THE LUCS-KDD RANDOM IMAGE GENERATOR: THE RanImgGen DIRECTORY

The LUCS KDD random image generator is intended for use with image mining software. It produces a sequence of images all measuring 120x80 pixels of the form shown in Figure 1.

Figure 1: Example Random Images.

The system comprises the following java source files (In alphabetic order):

2.1. Compiling

The software has been implemented in Java using the Java2 SDK (Software Development Kit) Version 1.4.0, which should therefore make it highly portable. The code does not require any special packages and thus can be compiled using the standard Java compiler:

javac *.java

2.2. Running the software

The software can be run in the usual manner:

java RandomImageGenApp

A GUI (Figure 2)will appear which is straightforward to use.

Figure 2: Random Image Generator Interface.

2.3. Documentation

The code can be documented using Java Doc. First create a directory Documentation in which to place the resulting HTML pages and then type:

javadoc -d Documentation/ *.java

This will produce a hierarchy of WWW pages contained in the Document directory.

The gebnerated random images comprise various predefined landscape features and are stored as a sequence of .bmp files. At a high level an image is structured as shown in Figure 2.

IMAGE ::= SKY, FOREGROUND

SKY ::= height, colour=white|blue|light gray, SKY_IMAGE_FEATURES

FOREGROUND ::= SEA | LAND | SEA_AND_LAND) | LAND_AND_SEA

SEA ::= colour=blue, SEA_IMAGE_FEATURES

LAND ::= colour=lightGreen, LAND_IMAGE_FEATURES

From Figure 2 it can be observed that an image comprises sky and foreground. The sky part of the image can contain a number of "sky image features". Currently the only sky features available are hills which "sit" on the horizon, an image may include between 0 and 3 hills. Hills come in three sizes: small. medium and large; and two colours: "light brown" and "lilac". Current possible sky colours "are light blue", "white", "light gray".

The foreground part of the image may comprise either: "land", "sea", "land and sea" or "sea and land" with the distinction between the last two being that the first named is the closest to the viewer. Sea is always "blue" and land "light green". The foreground may also contain between 0 and 5 foreground image features. Currently there are two land image features available (trees and houses) and one sea feature (sailing boats).

Image features are all sub classes of the class ImageFeature. Image features are built up from a number of primitives called image objects. Five kinds of image object are available:

1. Rectangle.
2. NE right angled triangle.
3. SE right angled triangle.
4. SW right angled triangle.
5. NW right angled triangle.

each represented by its own class. The NE, SE, SW and NW references refer to the orientation of the triangle, a NE triangle has the right angled corner to the NE etc. The triangle objects are all sub-classes of the class TriangleIO. The TriangleIO class in turn, and the rectangle class, are then sub-classes of the class ImageObject (See Figure 2).

                               ImageObject
                                    |
                     +--------------+---------------+
                     |                              |
                TriangleIO                      Rectangle
                     |
    +----------+-----+-----+-----------+
    |          |           |           |
NEtriangle SEtriangle SWtrainagle NWtraiangle

The available colours (currently there are 24 of these) are listed in the ImageUtilities.java file together with the features which they are used for. When creating a new image feature make sure to update the comments in ImageUtilities.java file listing what each colour is used for what.

To add a new colour to the ImageUtilities.java file:

To add a new feature follow the following steps.

1. Decide whether your image feature is going to be a sea or land feature.
2. Design your image feature in terms of rectangles and triangles, i.e. the available IOs (refer back to Section 2 for more detail). Do three (or more) versions: (i) distant, (ii) middle ground and (iii) close up. This is best done using squared paper with each square representing a pixel (the origin for each square/pixel is the top left-hand corner). To get some idea of scale a person will be 21 pixels high close up, 14 pixels high in the middle ground and 7 pixels high in the distance. When selecting colours for your feature the ist of available colours is in ImageUtilities.java>
3. Creata a java class file for your feature which must be a sub-class of the class Image Feature. It is probably easiest to make a copy of an existing feature class source file (Tree.java is quite simple). The Image Feature class contains two fields that your class will inherit:
   1. location --- a two element integer array for the X and Y coordinates describing your feature's location (usually the center of the base), and
   2. (ii) featureType --- a string describing the nature of your object and its variations, e.g "very large tree" (this is used for output purposes).
      These two fields must be set in your class.
      1. Hint: Look at the existing image feature classes (House, Tree, SailBoat) to get a feel for what methods your class will need to include. It needs at least a addIFtoBitMap method and a five argument constructor.
      2. Once your feature class is complete (and compiles correctly!) you are ready to include it within the generator. In the Foreground class include a label for your feature in the FEATURES field, and the same label in either the LAND_FEATURES field or the SEA_FEATURES field according to whether your feature is a land or a sea feature. Also include the label in the createSeaAndLandIF method and either the createLandIF method or the createSeaIF method as appropriate. You will need to study the Foreground class quite carefully to see how to include the labels.

      Now you will able to generate new images that include your feature!