Canopy structural parameters estimation with uav-based RGB imagery

In the past, agricultural activities were carried out with simple human-power and primitive tools but today, it's made in line with very advanced technologies. With the introduction of UAVs for civilian use, their use has become widespread, and many professions have started to include UAVs in their research. With the decrease in UAV costs, UAVs have also started to be used in precision agriculture applications. UAVs are used for many purposes such as biomass, spraying, disease detection, plant water stress analysis, weed detection, fertilization, irrigation, crop yield, soil erosion, classification, wild animal damage determination. As a result of observations and interventions by UAVs, soil and water resources are protected, spraying, irrigation and fertilization activities are kept under control, and maximum efficiency is achieved with low costs. When the literature studies were examined, it was seen that the canopy volume was related to issues such as yield, biomass, and phenotyping. Canopy structural parameters are closely related to plant health and growth, and with this study, it is aimed to contribute to precision agriculture practices and monitoring plant health and growth with high accuracy and precision. In this thesis, the images obtained from UAV were processed in Agisoft and Pix4D software working with the Structure from Motion algorithm and QGIS. In Agisoft Metashape and Pix4D, images of point cloud, Digital Surface Model, orthophoto and red, green and blue bands were obtained separately with medium and high precision. The resulting products obtained with different precision were transferred to the Quantum GIS. First of all, the R, G and B bands, which were created with high precision in Agisoft Metashape, were transferred to QGIS separately and the Excess Green Index (ExG) value was calculated, then a threshold layer was created for the classification of vegetation and other classes. Then, the soil pixels were masked and the Digital Surface Model was transferred to QGIS. Hence, canopy only and soil only raster layers were produced. Values such as mean canopy height, median soil height and canopy area derived from the "Zonal Statistics" tool were transferred to Excel. The canopy volume was obtained as a result of the calculations made with appropriate values in Excel. All processes were reiterated for products produced with different precision in Agisoft and Pix4D, and 4 different canopy structural parameter values were obtained. The obtained results were analyzed in SPSS software. According to the multiple linear regression analysis made between mean canopy height, mean canopy area, Excess Green Index and canopy volume in SPSS; It has been concluded that the mean canopy height is the most effective on the volume dependent variable. The Pearson correlation coefficient between volume and mean canopy height was found to be r=0,776 for Pix4D (medium), r=0,667 for Pix4D (high), r=0,767 for Agisoft (medium), r=0,764 for Agisoft (high). According to the results of multiple regression analysis, determination coefficient 〖(R〗^2) between volume and independent variables was found to be R^2=0.733 for Agisoft (medium), R^2=0.65 for Agisoft (high), R^2=0.741 for Pix4D (medium), R^2=0.684 for Pix4D (high). According to these results, it was concluded that data processing with medium precision would be more accurate and faster. As a result, it has been seen that RGB images obtained from the UAV are very useful and low cost in extracting the canopy structural parameters. Also, the Excess Green Index has been seen to be successful in extracting vegetation zones and coverage areas. It has been proven that UAVs can be used in precision agriculture applications, plant growth and development can be observe with this method.