Introducing an intelligent computerized tool to detect and predict urban growth pattern

Urban growth pattern is usually detected using spatial analysis. Spatial analysis is widely used in scientific research especially in the field of statistics, image processing and geoinformatics. In modeling urban growth, the analysis is mostly performed using statistical and mathematical techniques. With the advance computer technology, physical land (ground) situation for a place of interest can be represented in digital computerized form with the accurate and appropriate scale. In this way, measurement can be made on the digitized presentation for performing analysis. The change in land use is affected by many factors such as population growth, economic change, social structure, the change in rules and regulation, and many more. These influential factors have dynamic behaviors that require complex solutions. Much research has been undertaken to use several methods such as geographical information system (GIS) and cellular automata theory, to model the urban growth. Recently, an intelligent approach has been introduced that features dynamic behavior. Artificial Neural Network (ANN) has the capability to learn dynamic behavior and performs prediction based on its learning process. In this paper, we present an intelligent computerized tool, called DIGMAP-Detector. This tool is able to learn a pattern of urban growth based on at least two digital maps (with 4-bit/pixel bitmaps or 8-bit/pixel bitmap in Bitmap File Format (BMP)). Implemented using Java programming language, the tool reads digital map files with the size of 847 pixels length and 474 pixels width. Classification on the map with two independent binary classes (value 1 for urban and 0 for rural) are prepared using GIS software. By applying a cellular automata theory that considers the affect on a center pixel is influenced by its surrounding pixels (eight pixels), the tool uses a back propagation neural network to read the values of surrounding pixels as its input layer nodes and the center pixel as the output node. Several analyses are performed to determine the appropriate values for the neural network configuration before its learning engine starts to learn the pattern of dynamic urban changes based on the digital map patterns. When the neural network engine has learnt the pattern, prediction can be carried out to predict the missing years and future urban growth. With good prediction accuracy, urban planning and monitoring can be performed with maintaining good ecological and environmental system. In addition, better planning also gives benefit to economic values.