The recent developments of new deep learning architectures create opportunities to accurately classify high-resolution unoccupied aerial system (UAS) images of natural coastal systems and mandate continuous evaluation of algorithm performance. We evaluated the performance of the U-Net and DeepLabv3 deep convolutional network architectures and two traditional machine learning techniques (support vector machine (SVM) and random forest (RF)) applied to seventeen coastal land cover types in west Florida using UAS multispectral aerial imagery and canopy height models (CHM). Twelve combinations of spectral bands and CHMs were used. Our results using the spectral bands showed that the U-Net (83.80–85.27% overall accuracy) and the DeepLabV3 (75.20–83.50% overall accuracy) deep learning techniques outperformed the SVM (60.50–71.10% overall accuracy) and the RF (57.40–71.0%) machine learning algorithms. The addition of the CHM to the spectral bands slightly increased the overall accuracy as a whole in the deep learning models, while the addition of a CHM notably improved the SVM and RF results. Similarly, using bands outside the three spectral bands, namely, near-infrared and red edge, increased the performance of the machine learning classifiers but had minimal impact on the deep learning classification results. The difference in the overall accuracies produced by using UAS-based lidar and SfM point clouds, as supplementary geometrical information, in the classification process was minimal across all classification techniques. Our results highlight the advantage of using deep learning networks to classify high-resolution UAS images in highly diverse coastal landscapes. We also found that low-cost, three-visible-band imagery produces results comparable to multispectral imagery that do not risk a significant reduction in classification accuracy when adopting deep learning models.

Urban forests are often heavily populated by street trees along right-of-ways (ROW), and monitoring efforts can enhance municipal tree management. Terrestrial photogrammetric techniques have been used to measure tree biometry, but have typically used images from various angles around individual trees or forest plots to capture the entire stem while also utilizing local coordinate systems (i.e., non-georeferenced data). We proposed the mobile collection of georeferenced imagery along 100 m sections of urban roadway to create photogrammetric point cloud datasets suitable for measuring stem diameters and attaining positional x and y coordinates of street trees. In a comparison between stationary and mobile photogrammetry, diameter measurements of urban street trees (N = 88) showed a slightly lower error (RMSE = 8.02%) relative to non-mobile stem measurements (RMSE = 10.37%). Tree Y-coordinates throughout urban sites for mobile photogrammetric data showed a lower standard deviation of 1.70 m relative to 2.38 m for a handheld GPS, which was similar for X-coordinates where photogrammetry and handheld GPS coordinates showed standard deviations of 1.59 m and the handheld GPS 2.36 m, respectively—suggesting higher precision for the mobile photogrammetric models. The mobile photogrammetric system used in this study to create georeferenced models for measuring stem diameters and mapping tree positions can also be potentially expanded for more wide-scale applications related to tree inventory and monitoring of roadside infrastructure.

Capturing high spatial resolution imagery is becoming a standard operation in many agricultural applications. The increased capacity for image capture necessitates corresponding advances in analysis algorithms. This study introduces automated raster geoprocessing methods to automatically extract strawberry (Fragaria × ananassa) canopy size metrics using raster image analysis and utilize the extracted metrics in statistical modeling of strawberry dry weight. Automated canopy delineation and canopy size metrics extraction models were developed and implemented using ArcMap software v 10.7 and made available by the authors. The workflows were demonstrated using high spatial resolution (1 mm resolution) orthoimages and digital surface models (2 mm) of 34 strawberry plots (each containing 17 different plant genotypes) planted on raised beds. The images were captured on a weekly basis throughout the strawberry growing season (16 weeks) between early November and late February. The results of extracting four canopy size metrics (area, volume, average height, and height standard deviation) using automatically delineated and visually interpreted canopies were compared. The trends observed in the differences between canopy metrics extracted using the automatically delineated and visually interpreted canopies showed no significant differences. The R² values of the models were 0.77 and 0.76 for the two datasets and the leave-one-out (LOO) cross validation root mean square error (RMSE) of the two models were 9.2 g and 9.4 g, respectively. The results show the feasibility of using automated methods for canopy delineation and canopy metric extraction to support plant phenotyping applications.