Translator Disclaimer
11 December 1998 Methods for classification of agricultural fields in aerial sequences: a comparative study
Author Affiliations +
A comparative study of a selection of classification methods for agricultural fields in sequences of aerial images is presented. The image sequences are acquired by an RGB-CCD video camera which is assumed to be on board of an airplane, moving linear over the scene. The objects in the scenes being considered are agricultural fields. The classes of agricultural fields to be distinguished are determined by the type of crop, e.g. potatoes, sugar beet, wheat, etc. In order to recognize and classify these fields obtained from the aerial sequences of images, a common approach is in the use of surface texture. Textural features are extracted from the images to effectively characterize the vegetation. Methods based on Circular Symmetric Auto-Regression, Co-Occurrence Matrix and Local Binary Patterns are selected for the comparative study. The experiments are carried out with image sequences taken from a scaled model of a landscape and a selection from the Brodatz set. A few training images are used to set up the model bases for the three methods. The methods are tested using the same regions from other images of the sequence, and other sequences of images of similar fields. Comparison fa the methods is based on the confusion matrix. Sensitivity to variations in flight direction, variations in altitude and luminance conditions are being considered.
© (1998) COPYRIGHT Society of Photo-Optical Instrumentation Engineers (SPIE). Downloading of the abstract is permitted for personal use only.
Zweitze Houkes, Haijun Chen, and Jan-Friso Blacquiere "Methods for classification of agricultural fields in aerial sequences: a comparative study", Proc. SPIE 3499, Remote Sensing for Agriculture, Ecosystems, and Hydrology, (11 December 1998);

Back to Top