Scholarlink

<p>The development of numerous microfractures in fractured reservoirs is a key factor that influences the reserves and production of oil and gas in the formation. Electrical imaging logging is the most effective method for identifying and evaluating microfractures, but it still faces severe interference when microfractures are extensively developed. To objectively and accurately achieve rapid extraction of complex microfractures in electrical imaging, this paper introduces the Mask R-CNN image segmentation algorithm for fractures. However, existing algorithms often struggle in scenarios involving strip-shaped microfractures, mainly when microfractures are densely distributed. This study improved the Dynamic Snake Vision Mask R-CNN, aimed at improving microfracture extraction. The model incorporates dynamic snake convolution (DSConv) along with a residual network to enhance feature extraction. Additionally, a specialized data augmentation method is designed specifically for dense microfracture scenarios, which improves the model’s generalization capabilities. An enhanced dataset was utilized to train the model, and experimental results indicate that the improved model successfully detected microfractures that conventional models often missed. It also provided more accurate segmentation contours for certain microfractures. Notably, the average precision for small-sized objects improved significantly, rising from 0.591 to 0.850, and the model achieved an overall accuracy of 0.979 on the Archaean metamorphic rock buried-hill reservoir. This research offers substantial technical support for applying electrical imaging in the exploration of numerous complex microfractures.</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

Dynamic snake vision mask r-cnn: a method for microfracture detection in electrical logging imaging

  • 期刊论文

  • 发表日期：

    2025年10月28日

查看全文
  • Yihang Lu,
  • Sinan Fang,
  • Yi Yu,
  • Huimin Wu,
  • Guilan Lin,
  • Guohao Ma,
  • Zehan Wang

摘要

The development of numerous microfractures in fractured reservoirs is a key factor that influences the reserves and production of oil and gas in the formation. Electrical imaging logging is the most effective method for identifying and evaluating microfractures, but it still faces severe interference when microfractures are extensively developed. To objectively and accurately achieve rapid extraction of complex microfractures in electrical imaging, this paper introduces the Mask R-CNN image segmentation algorithm for fractures. However, existing algorithms often struggle in scenarios involving strip-shaped microfractures, mainly when microfractures are densely distributed. This study improved the Dynamic Snake Vision Mask R-CNN, aimed at improving microfracture extraction. The model incorporates dynamic snake convolution (DSConv) along with a residual network to enhance feature extraction. Additionally, a specialized data augmentation method is designed specifically for dense microfracture scenarios, which improves the model’s generalization capabilities. An enhanced dataset was utilized to train the model, and experimental results indicate that the improved model successfully detected microfractures that conventional models often missed. It also provided more accurate segmentation contours for certain microfractures. Notably, the average precision for small-sized objects improved significantly, rising from 0.591 to 0.850, and the model achieved an overall accuracy of 0.979 on the Archaean metamorphic rock buried-hill reservoir. This research offers substantial technical support for applying electrical imaging in the exploration of numerous complex microfractures.