RGBI images with UAV and off-the-shelf compact cameras: an investigation of linear sensor characteristics

UAVs are able to close the gap between ground-based (total station, Global Navigation Satellite System - GNSS) and airborne data acquisition. Usually, cameras in the visible range of light are used for this purpose. Until now, only a few sensors have been commercially available for the capturing of non-visible wavelengths, e.g., of the near infrared for vegetation analysis. Almost all common camera sensors can be used and be modified through small interventions for the near infrared to make a RGBI (red, green, blue, near infrared) sensor head. Vegetation parameters like NDVI can be calculated from this data. At the Frankfurt University of Applied Sciences, a sensor head has been developed to collect RGBI data. The sensor head consists of two Sigma DP2 FOVEON-sensor cameras. One of them is modified for the near infrared. This sensor head is mounted to an octocopter UAV with a maximum take-off weight of 3.5 kg. The cameras are triggered simultaneously. In the post-processing, the image channels are registered to each other by a self-developed software script and a radiometric adjustment based on reference targets in the field is carried out. The resulting multispectral images are processed to multispectral orthophoto mosaics in standard photogrammetric software. A spectrometer and various test panels are available for the investigation of linear characteristics of the cameras. This paper briefly presents the construction of the sensor head. In particular, the radiometric characteristics of the sensor should be considered. The Sigma cameras have a poor colour separation due to the installed FOVEON sensors. This is balanced by the algorithms implemented in the raw data software Sigma PhotoPro, offered by the manufacturer. However, the linear sensor characteristic is destroyed by gamma correction and white balance. Freely available raw data converters can process the Sigma data without these photographic corrections and preserve the linear characteristic, but the image quality suffers from the poor spectral separation.

搜集汇总

背景与挑战

背景概述

该数据集聚焦于使用无人机和市售紧凑相机获取RGBI（红、绿、蓝、近红外）图像，并研究传感器的线性特性；具体方法涉及构建由两个Sigma DP2 FOVEON传感器相机组成的传感器头（其中一个改装用于近红外），通过八旋翼无人机采集数据，并进行图像配准、辐射调整以生成多光谱正射影像。其特点在于探讨了传感器线性特性因相机内部处理（如伽马校正）而受损的问题，并强调了在植被分析（如NDVI计算）中的应用潜力。