a computational framework to generate sidewalk and road network representations from primitive geospatial information toward conflictless passage and traffic safety

Bạn đang xem bản rút gọn của tài liệu. Xem và tải ngay bản đầy đủ của tài liệu tại đây (5.09 MB, 80 trang )

<span class="text_page_counter">Trang 1</span><div class="page_container" data-page="1">

<i><b>---HUỲNH LÊ PHÚ TRUNG </b></i>

<b>A COMPUTATIONAL FRAMEWORK TO GENERATE SIDEWALK AND ROAD NETWORK </b>

<b>REPRESENTATIONS FROM PRIMITIVE GEOSPATIAL INFORMATION TOWARD CONFLICTLESS PASSAGE </b>

<b>AND TRAFFIC SAFETY </b>

Major: Engineering Physics Major code: 8520401

MASTER’S THESIS

HO CHI MINH CITY, January 2024

</div><span class="text_page_counter">Trang 2</span><div class="page_container" data-page="2">

Supervisor 1: PhD. Pham Tan Thi Supervisor 2: Prof. Hiroaki Wagatsuma

Examiner 1: Assoc. Prof. Cao Thanh Tinh

Examiner 2: PhD. Nguyen Trung Hau

This master’s thesis is defended at HCM City University of Technology, VNU- HCM City on January 28<small>th</small>, 2024

Master’s Thesis Committee:

1. Chairman: Assoc. Prof. Huynh Quang Linh 2. Secretary: PhD. Nguyen Xuan Thanh Tram 3. Examiner 1: Assoc. Prof. Cao Thanh Tinh 4. Examiner 2: PhD. Nguyen Trung Hau 5. Member: Prof. Phan Bach Thang

Approval of the Chair of Master’s Thesis Committee and Dean of Faculty of Applied Physics after the thesis being corrected (If any).

<b><small>CHAIR OF THESIS COMMITTEE DEAN OF FACULTY OF APPLIED SCIENCE </small></b>

</div><span class="text_page_counter">Trang 3</span><div class="page_container" data-page="3">

<i><b>THE TASK SHEET OF MASTER’S THESIS </b></i>

Full name: Huỳnh Lê Phú Trung Student ID: 1970503 Date of birth: January 15, 1997 Place of birth: Tiền Giang Major: Engineering Physics Major ID: 8520401

<b>I. THESIS TITLE (In Vietnamese): Một hệ thống tính tốn xây dựng mạng lưới vỉa hè và đường bộ từ thông tin không gian địa lý nguyên thủy hướng tới lối đi thơng thống và an tồn giao thơng</b>

<b>II. THESIS TITLE (In English): A Computational Framework to Generate Sidewalk and Road Network Representations from Primitive Geospatial Information Toward Conflictless Passage and Traffic Safety</b>

<b>III. TASKS AND CONTENTS:</b>

• Reconstruction of the sidewalk network.

• Reconstruction of the road network to assist the finding of conflict areas between vehicles and sidewalk passengers.

• Detection of conflict areas as a consistent extension of the road network.

<b>IV. THESIS START DAY: February 2023</b>

<b>V. THESIS COMPLETION DAY: December 2023VI. SUPERVISOR </b>

<b>Supervisor 1: PhD. Pham Tan Thi Supervisor 2: Prof. Hiroaki Wagatsuma </b>

SUPERVISOR HEAD OF DEPARTMENT

Hiroaki Wagatsuma

DEAN OF FACULTY OF APPLIED SCIENCE

</div><span class="text_page_counter">Trang 4</span><div class="page_container" data-page="4">

and Professor Hiroaki Wagatsuma for their guidance, good advices, great support and even challenges. I have learned a lot from them that is very valuable for both my knowledge and personal growth. I would not complete this thesis without their immense support. They always guilds me to the right direction. In addition, I would like to send my thanks to all members of Wagatsuma laboratory for their support and for the great time I have been there.

</div><span class="text_page_counter">Trang 5</span><div class="page_container" data-page="5">

national organization responsible for surveying and mapping the national land of Japan, various geospatial information has been prepared by different public/private organizations and those were provided depending on individual purposes. Since geospatial representations respectively have a certain degree of accuracy and quality, those are beneficial, while there exist differences in descriptions and measurements. For the solution of such inconsistencies, GSI have organized the standard description of the geospatial information called “basic map information (BMI),” and maintained by collecting update information of those organizations on road and building constructions. Due to the circumstances, BMI is an authentic data and compositive data simultaneously, which is not always suitable for routing information of moving entities. In other words, edges of building, road and sidewalk are very accurate in the sense of combinations of primitive lines in BMI; however, a single building, connected road and consistent sidewalk are represented in multiple lines and there are no guarantees to link together depending on the meaningful unit semantically.

In the aim to solve the problem, a computational framework was proposed in this study as an effective integration of methods in geometric computation. Model-free approaches ignoring geometric properties are not effective in the sense of computational costs. On the other hand, too much empirical and heuristic approaches may provide a specialized solution, but it does not provide general perspectives and contribute less to a new discovery in science and advancements in engineering fields. Therefore, a general scheme was introduced to form individual solutions in the form of an algorithm of computation, which are provided by an integration of matrix operations and methods in geometric computation. The proposed methods were evaluated in missions of reconstructions of sidewalk and road networks and applied to the detection of conflict areas vehicles and sidewalk passengers such as pedestrians and wheelchairs.

</div><span class="text_page_counter">Trang 6</span><div class="page_container" data-page="6">

trách nhiệm khảo sát và lập bản đồ quốc gia Nhật Bản, nhiều thông tin không gian địa lý khác nhau đã được các tổ chức công hoặc tư nhân khác nhau chuẩn bị và chúng được cung cấp tùy theo mục đích cá nhân. Vì các sự biểu diễn trong khơng gian địa lý có mức độ chính xác và chất lượng nhất định nên mặc dù bên cạnh các lợi ích thì vẫn tồn tại những khác biệt trong mơ tả và đo lường. Để giải quyết những mâu thuẫn như vậy, GSI đã thiết lập một sự mô tả tiêu chuẩn về thông tin không gian địa lý được gọi là “thông tin bản đồ cơ bản (BMI)” và được duy trì bằng cách thu thập thơng tin cập nhật của các tổ chức về đường bộ và cơng trình xây dựng. Mặc dù BMI là dữ liệu tổng hợp được xác thực, không phải lúc nào cũng phù hợp để sử dụng trong thông tin định tuyến của các thực thể chuyển động. Nói cách khác, các cạnh của tịa nhà, đường và vỉa hè rất chính xác theo nghĩa kết hợp các đường nguyên thủy trong dữ liệu BMI, tuy nhiên một tòa nhà riêng biệt, những đường được kết nối và các vỉa hè được thể hiện bằng nhiều đường và khơng có gì đảm bảo tính liên kết với nhau thể hiện ý nghĩa về mặt ngữ nghĩa. Với mục đích giải quyết vấn đề, một khung tính tốn đã được đề xuất trong nghiên cứu này, là sự tích hợp một cách hiệu quả của các phương pháp trong hình học tính tốn. Các phương pháp tiếp cận khơng có mơ hình bỏ qua các đặc tính hình học sẽ khơng hiệu quả về mặt chi phí tính tốn. Mặt khác, q nhiều phương pháp tiếp cận theo kinh nghiệm có thể cung cấp một giải pháp chuyên biệt, không cung cấp những giải pháp tổng qt và ít đóng góp cho những khám phá mới về khoa học và những tiến bộ trong lĩnh vực kỹ thuật. Do đó, một sơ đồ được đề xuất trong nghiên cứu này, hình thành các giải pháp riêng lẻ dưới dạng thuật toán, được cung cấp bằng cách tích hợp các phép tốn ma trận và các phương pháp trong hình học tính tốn. Phương pháp đề xuất đã được đánh giá trong các nhiệm vụ tái thiết mạng lưới vỉa hè và đường bộ, và áp dụng để phát hiện các khu vực xung đột giữa các phương tiện và đối tượng di chuyển trên vỉa hè như người đi bộ và xe lăn.

</div><span class="text_page_counter">Trang 7</span><div class="page_container" data-page="7">

I commit that this thesis “A Computational Framework to Generate Sidewalk and Road Network Representations from Primitive Geospatial Information Toward Conflictless Passage and Traffic Safety” is my research with support from my supervisors, PhD. Pham Tan Thi and Professor Hiroaki Wagatsuma. The research content and results have never been published in any previous research works.

Thesis author

</div><span class="text_page_counter">Trang 8</span><div class="page_container" data-page="8">

Chapter 2 Basement Technologies ... 5

2.1 Geographic Information System (GIS) ... 5

2.1.1 GIS data types ... 5

2.1.2 Coordinate System in GIS ... 6

2.2 OpenStreetMap (OSM) ... 8

2.2.1 OpenStreetMap data format ... 8

2.2.2 Sidewalk information in OSM ... 9

2.3 Basic Map Information from GSI ... 12

2.3.1 How to download map data for GSI ... 12

2.3.2 BMI data format ... 15

Chapter 3 Methods for Sidewalk Data Management ... 18

3.1 Sidewalk Network Reconstruction ... 18

3.1.1 Problems in BMI data to describe sidewalks ... 18

3.1.2 Proposed Method 1: Node connection algorithm to integrate consistent segments in the sidewalk data ... 20

3.1.3 Proposed Method 2: Flipping algorithm to represent a consistent sidewalk ... 24

</div><span class="text_page_counter">Trang 9</span><div class="page_container" data-page="9">

3.2.1 Extraction of road edge data ... 30

3.2.2 Geometric computation for target edge detection ... 31

3.3 Road network reconstruction ... 37

3.4 Conflict area identification ... 39

Chapter 4 Results and Discussion ... 43

4.1 Sidewalk network reconstruction ... 43

4.1.1 Integration of sidewalk data to represent a consistent path ... 43

4.1.2 Comparison of proposed methods ... 45

4.2 Sidewalk area reconstruction ... 46

4.2.1 Extraction of road edge extraction ... 46

4.2.2 Geometric computation for target edge detection ... 47

4.3 Road network reconstruction ... 52

4.3.1 Geometric computation for parallel line detection ... 52

4.3.2 Validation of road network reconstruction ... 56

4.4 Conflict area identification ... 58

Chapter 5 Conclusion ... 60

Publication ... 61

References ... 62

</div><span class="text_page_counter">Trang 10</span><div class="page_container" data-page="10">

Figure 2.2. Raster data ... 6

Figure 2.3. Geographic coordinate system (left) and Projected coordinate system (right) in GIS ... 7

Figure 2.4. OpenStreetMap data of an area of Kitakyushu, Japan ... 9

Figure 2.5. OpenStreetMap with sidewalk data ... 10

Figure 2.6. Three types of data from GSI ... 13

Figure 2.7. Select the area for download ... 14

Figure 2.8. Select the data for dowload ... 15

Figure 2.9. All data of basic items ... 15

Figure 2.10. BMI data displayed in QGIS ... 16

Figure 2.11. XML format of BMI data ... 17

Figure 3.1. Data structure of line segments for sidewalk representation in BMI data .. 19

Figure 3.2. Sidewalk is represented by separated line segments... 20

Figure 3.3. The flowchart of node connection algorithm ... 20

Figure 3.4. Sidewalks connect each other through end points ... 21

Figure 3.5. Sidewalk data of original BMI is displayed in MATLAB ... 21

Figure 3.6. List of end points ... 22

Figure 3.7. Connectivity matrix of line segments in BMI sidewalk ... 23

Figure 3.8. Connectivity between two line segments ... 24

Figure 3.9. Direction in line segments of BMI sidewalk ... 25

Figure 3.10. Flipping direction for the consistent line ... 26

Figure 3.11. Connection types of line segments in BMI data ... 28

</div><span class="text_page_counter">Trang 11</span><div class="page_container" data-page="11">

Map) ... 30

Figure 3.14. The idea for sidewalk area reconstruction ... 31

Figure 3.15. The concept of geometric computation for route maps ... 31

Figure 3.16. The sidewalk polygon reconstruction model ... 32

Figure 3.17. The target of sidewalk area reconstruction ... 33

Figure 3.18. Problems in sidewalk area reconstruction ... 34

Figure 3.19. Conditions to extract the target part for sidewalk area reconstruction ... 35

Figure 3.20. The design of algorithm for sidewalk area reconstruction ... 36

Figure 3.21. The road network reconstruction model ... 37

Figure 3.22. The design of algorithm for the road network reconstruction ... 38

Figure 3.23. Scenario 1 of overlapping of sidewalk and vehicle path (photos from Google Map and Street View) ... 39

Figure 3.24. Scenario 2 of overlapping of sidewalk and vehicle path (photos from Google Map and Street View) ... 40

Figure 3.25. Scenario 3 of overlapping of sidewalk and vehicle path (photos from Google Map and Street View) ... 40

Figure 3.26. Scenario 4 of overlapping of sidewalk and vehicle path (photos from Google Map and Street View) ... 41

Figure 3.27. The idea for conflict management ... 41

Figure 3.28. The design of algorithm for conflict area detection ... 42

Figure 4.1. The part 1 of map of original BMI data ... 43

Figure 4.2. The part 1 of map after applying proposed method ... 44

Figure 4.3. The part 2 of map of original BMI data ... 44

Figure 4.4. The part 2 of map after applying proposed method ... 45

</div><span class="text_page_counter">Trang 12</span><div class="page_container" data-page="12">

Figure 4.7. Sidewalk area reconstruction result for part 1 of the map ... 48

Figure 4.8. Sidewalk area reconstruction result for part 2 of the map ... 49

Figure 4.9. Sidewalk area reconstruction result for part 3 of the map ... 49

Figure 4.10 Sidewalk area reconstruction result for part 4 of the map ... 50

Figure 4.11. Sidewalk area reconstruction result for part 5 of the map ... 50

Figure 4.12. Sidewalk area reconstruction result for part 6 of the map ... 51

Figure 4.13. Sidewalk area reconstruction result for the whole map ... 51

Figure 4.14. Road network reconstruction result for part 1 of the map ... 52

Figure 4.15. Road network reconstruction result for part 2 of the map ... 53

Figure 4.16. Road network reconstruction result for part 3 of the map ... 53

Figure 4.17. Road network reconstruction result for part 4 of the map ... 54

Figure 4.18. Road network reconstruction result for part 5 of the map ... 54

Figure 4.19. Road network reconstruction result for part 6 of the map ... 55

Figure 4.20. Road network reconstruction result for the whole map ... 55

Figure 4.21. The road map from GSI [35] ... 56

Figure 4.22. The ground truth of road network from GSI in the pixel representation .. 57

Figure 4.23. The result of road network reconstruction in pixel representation ... 57

Figure 4.24. The validation result of road network reconstruction ... 58

Figure 4.25. Actual scene of the overlapping area in reconstructed data (photos from Google Map and Street View) ... 59

Figure 4.26. The results of conflict area detection ... 59

</div><span class="text_page_counter">Trang 13</span><div class="page_container" data-page="13">

Table 2.1. The statistic of highways and sidewalks in OSM dataset ... 10 Table 3.1. Connection types in BMI data ... 27 Table 4.1. Computational time (s) of Method 1 and Method 2 ... 45

</div><span class="text_page_counter">Trang 14</span><div class="page_container" data-page="14">

GSI Geospatial Information Authority of Japan

GIS Geospatial Information System

</div><span class="text_page_counter">Trang 15</span><div class="page_container" data-page="15">

The high-definition (HD) map is certainly one of the key technologies for all major automated driving projects [2]. The HD map can provide precise definition of road connected with semantic information for high-precision localization in 3D spaces [3], understanding the surroundings and making prediction for other vehicles’ movement [4], maneuver themselves [5]. In response to developments in the field of automated driving. The maps need to be refined more and more precisely to meet the high-quality requirements. For highly automated driving, vehicles have to understand complex environment about not only their lanes, but also other elements of map such as sidewalks, markings, traffic sign, … For the purpose of prediction of other road users, it is required for vehicles to get knowledge of all lanes where they can move to, especially true for sidewalk passenger, since their movement behaviors is difficult to predict with only using sensors [6]. Therefore, it is necessary to provide very detailed sidewalk information in maps to ensure the safety for automated driving.

Sidewalk is a vital infrastructure to provide safe paths for pedestrian and wheelchair users in their daily life for different purposes such as walking, transportation, economic

</div><span class="text_page_counter">Trang 16</span><div class="page_container" data-page="16">

development and environment improvement. Especially for people with disabilities, sidewalks play an important role in all their physical activities. Sidewalks are desirable to support them highly safety and mobility by reducing walking along roadway crashes. Roadways with sidewalks are 88.2% lower than as likely to have pedestrian crashes as them without sidewalks [7].

1.2 Related Past Works

In consideration of effective utilizations of geospatial information, automated preprocessing methods, computational tools and utilization frameworks have been proposed [8-30].

In order to conflict management for pedestrians and wheelchair for people with disabilities on the subject of automated driving, sidewalk information needs to be described in the clear and precise way. Nowadays, most of the map data is developed for use in the road networks that do not includes the sidewalk information [23]. There are many shortcomings in map system related to application for pedestrians and wheelchairs. Compare to the vehicle traveling in the road, pedestrians and wheelchair have more degrees of freedom in movement, thus they need a precise instruction to get the destination [24]. Sidewalk data is the key for routing and navigation for pedestrians and wheelchairs. Therefore, several studies focused on modeling sidewalk network to provide effective assistance for pedestrians and people with disabilities [25][26][27].

OpenStreetMap (OSM) is one of the available open map data that is widely used for application in routing and navigation. However, the sidewalk information still has shortcomings in the OSM data [28]. In previous study done by Mobasheri et al. [29], OSM and GPS traces data were employed to extract the geometry of sidewalk and construct sidewalk network. On the other hand, the raster data is also used for sidewalk extraction from street view imagery data source [30].

</div><span class="text_page_counter">Trang 17</span><div class="page_container" data-page="17">

1.3 Research Objectives

The purpose of the study is to provide a computational framework as the set of algorithms to solve crucial problems for reconstruction of navigational information for sidewalk passengers from the authentic primitive geospatial information. Particularly, following missions are target problems in this study:

⚫ Reconstruction of the sidewalk network

⚫ Reconstruction of the road network to assist the finding of conflict areas between vehicles and sidewalk passengers

⚫ Detection of conflict areas as a consistent extension of the road network

In the aim to solve the problem, a computational framework was proposed as an effective integration of methods in geometric computation. As proposed methods, a general scheme was introduced to form individual solutions in the form of an algorithm of computation, which are provided by an integration of matrix operations and methods in geometric computation. The proposed methods were evaluated in missions of reconstructions of sidewalk and road networks and applied to the detection of conflict areas vehicles and sidewalk passengers such as pedestrians and wheelchairs.

The thesis outcome is to validate proposed methods to be able to apply to the real geospatial data. Particularly, as a part of Kitakyushu city was selected to be the validation area, it included two square kilometers around the Wakamatsu Campus, Kyushu Institute of Technology known as Hibikino area, which is a well-constructed suburb having sidewalks and roads with variety of widths. In the computer experiment, proposed methods demonstrated that a significant reduction of the computation time in comparison to the simple model-fee approach for node-connection in the case of sidewalk network reconstruction, about 80% accuracy in the road network reconstruction and provide a solution to the detection of conflict areas vehicles and sidewalk passengers as a consistent extension of the road network.

</div><span class="text_page_counter">Trang 18</span><div class="page_container" data-page="18">

Those results will open doors for designs and operations of effective model-based solutions to treat geospatial information, which intermediate between model-free approaches and heuristic approaches in conditions of minimization of the computational cost and accuracy of the representation.

1.4 Document structure

This thesis will present the research in a meticulous structure designed for easy location of important information and reducing the possibility of missing information. The thesis is presented as follows:

⚫ Chapter 2: Basement Technologies gives basic insight on the GIS data and BMI data using in this Thesis. This data will be used as material for sidewalk reconstruction and conflict management.

⚫ Chapter 3: Proposed Methods for Sidewalk Data Management describes the development of the methods for individual missions of reconstructions of sidewalk and road networks and applied to the detection of conflict areas vehicles and sidewalk passengers

⚫ Chapter 4: Results and Discussion and describes the results and the validation of proposed methods. The limitation of proposed methods is also presented in this chapter.

<small>⚫ </small> Chapter 5: Conclusion and Future Perspectives summarizes the

accomplishments and contributions of the study and provides future perspectives for the research.<small> </small>

</div><span class="text_page_counter">Trang 19</span><div class="page_container" data-page="19">

Chapter 2 Basement Technologies

Basement technologies and materials were described in this chapter．

2.1 Geographic Information System (GIS)

2.1.1 GIS data types

A geographic information system (GIS) is a computer system for creating, analyzing, managing, displaying and storing geographical information. GIS connect all descriptive data to a map and integrate the location of them. Location can be represented in different ways such as address, latitude and longitude, … GIS can describe many kinds of information data in map such as streets, rivers, buildings and vegetation. GIS enables people to understand objects, relationship and geographic context. GIS provides a tool for mapping and analysis in different fields of sciences and technologies. GIS data is basically divided into two types: vector data [31] and raster data [32].

Vector data is usually used to represent real world features such as roads, trees, buildings, and so on. Three basic types of vector data are points, lines and polygons. A point feature has and is described by its X, Y and optional Z coordinate. The value of X and Y depend on the Coordinate Reference System being used. The Z value is used to describe the height. The point feature is used to describe geometry that consists of a single vertex such as trees, traffic lights, … A sequence of connected points is formed a polyline feature. The polyline feature represents road marking, curbs, … Polygon feature is also a sequence of connected points; however, the first vertex and last vertex are the same. Polygon feature describe geometry of lakes, buildings, …

Figure 2.1. Vector data

</div><span class="text_page_counter">Trang 20</span><div class="page_container" data-page="20">

Raster data is composed of a matrix of pixels where each pixel represents a specific geographical feature. Raster data is used to display features that are difficult to represent using vector data such as elevation, temperature.

Figure 2.2. Raster data

There are particular advantages and disadvantages of two forms of data. Since vector data represents data using vertices and paths, it is considered to be a good method for cartographic representation that gives high geographic accuracy. However, vector data is not very efficient to use for displaying continuous data. Raster data can be an efficient solution for continuous information representation due to its pixelated organization. While linear feature and paths are difficult to represent using raster data.

2.1.2 Coordinate System in GIS

A coordinate system is a system that use numbers for identifying the location of a point or geometric element on the earth. Spatial data is defined in both horizonal and vertical coordinate system. Longitude and latitude commonly used in coordinate system There are two types of horizonal coordinate systems: geographic coordinate system and projected coordinate system.

Geographic Coordinate System (GCS) uses a three-dimensional spherical surface to define the position on the Earth. GCS typical describe the location in decimal degrees that include degree of longitude and degree of latitude. The longitude ranges from +180<small>o</small>to -180<small>o</small> with the East-West direction, while latitude ranges from +90<small>o</small> to -90<small>o</small> with the North-South direction.

</div><span class="text_page_counter">Trang 21</span><div class="page_container" data-page="21">

Projected Coordinate System (PCS), as opposed to GCS, is a flat system for use dimensional representation of the Earth. The longitude and latitude coordinates are converted to planar coordinate x and y. The center horizontal line with East-West direction is referred to as the x axis and the center vertical line with North-South direction is referred to as the y axis. The intersection between x and y axis is the origin of coordinate system that has the coordinate of (0, 0). The horizontal lines above the origin have the positive value and those below the origin have the negative value. Similarly, the vertical lines in the right side of origin have the positive value and those in the left side of origin have the negative value. The value of coordinate in PCS is converted from three-dimensional Earth’s surface to two-dimensional flat surface using mathematical formulas. The transformation is referred to as map projection. Currently, there are many map projections is being used in around the world. Depending on the purpose for which the map will be used, a particular map projection is considered for suitable application.

two-Figure 2.3. Geographic coordinate system (left) and Projected coordinate system (right) in GIS

</div><span class="text_page_counter">Trang 22</span><div class="page_container" data-page="22">

2.2 OpenStreetMap (OSM)

2.2.1 OpenStreetMap data format

OpenStreetMap (OSM) is a Volunteered Geographic Information (VGI) project to create map of the world that is free to use. It is considered as the Wikipedia project for maps, where the community can contribute to build the maps around the world about the roads, trails, and much more. Currently, OpenStreetMap has more than 8 million registered users and growing everyday. As it stands now, OpenStreetMap has a high accuracy that is able to provide map data for numerous of websites and hardware devices. Whole world dataset of OpenStreetMap is free to access as an XML file called Planet.osm that is updated on weekly basis.

The OpenStreetMap data consists of the following three primitive data types:

- Node, which is defined as a point in the space associated with a node identifier, latitude and longitude coordinates.

- Way, which represents a line between two nodes, and associated with the way identifier and the two nodes identifiers of the two end points of the line. The line could be simply a road segment, part of a boundary of a building, city/country boundary, or part of a lake contour.

- Relation, which represents the relation between nodes, ways, or even other relation, and is used to express polygons. For example, to express the boundaries of a building, the nodes need to be defined, then the ways that connect nodes to each other, then a relation that connects the ways together to describe the building boundary. Since the Openstreetmap is a collaborative project where any volunteer can contribute, one building may be expressed in various relations depending on contributors. The relation could be separated or nested where each relation is composed of either ways or nodes.

All OSM elements (nodes, ways, relations) can have tags that describe the meaning

</div><span class="text_page_counter">Trang 23</span><div class="page_container" data-page="23">

of elements. A tag consists of two fields: a key and a value. The key is used to describe category or feature type. The value is used to describe the detail of corresponding key. Both fields are free form text. For example, the tag “highway = trunk” with a key “highway” and a value “trunk” is used to indicate the important highway in country’s system that are not a motorway; the tag “name = 本城バイパス” is used to describe the name of the highway.

Figure 2.4. OpenStreetMap data of an area of Kitakyushu, Japan 2.2.2 Sidewalk information in OSM

There are two ways to map sidewalks in OSM data: mapping sidewalks as separate way or mapping sidewalks as refinement of highways.

In the first method, a sidewalk could be a way that is separate from highway. This way is drawn by contributor and added tags “highway = footway” + “footway = sidewalk” to indicate a sidewalk. Disadvantages of this method is to link the sidewalk with associated road autonomously. A mapper adds the name of associated road as a tag into the line of sidewalk. Additionally, the “relation” tags are also required to assign road

</div><span class="text_page_counter">Trang 24</span><div class="page_container" data-page="24">

name to sidewalk.

In the second method, sidewalks are mapped by adding tag “sidewalk = ?” into the sections of an existing road. The values of this tag could be “both/ left/ right/ no” to indicate the position of the sidewalk relative to the road. In addition, the property of sidewalk is able to be added as a sub-tag in the road such as width, surface of the sidewalk or bicycle permission on the sidewalk. The drawback of this method is that detailed geometric information of sidewalks is not provided because sidewalks are attached with the main road and there is no standard view in OSM for the sidewalk data.

Figure 2.5. OpenStreetMap with sidewalk data

However, since OSM is a project that volunteers around the world could contribute, it is difficult to build a map data in a consistent way and there are still many lacks of information of sidewalk system. Table 2.1 gives the statistics of highway and sidewalk information in OSM data.

Table 2.1. The statistic of highways and sidewalks in OSM dataset Number of highways Number of sidewalks as

the highway refinement as the separated way <sup>Number of sidewalks </sup>> 198 millions > 2.5 millions > 2.6 millions

</div><span class="text_page_counter">Trang 25</span><div class="page_container" data-page="25">

As seen in the Table 2.1, number of highways is much larger than the total number of sidewalks generated by two above approaches. This statistic indicates that there is still many lacks of sidewalk information in the OSM which is considered as one of the largest open map data all over the world. Therefore, the enrichment and management of sidewalk data using for the map and especially for application in automated driving is the attractive issues in the research field of automated driving and mapping.

</div><span class="text_page_counter">Trang 26</span><div class="page_container" data-page="26">

2.3 Basic Map Information from GSI

As mentioned in the previous section, although OpenStreetMap is one of the largest open map data over the world, sidewalk still has lack of information to representing in OSM. Therefore, another open map data is considered for sidewalk reconstruction in this thesis.

The Geospatial Information Authority of Japan (GSI) is the national organization that conducts basic survey and mapping and instructs related organizations to clarify the conditions of land in Japan. In this study, the basic map information provided by GSI is used to reconstruct sidewalk geometry for pedestrian and wheelchair movements and manage the conflict in the scenarios of sidewalk traffic.

2.3.1 How to download map data for GSI

The data can be downloaded from the official website of GSI. There are three types of available basic map information: (1) basic items including benchmark for surveying, coastline, boundaries and representative points of administrative divisions, road edge, outline of building, … (2) digital elevation model and (3) geoid model. For the main purpose of this study, only the data of basic items is considered.

The steps to download are showed as follow.

Step 1: Access to the website of GSI and click the button shown in the red box for the data of basic items as shown in Figure 2.6.

</div><span class="text_page_counter">Trang 27</span><div class="page_container" data-page="27">

Figure 2.6. Three types of data from GSI <small> </small>

Basic Item

Digital Elevation

Geoid Model

</div><span class="text_page_counter">Trang 28</span><div class="page_container" data-page="28">

Step 2: A map of Japan appear and is divided into multiple areas. Select downloaded area by scrolling to enlarge and clicking the number of area for download the map data.

Figure 2.7. Select the area for download

In Figure 2.7, the selected box “503065” is the data of Kitakyushu area that is used in this Thesis. Then click the button “ダウンロードファイル確認へ” to check the download file. Note that the BMI data download service requires user registration.

Step 3: Download Basic Map Information data as Figure 2.8. The data contains man types of basic item of map data shown in Figure 2.9 that can be used for various research purposes. In this study, the Road composition and Road edge data will be used to analyze.

</div><span class="text_page_counter">Trang 29</span><div class="page_container" data-page="29">

Figure 2.8. Select the data for dowload

Figure 2.9. All data of basic items

The basic items of BMI data include various types of road components such as roads, sidewalks, rails, coastlines, … Each type is stored in a separated XML file that can be used for different application purposes. The target of this thesis is construction of sidewalks and conflict management between pedestrians/wheelchairs and vehicles traveling in the roads, thus two considered data types of basic items are road compositions which consists of sidewalk data and road edge for road reconstruction. 2.3.2 BMI data format

The road composition data consists of four types: sidewalks (歩道), median strips (分

</div><span class="text_page_counter">Trang 30</span><div class="page_container" data-page="30">

離帯), gutters (側溝), stormwater inlets (雨水桝). Sidewalks and median trips are displayed as the green line in map

The Road edge data consists of four types: straight roads (真幅道路), walking paths (徒歩道), road tunnels (トンネル内の道路), garden roads (庭園路等). Figure 2.10 is an example of an area with BMI data displayed in QGIS software [3].

Figure 2.10. BMI data displayed in QGIS

BMI data is stored in XML format. Coordinates of points forming line is stored in the posList. Additional information is also displayed such as time, name of office, … Unlike OSM data, the line in BMI data is not described by separated points with unique ID. The list coordinate included in data of lines will indicate the position of lines in map.

</div><span class="text_page_counter">Trang 31</span><div class="page_container" data-page="31">

Figure 2.11. XML format of BMI data

</div><span class="text_page_counter">Trang 32</span><div class="page_container" data-page="32">

Chapter 3 Methods for Sidewalk Data Management

Proposed methods for sidewalk data management and conflict detection were described in this chapter.

3.1 Sidewalk Network Reconstruction

3.1.1 Problems in BMI data to describe sidewalks

In the original BMI data, line segments are used to describe sidewalks. Problems for this representation are shown as follow.

⚫ As mention in previous chapter, the BMI data is stored in XML format for each line segments. the position of line segments in the map is described using a list of coordinates.

⚫ Furthermore, a same sidewalk is represented by multiple lines that make difficult to determine the connection of sidewalk and road networks.

Therefore, it makes difficult to manage the sidewalk as a consistent path to detect connection of sidewalks each other and between sidewalks and roads. The data structures of line segments in BMI to describe sidewalk information is showed in Figure 3.1, Each line segment consists of a list of coordinates that define the location of the points belong to line segments. For example, line a with n points will be represent with the structure of coordinates {(ax1, ay1), (ax2, ay2), (ax3, ay3), … (axNa, ayNa)} as seen in Figure 3.1a. Similarly, this structure is applied to line b, line c and all the line segments in the dataset.

In preparation of sidewalk network reconstruction, data of each line segment will be processed in the concept of matrix representation. Each coordinate now corresponds to a point on the line segment that extends from the starting point to the ending point as shown Figure 3.1b.

</div><span class="text_page_counter">Trang 33</span><div class="page_container" data-page="33">

Figure 3.1. Data structure of line segments for sidewalk representation in BMI data According to the necessity, it is required to connect all the road segments to form a consistent line representing the sidewalk. Sidewalk data need to be managed in the right way for fine treatment in detection of conflict areas between pedestrians and vehicles.

</div><span class="text_page_counter">Trang 34</span><div class="page_container" data-page="34">

3.1.2 Proposed Method 1: Node connection algorithm to integrate consistent segments in the sidewalk data

Since nodes in BMI data do not have unique ID, this makes difficult to determine whether two lines connect or not. Therefore, node connection is necessary to represent sidewalk consistently. Figure 3.2 shows an example of how the sidewalk is represented in BMI data. The same sidewalk (pink line) is divided into multiple line segments (blue, orange and green lines). The flowchart of node connection algorithm is shown in Figure 3.3.

Figure 3.2. Sidewalk is represented by separated line segments

Figure 3.3. The flowchart of node connection algorithm

</div><span class="text_page_counter">Trang 35</span><div class="page_container" data-page="35">

Firstly, the small area is selected for easy treatment. From the BMI data of this area, the line segments are extracted and stored as a data in MATLAB. Each line segments will be sequentially assigned a unique ID value for data management. As seen in format of BMI line data, each line segment consists of a list of nodes with coordinate (latitude and longitude). Line segments connects to each other through their two end points that are showed in Figure 3.4. Therefore, the positions of end points are necessary to describe the connectivity between line segments.

Figure 3.4. Sidewalks connect each other through end points

Figure 3.5. Sidewalk data of original BMI is displayed in MATLAB

According to the line data in BMI, the coordinates of end nodes shown in Figure 3.5

</div><span class="text_page_counter">Trang 36</span><div class="page_container" data-page="36">

are then extracted and stored as an array in MATLAB. Each line is represented by two end points that are then used for describe the connectivity between line segments.

Figure 3.6. List of end points

End points are determined to be coincident based on the array that are showed in Figure 3.6. If two points have the same coordinate (latitude and longitude), they are considered as connecting point of two lines which they belong to. Two those lines have connectivity. This connectivity is stored in a connectivity matrix of MATLAB as shown in Figure 3.7.

</div><span class="text_page_counter">Trang 37</span><div class="page_container" data-page="37">

Figure 3.7. Connectivity matrix of line segments in BMI sidewalk

In Figure 3.7, the connectivity matrix contains two values 0 and 1 that describe the connectivity between line segments through end points. When two nodes A and B have the same coordinate, the value of position [endnodeA, endnodeB] in the matrix will be set to 1 that means the line having node A is connected to the line having node B. In contrast, the value 0 describes that there is no connection between line segments. After that position of connection nodes is extracted to use for connect lines together as Figure 3.8.

</div><span class="text_page_counter">Trang 38</span><div class="page_container" data-page="38">

Figure 3.8. Connectivity between two line segments

Subsequently, two lines having the connectivity will be merged through the common point. The algorithm is iterative executed and thoroughly checked in numerous times until the sidewalks are no longer represented by many separated line segments. The consistent sidewalk data then is stored and managed as the new structure with the unique identification and position of connected nodes.

3.1.3 Proposed Method 2: Flipping algorithm to represent a consistent sidewalk

As introduced above, the first method works well in the sense of the minimization of the data management procedure. However, it does not satisfy the requirement of the sidewalk description in senses of the accuracy and the fine data management to manage conflict problems that will be described in the following sections.

Therefore, an improvement of the algorithm is considered for sidewalk network

</div><span class="text_page_counter">Trang 39</span><div class="page_container" data-page="39">

reconstruction in the purpose of reducing computational time, also fine management for conflict issues in the concept of matrix operations. As mentioned, the connection algorithm uses end points of line segments for connectivity detection and then merging line segments into a consistent path to describe sidewalk for BMI data. Nevertheless, The BMI data is well-refined data to form constructions in space by linked together partially, and line segments for sidewalk representation are also constructed like that.

The sidewalk in original BMI data can be described by one or many line segments. The important point is that these line segments are constructed one by one using directional line without a certain orientation. An example of line segments that do not follow in one direction can be seen in Figure 3.9.

Figure 3.9. Direction in line segments of BMI sidewalk

Each line segment representing the sidewalk is directed to management by a matrix that includes its nodes with coordinates. For example, line segment A treated as a matrix

ax1 ay1ax2 ay2ax3 ay3…

bx1 by1bx2 by2bx3 by3…

cx1 cy1cx2 cy2cx3 cy3…

</div><span class="text_page_counter">Trang 40</span><div class="page_container" data-page="40">

with coordinate [

𝑎𝑥1 𝑎𝑦1𝑎𝑥2 𝑎𝑦2

] , line sement B treated as a matrix with coordinate

𝑏𝑥1 𝑏𝑦1𝑏𝑥2 𝑏𝑦2

] and line sement B treated as a matrix with coordinate [

𝑐𝑥1 𝑐𝑦1𝑐𝑥2 𝑐𝑦2

] connect each other to form the same sidewalk. However, they are not described in the same direction in BMI data that make difficult to establish connection between them.

In consideration of the solution, a flipping algorithm is necessary for generally connecting line segments and management of sidewalk data in a consistent way. Moreover, the sidewalks are stored in the context of line data and each line segment will be assigned a unique ID for effective management.

Figure 3.10. Flipping direction for the consistent line

In the analysis of line segments in BMI data, the connection between two lines is divided into four types of connectivity as showed in the Table 3.1. The coordinates of all points are converted from longitude/latitude to equal-area coordinate x and y. The