Shaking table tests are widely used to evaluate seismic effects on railway structures,but accurately measuring rail displacement remains a significant challenge owing to the nonlinear characteristics of large displace...Shaking table tests are widely used to evaluate seismic effects on railway structures,but accurately measuring rail displacement remains a significant challenge owing to the nonlinear characteristics of large displacements,ambient noise interference,and limitations in displacement meter installation.In this paper,a novel method that integrates the Kanade-Lucas-Tomasi(KLT)feature tracker with an extended Kalman filter(EKF)is presented for measuring rail displacement during shaking table tests.The method employs KLT feature tracker and a random sample consensus algorithm to extract and track key feature points,while EKF optimally estimates dynamic states by accounting for system noise and observation errors.Shaking table test results demonstrate that the proposed method achieves an acceleration root mean square error of 0.300 m/s^(2)and a correlation with accelerometer data exceeding 99.7%,significantly outper-forming the original KLT approach.This innovative method provides a more efficient and reliable solution for measuring rail displacement under large nonlinear vibrations.展开更多
现有追求高压缩质量的高光谱图像压缩算法普遍存在计算复杂度高、离线式处理、嵌入式平台实现难度大等问题,目前很难得到实际应用。为解决以上问题,设计一种基于KLT和HEVC的嵌入式高光谱图像实时压缩方法。首先基于KLT去除谱间相关性,...现有追求高压缩质量的高光谱图像压缩算法普遍存在计算复杂度高、离线式处理、嵌入式平台实现难度大等问题,目前很难得到实际应用。为解决以上问题,设计一种基于KLT和HEVC的嵌入式高光谱图像实时压缩方法。首先基于KLT去除谱间相关性,然后基于HEVC去除空间相关性并完成量化编码的过程。基于NVIDIA Jetson TX1平台,设计并实现了CPU和GPU异构并行压缩处理系统。利用真实数据集对所设计算法和所实现平台进行了性能及可行性验证。实验结果表明:在相同压缩比下,与离散小波变换(DWT)+JPEG2000算法相比,该系统明显提升了重建精度,在峰值信噪比(PSNR)方面平均提高了1.36 d B;同时,相比CPU,在GPU中进行KLT计算也至多可缩短33%的运行时间。展开更多
基金The National Key Research and Development Program of China(No.2021YFB2600600,2021YFB2600601)the National Natural Science Foundation of China(No.52408456)+2 种基金China Postdoctoral Science Foundation(No.2022M720533)College Students’Innovative Entrepreneurial Training Plan Program(No.202410710009)Key Research and Development Program of Shaanxi,China(No.2024SF-YBXM-659).
文摘Shaking table tests are widely used to evaluate seismic effects on railway structures,but accurately measuring rail displacement remains a significant challenge owing to the nonlinear characteristics of large displacements,ambient noise interference,and limitations in displacement meter installation.In this paper,a novel method that integrates the Kanade-Lucas-Tomasi(KLT)feature tracker with an extended Kalman filter(EKF)is presented for measuring rail displacement during shaking table tests.The method employs KLT feature tracker and a random sample consensus algorithm to extract and track key feature points,while EKF optimally estimates dynamic states by accounting for system noise and observation errors.Shaking table test results demonstrate that the proposed method achieves an acceleration root mean square error of 0.300 m/s^(2)and a correlation with accelerometer data exceeding 99.7%,significantly outper-forming the original KLT approach.This innovative method provides a more efficient and reliable solution for measuring rail displacement under large nonlinear vibrations.
文摘现有追求高压缩质量的高光谱图像压缩算法普遍存在计算复杂度高、离线式处理、嵌入式平台实现难度大等问题,目前很难得到实际应用。为解决以上问题,设计一种基于KLT和HEVC的嵌入式高光谱图像实时压缩方法。首先基于KLT去除谱间相关性,然后基于HEVC去除空间相关性并完成量化编码的过程。基于NVIDIA Jetson TX1平台,设计并实现了CPU和GPU异构并行压缩处理系统。利用真实数据集对所设计算法和所实现平台进行了性能及可行性验证。实验结果表明:在相同压缩比下,与离散小波变换(DWT)+JPEG2000算法相比,该系统明显提升了重建精度,在峰值信噪比(PSNR)方面平均提高了1.36 d B;同时,相比CPU,在GPU中进行KLT计算也至多可缩短33%的运行时间。
