Seagrass is the subject of significant conservation research. Seagrass is ecologically important and of significant value to human interests. Many seagrass species are thought to be in decline. Degradation of seagrass populations are linked to anthropogenic environmental issues. Effective management requires robust monitoring that is affordable at large scale. Remote sensing methods using satellite and aircraft imagery enable mapping of seagrass populations at landscape scale. Aerial monitoring of a seagrass population can require imagery of high spatial and/or spectral resolution for successful feature extraction across all levels of seagrass density. Remotely piloted aircraft (RPA) can operate close to the ground under precise flight control enabling repeated surveys in high detail with accurate revisit-positioning. This study evaluates a method for assessing intertidal estuarine seagrass (Zostera muelleri) presence/absence and coverage density using multispectral imagery collected by a remotely piloted aircraft (RPA) flying at 30 m above the estuary surface (2.7 cm ground sampling distance). The research was conducted at Wharekawa Harbour on the eastern coast of the Coromandel Peninsula, North Island, New Zealand. Differential drainage of residual ebb waters from the surface of an estuary at low tide creates a mosaic of drying sediment, draining surface and static shallow pooling that has potential to interfere with spectral observations. The field surveys demonstrated that despite minor shifts in the spectral coordinates of seagrass and other surface material, there was no apparent difference in image classification outcome from the time of bulk tidal water clearance to the time of returning tidal flood. For the survey specification tested, classification accuracy increased with decreasing segmentation scale. Pixel-based image analysis (PBIA) achieved higher classification accuracy than object-based image analysis (OBIA) assessed at a range of segmentation scales. Contaminating objects such as shells and detritus can become aggregated within polygon objects when OBIA is applied but remain as isolated objects under PBIA at this image resolution. There was clear separability of spectra for seagrass and sediment, but shell and detritus confounded the classification of seagrass density in some situations. High density seagrass was distinct from sediment, but classification error arose for sparse seagrass. Three classifiers (linear discriminant analysis, support vector machine and random forest) and three feature selection options (no selection, collinearity reduction and recursive feature elimination) were assessed for effect on classification performance. The random forest classifier yielded the highest classification accuracy, with no accuracy benefit gained from collinearity reduction or recursive feature elimination. Spectral vegetation indices and texture layers substantially improved classification accuracy. Object geometry made a negligible contribution to classification accuracy using mean-shift segmentation at this image-scale. The method achieved classification of seagrass density with up to 84% accuracy on a three-tier end-member class scale (low, medium, and high density) when using training data formed using visual interpretation of ground reference photography, and up to 93% accuracy using precisely measured seagrass leaf-area. Visual interpretation agreed with precisely measured seagrass leaf area 88% of the time with some misattribution at mid-density. Visual interpretation was substantially faster to apply than measuring the leaf area. A decile class scale for seagrass density correlated with actual leaf area measures more than the three-tier scale, however, was less accurate for absolute class attribution. The research demonstrates that seagrass feature extraction from RPA-flown imagery is a feasible and repeatable option for seagrass population monitoring and environmental reporting. Further calibration is required for whole- and multi-estuary application.