Digitising and scanning for data entry Lecture 11: Digitising and scanning for data entry

So Far in This Course... We’ve dealt with data that has already been classified and transformed to a coordinate system. It’s been presented to us in a form suitable for spatial analysis. We haven’t had to consider a range of issues such as; Where we get our data? How do we get it into a GIS? What are our sources of error? How do we measure that error? How do we minimize that error?

Data Input and Associated Issues Data collection takes up a large proportion of the time and expense of a GIS facility. Four common methods of producing digital spatial (map) data. Digitizing maps with a digitizer (or on the screen). Scanning existing maps or air photos. Entry of coordinates from field measurements. Transfer from existing digital sources (such as from satellite imagery). Which to Choose? When?

Input and Processing of Data to Build a Raster Database.

Input and Processing of Data to Build a Vector Database. Extra steps in creating vector

1. Digitizing Use a digitizing tablet or digitize from the screen. Why digitize using the screen?

1. Digitizing Why digitize using the screen? You may have a raster coverage in digital format. you may only want to extract a few features may not be extractable by image processing methods - the spectral signature of the desired objects does not differ significantly from other features on the coverage. This can especially be the case for panchromatic (B/W) sattelite images. you can use logic, pattern recognition and context to identify features. You may not have access to a digitizer. You’ll be doing this type of digitization later in the semester.

Advantages of the Digitizing Tablet Some data are available only in paper form. Paper map more easily annotated to emphasise relevant features. More accurate than the screen Scanner not necessary If vector output required

Step 1 in Digitizing: Preparation Highlighting relevant information (especially where only certain features from a topographic map are being sought). Simply draw on the map or on a sheet(s) of tracing paper. This clarifies the information that the user is interested in. May involve incorporating information from different map sheets. Amendments to maps may be required, eg. temporal change. Road routes may have changed since the main map was produced. Easier to do this on paper than on the digital coverage

The Digitizer The digitizer is an electronic tablet. The user places a map or photograph on the digitizer. Behind it is a device that senses the location of a pointer. This pointer is used to trace the features on a map. Basically digitizing is a fancy, high tech method of tracing.

Step 2 in Digitizing: Digitizing. Then the features on the map are digitized. Crosshairs are placed over the map, and points are entered, either automatically or manually. The beginning and end of lines are entered, and lines automatically connect between points Object ID codes can be entered during this phase or augmented during the editing phase, depending on the program. Similarly, if digitizing from the computer screen, the mouse and cursor are used to trace features.

Georeferencing the coverage Tablet has its own independent coordinate system. On the screen can use the coordinate system of the map. If using a tablet, the coordinate system is registered to a coordinate system (eg. UTM) by cross-registering a number of coordinates on the tablet to known coordinates on the map eg. grid intersections, border markers, or other reliable locations on the map. Using these coordinates, transformation parameters are then computed, which allow the coverage to be registered. Georeferencing is done either prior to or after digitizing features

Transformations Three main processes must Occur 1. translation (object movement) 2. scale alteration 3. rotation This is called a Conformal Transformation. All objects retain their original shapes. Affine Transformations incorporate unequal x and y scale alterations and unequal rotations of the x and y axes. Affine transformations alter the shapes of objects This compensates for unequal shrinkage of maps, and is particularly important for digitizing from aerial photographs

Step 3 in Digitizing: Data editing Often there is error from digitizing, such as; gaps between lines that should meet lines that exceed the line they should meet (dangles) parallel lines that cross

Potential digitising errors in the real world!

Yet more potential digitising errors in the real world!

Editing the Digitized Coverage Data usually checked on the screen. Errors identified by the computer (dangling arcs, unclosed polygons) Edited by selecting the portion of arc to omit, snapping nodes together, labelling polygons etc.

2. Scanning Maps or Aerial Photos Scanning maps or air photos follows the same principles as satellite image analysis That markings on a map reflect light differently from areas that are blank and that That different features on an air photo reflect light differently from one another. Light reflectance is recorded in a grid made up of square pixels In other words, a raster coverage is created

Scanning Existing Maps Map must be high quality Map must be clean Must have lines 0.1mm wide or greater In the Scanning Process Range of issues such as contrast, brightness, and resolution must be sorted out in the scanning process Too little brightness = non-recognition of some features. Too much brightness = unwanted ‘information’ such as dirt and dust may be included. Contrast and Brightness are usually set by a mixture of trial and error, and judgement.

Scanning Resolution The chosen resolution depends on the width of the features, and the scale of the map. For example, if you have a 1:250,000 map, scanned at 300dpi Pixel size (cm) = Resolution of scanner (inches) * 2.5 * map scale Pixel size (cm) = (1/300) * 2.5 * 250,000 Therefore, each pixel is 2000cm wide. Pixel size (m) = 20m.

Post Scanning Image Analysis Scanned maps are rarely perfect. The image will contain smudges, defects, mistakes where map detail is complex. Therefore digital image analysis is necessary to improve the image and make it useable Can use highlighter on linework to create separate category extractable in image processing software (good for lineaments)

Thresholding or binarisation to classify different ranges of pixel values to two or more categories… In this example, all cells in the image with a grey scale value less than 80 are considered to be black, and given a value of 1. All other cells are considered to be white, and given the value 0.

Filtering By passing a neighbourhood filter over the coverage, small groups of pixels (that may represent dust specs etc.) are given the value of surrounding pixels Can use mean, median, modal filters The user decides what the threshold for the size of the group of pixels will be

To Recap on Scanning Any scanned map or air photo results in a raster image If a vector output coverage is required, this raster image must next be transformed (vectorized)

Vectorization of Lines Usually part of the post-scanning editing process A number of pixels forming a structure are registered (b) Skeletonization. All pixels along the line except along the centre are stripped off (c). Linearization. Pixels are connected one-by-one along the line. Where a line segment curve is greater than a defined threshold, the line segment is considered not to be straight, and a new line is started. Coordinates are calculated for the start and end points of the straight line segment, and an arc is then formed (d). Original map Scanned image Skeletonization Vector output

DIGITIZING VS SCANNING Digitizing-Advantages Equipment cheap Doesn’t require high map quality Less post-input analysis Already in vector format Digitizing-Disadvantages Very boring (quite unpleasant) Labour intensive Unsuitable for continuous data

Scanning-Disadvantages Scanning-Advantages Easy Very fast Suitable for continuous data Scanning-Disadvantages Maps must be of high quality, clean, and have clearly defined lines Raster output files may be very large Can’t link attribute data directly to features Improvement of data is labour-intensive Data may require conversion into vector format Large format scanners are VERY expensive

3. Entry of Data From Field Measurements Usually in the form of point data as locational coordinates. May be typed in at the keyboard May be loaded in a data file format eg. Excel, CSV. Applications Often represent survey locations (geology, soils, biological surveys). May represent ground control points for image rectification. With use of GPS, accurate to within 4m. Often used for interpolation

4. Transfer from existing digital sources (such as satellite imagery). Data already in a digital form In the case of remote sensing data, always comes in a raster form Steps in Processing Raw Satellite Data Georeferencing to a chosen coordinate system and rectification (affine transformation) Orthorectification for high resolution data Atmospheric correction Windowing Classification Classification accuracy check

Next Lecture Sources of error Measuring error Minimising error