Scholarlink

<p>Machine learning programs have been in use for various industrial processes for years now, primarily to analyze data and increase the efficiency of said processes. This contribution documents the attempt to use machine learning, specifically computer vision, to automate part of the evaluation process of testing rockfall protection kits.</p><p>The use of computer vision models for speed estimation is still a&#xa0;niche of this field of research. It aims to reduce the costs and complexities brought on by physical measuring instruments by replacing them through a&#xa0;combination of videos and software.</p><p>The main objective of this project was to design a&#xa0;program that could automatically process video material of rockfall protection kit tests by identifying the position, speed, and trajectory of a&#xa0;concrete block accelerated toward the barrier. Three models with different data inputs were trained and analyzed.</p><p>The results demonstrate that the use of computer vision AI for the specified task is possible, achieving an accuracy of 97–99% in a&#xa0;virtual test stand and deviations of less than 5% in real-world tests compared to manual evaluations.</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

KI-basierte Analyse von Blockbewegungen im Rahmen von Typenprüfungen flexibler Steinschlagschutzsysteme

  • 期刊论文

  • 出版模式：

    开放获取

  • 发表日期：

    2026年7月10日

查看全文
  • Oliver Perl,
  • Christian Heiss

摘要

Machine learning programs have been in use for various industrial processes for years now, primarily to analyze data and increase the efficiency of said processes. This contribution documents the attempt to use machine learning, specifically computer vision, to automate part of the evaluation process of testing rockfall protection kits.

The use of computer vision models for speed estimation is still a niche of this field of research. It aims to reduce the costs and complexities brought on by physical measuring instruments by replacing them through a combination of videos and software.

The main objective of this project was to design a program that could automatically process video material of rockfall protection kit tests by identifying the position, speed, and trajectory of a concrete block accelerated toward the barrier. Three models with different data inputs were trained and analyzed.

The results demonstrate that the use of computer vision AI for the specified task is possible, achieving an accuracy of 97–99% in a virtual test stand and deviations of less than 5% in real-world tests compared to manual evaluations.