In recent weeks the team at Cyprus University of Technology is engaged with the automation of the photogrammetric processing workflow of multispectral datasets that were acquired in the last couple of weeks. The core of the workflow is done using Agisoft Metashape photogrammetric package. While the software has a complete and simple User Interface, the team aims to utilize its API in order to implement a full automated process into the OenoWatch middleware. The middleware should compile a python script which will execute all the necessary processing steps as they are showcased in the flow diagram below.
Figure 1: Flow Diagram of processing steps for orthophoto generation. Note: Green boxes refer to processing steps while blue refer to products
In detail the script as a preprocessing step should read and import the images as well as calibrate their reflectance values. The calibration is done using data recorded from the irradiance sensor that accompanies the multispectral camera and is mounted on top of the drone. After these steps the script should execute the core photogrammetric processes. First an image alignment using Structure from Motion (SfM) and Bundle adjustment is done in order to obtain the relative camera positions (Fig 2.).
Figure 2: The image block with the derived sparse point cloud after image alignment and bundle adjustment, of one of the pilot flights as it was acquired by the UAV.
After the Image Block Alignment, the script executes a Dense 3D Point Cloud Generation. The script also executes a DEM generation as well as the Orthophotomosaic generation. The last step is the calculation of the indexed Orthophotomosaic. The script can calculate various indices such as NDVI or GNDVI and then it can export the final Orthophotomosaic with the index values (Fig 3). The values of the indices are from 0-1 and will later be used in order to identify potential illnesses or hazards that may affect the vineyards.
Figure 3: The derived Οrthphotmosaic NDVI index as a preliminary result.