Search

The need for timely, accurate, and interoperable geospatial information is steadily increasing. In this context, process-based image processing systems will be the initial segment for future’s automatic systems. A process-based system is believed to be a good approach for agricultural purpose because agricultural activities are carried out according to a periodic (annual) cycle. Therefore, a process-based image analysis procedure was designed for routine crop classification for an agricultural region in Kırklareli, Turkey. The process tree developed uses a multi-temporal image data set as an input and gives the final crop classification as an output by using an incremental rule set. The test data set was composed of five images of Satellite Pour l’Observation de la Terre 4 (SPOT 4) data acquired in 2007. Basically, image objects were first extracted and then classified. A rule set was structured depending on class definitions. A decision-based process was executed and formed a multilevel image classification system. The final classification was obtained by merging classes from the appropriate levels where they were extracted. To evaluate the success of the application the accuracy of the classification was assessed. The overall accuracy and kappa index of agreement was found to be 80% and 0.78, respectively. At the end of the study, problems of segmentation and classification operations were discussed and solution approaches were outlined. To assess the process in terms of its scope for automation, the efficiency and success of the rule set were also discussed. [ABSTRACT FROM AUTHOR]