Evaluating Mesquite Distribution Using Unpiloted Aerial Vehicles and Satellite Imagery

Using drones to detect and quantify wild pig damage and yield loss in corn fields throughout plant growth stages

“…Imagery captured by the drone was processed in Pix4Dmapper (Pix4D S. A., Prilly, Switzerland), which enabled us to stitch overlapping images together to create 2‐D orthomosaics and 3‐D models of each flight (DiMaggio et al 2020, Page et al 2022). Pix4Dmapper uses the structure from motion algorithm to create 3‐D photogrammetric meshes and 3‐D point cloud datasets (X, Y, Z), generating a digital surface model (DSM) and digital terrain model (DTM; Kuzelka and Surovy 2018, DiMaggio et al 2020, Page et al 2022). The DSM represents height values of the vegetation canopy and the DTM depicts elevation values of the terrain (Jimenez‐Jimenez et al 2021).…”

“…The 25‐class raster and the RGB orthomosaic were used as input for the Export Training Data for Deep Learning tool, with the metadata format set on classified tiles. The output creates a folder containing images, labels, maps, and stats of the created training data or image chips identified in the drone imagery (2.9 cm 2 ; Page et al 2022). We applied the Train Deep Learning Model tool (Esri 2022) to the RGB orthomosaic and image chips using the following parameters: max epochs (amount of times the dataset will be passed back and forth through the neural network) set at a default value of 20, model type was u‐net pixel classification, batch size of training samples processed at a time is the default size of 2, backbone model is the default ResNet‐34 (preconfigured model with more than 1 million images and is 34 layers deep), and validation percentage at a default value of 10 (10 percent of the training samples will be used to validate the model; ESRI 2022).…”

“…Flight altitude was 100 m above ground level (AGL) at a speed of 3–5 km/h, −90◦ camera angle, with 80% image overlap and sidelap (DiMaggio et al 2020). All flights took place on sunny, clear days with winds <16 km h −1 and were conducted between 0930 and 1430 to minimize shadows (DiMaggio et al 2020, Page et al 2022).…”

“…The DSM represents height values of the vegetation canopy and the DTM depicts elevation values of the terrain (Jimenez‐Jimenez et al 2021). To calculate the height of the vegetation, we subtracted the DSM from DTM to get a normalized DSM (nDSM; Mayr et al 2018, Page et al 2022).…”

“…We conducted drone flights April through August We conducted flights throughout 4 corn growth stages to determine if different types of crop damage could be detected and quantified from drone imagery (Table 1). The 4 corn growth stages are establishment, vegetative, blister-milk, and dent-mature (Nafziger 2009). Wild pig damage differs during distinct corn growth stages (Figure 2); wild pigs use their snout to root up seeds and seedlings during the establishment stage, then trample plants in the latter reproductive stages (R2 blister-milk and R6 dent-mature stages) once kernels are beginning to form and provide maximum nutritional value (Schley et al 2008, Bleier et al 2017, Boyce et al 2020.…”