تفسیر تصویر معنایی پروفیل های پرتو گاما در اکتشاف نفت
|کد مقاله||سال انتشار||مقاله انگلیسی||ترجمه فارسی||تعداد کلمات|
|20135||2011||11 صفحه PDF||سفارش دهید||محاسبه نشده|
Publisher : Elsevier - Science Direct (الزویر - ساینس دایرکت)
Journal : Expert Systems with Applications, Volume 38, Issue 4, April 2011, Pages 3724–3734
This paper presents the S-Chart framework, an approach for semantic image interpretation of line charts; and the InteliStrata system, an application for semantic interpretation of gamma ray profiles. The S-Chart framework is structured as a set of knowledge models and algorithms that can be instantiated to accomplish chart interpretation in all sorts of domains. The knowledge models are represented in three semantic levels and apply the concept of symbol grounding in order to map the representation primitives between the levels. The interpretation algorithms carry out the interaction between the high-level symbolic reasoning, and the low-level signal processing. In order to demonstrate the applicability of the S-Chart framework, we developed the InteliStrata system, an application in Geology for the semantic interpretation of gamma ray profiles. Using the developed application, we have interpreted the charts of two gamma ray profiles captured in petroleum exploration wells, indicating the position of stratigraphic sequences and maximum flooding surfaces. The results were compared with the interpretation produced by an experienced geologist using the same data input. The system carried out interpretation that were compatible with the geologist interpretation over the data. Our framework has the advantage of allowing the integration of existing domain ontologies with domain independent visual knowledge models and also the ability of grounding domain concepts in low-level data.
Human image interpretation is a process based on the combination of memorized visual knowledge and interpretation algorithms. The reasoner is able to compare, in a flexible and creative way, patterns captured from the domain with his/her own knowledge base of abstractions created from the previous seen visual patterns. The recognized objects support the semantic interpretation of the whole scene. We consider that a successful image interpretation system requires the capability of applying visual knowledge to support the recognition and interpretation of objects in the domain. This is even more valid in knowledge intensive domains, where experts employ both visual and conceptual knowledge to infer new information from pictorial data (e.g., from X-ray images in medicine). An image interpretation system can increase significantly the capability of extracting meaningful information from images by combining low-level pixel processing with conceptual models of visual and domain knowledge. Indeed, since the origins of the first systems based on this concept (like VISIONS (Hanson & Riseman, 1978), SIGMA (Matsuyama, 1987)) a great research effort has been made in proposing frameworks that integrate and minimize the semantic gap (Smeulders, Worring, Santini, Gupta, & Jain, 2000) between explicit knowledge models and low-level image processing algorithms. These so called semantic interpretation systems are being used with relative success in areas like biology ( Hudelot, Maillot, & Thonnat, 2005), traffic control ( Fernyhough, Cohn, & Hogg, 2000) and robotics ( Chella, Frixione, & Gaglio, 2001). The main focus of recent research is related to the interpretation of images captured from natural scenes. However, in some specialized domains, the interpretation of images produced artificially (e.g., charts, diagrams, etc.) is a common and critical task (like in Geology (Serra, 1984), Economy (Oberlechner, 2001) and Medicine (Berger et al., 2005)). One typical type of artificial image is the line chart, like the one depicted in the Fig. 1. Line charts are a common visual representation of data sets like space–time series. Since they can be easily plotted from raw data, they are extensively used in many time-critical problems, such as the diagnostic of heart attack, the definition of strategies in petroleum well production, or the identification of faults in monitoring systems. Experts can better visualize trends and patterns hidden in the data when it is projected as line charts. The expert relates them to objects in his/her domain to infer new information. An interpretation system that could automate (or semi-automate) the interpretation process of line charts would incorporate the specific knowledge applied by experts when recognizing trends and patterns on the graph. Full-size image (15 K) Fig. 1. Line chart example, depicting the variation of the gamma radiation along an exploration well. Figure options The reasoning process for visual chart interpretation close resembles the interpretation of real-world images. However, it is not possible to apply the semantic image interpretation methods directly to chart interpretation for two different reasons. First, the visual features that are significant in charts are not modeled in the usual image interpretation methods. Examples of features in line charts are curve patterns, points descriptors and other geometric features. The second reason is the evident difference in the raw data itself. Low-level primitives like pixels, regions and color codes are not representative of the low-level chart data. These differences are enough to encourage the proposal of a new framework for semantic interpretation of line charts. This paper is divided in two parts. In the first part, we the present the S-Chart, a semantic image interpretation framework specialized in the extraction of meaningful information from chart-like data. The objects we are looking for can be represented in a knowledge model (such as a domain ontology, for instance). The chart-like data is the one-dimensional signal that can be projected as a line chart. The framework consists of a set of models of visual knowledge conceived for representing the semantic objects that can be visually recognized in charts and the interpretation algorithms. The input of these algorithms is the raw signal, or the measure of some indicator (such as a gamma ray log, as in our application) against some other dimension, such as time, depth, distance, etc., which can be expressed as a chart. The output is the set of the instances of the visual knowledge model that represents the semantic content that can be extracted from the signal. The meaning of the recognized instances is given by the knowledge model itself. Furthermore, the framework is conceived to be domain-independent and it is meant to be applied to problems that require visual knowledge to be solved. It works by associating the recognized pattern on the chart with its high-level representation in the knowledge model. All the process is driven by the knowledge model. In the second part of the paper, we present the InteliStrata system, an implementation of the S-Chart framework for semantic interpretation of well log charts in the domain of sequence stratigraphy, a sub area of Geology. The task of the InteliStrata system is to infer the presence of particular geological features based on the recognition of significant visual features in the gamma ray logs. The system employs knowledge models (domain and visual) to carry out this interpretation. Our approach offers new, domain-independent primitives to support chart interpretation compatible with the existent primitives applied in image interpretation. We also show how semantic interpretation can be achieved using standardized representation formalisms, like OWL (McGuinness & Harmelen, 2004) and SWRL (Horrocks et al., 2004).
نتیجه گیری انگلیسی
This paper presents the S-Chart framework for semantic interpretation of line charts, which consists of a visual knowledge model, a symbol grounding model and an interpretation algorithm. The visual knowledge models provide the necessary primitives for visual knowledge representation regarding line-chart interpretation. They are divided in three semantic levels. The model is provided with the OWL formalism, contributing to its use in real-world scenarios. It also offers modeling primitives to explicitly integrate symbolic models with low-level raw data. The symbol grounding model provides the mapping link between primitives in different semantic levels. We have defined four relations to represent the mapping between interpreted instances. We also have defined the symbol detector primitive, a constrained SWRL rule that model the mapping between primitives. The presented algorithms carry out the interpretation by generating and testing of hypothesis. It has been shown how they can take advantage of the structure of the visual knowledge models to generate new visual hypotheses and how the symbol detectors are used to infer new knowledge. The algorithms also coordinate the symbol interpretation with the signal processing algorithms. We also described the InteliStrata system, an application of the S-Chart framework for interpretation of well logs in the domain of sequence stratigraphy. We demonstrate how the abstract models and algorithms of the S-Chart framework can be implemented in an useful interpretation system. It has been shown how the system can rely on third-party software, minimizing the development effort. We also compare the interpretation of a gamma ray log made by an expert in the domain and the interpretation done by the system. The results are encouraging and show the effectiveness of the S-Chart approach for line chart interpretation. Further research on the integration of the visual knowledge models with other reasoning algorithms is in progress.