Remote sensing image semantic segmentation, which aims to realize pixel-level classification according to the content of remote sensing images, has broad applications in various fields. Thanks to the superiority of deep learning (DL), the semantic segmentation model based on the convolutional neural network (CNN) dramatically promotes the development of remote sensing image semantic segmentation. Due to the high resolution, comprehensive coverage, extensive data, and sizeable spectral difference of high-resolution remote sensing images (HRRSI), the existing GPU is not suitable for directly semantic segmentation of the whole image. Cutting the image into small patches will lead to the loss of context information, resulting in the decline of accuracy. To address this issue, we propose the multiscale context self-attention network (MSCSANet). It combines the benefits of the self-attention mechanism with CNN to improve the segmentation quality of various remote sensing images. The MSCSANet extracts multiscale features from multiscale context images to solve the problem of feature loss caused by image segmentation. In addition, in order to make use of the feature of large-scale context, the multiscale context patches are used to guide the local image patch to focus on different fine-grained objects to enhance the feature of the local image patch. Moreover, considering the limited computing resources, we designed a linear self-attention module to reduce the computational complexity. Compared with other DL models, our proposed model can enhance the ability of multiscale features in complex scenes, and realizes improvements of 1.56% mean intersection over union (MIoU) on the Gaofen Image Dataset and 1.93% MIoU on the ISPRS Potsdam Dataset, respectively.
The types of fractured-vuggy reservoirs are diverse, with dissolution holes and fractures of different scales as the main reservoir spaces. Clarifying the connectivity between wells is crucial for improving the recovery rate of fractured-vuggy reservoirs and avoiding problems of poor water- flooding balance and serious water channeling. A traditional dynamic connected model hardly describes the geological characteristics of multiple media, such as karst caves and fractures, which cause multiple solutions from the calculation. The static analysis is the basis for connectivity evaluation. In this study, we designed an intelligent search strategy based on an improved A* algorithm to automatically find a large-scale fractured-vuggy connected path by seismic multi-attribute analysis. The algorithm automatically evaluates the interwell-connected mode and clarifies the relationship between the static connected channel and the fractured-vuggy space configuration. Restricted by various factors, such as seismic identification accuracy, a static connectivity study can hardly determine the filling and half-filling inside the channel effectively, even if it can identify the main connectivity channels. An injection-production response analysis based on dynamic production data can more accurately reflect the reservoir’s actual connectivity and degree of filling. This paper further studies dynamic response characteristics based on multifractals combined with production data. To reduce the evaluation uncertainty, we combined the static and dynamic connected analysis results to comprehensively evaluate the main connected modes, such as large fracture connectivity, cavern connectivity, and fractured-vuggy compound connectivity. We use the Tahe oilfield as an example to carry out an automatic evaluation of the connected pattern. The comprehensive evaluation results of the new algorithm were basically consistent with the tracer test results and can better reflect the interwell space-configuration relationship. Our model has certain guiding significance for the adjustment of working measures during waterflooding in fractured-vuggy reservoirs.
Fractured-vuggy carbonate reservoirs have complex geological structures including pores, caves and fractures, which causes frequent working system adjustments and makes the production prediction extremely challenging. Currently, the most widely used methods in such prediction are the water drive characteristic curve methods and machine learning based models. However, frequent working system adjustments (such as well shut-in) could make water drive characteristic curve unstable, which provides unsatisfactory accuracy of the prediction model. In this paper, by integrating the variational mode decomposition (VMD) and gated recurrent unit (GRU), a novel machine learning based prediction model termed VMD-GRU is proposed to address the limitations of the water drive characteristic curve methods. The time-series production data are firstly decomposed by using VMD into several sub-series that represent different characteristics of the data. GRU is used to establish autoregression model, which can extract the inner characteristics of each sub-series and make prediction. The final prediction outputs are obtained by aggregating prediction result of each GRU model. The proposed VMD-GRU model is verified with the real-world production data from the Tahe oilfield of China. The experimental results demonstrate that the proposed VMD-GRU model outperforms the existing production prediction models for fractured-vuggy carbonate reservoirs.
15 August 2024 in Environmental Monitoring and Assessment