2.2.2. Data Analysis and Shoreline Detection Techniques

Various indices were calculated using the spectral bands of the orthomosaics in order to separate water from other areas (sand, land vegetation, etc.). Indices were calculated to simplify complex data amalgamations, e.g., orthomosaic images. The NDWI and NDVI were used to highlight the water from other surfaces during the analysis of multispectral images. These indices were derived from the NIR, Green (G) and Red (R) spectral bands. They are dimensionless indices that describe the difference between visible and near-infrared reflectance of surfaces. The orthomosaics created from RGB images did not have an NIR band available. For this reason, the VARI and NGRDI indices were used in the same application to highlight the pixels that represented water. Equations of the aforementioned indices can be found in the literature that was reviewed in Section 1.

The resulting raster images of various indices were thoroughly analyzed. The index values of four different surface conditions in the study area were compared. The thresholding technique was used to segment images into several classes—water, aquatic macrophytes, land vegetation, and sand. Generally, a threshold limits can be selected by trial and error (manually) or using the peak–valley method of histogram segmentation [35]. In this instance, index value limits were set manually for each surface. The resulting segmented images were used to detect the shorelines.

Two different areas were analyzed to evaluate the performance of different approaches. The first area was a wavy shoreline with steep banks. The water in that area was mostly free from aquatic vegetation, so the shoreline, between land vegetation that covers the banks and the water, was well-defined. The more challenging part was between the bends of the shore, which were shaded from the sun. The second area was the opposite—the banks were very shallow, covered with sand, and clear from vegetation. One part of the sandy beach was covered with silt. Thresholding and unsupervised classification were carried out for the raster images containing NGRDI, VARI, NDVI, and NDWI values.