A Conceptual Framework for the Automated ...

This article has been accepted for publication in a future issue of this journal, but has not been fully edited. Content may change prior to final publication. Citation information: DOI 10.1109/ACCESS.2016.2594259, IEEE Access JOURNAL OF LATEX CLASS FILES, VOL. 14, NO. 8, AUGUST 2015

1

A Conceptual Framework for the Automated Generalization of Geological Maps Based on Multiple Agents and Workflow Junqiang Zhang, Chonglong Wu, and Lizhe Wang, Senior Member, IEEE

Abstract—For the complex semantic features of objects in a geological map, maintaining a hierarchy between different geological objects is a big challenge when automatically generalizing the geological map. The typical methods focus on automation of geological map generalization or pay more attention to the hierarchical relation between geological objects, which reduces the accuracy of the geological map generalization result. Therefore, a conceptual framework that focuses on both the automated process and the geological objects is particularly important in developing efficient software designed for automated generalization of geological maps. In this paper, we design a compound conceptual framework for automated generalization of geological maps based on multiple agents and workflow. In this framework, the process is divided into three stages: structure analysis, map generalization, and style standardization. The map objects in the source geological map are abstracted as diverse agents with different properties and behaviors, and the agents can communicate with each other when they are activated. Thus, the relationship of the map objects are coordinated in geological map generalization, avoiding the conflict between the different operation levels. Workflow technology is used to manage the automated process. We discuss the task, modeling method, and specific operation in every stage based on the current conceptual framework and the characteristics of a geological map. Finally, we use a simple geological map for experimental studies that verify the proposed conceptual framework. The result shows that it is advantageous to design the software for automated generalization of geological maps based on the proposed compound conceptual framework. Index Terms—Geological map, Map generalization, Geological map generalization.

I. I NTRODUCTION geological map is a complex document that attempts to describe the geological structure at and below the earth’s surface. The maps use legends to divulge a wealth of information, including the distribution of surface petrographic, lithology, geological ages, geological structure, magmatic activity, and mineral deposits [1]. These maps are the basis of basic geosciences research, planning, design and construction of buildings, mineralization regularity, and mineralization prediction. Geological maps with different scales of the same region are an essential tool for analyzing the geological conditions pattern in the region. The mapping process from a largescale geological map to a medium- or small-scale map is a process of extracting and integrating information, which is

A

Corresponding author: Lizhe Wang ([email protected]). Manuscript received May 28, 2016; revised July 22, 2016.

called geological map generalization for short [1]. This is a tedious process with a heavy workload. Too much work must be done, such as maintaining logical and unambiguous relations between map objects, showing the salient features in a simplified form, and removing unnecessary detail. The traditional method for generalizing geological maps is a manual cartographic process [2] [3]. The method is timeconsuming, inefficient, low precision, and labor intensive. With the advance of computer information systems (for example, computer-aided drafting and geographic information science) [4] [5], scientists started to look for ways to automate the map generalization process. At present, the two main research directions are generalization-related algorithms, the related theoretical foundation, and it’s practical application [2] [6] [7]. The conceptual framework conducts the process and algorithms is one of the theoretical foundations of this work. Some fruitful results have been achieved in research on the conceptual framework of automated geological map generalization. Based on the literature reviewed, all frameworks can be broadly classified as process-oriented conceptual frameworks and object-level-oriented conceptual frameworks. The former seeks to structure the entire generalization process conceptually. Thus, this method provides a complex generalization process based on the characterization of the structure and semantics of the map object. The advantage of this conceptual framework is a clear train of thought, explicit structure, and easy-to-achieve procedures [8] [9] [10]. The latter method addresses the level of map objects; the content is divided into different levels. At every level, appropriate knowledge and rules are required. Therefore, changing the domain expert and cartographic knowledge into digitized form is difficult. The benefit of this conceptual framework is a determined target, a clear hierarchy, and a specific operation [11] [12] [10]. However, the typical frameworks mainly focus on the automation process of general map generalization; little attention has been paid to generalization of geological maps. The only existing framework is not concerned enough with the hierarchical relation between geological objects. This causes the semantic error in the result map. It also ignores the coordination between the hierarchical relationship and the automated process of geological map generalization, which leads to a conflict between the different levels of operation and the low level of automation. To solve these problems, we designed a compound conceptual framework for automated generalization of geological

2169-3536 (c) 2016 IEEE. Translations and content mining are permitted for academic research only. Personal use is also permitted, but republication/redistribution requires IEEE permission. See http://www.ieee.org/publications_standards/publications/rights/index.html for more information.


maps based on workflow and multi-agent technology, which uses the two current frameworks within one conceptual framework. In this framework, geological objects are abstracted as agents with different properties and behavior, communication among agents based on generalization constraints is used to maintain the hierarchy between different geological objects [13] [14], and workflow technology is introduced to conduct the whole process of geological map generation. The innovation of this paper is that we put forward a compound conceptual framework for the automated generalization of geological maps. With this conceptual framework, the hierarchical relationship is reserved by the communication among the agents, and specific algorithms and operations are conducted with workflow technology. All this makes the operation coordination and ample data sharing realized. Therefore, maps can be generated effectively with more accurate results. The rest of this paper is organized as follows. We first summarize related work on geological map generalization in Section 2. Then the objectives and problems are defined in Section 3. We discuss the overall structure and implementation of the framework in Section 4. In Section 5, the experimental results and discussion are given. Section 6 concludes the paper and suggests opportunities for future work. II. R ELATED W ORK Related work includes two aspects, the existing framework and the characteristics. In this section, a brief review of the two aspects of related work are given. A. Characteristic of Geological Map Generalization Compared with general maps, geological objects on geological maps not only have spatial characteristics but also obvious thematic attribute characteristics. Therefore, the generalization of geological maps and general maps share dissimilarities as well as commonalities, mainly reflected in the following aspects. (1) The purpose. General maps are generalized mainly to solve conflicts between the limited mapping space and the rich mapping content caused by the reduced scale. However, geological maps are generalized not only to coordinate the space and content of a map but also to infer geological evolution process via the spatial distribution of geological objects with different scale. (2) The target. The proportion of points, polylines, and polygons in a general map is relatively balanced. However, in geological maps, polygons occupy a big portion. In addition, the polygons are distributed throughout the space, with no gaps and no overlaps. Consequently, the generalization of a geological map should pay attention to polygon objects particularly [15]. (3) The content. In general map generalization, the major operations are as follows: preserve and delete map objects, simplify geometric features, and solve location conflicts between map objects. However, in a geological map, the main task is to solve the semantic synthesis in addition to the integrated treatment of geometric characteristics, such as lithology synthesis and stratum merger with the adjacent

2

objects. This leads to different ideas, methods, and procedures compared with general map generalization. (4) The level. The generalization level for all kinds of map objects on a map is almost identical in general map generalization. However, the rule changes for the generalization of a geological map. For example, we often reduce the generalization level of a fault to reserve the necessary semantic information. To highlight some geological features, we need to increase the level of map generalization on non-focus geologic elements [16]. This analysis shows that the geological map generalization is unique in addition to the commonness of the general map generalization. We can get efficient and high-quality cartographic generalization results only when we integrate the spatial and semantic features of the geological factors. B. The Existing Framework for Map Generalization As an important foundation of automated map generalization, the conceptual framework has attracted much attention from cartographers over the past three decades. Several conceptual frameworks have been proposed. We distinguish these frameworks into two types, process-oriented and object-leveloriented conceptual frameworks. Some of the most influential conceptual frameworks will be explained briefly. The best representative process-oriented conceptual framework is put forward by Brassel and Weibel [17]. In this framework, the entire process is distinguished into five stages of processing. Every stage has its corresponding objective and work content. This conceptual framework advocated a complex generalization process based on the characterization of the structure and semantics of a map object. The impact of this conceptual framework in the field of automated cartographic generalization has been profound. However, excessive dependence on the structure and semantics of map objects hindered the development of this conceptual framework. The advantage of this conceptual framework is a clear train of thought, an explicit structure, and easy-to-achieve procedures. However, frequent interactions and strong subjectivity are its disadvantages, for the result depends on the cartographer’s experience too much [18]. According to the object-level-oriented conceptual framework, the content of map generalization can be divided into different levels. At every level, appropriate knowledge and rules are required. Therefore, changing the domain expert knowledge and cartographic knowledge into a digitized form is the main difficulty [19]. A simple way is the human interaction method. Another way is dividing the generalization process into a series of conditions and actions and then using clear rules to transfer knowledge, which was called the rule-based method [20]. Both ways have disadvantages. An evaluation of interactive generalization systems based on the human interaction method showed very low productivity [21] [22], and the generalization results highly depend on the user’s skills [17]. Some weaknesses of rule-based systems are the difficulty of acquiring and formalizing cartographic rules and the large number of rules required to describe conditions and actions between map objects sufficiently. The advantage of this



conceptual framework is a determined target, a clear hierarchy, and specific operations. However, the conflict between the different levels of each operation is not easy to reconcile, and the level of automation is not high.

III. O BJECTIVES AND P ROBLEMS D EFINITION A. Research Objectives Based on this discussion, the overall objective of this paper is to establish a generalization framework for geological maps applicable to vector data, including the technical modeling method on attributes and geometrical generalization of geological maps. The conceptual framework should be able to coordinate the relationship of the different map objects in geological map generalization so that conflicts between the different levels of operations can be avoided. Finally, the conceptual framework should improve the automation efficiency as much as possible. The goals of this research are to do the following: •

• •

design a conceptual framework for automatic generalization of geological maps based on the existing conceptual framework of general map generation and the characteristics of geological map generalization; implement the process with appropriate technology; and reconcile conflicts between the different levels during the process.

B. Research Problems The following problems are addressed to fulfill these specific research objectives. Specific Objective 1: Design a conceptual framework for automated generalization of geological maps based on the existing conceptual framework of general map generation and the characteristics of geological map generalization. • • •

What architecture is used to model the process? How is each relatively independent and yet closely linked? What are the necessary steps for a satisfactory visualization product?

Specific Objective 2: Implement the process. • • •

What is the main task in every stage of the process? What are the appropriate generalization operations and algorithms for every stage? How are the operations and algorithms performed interactively or automatically?

Specific Objective 3: Reconcile the conflicts between the different levels of each operation in the process of geological map generalization. • •

How can the hierarchical relation be maintained between geological objects during the process? How can conflicts between the different levels of operation be avoided?

3

IV. C OMPOUND C ONCEPTUAL F RAMEWORK FOR AUTOMATED G ENERALIZATION OF G EOLOGICAL M APS We put forward a compound conceptual framework for the automated generalization of geological maps based on multiagent and workflow technology. This conceptual framework overcomes the weaknesses of the process-oriented conceptual framework by acquiring the strong points of the object-leveloriented conceptual framework. Therefore, it has characteristics of both types of conceptual frameworks. In this conceptual framework, we call the geological map before generalization the source geological map and the resulting map the target geological map for the convenience of the narrative. In the next section, we describe the structure and implementation of this conceptual framework. A. The Overall Structure of the Compound Conceptual Framework Figure 1 is an overall structure diagram of the compound conceptual framework. The entire process can be divided roughly into three stages: the source geologic map structure analysis phase, the source geological maps generalization phase, and the target geological map format standardization phase. We will describe the main task of every stage in the following section. (1) Source geological map structure analysis phase This phase has the characteristics of object-level-oriented conceptual framework, which is similar to the structure recognition stage of process-oriented conceptual framework, but the operation target here refers only to the map object group [8] [23]. The objective of this step is to analyze the spatial and semantics relationship between object class level and the level of groups of map objects based on the purpose, type, and scale of the geological map as well as the characteristics of the mapping area and the geological data. The specific work contents of this phase are: leftmargin=*,labelsep=4mm • to determine the standard for the map object group for aggregation and elimination; • to define the aggregation hierarchy of map objects; and • to summarize the characteristics in quantity and quality of the map objects. The summary of quantity characteristics refers to expanding the numerical interval range of the indicators of the feature class. The number for classification of the geophysical field and isobaths, for example, should decrease as the map scale decreases. The summary of the quality characteristics refers to replacing detailed classification with rough classification. For instance, faults should be described with type differences, and the exact nature of the occurrence and stress on a map with a large scale; however, only a construction line can be used to represent it. What the structure analysis phase gets is a series of rules and data categories, which we call Type I constraints. Type I constraints can form semantic characteristics of map objects or the repeated communication between cartographers and a geological engineer. As these constraints reveal the importance of various types of geological factors and treatment methods,



4

Original Map

constraint TypeⅠ

constraint TypeⅡ

Task

-define the retention or rejection of map object -determine the aggregation hierarchy for object -summarize the characteristics in quantity and quality of the map object

Structure Analysis

-carry out the map generation at different levels from geometrical and semantic properties of map object to ensure a clear map and reasonable content

Map Generalization

possible backtrack

constraint Type Ⅲ

-from the angle of map clarity and classified visualization to assurance the transmission efficiency of map information

Format Standardization

Final Map

Fig. 1. The overall structure of the compound concept model for geological map generalization based on multi-agent and workflow technology.

purpose

type

scale

Object Object class class level level

Spatial Spatial relationship relationship

Object Object group group level level

Smantics Smantics relationship relationship

Geological Geological Map Map

Fig. 2. The process of structure analysis phase.

the constraints are very important to the processing target and the specific algorithm in the following stage. (2) Source geological map generalization phase The task of this phase is to generalize the map with different operations and algorithms for different levels of geological map objects. Concretely speaking, it consists of three levels of geological map objects: large (whole map), medium (feature classes), and small (a single feature). The main operations performed are aggregation, fusion, classification, shrink, shift, enhancement, exaggeration, mergers, refinement, simplification, smooth, and typifying. The meaning of these operations

may differ for different levels of map objects. The goal at this stage is to generate geological data that comply with the predetermined scale and with reasonable spatial and semantic relationships. The entire process of this phase has obvious characteristics of process orientation. Inevitably, this process will be affected by a series of constraints, which is called Type II constraints. Such constraints can be divided into geometric, topological relation, and process constraints [7] [24]. The geometric constraint is influenced by the change in the scale between the source geological map and the target geological map. The topological relation constraint refers to keeping the



relative position of the map objects after the dimensional change, and the process constraints mean the algorithm order in the process of geological map generalization. Of course, the processing of data in this phase is influenced by Type I constraints, such as the aggregation hierarchy between different map objects. (3) Target geological map format standardization phase This stage processes the data delivered from the previous phase in the visualization angles to obtain a delicate geological map with rich information and good visual effect. Therefore, it needs more interactions to integrate a huge amount of cartographic knowledge. The main tasks include text labeling, symbolization of point, line and polygon objects, and a density check of the map content. Text labeling mainly concerns the location, size, and style of text labels and the conflict between different labels. The symbolization of map objects refers to the color, geometry shape, line width, and fill pattern. The density check of map content is aimed to maintain balance between map content and map space, which affect the overall visual effect of a map. The rules, which are called Type III constraints in this stage, can be acquired from the geological map mapping standardized file. At this stage, we use workflow tools to manage the symbolic operation of various map objects in order to ensure the correct order and automation of all symbol operations [16] [25]. Separating visualization from map generalization meets the diverse needs for users of the geological map. The requirements of different users (nonprofessional users and domain experts) are not the same. Furthermore, it is necessary to determine the symbol style of the final geological map according to the user groups and the output medium. The reason is that the media used for geological map is becoming more diverse, such as paper, screen, and web application, even in the form of web services [26]. Only by adapting to these new media will geological maps last [27]. Another reason is the pursuit of a better result. We design different styles of the style template to meet the basic needs of users and meet the individual needs of users through customized style made by flexible operations. B. The Implementation Figure 2 shows the implementation of the compound conceptual framework and the generalization process modeling. The implementation procedure starts with the structure analysis phase. The input data usually are geological thematic data with a topographic map as the base map, for example, faults and lithologic units. The result data of this stage make up the geological object catalog, which indicates the retention or deletion of a geological object. We can gain such information from structure analysis of the input data or the manual input method. During map object classification, possible data analysis methods are data gradual optimization approaches or data clustering algorithms with the metrics of entropy measures. When polygonal map objects are generalized, the hierarchy for reallocating the space gained from the deleted elements based on the priority of the polymerization is necessary. Both outputs (geological object catalog and hierarchy) require expert interaction to determine the merge mode. Therefore, the first phase

5

of the model described in this paper is a human-computer interaction stage that emphasizes the application of knowledge. Due to the advantages in the combination of human-computer interaction and automation, we use a workflow system for managing the interactive operations involved in this stage to increase the flexibility and degree of automation of the operation. When the two outputs (geological object catalog and hierarchy) of the first phase are transferred to the source geological map generalization phase, they are used to generate a legible map. Before the generalization, the geometry dimension catalog should be clear. That is to say, the legibility thresholds of the smallest object size and the smallest separation distances should be defined in accordance with the target map scale. These data, the rules, and data categories from Type I constraints, together with the topological and procedural constraints from Type II constraints, constitute all data used in this stage. Then we characterize the objects and object groups with a series of constraints. At this time, the map object or object groups are objects with different attributes and behavior, which are called agents. Thus, the activated agents complete the cartographic generalization process successfully with multi-agent technology. When the properties and behavior of an agent violate a constraint (such as a map object is too large, is too close to another object, crosses other objects and so on), which is called a generalization con?ict [28], a series of algorithms and operations is needed to solve the generalization con?icts. Then, we should verify the result of this step in accordance with the constraint. If the evaluation result is not inconsistent with the related constraints, meaning the generalization conflict has been resolved, then this current state of agent is stored, and the mode of the agent changes to passive. If the conflict still exists, a new plan should be made and carried out until the state of the agent is perfect. At last, we can get a clear, narrow-scale geological map. The specific synthesis procedure is shown in Figure 3. The main task of the target geological map format standardization phase is the text labeling, symbolization (symbol, color, size) of points, symbolization (color, width, style) of line object, color and a fill pattern allocation of polygons, and the legibility check of the entire map. Therefore, at least four types of catalogs are needed: a text style catalog, a point symbol catalog, a line style catalog, and a color and fill pattern allocation catalog for the polygons. Most of these catalogs can be obtained with the current mapping standard. If no predefined catalogs exist, the definitions can be made by the use of a symbol brewer or color brewer [29]. This process requires constant testing in accordance with the knowledge of geology and cartography. In this stage, workflow technology can combine human-computer interaction and an automated algorithm. In summary, at different stages of the framework, repeated decision-making and the assessment of results are needed. Geological experts and cartography specialists should participate to ensure the semantics and illustrated effect of cartographic generalization. At the same time, the map must be implemented in conjunction with interactive decision support algorithms. In this case, the use of workflow management has significant



6

Original Map Process modeling object catalog object hierarchy

Human-Computer Interaction model way

Structure Analysis workflow management system

object summary

constraint based model way legible map

Map Generalization Multi-Agent system

possible backtrack Constraint and rules based model way

text color


symbol

workflow management system

Final Map

Fig. 3. The implementation of every phase in the compound concept model for geological map generalization.

advantages [24]. For the rules treatment of knowledge and the drawing constraints, a high degree of automation based on the multi-agent system in the second phase can be achieved [16]. Thus, the composite conceptual model of geological map generation can also be called as a multi-agent and workflowbased geological map generation conceptual model.

C. The Geological Map Agent and Workflow (1) The classification of geological map agent The compound conceptual framework divides the geological map generalization agent into generalization object agent and relationship agent. The generalization object agent can be instantiated as feature agent, target agent, and group agent. The classification and inheritance is shown in Figure 4 as follow. A feature agent contains one type of map element. It is responsible for the ordinary operations (such as selection and combination) in the macro level and the maintenance of index and communication inside the agents.

A target agent contains one map target. It is responsible for the ordinary operations in microcosmic level. It can compose a collaborated group with other target agents. A group agent contains the index of some target agents that have proximity relation or incidence relation. It mainly handles conflicting target agents. (2) Use workflow to manage the generalization process We made customized corresponding role model action condition and action. Then, we stratified process definition tools, process analytical tools, and process running tools based on workflow technology to realize the coupling between flow and flow work. On that basis, we divided the task of geological map generalization into a series of task units, in which every task unit will be realized by an agent. Finally, we use OSWorkflow to organize and manage all the agents ( Figure 5 ). In figure 5, the RA, FA and OA stand for the relationship agent, feature agent and object agent respectively. The OA agent includes target agent and group agent. As can be seen form the figure, the RA is the start of map generalization. Then, the FA



7

Geologica map generalization agent

Generalization target agent

Feature agent

Relationship agent

Target agent 1

*

1

Group agent

*

Fig. 4. The classification of geological map agent.

can be gained by the spatial relationship querying. Followed that, the specific generalization operation is implemented on OA, when the operation is finished, the FA and RA will be informed. This process cycles continuously until all the RA, FA and OA are generalized. V. E XPERIMENTAL S TUDY WITH THE C OMPOUND C ONCEPTUAL F RAMEWORK The data of this experiment is from a simple geological map of a local region in the Hubei province of China. The main targets of the experiment include keeping the map orderly when the scale changes from 1 : 50,000 to 1 : 250,000, keeping as much information as possible, and taking appropriate measures to ensure the quality of the map generalization. A. Introduction of the Experimental Zone Figure 6 shows the experiment of the compound conceptual framework for geological map generalization. The map range is 3.6 × 2.3 Km, and the scale is 1 : 50,000 ( Figure 6A). The geological map contains only geological lithology and structure information to reduce the difficulty. The symbolic polygons in different colors on the map represent different lithology units. The lithology units marked with K11 , K21 and K31 have the same paternal lithology unit K1 . T11 and T21 have the same paternal lithology unit T1 . T2 is the paternal lithology unit of T12 . The polygon marked with r3−1 is granite of the 5 Late Yanshan stage. Red lines stand for different faults in the experimental zone. B. Problems with Existing Approaches When the current frameworks are used to generalize this geological map, sequential application of these algorithms is a tedious and time-consuming procedure, which also can result in loops of operations. In addition, the hierarchical relation between geological objects can not be reserved. Furthermore, every step requires a number of parameters be input by the user, and therefore,the level of automation is low.

C. Stages of Geological Map Generalization We design the specific operations applied to automated generalization of the simple geological map in the experimental zone based on the compound conceptual framework (Figure 7). The process starts with the data input. The input data include geological thematic data and terrain data. According to Steiniger [18], three other types of information have to be given: (1) information on geological map patterns is necessary when there is a big difference between the original map scale and the target map scale; (2) The interdependent relationship among object class should be known; and (3) dependencies on professional themes are necessary. The information describes the potential of a structure or pattern, obtained from field experts or the original geological map [27]. In the following structure analysis phase, the number of map units and the hierarchy of all map units are defined, and the geological map pattern also should be identified as the definition of the geological map object group. The definition of an aggregation hierarchy is connected closely to the previous task of defining the number of map units. In the source geological map generalization phase, additional data are needed apart from the results of the first phase. A series of object class catalogs are defined. For example, to select data from topographic maps, a topographic object classes catalog is defined. A catalog of graphical minimal dimensions is needed to determine the minimum area parameters of lithology polygons. A clear definition of topological relationships of map objects is required when dealing with problems like polygons and must be generalized for point objects (particularly boreholes, moraine, spa spot) to ensure that the point symbols will still be inside the attributive geological area. When the output data of the source geological maps generalization phase are passed on to the target geological map format standardization phase, we should develop retention



RA1

FA1

8

OA1

FA2

OA2

Select Related FA2 Select FA1 Query spatial relationship Inform spatial relationship

Delete Delete Completing

Inform Selection Inform complete selecting Query spatial relationship Inform spatial relationship

Delete Delete Completing Inform merging result

Fig. 5. The management of map agents.

Fig. 6. Screenshots of the geological map before and after map generalization. (A) The original geological map with the scale of 1 : 50,000. (B) The resulting geological map after generalization (1 : 250,000).

strategies for some objects in connection with the problem of text symbols and line symbols processing and display. Generally speaking, a geological map should contain at least the following five information layers [26]: • full topographic information (used as the base map); • lithology and its boundary layer (polygon layer, visualized with colors and filling pattern); • point symbols (moraines, boreholes, etc.); • geological structure layer (represented by line symbols); and • related text information. Specific to these experimental data, the map information mainly includes lithological information and construction information. When there is more information to be displayed in the same location, it is necessary to display important objects selectively and eliminate less important objects. In addition, a check for information density should be performed in addition

to labeling, symbolization, and color and fill pattern allocation. Therefore, the visualization of the final geological map is much more complicated than other maps. D. Implementation of Automated Geological Map Generalization In accordance with the compound conceptual framework presented in this paper, we first define the retention catalog and the aggregation hierarchy of the original geological map by considering regional geological background. For example, the stratum is synthesized into a series unit, deleting the faults displaced less than 250 and shorter, less than 1,000 m, and so on. These catalogs and rules are managed as workflow. Then, the lithology polygons and faults are taken as agents with merging, shift, deleting, etc. As the integrated process management module of the planar lithology unit, the lithology agent has the property of area, lithology, degree, etc.



Geological Data

Topographic Data

9

Pattern Information

Category Basis

Operation

map unit hierarchy

-determining map unit -define the priority of map units -pattern recognition

Structure Analysis

map pattern

legible map

-characterization and generalization (eliminate, integration, simplification,smoothing,exaggerat ed,displaceme-nt,reclassification, etc.)

Map Generalization

possible backtrack text color


symbol

-line symbolized -Annotation Processing -color and texture allocation -check of information density

Final Map

Fig. 7. The operation in every stage of the geological map generalization in the experimental zone.

According to the mathematics method, a lithology agent is a four-tuple, A={rule,action,state,communication}. The rule represents the rules that should be followed, such as the area threshold of the reserved polygon. Action is the general designation of behavior in polygon generalization, such as the merger, reservations, simplify to point, and so on. State represents the states of the polygon in the generalization process. Communication stands for the information exchange between different agents. The definition of the lithology agent is shown in Figure 8, the main criterion is a threshold area when the lithology agent is generalized, the area of a lithology is taken as the information exchange communicated with other lithology agents. The fault agent designed with the same method is shown in Figure 9, while the main criterion is a threshold length when the fault agent is generalized, the length of a fault is taken as the information exchange communicated with other fault agents. When the geological map generalization starts, the agents are active. The operations (simplification, delete, merge, etc.) are performed under the constraint of all rules. When the state of the agent satisfies a given threshold, it will be the end of life for the agent. The collaboration of the agents in this geological map is a interactional and restrictional process (Figure 10). It is to deal with the passive or active operation of the element, and the influence or dependent relationship is essential. The detail process can be described as follows. (1) Send map generalization instructions. The relationship

agent sends commands to the fault agent and lithology agent to complete the selection and combination task when it detects the task to be performed. (2) Implementation of the generalization operation. After receiving the instruction, the feature agent queries the spatial relationship through the relationship agent, and then does the operation of deleting fault agent and lithology agent based on the spatial relationship. (3) Complete operation. When the generalization task is finished, the fault agent and lithology agent send the deleting result to the relationship agent. Then the relationship agent will reconstruct the spatial relationship and perform a new detection. In the process, an iterative algorithm is used to combine multiple agents. When the program begins, the agent marked with K11 , T11 , T12 is treated as a seed, and then, the property of the adjacent agent is judged individually. They are merged into one agent if their parent agent is identical. Following that, the new agent is treated as a seed and continues to iterate until it cannot merge. Finally, we model the operations of polygon attribute consolidation, symbol classification, and so on, in order to automate the operation. Figure 6B shows the target geological map at a scale of 1 : 250,000. The overall process not only integrates geological knowledge but also achieves the maximum degree of automation, which indicates that conducting geological map generalization automatically under the guidance of this concept model is effective.



10

< Lithology agent >:: = < rule，action，state，communication > :: = < Area threshold >< Aggregation Rules > < action >:: = < Behavioral repertoire > < state >:: = < Area >< lithology > < communication >:: = < information exchange > Fig. 8. The definition of the lithology agent.

< Fault agent >:: = < rule，action，state，communication > :: = < Length threshold >< Aggregation Rules > < action >:: = < Behavioral repertoire > < state >:: = < property > < communication >:: = < information exchange > Fig. 9. The definition of the fault agent.

RA

Fault FA

Fault OA

Lithology FA

Lithology OA

Select lithology when complete selecting falut Select Falut Query spatial relationship Inform spatial relationship


Inform Selection Inform complete selecting Query spatial relationship Inform spatial relationship


Inform merging result

Fig. 10. The collaboration of agents in experimental geological map.

E. Result Evaluation

TABLE I T HE COMPARISON

Our validation procedure was split into three aspects: validation of hierarchical relationship between geological objects, semantic correctness in generalization results, and the computational time. We employed some of the methods traditionally used to evaluate generalization results. We marked all modified lithology with a distinct color different from the common background. A close examination of it shows that the alterations are semantically correct and they are uniformly spread throughout the experimental region. Then, we compared the generalized fault with the geological structure map of the target scale (1 : 250,000). The latter was produced manually according to well-established nationwide standards and, therefore, can be considered as a point of reference. The results from a one-by-one comparison for the experimental region yielded an overall similarity over 98.3%. This implies that the achieved level of generalization can be considered as appropriate.

OF LITHOLOGY CLASS BETWEEN AUTOMATIC GENERALIZATION RESULTS AND MANUAL GENERALIZATION RESULTS IN TARGET SCALE MAP.

Before generalization

After generalization

In target scale map

K11

K1

K1

K21

K1

K1

K31

K1

K1

T11

T1

T1

T21

T1

T1

T12

T2

T2

r3−1 5

r3−1 5

r3−1 5

The class change of lithology units are presented in Table 1. The table cites the name of objects found on the source map to those in the generalized result. As seen from the table, the



deviation is almost negligible. This implies that the semantic correctness of the generalization result is reliable. Finally, we compared the computational time of the experimental geological map between the proposed approach and a traditional manual approach. The former is 1.3 hour, while the latter is 2.5 days. This shows the proposed approach has a significant advantage in computational speed. The evaluation result shows a better performance of our proposed approach compared with the traditional manual approach. VI. D ISCUSSION AND C ONCLUSION With the continuous development of means of acquiring geological data, the update rate of geological data continues to accelerate. The generalization of geological maps with consistent semantics based on the same geological data is necessary. Due to the complexity of geological objects and geological semantics, the automation of geological maps requires further study. The main barrier is how to keep the hierarchical relation and automation in the process of geological map generalization. This paper put forward a compound conceptual framework for automated generalization of geological maps based on the research results of general map generalization and the characteristics, cartographic rules, and evaluation criteria of geological maps. This compound conceptual framework has the ascendancy of the process-oriented conceptual framework and object-level-oriented conceptual framework. In this conceptual framework, the process is divided into three stages, including the source geologic map structure analysis phase, the source geological maps generalization phase, and the target geological map format standardization phase. In the whole process, the geological objects are abstracted as agents with different properties and behavior, and the agents can communicate with each other when they are activated. Therefore, it can coordinate the relationship of the different map objects in geological map generalization and avoid the conflict between the different levels of operations. Finally, for the introduction of workflow, the conceptual framework improves efficiency of automation for map generalization. The results of the application examples verify the validity of this method. However, automation of geological map generalization is based on automated processing of spatial relationships and large amounts of knowledge rules, which need further research. ACKNOWLEDGMENT The authors would like to thank the anonymous reviewers for the constructive suggestions. R EFERENCES [1] T. C. Downs and W. A, “An integrated approach to the generalization of geological maps,” The Cartographic Journal, vol. 39, no. 2, pp. 137– 152, 2002. [2] L. R. J. Barnes J W, Basic geological mapping. John Wiley & Son, 2013. [3] B.-C. G. F, Geographic information systems for geoscientists: modelling with GIS. Elsevier, 2014.

11

[4] D. J. D. D. M. W. R. W. Clegg P., Bruciatelli, “Digital geological mapping with tablet pc and pda: A comparison,” Computers & Geosciences, vol. 32, no. 10, pp. 1682–1698, 2006. [5] G. C. Jackson I, “Digmapgb-the digital geological map of great britain,” Geoscientist, vol. 13, no. 2, pp. 4–7, 2003. [6] S. Lee and S. Lee, “Map generation and updating technologies based on network and cloud computing: A survey,” International Journal of Multimedia and Ubiquitous Engineering, vol. 8, no. 4, pp. 107–114, 2013. [7] P. I. Lee S, Hwang J, “Application of data-driven evidential belief functions to landslide susceptibility mapping in jinbu,” Korea, no. 100, pp. 15–30, 2013. [8] W. S. KONG Fanqiang, YANG Jun, “Generalization of land use raster data based on semantic,” Bulletin of surveying and Mapping, no. 4, pp. 61–64, 2014. [9] H. Z. SONG Ying, “Knowledge acquisition model for cartographical generalization based on granular computing,” Geomatics and Information Science of Wuhan University, vol. 34, no. 6, pp. 748–751, 2009. [10] S. Steiniger and R. Weibel, “Relations and structures in categorical maps,” The 8th ICA Workshop on Generalisation and Multiple Representation, 2005. [11] C. M. Domingo, “A reference network model for large-scale distribution planning with automatic street map generation,” Power Systems, IEEE Transactions on 26, no. 1, pp. 190–197, 2011. [12] S. R. W. e. a. Grabler F, Agrawala M, Automatic generation of tourist maps. ACM, 2008. [13] S. Lamy, A. Ruas, Y. Demazeau, M. Jackson, W. Mackaness, and R. Weibel, “The application of agents in automated map generalisation,” in 19th international cartographic conference, vol. 13. Citeseer, 1999. [14] C. B. Jones, R. Purves, A. Ruas, M. Sanderson, M. Sester, M. Van Kreveld, and R. Weibel, “Spatial information retrieval and geographical ontologies an overview of the spirit project,” in Proceedings of the 25th annual international ACM SIGIR conference on Research and development in information retrieval. ACM, 2002, pp. 387–388. [15] B. R. Smirnoff A, Paradis S J, “Generalizing surficial geological maps for scale change: Arcgis tools vs. cellular automata model,” Computers & Geosciences, vol. 34, no. 11, pp. 1550–1568, 2008. [16] V. Kaplanoglu, “Multi agent based approach for single machine scheduling with sequence dependent setup times and machine maintenance,” Applied Soft Computing, vol. 23, pp. 165–179, 2014. [17] K. E. Brassel and R. Weibel, “A review and conceptual framework of automated map generalization,” International Journal of Geographical Information System, vol. 2, no. 3, pp. 229–244, 1988. [18] S. Steiniger and R. Weibel, “A conceptual framework for automated generalization and its application to geologic and soil maps,” in Proceedings of XXII Int. Cartographic Conference. Citeseer, 2005. [19] Y.-n. XU, Y. ZHAO, L.-p. LIU, and X.-d. SUN, “R´enyi entropy-based saliency map generation and target detection,” Optics and Precision Engineering, vol. 3, p. 046, 2010. [20] M. Ivanov and J. Head, “Formation and evolution of lakshmi planum, venus: assessment of models using observations from geological mapping,” Planetary and Space Science, vol. 56, no. 15, pp. 1949–1966, 2008. [21] S. Lafay, A. Braun, D. Chandler, M. Michaud, L. Ricaud, and S. Mustiere, “Automatic mapping and innovative on-demand mapping services at ign france,” Cartography and Geographic Information Science, vol. 42, no. 1, pp. 54–68, 2015. [22] J. Renard and C. Duchene, “Urban structure generalization in multiagent process by use of reactional agents,” Transactions in GIS, vol. 18, no. 2, pp. 201–218, 2014. [23] R. N. C. V. D. A. W. B. P. J. M. A. J. Mans, R. S., “Simulation to analyze the impact of a schedule-aware workflow management system,” Simulation, vol. 86, no. 8-9, pp. 519–541, 2010. [24] A. H. A. R. M. Cichocki, A., Workflow and process automation: concepts and technology. Springer Science & Business Media, 2012. [25] S. C. Dominy and I. M. Platten, “Grade control geological mapping in underground gold vein operations,” Applied Earth Science: Transactions of the Institutions of Mining and Metallurgy: Section B, vol. 121, no. 2, pp. 96–103, 2012. [26] M. W. A. Downs T C, “An integrated approach to the generalization of geological maps,” The Cartographic Journal, vol. 39, no. 2, p. 9, 2002. [27] M. S. e. a. Hewson R D, Cudahy T J, “Seamless geological map generation using aster in the broken hill-curnamona province of australia,” Remote Sensing of Environment, vol. 99, no. 1, pp. 159–172, 2005. [28] U. A. M. Grunsky, Eric C. and D. Corrigan, “A study of the lake sediment geochemistry of the melville peninsula using multivariate



12

methods: Applications for predictive geological mapping,” Journal of Geochemical Exploration, vol. 141, pp. 15–41, 2014. [29] A. V. N. R. Stoter J, Post M, “Fully automated generalization of a 1:50k map from 1:10k data,” Cartography and Geographic Information Science, vol. 41, no. 1, pp. 1–13, 2014.

JUNQIANG ZHANG received the M.S. and Ph.D. degrees in Geoscience information engineering from China University of Geosciences, Wuhan, Hubei, China, in 2012 and 2015, respectively. From 2015 , he was a Postdoctoral Fellow with Institute of Remote Sensing and Digital Earth, Chinese Academy of Sciences, Beijing, China.His research interests include processing of geoscience data and automatic drawing of geological maps.

CHONGLONG WU obtained B.S.(1968) and M.S.(1982) degrees in China University of Geosciences. He worked in Guizhou Metallurgical Bureau from 1968 to 1973 and Sichuan steel plant from 1973 to 1979. He is professor on Geoscience Information Engineering since 1991 at China University of Geosciences. His current research works mainly include theory and methods of geology and mineral resources information systems, digital land engineering, mining exploration and development of information technology, simulation of geological processes.

LIZHE WANG received the B.E. and M.E. (magna cum laude) degrees from Tsinghua University, and the D.Eng. degree from the University of Karlsruhe, Germany. He is currently a Professor with the Institute of Remote Sensing and Digital Earth, Chinese Academy of Sciences, and a ChuTian Chair Professor with the School of Computer Science, China University of Geosciences. His main research interests include high-performance computing,e-science, and spatial data processing. He serves as an Associate Editor of the IEEE TRANSACTIONS ON COMPUTERS and the IEEE TRANSACTIONS ON CLOUD COMPUTING. He is a fellow of IET and the British Computer Society.


A Conceptual Framework for the Automated ...

A Conceptual Framework for the Automated ...

Suggest Documents

a conceptual framework for automated generalization and ... - CiteSeerX

A Conceptual Framework for Adaptationâ

A CONCEPTUAL FRAMEWORK FOR SUSTAINABLE ...

A conceptual framework for the neurobiological ... - WordPress.com

A Conceptual Framework for the Design of

DEVELOPING A CRM CONCEPTUAL FRAMEWORK FOR THE ...

Developing a Conceptual Framework for Considering the ...

A Conceptual Framework for the Analysis ...

Towards a conceptual framework for understanding the ...

A Conceptual Framework for Understanding the

Towards a conceptual framework for the