Application of Faster R-CNN in Reservoir Fractures Identification
Fracture detection and parameter extraction is the key and difficult point for reservoir.The development of the regional convolution neural network(R-CNN)provides a new idea for computer automatic identification of fractures.The R-CNN converts classification problems into regression problems.The Fast R-CNN incorporates the per-class bounding-box regression into the training process,which could shorten the time spent in the per-class bounding-box regression.By changing the way to output the bounding-box,the Faster R-CNN could greatly shorten the time spent in the output bounding-box.In this paper,the preliminary application of Faster R-CNN in logging imaging identification of fractures presents excellent results.The time for detection of fracture area and parameters is 0.041s.It is verified that Faster R-CNN is an effective method to identify fractures and extract fracture parameters.
国家科技期刊平台
NSTL
万方数据
