Geographical Data Management for Web Applications

Web Information Systems
prof. Athena Vakali

Geographical Data Management for
Web Applications
Symeon Papadopoulos

Informatics Department, AUTh
13 Dec 2011

Overview

• Introduction
• Representation & storage of geographical data
• Geographical data management with Java (JDBC)
• Text Indexing for Geographical Data
• Google Maps API
• Use Cases: Geographical Data Clustering + Heat maps


The Era of Geo

• 350K websites make use of Google Maps API (2010)

• GPS-enabled mobile phones > 200M (2010)

• Foursquare users: 5M (2010)  15M (2011)

• The HotPotato check-in service was acquired by
Facebook for $10Μ (2010)


Categories of Geo-Services

• Primary services (providers)
– Google Maps, Bing Maps, Yahoo Maps, MapQuest,
OpenStreetMap

• Derivative services (mash-ups)

• Location-based social networks
– Foursquare, Gowalla (acquired by Facebook), Brightkite

• Intelligent geo-based services
– Analytics, Route planning, Real estate


GIS

• GIS offer sophisticated software for geographical
data management and processing.
• They offer many means of representation, search
and rendering.
• They mostly target geo-centered applications:
– Urban planning, environmental studies, etc.
• Popular tools: ArcGIS, MapInfo, GeoMedia, κλπ.
• GIS will not be discussed in this lecture.


Representation and Storage of
Geographical Data

Lat, Lon

• Geo coordinates constitute the basic element of any
geo-enabled service.

• A coordinate element is defined by the following:
Latitude Longitude


Lat, Lon (ΙΙ)

• Traditional writing:
40° 38′ 0″ N, 22° 57′ 0″ E

• Decimal representation: 40.633333, 22.95

• Important note for storage:
Precision of 0.00001 in the decimal representation
corresponds to approximately 1m distance (near the
equator)


Altitude, Bearing

• Two additional variables related to geographical
points (still not very common):
– Altitude: vertical distance of point from the level of sea
– Bearing: direction to which the user/device is pointed

• The increasing use of GPS-enabled devices is
expected to increase the importance of these
variables, especially in Augmented Reality (AR)
applications.


Distance calculation (lat1, lon1)

d
• Haversine formula:
– Δlat = lat2 – lat1, Δlon = lon2 – lon1
– a = sin2(Δlat/2) + cos(lat1)  cos(lat2)  sin2(Δlon/2) (lat2, lon2)
– c = 2  atan2(sqrt(a), sqrt(1-a))
– d=Rc [R: earth radius (6 371 km)]

• In JavaScript: More accurate and computational expensive
– var R = 6371; // km alternative: Vincenty formula
– var dLat = (lat2-lat1).toRad();
– var dLon = (lon2-lon1).toRad();
– var a = Math.sin(dLat/2) * Math.sin(dLat/2) +
Math.cos(lat1.toRad()) * Math.cos(lat2.toRad()) *
Math.sin(dLon/2) * Math.sin(dLon/2);
– var c = 2 * Math.atan2(Math.sqrt(a), Math.sqrt(1-a));
– var d = R * c;


Geo in DB

• Representation in table form
• e.g. List of Points of interest  Table of Pois

• double  slightly faster
• DECIMAL  higher accuracy Field type:
double ή DECIMAL(13,9)


Use cases

• Query 1: What points of interest can I see near point
of interest X?
– Online city exploration in cities (e.g. recommend new Pois
based on proximity)

• Query 2: What points of interest are there near me?
– Valuable within location-based applications (e.g. I need
recommendations for sightseeing, eating, etc. based on
my current location)


Geo in DB – Proximity Queries (I)

• Implementation 1 – Compute distance by use of SQL and
return ordered results:
• set @lat0 = 40.645466; set @lon0 = 22.858342;
set @dist = 10;
• SELECT *, 6371  2  ASIN ( SQRT( POWER(
SIN(@lat0 – abs(dest.lat))  pi()/180 / 2), 2) +
COS (@lat0  pi()/180 )  COS(abs(dest.lat)  pi()/180)  POWER(
SIN(@lon0 – dest.lon)  pi()/180 / 2), 2) )) as distance
FROM pois dest
HAVING distance < @dist Haversine formula

ORDER BY distance LIMIT 10;

Geo in DB – Proximity Queries (II)

• Problems:
– Painfully slow!
For tables of even modest sizes (~1000), response might
take as long as 8-10 seconds.

– Distance computations are not reused in successive
invocations of the same query, since caching computations
is not supported

– Not an option for web applications.


Geo in DB – Proximity Queries (IIΙ)

• Alternative implementation:
– If, instead of nearby Pois, we ask for Pois that lie within a
certain radius from a fixed point (e.g. current location)?

300m

Aghia 300m
Sophia


Geo in DB – Proximity Queries (IV)

• Bounding box query:
– SELECT * FROM pois WHERE
pois.latitude > @min_lat AND pois.latitude < @max_lat
AND pois.longitude > @min_lon AND
pois.longitude < @max_lon;
(@max_lat, @max_lon)

– The question now is how
to determine
min_lat, min_lon,
max_lat, max_lon.

(@min_lat, @min_lon)


Geo in DB – Proximity Queries (Vα)

• Bounding box computation (top and bottom):
• 1°difference in latitude ~ 111km (110.567 near the
equator, 111.699 near the poles) 
– 300m  ~0.0027° latitude (since our fixed point is
located at approximately 45°)
@max_lat
– @min_lat = @lat0 – 0.0027
– @max_lat = @lat0 + 0.0027

e.g. for lat0 = 40.632844 (Aghia Sophia) @lat0
@min_lat ~ 40.630144
@max_lat ~ 40.635544
@min_lat


Geo in DB – Proximity Queries (Vβ)

• Bounding box computation (left and right):
• 1°difference in longitude  depends a lot on the latitude of
the fixed point, e.g ~111.320 km near the equator and 0 near
the poles.
Distance (km) for 1°
– Table with distance correspondences difference in longitude
(the more lines, the better the achieved
approximation – for intermediate values
one needs to use interpolation)
– 300m  ~0.003805° latitude since the
fixed point lies at latitude~45°
– @min_lon = @lon0 – 0.003805
– @max_lon = @lon0 + 0.003805
– e.g. for lon0 = 22.947094 (Aghia Sophia) @max_lon
@min_lon ~ 22.943289
@max_lon ~ 22.950899 @min_lon @lon0


Geo in DB – Proximity Queries (VI)

• In summary:
– During the table creation, we make sure to index the
latitude and longitude columns (conventional B-tree).

– We convert the proximity query to a bounding box query
by use of geographical approximations (that can be made
sufficiently precise if needed).

– In this form, the SQL query can be executed really fast
thanks to the indexes.


Geo-enabled DBs

• PostGIS (Extension for PostgreSQL)
– Support for “geographical objects”:
• POINT (X Y)
• LINESTRING (0 0, 1 1, 1 2)
• POLYGON (0 0, 4 0, 4 4, 0 4, 0 0)
• MULTIPOINT, MULTIPOLYGON, GEOMETRYCOLLECTION
– Support for tables and calculations (e.g. proximity queries)
directly through the query engine.

• GeoServer (GIS)
– Complete Java framework for search of geographical
objects and map rendering based on a server-client
architecture.


Geographical data management in
Java (JDBC)

Typical lifecycle of geographical data

Relational model
specification

JDBC

Geo data source Transformation to
RDBMS storage
(e.g. Wikimapia) records
JDBC
UPDATE READ

Processing

JAVA


Data model specification

• Very simple example: table with id + 3 columns
– Name: VARCHAR (100)
– Latitude: DOUBLE
– Longitude: DOUBLE

Pois


Input data + transformation

parsing
<name>Benaki museum</name>  “Benaki Museum”
<lon>23.70825</lon>  23.70825
<lat>37.974515</lat>  37.974515

Record
name latitude longitude
Benaki museum 23.70825 37.974515


Record insertion with JDBC (1)

• Step 1 Connection Creation
(A) Load the appropriate JDBC driver
E.g. in the following example: JDBC_DRIVER = “com.mysql.jdbc.Driver”
Class.forName(JDBC_DRIVER);

(B) Connection initialization
Τhe variable dbURL has the form: protocol + vendor + server + port nr + dbName
+ params
e.g. jdbc:mysql//myserver.gr:3306/testdb/?
useUnicode=true&characterEncoding=UTF-8
Connection con = DriverManager.getConnection(dbURL, username, password);

throws exceptions (InstantiationException, IllegalAccessException,
ClassNotFoundException, SQLException)

ATTENTION: Connection objects should be reused!


Record insertion with JDBC (2)

• Step 2 Creation and execution of statement
(A) Statement creation
String sqlInsert = “INSERT INTO Pois VALUES (‘Benaki museum’,
23.70825, 37.974515)”;
Statement insertStmt = con.createStatement();

(B) Statement execution (throws SQLException)
insertStmt.executeUpdate(sqlInsert);

Alternatively
PreparedStatement prepInsert = con.prepareStatement(sqlInsert,
Statement.RETURN_GENERATED_KEYS);
prepInsertStmt.executeUpdate();
ResultSet rs = prepInsertStmt.getGeneratedKeys();
long recordId = -1;
Record database key  could
if (rs.next()) recordId = rs.getLong(1);
be useful later (better keep it
than query for it)

Wrapper method creation

• Encapsulation of process for easy reuse
public long insertPoi(Connection con, String poiName, double lat, double lon) {
String sqlInsert = “INSERT INTO Pois VALUES (‘” + poiName + “’,”
String.valueof(lat) + “,” + String.valueof(lon) + “)”;
PreparedStatement prepInsert = con.prepareStatement(sqlInsert,
Statement.RETURN_GENERATED_KEYS);
prepInsertStmt.executeUpdate();
ResultSet rs = prepInsertStmt.getGeneratedKeys();
long recordId = -1;
if (rs.next()) recordId = rs.getLong(1);
return recordId;
}


Batch insertion

• In case of massive insertions, it is better to make use of the
batch mode of insertion in order to significantly speed up
the process.

con.setAutoCommit(false); Very important!!
Statement batchInsert = con.createStatement();
for (int i = 0; i < 10,000; i++){
batchInsert.addBatch(“INSERT INTO Pois VALUES (‘“ +
names*i+ + “’,” + String.valueof(lats*i+) + “,” + String.valueof(lons*i+) + “)”;
}
int[] updateCount = batchInsert.executeBatch();
con.commit();
con.setAutoCommit(true);


ResultSet structure

Initial cursor position C1 C2 C3 C4 C5 C6 C7 ….
R1 22 ab 0.1
R2 27 cd 0.5
rs.next() R3 28 gc 1.2
R4 40 ob 3.5
R5 51 gz 3.2

rs.absolute(3)
rs.getInt(1) rs.getDouble(3)
rs.getInt(“C1”) rs.getDouble(“C3”)

rs.getString(2) (faster)
rs.getString(“C2”)


Record reading

• Use SELECT statement and iterate through ResultSet
(A) SELECT statement creation
Statement selectStmt = con.createStatement();

(B) Form and execute SELECT statement
List<Poi> pois = new ArrayList<Poi>();
String sqlSelect = “SELECT * FROM Pois”;
ResultSet rs = stmt.executeQuery(sqlSelect);
while ( rs.next() ) {
long poiId = rs.getLong(“id”);
String poiName = rs.getString(“name”);
double lat = rs.getDouble(“latitude”); Candidate wrapper method
Poi extractPoiFromResultSet(ResultSet rs)
double lon = rs.getDouble(“longitude”);
Poi poi = new Poi(poiId, poiName, lat, lon);
pois.add(poi);
}


Record updates

• Suppose I have collected data about the popularity of a Poi
(e.g. count the number of results from a search engine) and I
have its id at my disposal:
public void updatePoiPopularity(Connection con, long poiId, int popularity) {
PreparedStatement updateStatement = con.prepareStatement(
“UPDATE Pois SET poi_popularity = ? WHERE id = ?”);
updateStatement.setInt(1, popularity);
updateStatement.setLong(2, poiId);
int n = updateStatement.executeUpdate();
}
NEW COLUMN
name latitude longitude poi_popularity
…. … … …


Example Design WikimapiaWrapper PopWrapper

PoiDbWrapper
• Task separation
List<Poi> pois = WikimapiaWrapper.collectPois(“Thessaloniki”);
for (int i = 0; i < pois.size(); i++){
long poiId = PoiDbWrapper.insertPoi(con, pois.get(i).getName(),
pois.get(i).getLatitude(), pois.get(i).getLongitude());
pois.get(i).setId(poiId);
}
….
Map<Poi,Integer> poiPopMap = PopWrapper.collectPoiPopularities(pois);
for (Entry<Poi,Integer> poiPopEntry : poiPopMap.entrySet()){
PoiDbWrapper.updatePoiPopularity(con,
poiPopEntry.getKey().getId(), poiPopEntry.getValue());
}


Text Indexing for Geographical Data

Full-text Indexing

• Extremely valuable technology for applications requiring full-
text search features.

• Classical DB systems do not offer a standard means of
indexing and searching large amounts of text.

• Established technology: Inverted indices

• Popular implementations:
– Lucene, Solr (Java)
– Sphinx (C++)
– OpenFTS (PostgreSQL)
– Lemur (C++)


Text queries in RDBMS

• Operators LIKE and RLIKE
– SELECT * FROM pois WHERE name LIKE ‘%museum%’
– SELECT * FROM pois WHERE name RLIKE ‘museum’

• Conventional indexing structures in RDBMS are not
particularly helpful in quickly responding to such queries
(might require visiting every record  too slow for large
tables)

• Each vendor offers own solution, e.g. FULLTEXT INDEX από
MySQL.


Inverted index (I)

• Starting point: Document collection (corpus)

• Text processing 
Dictionary: set of unique terms appearing in the
documents of the collection + additional information
(stats, stems, stop words)

• Index: Document retrieval mechanism based on the
terms contained in the documents


Inverted index (II)

• Inverted index: Set of inverted lists.

• Inverted list: A list containing information about the
occurrence of a term within the documents of the collection –
represented as a list of postings.

• Posting: Occurrence of a term in a certain position of a
document. Apart from the document, where the term occurs,
the posting also contains its position in the text  support for
phrase search (increases the size of the index and complexity
of the query mechanism).


Inverted index (IIΙ)

Example
• Document collection:
– Document 1: “White Tower”
– Document 2: “Byzantine Museum”
– Document 3: “Archeological Museum”

• Inverted lists of postings:
– “Archeological”: {(3,1)}
– “Byzantine”: {(2,1)}
– “Museum”: {(2,2), (3,2)}
– “White”: {(1,1)}
– “Tower”: {(1,2)}
Document id
Term position within the document

Inverted index (IV)

Query example
• Query: “Museum”  documents 2 and 3

• Query: “Byzantine Museum” 
– Result 1: document 2 (contains both terms)
– Result 2: document 3 (contains only the term “museum”)

• “Archeological”: {(3,1)}
• Document 1: “White Tower” • “Byzantine”: {(2,1)}
• Document 2: “Byzantine Museum” • “Museum”: {(2,2), (3,2)}
• Document 3: “Archeological Museum” • “White”: {(1,1)}
• “Tower”: {(1,2)}


Solr (text search engine)

• Complete framework for indexing and searching massive
document collections.

• Based on the Lucene library.

• Flexible document representation structure.

• Flexible query structure.

• Can be installed as a servlet in an appropriate container (e.g.
Tomcat, Jetty)  Possibility for use over HTTP


Solr basics (Ι)

• Every document consists of a set of fields (similar to Lucene)

• Every field can be of different type (e.g. text, number, etc.)

• Every field can processed (e.g. tokenized, normalized) and
indexed in a different way.

• Every field or combination of fields can be queried in a
different way  flexible query system


Solr basics (ΙΙ)

Example:
– id: 13424
– name: White Tower of Thessaloniki
– latitude: 40.626369
– longitude: 22.948428
– history: The tower, which once guarded the eastern end of the city's sea
walls, was for many years attributed to Venice, to which the Byzantines ceded
Thessaloniki in 1423.
– wikipedia_link: http://en.wikipedia.org/wiki/White_Tower_of_Thessaloniki

Xml representation suitable fοr use by Solr:
<add><doc>
<field name=“id”>13424</field>
<field name=“name”>White Tower of Thessaloniki</field>
...
</doc></add>


Solr: index management

• Addition of documents: java -jar post.jar lefkos_pirgos.xml
– The xml file should follow the structure of the example in the previous
example.
– A list of files can be provided as input (whitespace separated).
– If a document with the same id (unique key) is provided as input, then
the new document will replace the existing.

• Deletion of documents:
java -Ddata=args -Dcommit=no -jar post.jar
"<delete><id>13424</id></delete>"

• Committing changes: java –jar post.jar

• The above operations (and more) are available through a Java
(not only) API as well.

Solr: Queries

HTTP API: Base URL: http://server:8983/solr/select/?

Return fields “name”, “id” of all entries containing ‘museum’.
• q=museum&fl=name,id

As above but also return relevance score for each entry.
• q=museum&fl=name,id,score

As above but return all stored field.
• q=museum&fl=*,score

Return entries in JSON format.
• q=museum&wt=json

Return all entries containing the keyword ‘pirgos’ in their field “name”.
• q=name:pirgos


Challenges

• Text indexing and search frequently encounters the following
two challenges:

– Synonyms, writing variations: users search for something using a form
different than the “reference” form.
e.g. White Tower, Λευκός Πύργος, λευκός πύργος, lefkos pyrgos, ... 
Solution: Attempt to collect and index many different lexical forms of
the same term.

– Polysemy/Ambiguity: A single term refers to more than one concepts.
e.g. the term museum may refer to a number of real-world museums
Solution: Appropriate interfaces are necessary to make the user
aware of the different senses/contexts of a term and ask for query
reformulation/specification (could also come up with suggestions).


Synonyms

• Strategy 1 (simple but not always available):

– Use dictionary/thesaurus/gazetteer (e.g. Geonames)
Contains alternative names and variations for a large number of Pois
and locations.

– Addition of a “synonyms” field in the index

– Include the additional field in the search
…but give a higher weight to the “name” field.


Related terms

• Strategy 2 (more complex but generally applicable):
– Issue query to search engine or other content application (e.g. Flickr)
• Use Poi name as query.

– Process search results in two ways:
• Find terms or term sequences that frequently occur in the same context (e.g. web
page, title of photo, etc.).
• Find terms that are lexically similar to the query term (e.g. based on the
Levenshtein distance) .

– Addition of a “similars” field in the index.
– Include the additional field in the search
• but give higher weight to the “name” field


Licensing

• The Google Maps service is free for web pages that
are freely available. Charges apply for commercial
usage of the API.

• Since V3, the API does not require a key.


Google Maps API (I)

• The map service is based on the notion of map tiles,
which, when stitched together, provide an
impression of a continuous map.


Google Maps API (II)

• There are different types of map types, e.g. ROADMAP,
SATELLITE, HYBRID, TERRAIN. Those are called base map tiles.
ROADMAP HYBRID

HYBRID TERRAIN


Google Maps API – Coordinates (I)

Coordinates:
• World: Point 0,0  lat ~ 85°
lon 180°

• Pixel: pixelCoord = worldCoord  2zoomLevel

The API first determines the visible area of the map on the client
(viewport) based on the zoom level, the map center and the
size (in pixels) of the DOM element in the client web page.


Google Maps API – Coordinates (II)

Coordinates:
• Tile coordinates:
– With every scroll or zoom to a new viewport, the API needs to render
a new area on the map. For this to happen, it needs to translate the
pixel coordinates into appropriate tile coordinates in order to load the
appropriate tiles. At each zoom level the tile coordinates are different.

– e.g. at zoom level 2 (beside) 16 tiles
need to be indexed. The tile containing
Greece has coordinates (2,1).
At zoom level 3, 16 x 4 = 64 tiles need to
to be indexed, and so on.


The “hello world” of Google Maps
• The map object is integrated to any web page as a DOM element.
<html>
<head>
<meta name="viewport" content="initial-scale=1.0, user-scalable=no" />
<style type="text/css"> html { height: 100% } body { height: 100%; margin: 0px; padding: 0px }
#map_canvas { height: 100% }</style> Library import
<script type="text/javascript" src="http://maps.google.com/maps/api/js?sensor=false"></script>
<script type="text/javascript">
function initialize() { Map initialization and centering
var latlng = new google.maps.LatLng(40.66, 22.95);
var myOptions = { zoom: 12, center: latlng,
mapTypeId: google.maps.MapTypeId.ROADMAP };
var map = new google.maps.Map(document.getElementById("map_canvas"), myOptions);
}
</script>
</head>
<body onload="initialize()"> Map canvas DOM element
<div id="map_canvas" style="width:100%; height:100%"></div>
</body>
</html>
Spans the whole page body


Basic objects

• google.maps.LatLng
– Initialization: var latlng = new google.maps.LatLng(lat_p, lon_p);
– Very useful object, e.g. for marker placement.
– In case we don’t know the geo coordinates, but know the street name,
we could make use of a Geocoding service.

• google.maps.Map
– Initialization: var mapx = new google.maps.Map(domElement,
options);
– Basic parameters: zoom level, center, mapType (ROADMAP, HYBRID,
etc.)
– It is possible to have more than one map objects on the same page.


Zoom level

• Very important to select the proper zoom level depending on
the application needs.
zoom = 4 zoom = 8 zoom = 12

zoom = 16 zoom = 18 zoom = 20


More basic objects

• google.maps.Marker
– Initialization:
var latlng = new google.maps.Marker({
position: latlng,
map: mapx,
title: “POI Title”
});

• Very useful object!
• Use it for marking Pois.


Events on the map UI

• Interacting with the map, as in every UI, triggers a
variety of events.

• There are two basic event categories:
– User/UI events: Objects on the map respond to certain
user actions.
– MVC events: Triggered by changes in the state of the map,
e.g. change of value for some parameter (zoom level).


Map events

• UI events
Look like common DOM events but are separately triggered in
order to prevent incompatibility with different browsers.

e.g. the Marker object responds to the following events:
– “click”,
– “dblclick”,
– “mouseup”,
– “mousedown”,
– “mouseover”,
– “mouseout”.


Event listening (I)

• addListener()
– Object that “listens”
– event that is “listened” MVC event
– Method that is invoked
as soon as the event occurs
• Example:
google.maps.event.addListener(map, 'zoom_changed', function() {
setTimeout(moveToDarwin, 3000); // fire moveToDarwin() after 3 secs
});
google.maps.event.addListener(marker, 'click', function() {
map.setZoom(8);
}); UI event

function moveToDarwin() {
var darwin = new google.maps.LatLng(-12.461334, 130.841904);
map.setCenter(darwin);
}

Event listening (II)

• Argument passing
– Only for UI events
– Access properties and their values as in any JavaScript object

• Example:
google.maps.event.addListener(map, 'click', function(event) {
placeMarker(event.latLng);
});
function placeMarker(location) {
var marker = new google.maps.Marker({
position: location, Marker creation at the point clicked by
map: map the user.
});
map.setCenter(location); Map centering on the same point.
}


Event listening (IΙI)

• Read and write properties within MVC events
– The value that we read may be different than the value at the time of
event triggering (usually we care only for the current value).

• Example:
google.maps.event.addListener(map, 'zoom_changed', function() {
zoomLevel = map.getZoom();
alert("Zoom: " + zoomLevel);
if (zoomLevel == 0) {
map.setZoom(10);
}
});


Event listening (IV)

• Listen to common DOM events
– addDomListener()
window, document.body, etc.
DOM element
• Example: named elements
<script>
event name
function initialize() {
// map initialization method to be invoked
}
google.maps.event.addDomListener(window, 'load', initialize);
</script>
<body>
<div id="map_canvas"></div>
</body>


Control elements (I)

• UI elements that enable users to interact with the
viewport on the map.

• Navigation control: pan & zoom

• Scale control

• MapType control


Control elements (II)

• Determine active controls by use of approprate arguments in the map
object options (only at create time):
{
navigationControl: boolean,
mapTypeControl: boolean,
scaleControl: boolean
}

• Example:
function initialize() {
var myOptions = {
zoom: 4,
center: new google.maps.LatLng(-33, 151),
navigationControl: false,
scaleControl: true,
mapTypeId: google.maps.MapTypeId.ROADMAP
}
var map = new google.maps.Map(document.getElementById("map_canvas"),
myOptions);
}


Control elements (IΙI)

• Control elements offer additional options:
– google.maps.NavigationControlStyle.SMALL  mini-zoom form
– google.maps.NavigationControlStyle.ZOOM_PAN  standard form of zoom & pan
– google.maps.MapTypeControlStyle.DROPDOWN_MENU  select map types from
dropdown menu

• Example:
var myOptions = {
...
mapTypeControl: true,
mapTypeControlOptions: {
style: google.maps.MapTypeControlStyle.DROPDOWN_MENU
},
navigationControl: true,
navigationControlOptions: {
style: google.maps.NavigationControlStyle.SMALL
},
....
}
var map = new google.maps.Map(document.getElementById("map_canvas"),
myOptions);


Control elements (IV)

• Control elements can be placed in different positions on the
map:

navigationControl: true,
navigationControlOptions: {
style: google.maps.NavigationControlStyle.ZOOM_PAN,
position: google.maps.ControlPosition.TOP_RIGHT
}


Control elements (V) custom control

• It is possible to create custom
control elements.
• Control elements are
essentially div elements.
• In order to create them, you need:
– Define their appearance by use of CSS.
– Determine their interactions by use of event handlers.
– Create div element and add it to controls property of the
Map object.


Overlays

• Objects with specific coordinates.
Markers Icons

Polylines Polygons InfoWindow


Overlays - Markers

• Markers are initialized with their position, the map,
and a title as arguments (the title appears in a tooltip
upon mouse hover).

• It is also possible to animate markers with two
predefined animation schemes by use of the method
marker.setAnimation(method):
– google.maps.Animation.DROP: Drops and stops.
– google.maps.Animation.BOUNCE: Bounces.
– null: Static.


Overlays - Icons

• Markers with a different look.
• Set the icon property.
• Example:
var image = `beachflag.png’;
var myLatLng = new google.maps.LatLng(44.890, 22.27);
var beachMarker = new google.maps.Marker({
position: myLatLng,
map: map,
icon: image
});
• Options for more complex appearance (e.g. by use of
shadow and stacking).

Polylines

• Useful for presenting routes, directions, etc.
• Example:
var flightPlanCoordinates = [
new google.maps.LatLng(37.772323, -122.214897),
new google.maps.LatLng(-18.142599, 178.431),
new google.maps.LatLng(-27.46758, 153.027892)
];
var flightPath = new google.maps.Polyline({
path: flightPlanCoordinates,
strokeColor: "#FF0000",
strokeOpacity: 1.0,
strokeWeight: 2
});
flightPath.setMap(map);


Polylines - Processing

• It is possible to dynamically modify a polyline.

• Read and modify the points in a line:
– var path = poly.getPath();
– var latlng_idx = path.getAt(idx)  read element on
position idx
– path.insertAt(idx)  insert element in position idx
– path.removeAt(idx)  delete element on position idx
– path.push(latlng)  add element at the end of the line


Polygons

• Polygons cover an area.
• Example:
var triangleCoords = [
new google.maps.LatLng(25.774252, -80.190262), Same start and
new google.maps.LatLng(32.321384, -64.75737), end coordinate
new google.maps.LatLng(25.774252, -80.190262) (optional)
];
bermudaTriangle = new google.maps.Polygon({
paths: triangleCoords,
strokeColor: "#FF0000",
strokeOpacity: 0.8,
strokeWeight: 2,
fillColor: "#FF0000",
fillOpacity: 0.3
});
bermudaTriangle.setMap(map);
• Same behaviour as Polylines with respect to processing their elements
(e.g. getPath(), etc.).

Info Window (Ι)

• Very important for providing information for a Poi
upon click.
• May contain: (a) simple text, (b) HTML code, or (c)
DOM element.
• Initialized by use of the following elements:
– content: String (simple text/HTML), DOM node
– position: If the Info Window is to be opened on top of a
marker, then the marker position is provided.
– maxWidth: Maximum width in pixels (autowraps text).


Info Window (ΙΙ)

• Example:
var contentString = '<div id="content">'+
'<div id="siteNotice">'+ '</div>'+
'<h1 id="firstHeading" class="firstHeading">Uluru</h1>'+
'<div id="bodyContent">'+
'<p><b>Uluru</b>, also referred to as <b>Ayers Rock</b>, is a large ' +
'sandstone rock formation in the southern part of the '+
'Northern Territory, central Australia. It lies 335 km (208 mi) '+ .... + </p>'+
'<p>Attribution: Uluru, <a
href="http://en.wikipedia.org/w/index.php?title=Uluru&oldid=297882194">‘ +
'http://en.wikipedia.org/w/index.php?title=Uluru</a> (last visited June 22, 2009).</p>‘ +
'</div>‘ + '</div>';

var infowindow = new google.maps.InfoWindow({
content: contentString
});
var marker = new google.maps.Marker({
position: myLatlng,
map: map,
title:"Uluru (Ayers Rock)«
});
google.maps.event.addListener(marker, 'click', function() {
infowindow.open(map,marker);
});


Layers

• Collections of information that are overlaid on top of
base map tiles and modify its appearance. They are
treated as a single object.

• Existing layers:
– KmlLayer, GeoRSS
• Presentation of feed of geotagged posts on a map.
– TrafficLayer
– BicyclingLayer


Geocoding

• Service for converting between street names and
geographical coordinates.
• Rate limit: 2500 calls / day
• Base URL
http://maps.googleapis.com/maps/api/geocode/output?par
output: json / xml
par 
required: address OR latlng, sensor
optional: bounds, region, language


Use Cases:
Geographical Data Clustering + Heat Maps

Geo-Clustering (1)

• Represent each location p = (lat, lon) as a feature
vector: x = [x1 x2]

• Apply some density-based clustering method.
– Prefer methods that do not require setting the number of
clusters as parameter (e.g. k-means is not really
appropriate). Recommend methods: BIRCH and DBSCAN
(they require setting other parameters).


Geo-Clustering (2)

• Example of using the BIRCH algorithm on a set of
geotagged Flickr photos in the vicinity of
Thessaloniki.

CENTER +
40 ΕKKLISIES +
ANO POLI

NEA PARALIA
KARABOURNAKI
ARETSOU

d = 0.0015 d = 0.0030


Geo-Clustering (3)

• Example of using clustering in a web application.

http://www.clusttour.gr

markers + polygons + infowindow


Heat maps

• A visual analytics tool.

• Useful for visualizing location-dependent quantities,
e.g. population density, housing prices, crime rates,
pollution, etc.

• Requires knowledge or inference of quantity
distribution over space.

• Overlay on top of the map.


Heat maps

• Example: geotagged photos collected from Flickr based on the
query “sziget festival”)

GRID


Heat maps

• Counting frequency in each cell + smoothing based on
adjacent cells + color coding


Heat maps (1)

Languages in Twitter

http://flowingdata.com/2011/10/27/language-communities-of-twitter/


http://flowingdata.com/2011/12/07/every-death-on-the-

Heat maps (2)
road-in-great-britain/

Deadly car accidents in the UK


http://www.flickr.com/photos/walkingsf/5925793753/in/set-72157627140310742

Heat maps (3)

La Grande Arche de la Defense

Versailles

Orly Airport
Twitter (blue) +
Flickr (orange)


http://tagmaps.research.yahoo.com/worldexplorer.php

Tag maps Rome

London


Thank you!
Symeon Papadopoulos
papadop@iti.gr

Geographical Data Management for Web Applications

Empfohlen

Empfohlen

Weitere ähnliche Inhalte

Ähnlich wie Geographical Data Management for Web Applications

Ähnlich wie Geographical Data Management for Web Applications (20)

Mehr von Symeon Papadopoulos

Mehr von Symeon Papadopoulos (20)

Kürzlich hochgeladen

Kürzlich hochgeladen (20)

Geographical Data Management for Web Applications