Precise delineation of a salt dome’s geobody in seismic data requires intelligent integration, image fusion or combination of seismic attributes using advance methods. There are various attribute integration methods available and many other are still under development. In this study, we introduce a new strategy for feature extraction from seismic images followed by their combination at the data level and subsequent information integration on seismic image. The aim of the presented study was to introduce an efficient method for image segmentation using the ordered weighted averaging (OWA) and the logistic function methods. In other competitive methods, the combination of information is used by employing various weighting functions whereas in the OWA method the specific weights are defined according to the importance of the characteristics of the target under investigation, which can be enhanced in the extracted seismic attributes. Then, the images of seismic attributes are combined by the logistic function method to distinguish the target geobody from the rest of the image. In the logistic function method, the attributes are combined with modified equations with the fuzzy gamma and geometric mean operators. This strategy can define the boundaries of the target and distinguish the geobody in the seismic image. The methodology was applied to synthetic and real field datasets, which contain a salt dome. For comprehensive comparison of the performance of the proposed method with the OWA and the logistic function methods, their various modifications of competitive methods were also applied to the same datasets. The OWA with pessimistic and optimistic weighting algorithms were both applied to fuzzy and binary models. The modified fuzzy logistic function was also applied to the fuzzy and binary models whereas the modified geometric averaging was applied to the datasets. The results were compared qualitatively and quantitatively. For the synthetic data, a synthetic model was used as the base model for pixel-by-pixel comparison with the true model and the binary models. The accuracy is the ratio of pixels correctly selected as the target in the final binary model to the correct pixels of the target in the true model. For the field data example, however, because there was no true model available, an expert interpreted model was used as the base model. The qualitative comparisons of results for the synthetic and field datasets show that the OWA method can better identify the target under investigation in the seismic image. In the quantitative comparison, the OWA pessimistic method presented 99.00% and 94.6% accuracy in the synthetic and real field datasets, respectively.

在地震数据中精确描绘盐丘地质体需要使用先进方法进行智能整合、图像融合或地震属性组合。有多种属性集成方法可用，许多其他方法仍在开发中。在这项研究中，我们介绍了一种从地震图像中提取特征的新策略，然后在数据级别将它们组合起来，然后在地震图像上进行信息整合。本研究的目的是介绍一种使用有序加权平均 (OWA) 和逻辑函数方法进行图像分割的有效方法。在其他竞争方法中，通过采用各种加权函数来组合信息，而在OWA方法中，具体的权重是根据被调查目标特征的重要性定义的，可以在提取的地震属性中得到增强。然后，将地震属性的图像通过逻辑函数方法进行组合，以区分目标地质体与图像的其余部分。在逻辑函数方法中，属性与带有模糊伽马和几何平均算子的修正方程相结合。该策略可以定义目标的边界并区分地震图像中的地质体。该方法应用于包含盐丘的合成和真实现场数据集。为了全面比较所提出的方法与 OWA 和逻辑函数方法的性能，它们对竞争方法的各种修改也应用于相同的数据集。具有悲观和乐观加权算法的 OWA 都适用于模糊和二元模型。修改后的模糊逻辑函数也适用于模糊和二元模型，而修改后的几何平均则适用于数据集。对结果进行了定性和定量比较。对于合成数据，使用合成模型作为基础模型，与真实模型和二元模型进行逐像素比较。准确率是最终二值模型中正确选择为目标的像素与真实模型中目标的正确像素之比。对于字段数据示例，但是，由于没有可用的真实模型，因此使用专家解释模型作为基础模型。合成数据集和现场数据集结果的定性比较表明，OWA方法可以更好地识别地震图像中的调查目标。在定量比较中，OWA 悲观方法在合成和实际现场数据集中的准确率分别为 99.00% 和 94.6%。