the problems, one optical flow algorithm based on deep learning is introduced to measure the global velocity of fluorescent oil film. The test results of numerical simulation show that the Average Endpoint Error (AEE) of HS, LK and Flownet2 are 2.4604 pixels, 2.3898 pixels and 0.9916 pixels in terms of warping 0.1s respectively, and the error of FlowNet2 is always the lowest under the conditions of large displacement. Therefore, the algorithm is more robust to large displacement than the others that is currently being used, which has certain exploration significance and engineering reference value.

In global skin-friction measurement of aircraft, the fluorescent oil film method can characterize the distribution of skin friction well. However, in an actual wind tunnel test, the wing of the aircraft will inevitably produce corresponding vibrations due to the influence of wind, which will change the relative position between fluorescent oil film and UV (ultraviolet) excitation light source (position fixed). This also directly affects gray value imaging of fluorescent oil films. Based on this, a mathematical model is established to judge the stability of the gray value of fluorescent oil film in this vibrational environment; then, the model can be solved to obtain the vibrational range constraint that enables the gray value of fluorescent oil film to be stabilized. In order to simplify the calculation process, the light vector angle is used to describe the constraint, which also makes the results more intuitive. Through experimental analysis and demonstration, the prediction accuracy of this model can reach 95.61%, which has certain practical engineering application significance.

为了快速计算分析利用视频测量方法测得的高速风洞试验密度场在扰动流场作用下的实验数据,针对密度场的数值求解问题,经过光线偏折理论分析密度场得到的二阶偏微分方程,对其研究实现了CPU串行有限元法求解.在此基础上提出了基于GPU的快速有限元求解密度场的方法,该方法经过对串行有限元法求解过程效率分析后,将耗时的神经网络拟合、总刚度矩阵和总载荷向量的求解进行了基于GPU的并行加速.实验结果表明:在精度满足实际工程要求的前提下,相对于CPU串行求解方法,所提方法可大大提高求解效率,且随着网格剖分成倍加密,其加速比成倍增加.

针对荧光油膜灰度与厚度关系数据采集方法较为繁琐、耗时耗力这一问题,研究了基于Hankel阵的系统辨识算法,并在此基础提出了Hankel阵误差修正模型和Hankel阵高阶迭代误差修正模型等两种改进方法,利用了极少数据量建立模型,实现了对其余未知荧光油膜厚度值的预测,且保持了较高的精度.试验结果表明:基于极少数据量建模以预测该数据量外的数据点这一特殊背景下,插值法的外插能力显得并不适用.而传统Hankel阵预测模型的预测精度为76.69％,Hankel阵误差修正模型和Hankel阵高阶迭代误差修正模型的预测精度分别为85.69％和89.25％,较之传统方法预测精度分别提高了9％和12.56％,为荧光油流技术领域针对荧光油膜灰度与厚度建模问题提供了一种可行技术路线,具有一定的实际工程应用意义.

Aiming at improving the accuracy and stability of the pose estimation when the target points located in the planar, quasi-planar and quasi-linear case. In this study, we propose an iterative solution for singular configuration of target points. The main idea of the algorithm is to select two farthest points as the basic reference points, and divide n points into n-2 three-point sets. Then, the auxiliary points are constructed according to the geometric relationship of the three-point set, aiming to increase the geometric constraints of the perspective similar triangle algorithm, and obtain a more accurate initial value. Finally, the simplified EPnP algorithm is combined with Gaussian Newton algorithm for optimization. Experiments conducted on synthetic data and real images show that when the number of planar target points n=4, the average image re projection error of this algorithm is 0.003 mm, compared with the orthogonal iterative algorithm, EPnP algorithm and IEPnP algorithm, which is 0.062 mm, 0.324 mm and 2.238 mm respectively, this algorithm effectively improves the accuracy and stability of the pose estimation of the target point in singular configuration.

With the improvement of public security awareness, video anomaly detection has become an indispensable demand in surveillance videos. To improve the accuracy of video anomaly detection, this paper proposes a novel two-stream spatial-temporal architecture called Two-Stream Deep Spatial-Temporal Auto-Encoder (Two-Stream DSTAE), which is composed of a spatial stream DSTAE and a temporal stream DSTAE. Firstly, the spatial stream extracts appearance characteristics whereas the temporal stream extracts the motion patterns, respectively. Then, based on the novel policy joint reconstruction error, this model fuses the spatial stream and the temporal stream to extract spatial-temporal characteristics to detect anomalies. Furthermore, since the optical flow is invariant to appearances such as color or light, we introduce optical flow to enhance the capability of extracting continuity between adjacent frames and inter-frame motion information. We demonstrate the accuracy of the proposed method on the publicly available standard datasets: UCSD, Avenue and UMN datasets. Our experiments demonstrate high accuracy, which is superior to the state-of-the-art methods.

Recently, crowd counting has drawn widespread attention in computer vision, but it is extremely challenging because of the varying scales and densities. Many existing methods focus on improving the multi-scale representation by utilizing multi-column or multi-branch architectures with different kernel sizes. However, such networks cannot extract the feature maps with large receptive fields due to limitation of depth. In addition, the importance of utilizing the multi-level feature information in a deep network is ignored. In this paper, we propose a multi-scale and multi-level features aggregation network (MFANet) for accurate and efficient crowd counting, and it can be trained by end-to-end. A vital component of the network is the scale and level aggregation module (SLAM), which can extract multi-scale features and make full use of multi-level feature information for more accurate estimation. When six SLAMs are stacked together and applied to our network, our method can achieve the best performance. Furthermore, we introduce a new loss function called normalized Euclidean loss (NEL) to balance the contribution of all samples to network training. To demonstrate the performance of the proposed method, extensive experiments are conducted on four benchmark crowd counting datasets, including ShanghaiTec Part A/B, UCF-CC-50, Mall, and UCF-QNRF. Experimental results show that our MFANet achieves state-of-the-art performance in crowd counting and crowd localization.

针对高分辨率视频测量(VM)中标记点难以实时识别的问题,提出基于运动估计的图像标记点实时识别方法,通过标记点运动估计模型获得标记点在时序图像中的预估坐标,根据各预估坐标对图像并行分割获得对应搜索域,并采用增量更新的Freeman链码提取标记点轮廓信息,从而达到实时识别标记点的目的.运动估计实验表明:该模型能实现风洞环境中图像标记点运动估计,最大误差仅为5.597像素;标记点并行识别实验表明:利用现场可编程门阵列(FPGA)实时识别标记点,处理效率达到2GB/s,应用前景广阔.

某流场中,密度投影场与扰动引起的光线偏移角之间满足泊松方程的关系,且该泊松方程的源项是一系列离散的偏移角,无法利用现有方法直接求解.在有限元法的基础上,利用神经网络拟合源项中偏折角与采样坐标的关系,得到剖分网格点上的源项;同时,针对神经网络的加入使得有限元法中单元载荷向量的求解过于耗时的问题,在求解源项时,论文将三角单元整体预测的二重积分表达式近似替换为三角单元结点预测的常数表达式.仿真实验表明,引入神经网络拟合偏折角,相对于传统的插值方法,可以取得更高精度的结果;而且在相同误差下,提出的算法大大提升了运算速度.将该算法应用到真实流场中,得到的该密度投影场的特性与搭建的真实环境的结果相似,进一步说明所提算法的有效性.

In the wind tunnel flow field, the relationship between the density projection field, caused by disturbance, and the offset angle of light meets the Poisson equation. However, the source term of the Poisson equation is composed of a series of measured offset angles, which makes it can not be solved effectively by the existing methods. On the basis of the finite element method (FME), we established the fitting formula between the offset angles in the source terms and their corresponding coordinates by employing the Genetic algorithms and back-propagation (GA-BP) neural network. Meanwhile, when the element load vectors were solved, the double integral expression of the entire triangle element prediction was approximately replaced by the constant expression of the triangle vertex element prediction. Simulation experiments demonstrate that compared with the traditional interpolation, the neural network can achieve higher fitting precision. And the proposed algorithm greatly improves the operation speed under the same solution error. By applying the proposed algorithm to the real flow field, the obtained features of the density field are similar to those obtained in the real environment. These imply that the proposed method provides a new useful tool for the study of the density projection field.

风洞试验中模型迎角的精准测量是降低阻力系数误差的重要途径之一,为此,提出了基于单应性矩阵的模型迎角单目视频测量方法.该方法通过两个单应性矩阵,获取试验过程中相机实时位姿和标记点物方空间位置坐标,应用坐标旋转关系,完成试验模型的迎角测量.数值仿真试验结果表明:迎角测量误差与待测标记点到风洞壁板间的距离偏差近似为线性关系,因此,当标记点不满足共面条件时,可根据该特点进行测量误差修正.静态标定和风洞迎角测量试验结果表明:修正系统误差后,迎角实测数据的测量准度在0.01°以内,精度不超过0.012°.本文方法易于实施,工程实用价值强.

风洞试验模型在气流脉动作用下小幅振动,导致光流法从荧光油膜时序图像中解得的荧光油膜路径运动速度含有模型运动速度,降低了荧光油膜全局摩阻测量准度.为此,提出试验模型表面的荧光油膜路径运动速度测量方法,将模型表面的背景纹理(如人工网格线或其他典型特征)作为基准,利用图像相关法离散匹配,获得相邻时序图像中背景纹理的(几何位姿)映射矩阵;基于模型运动的连续性,推导了映射矩阵的全局优化方程,并结合光流法,实现了模型振动与其表面荧光油膜路径运动的解耦.Oseen涡对的荧光油膜路径运动速度场仿真试验结果表明:在给定的平移旋转条件下,本文方法的计算结果(沿Oseen涡核连线分布的测量速度)与理论值的最大相对误差为4.1％,较无平移旋转条件下的光流计算结果最大相对误差仅增加0.6％.2m量级高速风洞某空腔试验与机翼试验的荧光油膜路径运动速度测量结果进一步显示:本文方法测得的流动现象正确,能得到定量、清晰的表面摩擦应力线图谱与油膜路径运动速度场,较传统方法优势明显,工程应用价值大.

The complex illumination conditions in the high-speed wind tunnels can easily lead to the flicker of the videogrammetric measurement(VM) test image sequences,which affect the accuracy of the measurement result.As the flicker contains both the global change of the gray level caused by the illumination and the local change caused by the object motion,deformation and so on,it is difficult to apply the compensation methods based on affine transformation model (linear or non-linear) between pixels or between blocks of the images.Based on the scale space theory,the scale-time equalization(STE) method was proposed by Delon(2006) and applied to correct flicker in video and movie.The STE method performs the Gaussian convolution in the time dimension of each gray value in the histograms to obtain the target histogram of each image,and then uses histogram matching to get the flicker compensated images.Three experiments with different flicker and interference were conducted:the camera calibration plate image under the fluorescent lighting,the deformation test image of the wind tunnel model and the cavity oil film test image with diffusion.The experimental results show that the STE method is suitable for the global flicker compensation of the VM test image sequence.Compared with the model-based methods,the flicker compensation effect is not dependent on the reference image,and the robustness to the interference factors such as image jitter,local motion or deformation is strong.Meanwhile,the method has the advantages of simple procedures and small computation cost,and therefore it has high engineering application value.

The captive trajectory simulation (CTS) system can be operated in two modes, the position-control technique and the velocity-control technique. The position-control technique with closed-loop is employed in existing CTS at CARDC and yield satisfactory results. Its disadvantages are low efficiency and off-trajectory collisions caused by positioning the store model using move and pause technique. So, a velocity control technique based on time-space transform for CTS testing is presented in this paper. This technique uses time-space transform to reduce the requirement of acceleration and velocity capabilities for CTS rig, and establishes an error control loop for forces and moments to automatically adjust the time-step for solving the store model equations of motion. It allows the store model to move continuously along the trajectory to eliminate off-trajectory collisions, and dynamically generates the proper velocity-scaling and time-step to provide more efficient operation. The simulation and wind tunnel tests results have demonstrated that the velocity control technique is feasible. Its trajectories compare favorably with those obtained by the position-control, while the time occupation can be shortened about 50%, and in the same time a greater number of point on the trajectory can be obtained. Therefore, the velocity-control technique has a broad application prospects.

In the practical applications for camera pose estimation, the coordinates of reference points inevitably contain measurement errors, and the magnitude of the errors will not always be the same. If the camera pose is estimated directly without distinguishing the errors, the estimation result may be very different from the true value. Therefore, the weighted orthogonal iterative algorithm is proposed based on the widely used orthogonal iterative algorithm. In this algorithm, the weighted collinear error is taken as the objective function. In each iteration, the weight coefficients are determined according to the re-projection errors in image, and the camera pose estimation results are optimized by the coefficients. This algorithm satisfies the conditions of global convergence, and has the advantages of high precision and good robustness. The experimental results show that the proposed algorithm is effective. The estimation results of the proposed algorithm are significantly better than those of the orthogonal iterative algorithm, when the coordinates of reference points contain different errors. It shows that the proposed algorithm has strong engineering practical values.

针对现有的采用Booth算法与华莱士(Wallace)树结构设计的浮点乘法器运算速度慢、布局布线复杂等问题,设计了基于FPGA的流水线精度浮点数乘法器.该乘法器采用规则的Vedic算法结构,解决了布局布线复杂的问题;使用超前进位加法器(Carry Look-ahead Adder,CLA)将部分积并行相加,以减少路径延迟;并通过优化的4级流水线结构处理,在Xilinx(R)ISE 14.7软件开发平台上通过了编译、综合及仿真验证.结果证明,在相同的硬件条件下,本文所设计的浮点乘法器与基4-Booth算法浮点乘法器消耗时钟数的比值约为两者消耗硬件资源比值的1.56倍.

To quantify the structures of high-speed flow over a cavity,small circle points with equal space in the row and column are used as background for background oriented schlieren (BOS),and image processing techniques of mark points in videogrammetry measurement (VM) are also employed to break the limits of cross-correlation in existing BOS.The expres sions for computing refraction angle and displacement of nonparallel beams are derived.The fields of optical path difference (OPD) and refraction displacement when the beams from the small circle points to the center of the camera is crossing the flow are accurately calculated based on VM collinear equations.The measuring data on flow over the cavities in FL-21 wind tunnel demonstrates that the OPD differences no more than 1 μm and refraction angle about 1 μrad can be perceived distinctly,and the structures of waves/vortices/shear layer are quantified.The method proposed can provide a new way to measure aero-optic effects and visualize the complex flows.With simple optical system and no expensive coherent sources,the method has great application potential.