Evaluation of Spatial Filtering on the Accuracy of Wheat Area Estimate,
Abstract
The objective of this study was to select the optimal combination of threshold values for the 3x3 pixel spatial filtering for post-classification implanted in Image-100 system of INPE and to evaluate the effects of this procedure on the accuracy of wheat area which was estimated by Image-100 system using a hybrid classifier and LANDSAT digital data obtained from a single pass. An area of 800 sq km in Cruz Alta which is one of the most important municipals for wheat production in southern Brazil, was selected for this study. Different threshold combinations: (1,2), (2,1), (2,2), (2,3), (3,2), (3,4), and (3,5) were employed in the spatial class filtering for the whole study area after wheat was classified. Alphanumeric theme prints of classification results with and without employing spatial filtering were compared using aerial photographic mosaic as 'ground information'. The combination of (2,2) was then selected as the best threshold values in spatial filtering and was applied to five test sites (approx=40 km each) with different wheat densities for a quantitative evaluation of the accuracies of wheat area estimates. T tests showed that filtering with threshold values (2,2) significantly decreased errors of commission and omission, also, the accuracy in area estimate was improved from the over estimate of 4.5% to 2.7% and the root-mean-square error decreased from 126.18 ha to 107.02 ha.
Document Details
- Document Type
- Technical Report
- Publication Date
- Jun 01, 1982
- Accession Number
- ADP002039
Entities
People
- A. M. De Lima
- M. A. Moreira
- S. C. Chen