Scholarlink

<p>Underground coal mine search tasks play a critical role in safety production, emergency rescue, and resource management. However, the complex environment of underground coal mines, characterized by uneven lighting and dust interference, imposes stringent requirements on accuracy and efficiency of search tasks. To address these challenges, this paper proposes a novel search method for underground coal mine. Firstly, a database is constructed. Secondly, the Fourier Transform is applied to obtain high-frequency and low-frequency images. The Markov Transition Field is then utilized to generate the Markov maps of the low-frequency image. Subsequently, feature extractors are employed to extract features, and Hill Diversity is applied to binarize the features. Finally, based on binarized features, the similarity measures are calculated, and the final search results are obtained by sorting the similarity measures. Experimental results show that, compared to directly extracting features from ShuffleNetV2, VGG16, ResNet18, DenseNet121, AlexNet, ConvNeXt, SqueezeNet, GoogleNet, EfficientNet, MnasNet, RegNet, and MobileNet to search, the proposed method which optimizes both the pre-processing and post-processing stages of feature extraction achieves superior performance, with an average improvement of 4.47% in accuracy and 82.97% in efficiency.</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

Coal mine underground target search based on markov transition field and hill diversity

  • 期刊论文

  • 发表日期：

    2025年6月11日

查看全文
  • Jiaojuan Wang,
  • Hailong Yao,
  • Minghan Guo,
  • Tongyang Dao,
  • Hongbin Dou,
  • Xuefeng Zhu

摘要

Underground coal mine search tasks play a critical role in safety production, emergency rescue, and resource management. However, the complex environment of underground coal mines, characterized by uneven lighting and dust interference, imposes stringent requirements on accuracy and efficiency of search tasks. To address these challenges, this paper proposes a novel search method for underground coal mine. Firstly, a database is constructed. Secondly, the Fourier Transform is applied to obtain high-frequency and low-frequency images. The Markov Transition Field is then utilized to generate the Markov maps of the low-frequency image. Subsequently, feature extractors are employed to extract features, and Hill Diversity is applied to binarize the features. Finally, based on binarized features, the similarity measures are calculated, and the final search results are obtained by sorting the similarity measures. Experimental results show that, compared to directly extracting features from ShuffleNetV2, VGG16, ResNet18, DenseNet121, AlexNet, ConvNeXt, SqueezeNet, GoogleNet, EfficientNet, MnasNet, RegNet, and MobileNet to search, the proposed method which optimizes both the pre-processing and post-processing stages of feature extraction achieves superior performance, with an average improvement of 4.47% in accuracy and 82.97% in efficiency.