BORDERS DETECTION METHOD BASED ON A MODIFIED SOBEL ALGORITHM FOR CROP CLASSIFICATION MAPS

Obtaining reliable and accurate crop classification and land cover map based on satellite data, in particular high resolution data, is one of the most important tasks in remote sensing. Such maps provide basic information for many other applied problems and are vital in remote sensing studies. Despite of which machine learning methods were utilized for maps obtaining: traditional (Random Forest, Support Vector Machine, Multi-layer perceptron, logistic regression) or state-of-the-art approaches (autoencoder, convolutional neural network, recurrent neural network) there is some noise (single pixels or groups and clusters of pixels that wrong classify) on such maps. There are traditional methods for noise reduction, however these methods do not take into account image semantics. Therefore, they are not effective for filtration land cover and crop classification maps based on satellite images. The most complicated task in the filtering such maps is to preserve edges and boundaries between different agricultural fields. Often these boundaries are small and common filters consider them like a noise and remove them. Therefore, final crop classification map after filtration using common methods is smoothed and all edges are loosed. Thus, in this paper we proposed new method for boundaries identification on the crop classification map based on modified Sobel algorithm. It is impossible to use gradient based methods for boundaries detection because important peculiarity of the crop classification map that it has finite discrete set of pixel values. We proposed modification of Sobel algorithm based on using additional steps of processing. These steps consist of convolution with structural element (square), threshold filter (considers all objects that have square less than threshold as a noise and remove them) and morphological closing operation for boundaries detection between agricultural fields but not for other changes in pixel values identification. Accuracy and efficiency of this method with the proposed filtration method have been tested on the independent set and using the visual comparison with the results of utilizing common filters