The Potential of Imaging from Ultra-Light Aircraft for Urban Tree Inventories: Case Study in Kaunas, Lithuania
Résumé
The presentation is aimed to introduce imaging system based on simultaneous use of frame-type Rikola hyperspectral and RGB/CIR cameras applied for inventories and monitoring of urban green spaces. The system is run from a manned ultra-light aircraft, equipped for aerial photography missions. An aerial imaging campaign over the city of Kaunas, Lithuania was carried-out in summer of 2015 and 2016, acquiring narrow-bands measurements in the 500-900 nm spectral domain at 16 and 60 bands respectively, with a full width at half maximum ~10 nm. Simultaneously, aerial images were shot using Nikon D800E camera, converted and reconfigured to capture in NIR, Red+NIR and Green+NIR bands. Several methodological solutions for image calibration, co-registering of separate image bands, atmospheric correction and photogrammetric processing are investigated. Information available from the inventory of green spaces in Kaunas city was used to guide the selection of trees for identification studies and validation. Six urban deciduous tree species were separated using tree crown level statistics, extracted from 16 visible-near infrared spectral bands hyperspectral images, and discriminant analyses with an overall classification accuracy of 63.1 %. Fusion of hyperspectral and color-infrared images increased the classification accuracy by 3%. The identification accuracy of three tree health condition classes ranged from poor to moderate. The influence of increased number of hyperspectral bands on tree species ability as well as more advanced image processing solutions will be tested using materials from year 2016 campaign. The presentation will also discuss the potential of combining images from different platforms, ranging from unmanned aerial vehicles to space satellite systems, including the technological and economic reasoning.