Vegetation Removal From Uav Derived Dsms, Using Combination of RGB and Nir Imagery

Abstract

Current advancements on photogrammetric software along with affordability and wide spreading of Unmanned Aerial Vehicles (UAV), allow for rapid, timely and accurate 3D modelling and mapping of small to medium sized areas. Although the importance and applications of large format aerial overlaps cameras and photographs in Digital Surface Model (DSM) production and LIDAR data is well documented in literature, this is not the case for UAV photography. Additionally, the main disadvantage of photogrammetry is the inability to map the dead ground (terrain), when we deal with areas that include vegetation. This paper assesses the use of near-infrared imagery captured by small UAV platforms to automatically remove vegetation from Digital Surface Models (DSMs) and obtain a Digital Terrain Model (DTM). Two areas were tested, based on the availability of ground reference points, both under trees and among vegetation, as well as on terrain. In addition, RGB and near-infrared UAV photography was captured and processed using Structure from Motion (SfM) and Multi View Stereo (MVS) algorithms to generate DSMs and corresponding colour and NIR orthoimages with 0.2 m and 0.25 m as pixel size respectively for the two test sites. Moreover, orthophotos were used to eliminate the vegetation from the DSMs using NDVI index, thresholding and masking. Following that, different interpolation algorithms, according to the test sites, were applied to fill in the gaps and created DTMs. Finally, a statistic analysis was made using reference terrain points captured on field, both on dead ground and under vegetation to evaluate the accuracy of the whole process and assess the overall accuracy of the derived DTMs in contrast with the DSMs.

An Accurate Vegetation and Non-Vegetation Differentiation Approach Based on Land Cover Classification

“…In some applications, such as construction planning, land surveying, etc., an accurate digital terrain model (DTM) is essential [24]. However, the terrain may be covered by vegetation (trees, shrubs, grass).…”

The Use of Rolling Stock as a Carrier of Sensors in the Field of Environmental Monitoring

“…W zadaniach związanych z monitoringiem środowiska naturalnego stosowane są sensory wykorzystujące promieniowanie elektromagnetyczne [1][2][3], fale akustyczne [4] i zapachy [5]. Umiejscowienie sensorów używanych w tym procesie na nośnikach satelitarnych (Landsat [6], Eumetsat [7]), rakietach sondujących [8], balonach stratosferycznych [9], platformach lotniczych [3,10] jest często kosztowne, wymaga przeprowadzenia procedury planującej i przeprowadzającej kampanię pomiarową. Z drugiej strony umiejscowienie sensorów stacjonarnie: bezpośrednio na powierzchni [11] lub na platformach [12] ogranicza znacznie użyteczny zasięg pozyskiwanych w ten sposób danych.…”

Tree, Shrub, and Grass Classification Using Only RGB Images

“…Land cover classification has been used in change monitoring [1], construction surveying [2], agricultural management [3], digital terrain model (DTM) generation [4], and identifying emergency landing sites for UAVs during engine failures [5,6]. Some other uses of land cover classification are for biodiversity conservation [7], land-use [8], and urban planning [9].…”

“…In emergency landing of unmanned air vehicles (UAVs), it is critical to land on grassland rather than on trees or shrubs [5,6]. Removing tall vegetation from the digital surface model (DSM) such as trees and shrub is an important step in developing an accurate digital terrain model (DTM) [2]. Traditionally, normalized difference of vegetation index (NDVI) has been used for vegetation detection.…”