Agricultural information processing systems that utilize GIS-compatible paddy field maps have been proposed. Paddy field maps are usually created manually, but this method is time consuming. We propose a method for the automatic creation of maps. To create paddy field maps, it is necessary to extract and vectorize the field borders. The conventional method for the extraction of borders uses an edge-detection approach combined with edge tracking, but gaps where edges cannot be detected lower the rate of identification. As an alternative, we used evolutionary image processing to create a wood-structure filter from a false-color aerial photograph with a resolution of 2,048 × 2,048 pixels. Our method, designed to automatically create paddy field polygons by evolutionary image processing, scans a paddy field repeatedly to create tone-corrected images with different maximum and minimum pixel values. When the filter was applied to 18 test images that did not overlap, the rate of the identification of fields in which plants were growing was 55.1% ± 24.9%, and the time required to create a field map was 36.5% of the time required to manually create the same map. When the filter was applied to 55 test images that overlapped each other by 50%, the rate of identification was 64.4% ± 20.2%, and the time required was 34.4% of that required to manually create the same map. Thus, the method reduced the time required to create paddy field maps to about a third.
View full abstract