(GIS導論-22)Address Geocoding and Geotagging！

版權宣告：
(1)ArcView, ArcGIS為ESRI公司所有, http://www.esri.com/index.html, 編寫過程使用NTU GIBP授權版本。
(2)Google Earth為google公司所有, 編寫過程使用Google Earth Free版本, 1600 Amphitheatre Parkway., Mountain View, CA., 94043 USA. http://earth.google.com/。(以下簡稱GE)。

參考網站：
(1)The University of Washington, ESRM 250.
網址：http://gis.washington.edu/esrm250/main/syllabus.html
(2)麻省理工學院「開放式課程網頁」主要課程列表
網址：http://www.myoops.org/twocw/mit/Global/all-courses.htm
(3)Geographic information system - Wikipedia, Retrieved on 2008-08-01.
網址：http://zh.wikipedia.org/w/index.php?title=%E9%A6%96%E9%A1%B5&variant=zh-tw
(4)Christopher J. Date, An Introduction to Database Systems, -  Wikipedia, the free encyclopedia
網址：http://en.wikipedia.org/wiki/Christopher_J._Date

Geotagging, is the process of adding geographical identification metadata to various media such as websites, RSS feeds, or images and is a form of geospatial metadata. This data usually consists of latitude and longitude coordinates, though it can also include altitude, bearing, and place names.

Sometimes the term geocoding is also used for the geotagging process, though it is generally referred to as being the process of taking non-coordinate based geographical identifiers, such as a postal address, and assigning geographic coordinates to them (or vice versa).

Geotagging can help users find a wide variety of location-specific information. For instance, one can find images taken near a given location by entering latitude and longitude coordinates into a Geotagging-enabled image search engine. Geotagging-enabled information services can also potentially be used to find location-based news, websites, or other resources.

Geotag techniques

The base for geotagging is positions. The position will in almost every case, be derived from the global positioning system, and based on a latitude/longitude-coordinate system that presents each location on the earth in a coordinate spanning from 180° west through 180° along the Equator and 90° north through 90° south along the prime meridian.

GPS-formats

GPS co-ordinates may be represented in text in a number of ways, with more or fewer decimals:

Using geotagging

Geotagging works two ways: First it tells people rather precisely where the content of a given media is located, but on some media platforms (such as Google Earth) it also gives the reverse ability: Showing people relevant media to a given location.

JPEG-photos

With JPEG-photos the geotag-information is typically embedded in the metainformation (typically as EXIF-data or XMP-data. These data isn't visible in the picture it-self, but are read and written by special programs and most digital cameras and modern scanners. The format used is degrees with decimals, and plus/minus indicator and can be read in detail by many programs, i.e. the cross-platform freeware tool ExifTool. A interpreted readout for a photo would look like this:
GPS Latitude : 57 deg 38' 56.83" N
GPS Longitude : 10 deg 24' 26.79" W
GPS Position : 57 deg 38' 56.83" N, 10 deg 24' 26.79" W

While the more uninterpreted EXIF-data looks like this:
GPSLatitude : 57.64911
GPSLongitude : 10.40744
GPSPosition : 57.64911 10.40744

Or even more:
GPS information:
GPSVersionID:2.0.0.0
GPSLatitudeRef:N
GPSLatitude:57 38 56.83
GPSLongitudeRef - E
GPSLongitude - 10 24 26.79

HTML-pages:

(1)ICBM-method

The GeoURL standard requires the ICBM tag:

The similar Geo Tag format allows the addition of placename and region tags:

(2)RDF-method

The RDF-method is defined by W3 Group and presents the information in RDF-tags[5]:
55.701; 12.552

(3)Microformat

The Geo microformat allows coordinates within HTML pages to be marked up in such a way that they can be "discovered" by software tools. Example:
50.167958; -97.133185

which might display as:
50.167958; -97.133185
(and gives a live Geo microformat on this page).

A proposal has been developed to extend Geo to cover other bodies, such as Mars and the Moon. An example is the Flickr photo-sharing Web site, which provides geographic data for any geocoded photo in all of the above-mentioned formats.

Geotagging in tag-based systems

No industry standards exist, however there are a variety of techniques for adding geographical identification metadata to an information resource. One convention, established by the Web site called GeoBloggers and used by more and more sites i.e. the photosites Panoramio and Flickr and the social bookmark-site del.icio.us, enabling them to be found via a location search. All sites allow users to add metadata to an information resource via a set of socalled machine-tags (see folksonomy).

geotagged
geo:lat=57.64911
geo:lon=10.40744

where latitude and longitude are the geographic coordinates of a particular location. These are expressed in decimal degrees in the WGS84 datum, which has become something of a default geodetic datum with the advent of GPS.

Using three tags works within the constraint of having tags that can only be single 'words'. Identifying geotagged information resources on sites like Flickr and del.icio.us is done by searching for the 'geotagged' tag, since the tags beginning 'geo:lat=' and 'geo:lon=' are necessarily very variable.

A further convention proposed by FlickrFly adds tags to specify the suggested viewing angle and range when the geotagged location is viewed in Google Earth:
ge:head=225.00
ge:tilt=45.00
ge:range=560.00

These three tags would indicate that the camera is pointed heading 225° (south west), has a 45° tilt and is 560 metre from the subject.
Both Panoramio (which is focused on showing geotagged pictures of the world) and Flickr, has the generated and place a picture from JPEG-metadata coordinates (as described above).

Geoblogging

Geoblogging attaches specific geographic location information to blog entries via geotags. Searching a list of blogs and pictures tagged using geotag technology allows users to select areas of specific interest to them on interactive maps.

The progression of GPS technology, along with the development of various online applications such as Flickr, has fueled the popularity of such tagged blogging.

With the advent of GPS Phones and GSM localization emerged Moblog which enable tag the blog posts with exact position of the user.

關於投影與座標請參考：
(1)(GIS導論-02)投影與座標(2)：http://blog.xuite.net/lwkntu/blog/19564845
(2)Wikipedia:WikiProject Geographical coordinates - Wikipedia, the free encyclopedia：http://en.wikipedia.org/wiki/Wikipedia:WikiProject_Geographical_coordinates

Auto-geotagging 

Auto-geotagging or automatic geotagging was first coined by MapWith.Us and refers to automating the process (U.S. patent pending) of acquiring media, associating location with the media, transferring the media to an online map and publishing the media in real time. Auto-geotagging with a media acquisition device (e.g. photographic or video camera) requires communication with location acquisition (e.g. GPS, wi-fi triangulation) and wireless data transfer devices. Modern cell phones, aGPS, and wireless data transfers data into one device. Auto-geotagging is sometimes referred to as "mobile geotagging", but this does not imply automation.

Location Acquisition

Geotagging is gaining popularity with photographers, however, most cameras do not possess the capability of determining location. Often photographers rely on external GPS receivers to determine location. Acquiring location from a stand alone GPS unit requires a lock from at least three satellites (for position) and usually requires up to 60 seconds, however, acquisition time is decreasing rapidly with hardware improvements. A new twist on conventional GPS receivers uses cell tower location and one global positioning satellite to obtain a faster lock on location. This technology known as assisted GPS (aGPS) is becoming more popular in cellular phones since it leverages cell tower locations. An alternative to GPS is WiFi triangulation which uses the MAC addresses of nearby wireless access points to determine position. Auto-geotagging relies on media acquisition devices that contain GPS, aGPS, or WiFi.

Real Time Media Transfer

Wireless data transfer is essential for auto-geotagging because it allows for real time mapping of media. Transferring images from cell phones to social networking sites is gaining popularity (see Facebook). Facebook and other sites promoting real-time media transfer have not attempted to associate location with the media since this requires mapping technology. Applications for real time mapping include travel, real estate, geo-social networking, people tracking, security, and geo-fencing.

Online Mapping

Geotagging becomes useful and relevant if you can present the geotagged media via a map. Advanced online mapping tools (e.g. MapWith.Us) allow auto-geotaggers to present data in a public, private, or protected venue. Associating a location with a media object on a map is currently driving the auto-geotagging market. However, presenting the data using online maps is a challenging problem, especially when combined with collaborative mapping (see collaborative mapping).

Privacy Concerns

Since auto-geotagging provides real-time location of the person operating the auto-geotagging device, it is possible to track that person. However, one solution to this problem is (by default) to restrict online access to maps that are auto-geotagged. The geotagger can then decide at their discretion to change access privileges.

Note：關於數位相片的地理標記法，請參考下面文章：
(1)(GE-7)GPS的gpx檔＋照片定位＋Google Earth的kml檔：http://blog.xuite.net/lwkntu/blog/13586958
(2)(GPS運用-4)DC與GPS結合：http://blog.xuite.net/lwkntu/blog/17537742

Address Geocoding

Nearly all of the data we have been dealing with throughout these lessons have been spatially explicit. Either the vector or raster data have been stored in such a way as to retain georeferencing information internally, or tabular data have been used that contain explicit fields containing X and Y coordinates. A very large body of intrinsically spatial data are not stored with explicit X and Y locations. These data are stored with addresses.

Addresses are locational data, but they do not contain Cartesian coordinates. Addresses only work as locational pointers if an underlying street-address-range framework exists. For a traveler to an unfamiliar city, a map with street names and block numbers is usually a necessity in order to locate specific addresses. The questions, "Where is Pekoe and Sepulveda?" or "Where is 10 Downing Street?" can only be answered by referencing a map.

The reason address location is such an important issue in GIS is that a large amount of data is stored by address, rather than by X and Y locations. How many students know the latitude and longitude, state plane, or UTM coordinates of their own house? How many know their own address? How do you let your tires do the walking after your fingers have done the walking? Address geocoding is an answer to this problem.

Address values & styles

Address values may come in many different formats. In general, they must contain at least:
1.house or building number (e.g., 351)
2.street name (e.g., Lee)

Address values & styles

Address values may come in many different formats. In general, they must contain at least:
1.house or building number (e.g., 351)
2.street name (e.g., Lee)

Optional items include
1.direction (e.g., North, No. N., N)
2.street type (e.g., Street, Str., St., St)

The more complete an address, the better the possibility of making a match. Take these addresses, for example:
941 Allison
941 E Allison
941 E. Allison St.
941 East Allison Street

Each address represents the same physical location, but they decrease in ambiguity as they are listed. The less ambiguous the address, the greater the possibility of making a match between the address and the on-street location. Some common problems occur where data are ambiguous in nature. Does the St. in this example mean Street or Saint?

Different cities and jurisdictions often use different methods of recording addresses. Some places, like the City of Seattle, use a convention to distinguish street and avenue addresses from each other. For example, here are 2 addresses:
3508 Northeast 35th Street
3508 35th Avenue Northeast

For street addresses, the direction is a prefix to the road name, while for avenue addresses, the direction is a suffix to the road name. How does your individual jurisdiction treat the same features? How about roads such as Courts, Circles, Boulevards, Lanes, Drives, Highways, Pikes, ...?

The important thing to bear in mind is that the address values for individual places should follow the same format as the address ranges attached to the roadways.

Address databases are generally stored in tabular format, either in a stand-alone relational database, a spreadsheet, or as ASCII files. Before the development of automated address geocoding methods, placing these addresses in a spatial context was tedious and time-consuming.

Address range values

What makes an address matchable to a location is an interpolation along a line. The line itself is coded for the road name and a range of addresses.

Here is a selected roadway from the "City of Seattle street centerline database".

Attribute data for this street are listed in the Identify Results dialog:

As you can see, the street is coded quite extensively. It contains attribute values for:

These attributes can be interpreted as shown in this diagram:

One of the central tenets of the Arc-Node topology vector data model is the idea of the directionality of lines. Because each arc has a starting and ending node, it has a right and left side. The arc can be coded for its starting and ending address for each side. In addition, the arc can be coded for its left and right zip code.

In this example, we see that the selected segment of South Orcas Street contains left-hand addresses ranging from 4200 to 4298, and right-hand addresses ranging from 4201-4299.

Addresses are placed on this view according to a linear interpolation between the starting and ending addresses. If an address is even, it is placed on the left (north) side of the street. If the address is odd, it is placed on the right (south) side of the street.

For example, take the address 4220 S. Orcas St. It will be placed on the north side of the street, at a distance of approximately 20% of the length of the arc west of the intersection of 42nd and Orcas.

Making a theme matchable

A match occurs when an address or a list of addresses matches similar attribute values (address ranges) along an arc, such as in the previous example.

Before a match can occur, the structures of the road attribute table and address tables must match. To follow the above example, here is a view of each different table.

Addresses Table

Road Attribute Table

Note that the Addresses Table contains both the complete address of the house as well as the individual parts of the address, including the house number, direction, street name, street type, and Zip code. If your address table does not contain the Address field, which holds the full street address (number, direction, name, and type), you will need to create a new field that concatenates the contents of the individual fields into one.

When the tables' structures are set, the road network theme is made matchable. This is achieved by setting the Geocoding Properties in the theme's Property Sheet.

The selected Address Style sets up associations between the fields necessary fir the style and fields existing within the attribute table for the road network theme. When these geocoding properties are set up, special indexes are created for the line theme.

Here, you can see that there are actually 7 files now associated with the seastr.shp shapefile, 2 of these being geocoding index files:

The geocoding process

After the initial data are prepared, most of the difficult work has been done. Geocoding proceeds by selecting View > Geocode Addresses from the menu.

The Address table is specified, along with its Addresses, Zone, and Display Fields. An Offset Distance can be specified so that the geocoded points do not lie on top of the road line, but are set back from the centerline by a constant distance (in map units). Alias Tables can be used to associate explicit addresses with place names (such as parks, buildings, hospitals, etc.). The user can also select an output shapefile name and directory. The default filename is geocdn.shp, which increments by 1, and is placed in the current working directory.

Geocoding Preferences can be altered to allow for some leeway in spelling and matching scores. It is also possible to specify interactive review if a single street location matches with multiple address records, or if there are no matches for a given address.

After Geocoding Preferences are selected, Batch Matching will attempt to match any address records to road locations. This method is preferred, and will work better with data sets which are more explicit and complete.

Interactive Matching allows the user to match individual address records. This method is especially useful if the data quality is not very good; it allow the user to tweak individual records to adhere to the address field standards.

When the matching process has finished, a point theme is created. Here is the single address in our example from above after matching. Because of the high data quality for the data sources, the match completed without any errors.

The newly created shapefile will contain all the attributes from the original table, as well as a series of attributes related to the geocoding process (these attributes start with the prefix Av_).

Fixing errors (rematching)

When errors do occur in geocoding, as they inevitably will, the user has the opportunity to attempt to match those records that did not match. There are a few different tacks that can be taken.

By altering the Geocoding Preferences to make the matching less stringent in terms of spelling or match scores, it may be possible to match more records. After loosening the constraints, Batch Rematch can be attempted to match more of the records that did not match on the first attempt. If you suspect that the reason matches did not occur is because of low-quality data, and if you have relatively few unmatched addresses, you can use Interactive Rematch to tweak those individual records until they match.

Once you have matched an acceptable number of addresses, you have created a spatially explicit theme out of the combination of a road network and a series of addresses. Address geocoding has obvious applications in any discipline in which the spatial location of addresses is needed, from business marketing, to real estate, to emergency services and law enforcement.