Scholarlink

<p>In surface roughness evaluation of machined surfaces by image processing, many flaws need to be addressed such as image size, resolution, lighting, clarity, and image processing methodology. In this investigation, a new methodology is adopted to overcome the above flaws by calibrating the measurement process on a single cast iron faced specimen by taking images at various spots, considering each spot with 16 adjacent images (of size 4 × 3&#xa0;mm) stitched together to assess the larger area (of size 16 × 12&#xa0;mm) of the specimen surface. To evaluate the surface roughness, the stitched images of various specimen surfaces are 1D wavelet transformed using coiflet wavelets for first frequency in the order of six-level decomposition. It is found that the 4th order decomposition gives good correlation levels. The means of the 4th-order detail coefficient have 95% confidence that the true mean falls between 1.25406E-18 to 3.08743E-18. Though the 5th-order detail coefficient gives good correlation coefficients which is more than 0.8, their true mean does not fall between confidence intervals. So, it is concluded that this image stitching MATLAB algorithm achieves an enhanced surface roughness evaluation through wavelet transformation of stitched images.</p>

错误：搜索内容不能为空，请输入英文关键词
错误：关键词超出字数限制，请精简
高级检索

Enhanced in-situ surface roughness evaluation of faced cast iron surfaces by image stitching

  • 期刊论文

  • 发表日期：

    2025年1月5日

查看全文
  • J. Mahashar Ali,
  • H. Siddhi Jailani

摘要

In surface roughness evaluation of machined surfaces by image processing, many flaws need to be addressed such as image size, resolution, lighting, clarity, and image processing methodology. In this investigation, a new methodology is adopted to overcome the above flaws by calibrating the measurement process on a single cast iron faced specimen by taking images at various spots, considering each spot with 16 adjacent images (of size 4 × 3 mm) stitched together to assess the larger area (of size 16 × 12 mm) of the specimen surface. To evaluate the surface roughness, the stitched images of various specimen surfaces are 1D wavelet transformed using coiflet wavelets for first frequency in the order of six-level decomposition. It is found that the 4th order decomposition gives good correlation levels. The means of the 4th-order detail coefficient have 95% confidence that the true mean falls between 1.25406E-18 to 3.08743E-18. Though the 5th-order detail coefficient gives good correlation coefficients which is more than 0.8, their true mean does not fall between confidence intervals. So, it is concluded that this image stitching MATLAB algorithm achieves an enhanced surface roughness evaluation through wavelet transformation of stitched images.