Scholarlink

<p>The identification of cracks in materials and surfaces is paramount for averting potential disasters and ensuring structural integrity. To address this critical need, an innovative image processing model is developed and employed using Convolutional Neural Networks (CNNs). The proposed model utilizes webcam technology to continuously capture images, which are then analysed to detect cracks. Through rigorous training on extensive datasets containing both positive and negative examples, comprising 20,000 images each, our model achieves exceptional accuracy, consistently reaching a confidence level of 93–97%. Following thorough testing with diverse images sourced from the internet, our model demonstrates robust performance in crack detection. Building upon this success, we have expanded its capabilities to access live video feeds from devices’ webcams, enabling real-time monitoring and prediction of cracks. Upon detecting a crack, the model promptly stores the corresponding image, facilitating timely intervention and maintenance actions. This innovative approach not only enhances efficiency in crack detection but also empowers manufacturers to proactively address potential vulnerabilities in materials and parts. By leveraging cutting-edge technology and advanced machine learning techniques, our model contributes significantly to mitigating risks associated with structural weaknesses and ultimately safeguarding against catastrophic failures.</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

Real-Time Detection of Cracks and Structural Weaknesses in Materials and Surfaces Using Improved Deep Layered CNN

  • 期刊论文

  • 发表日期：

    2025年5月6日

查看全文
  • K. Amudhan,
  • K. Vignesh Saravanan,
  • R. Venkatesh,
  • M. Ramnath

摘要

The identification of cracks in materials and surfaces is paramount for averting potential disasters and ensuring structural integrity. To address this critical need, an innovative image processing model is developed and employed using Convolutional Neural Networks (CNNs). The proposed model utilizes webcam technology to continuously capture images, which are then analysed to detect cracks. Through rigorous training on extensive datasets containing both positive and negative examples, comprising 20,000 images each, our model achieves exceptional accuracy, consistently reaching a confidence level of 93–97%. Following thorough testing with diverse images sourced from the internet, our model demonstrates robust performance in crack detection. Building upon this success, we have expanded its capabilities to access live video feeds from devices’ webcams, enabling real-time monitoring and prediction of cracks. Upon detecting a crack, the model promptly stores the corresponding image, facilitating timely intervention and maintenance actions. This innovative approach not only enhances efficiency in crack detection but also empowers manufacturers to proactively address potential vulnerabilities in materials and parts. By leveraging cutting-edge technology and advanced machine learning techniques, our model contributes significantly to mitigating risks associated with structural weaknesses and ultimately safeguarding against catastrophic failures.