Integration of high resolution satellite imagery and LiDAR data for forest damage detection Hitoshi Taguchi, Yuichiro Usuda, Hiromichi Fukui and Kuniaki Furukawa tagchan@iis.u-tokyo.ac.jp, usuyu@bosai.go.jp hfukui@sfc.keio.ac.jp, furu@forest.rd.pref.gifu.jp 0-0 -- 0-000 - -0 0- - Institute of Industrial Science, University of Tokyo, -- Komaba, Meguro, Tokyo, -0, Japan National Research Institute for Earth Science and Disaster Prevention, - Tennodai, Tsukuba, Ibaraki, 0-000, Japan Keio University, Graduate School of Media and Governance, Endo, Fujisawa, Kanagawa -0, Japan Gifu Prefectural Research Institute for Forests, - Sodai, Mino, Gifu, 0-, Japan
0 Abstract Fallen (i.e. snow damage and wind thrown) and withering (i.e. disease and insects) of trees in abandoned forests are one of the major problems in forestry. However the current investigation method relies on a ground survey, which is difficult to grasp the conditions extensively. Recently, usage of high spatial resolution satellite imagery and LiDAR (Light Detection And Ranging) data are anticipated as an effective solution for the forest monitoring. High resolution satellite imagery is effective for detecting withered and fallen damage, although this data has a difficulty in distinguishing between withered and fallen damage. Digital Surface Model (DSM) and Digital Elevation Model (DEM) which are made from LiDAR data are effective for detecting fallen damage, although this data has a difficulty in detecting withered damage. In the developing method, integration of high resolution satellite imagery and LiDAR data were utilized to detect two types of damage separately at same time. Multinomial Logit Model (MLM) was utilized for integrated processing. Red, NIR channel and gap areas detected by DSM and DEM were dependent variables for MLM. This method was examined on the IKONOS Multispectral Imagery and LiDAR data in the test area. Accuracy assessments were conducted from the aspect of omission (User s accuracy) and commission (Producer s accuracy). In withered damage detection, % and % of pixels were correctly detected, respectively. In fallen damage detection, % and % of pixels were correctly detected, respectively. From these results, this method was demonstrated that integration of two data can detect fallen and withering damage in high accuracy. 0 Keyword High resolution satellite imagery; LiDAR data; Fallen damage; Withered tree; Integration
0 0 ) LiDAR (Light Detection And Ranging) ) ) LiDAR Digital Surface Model( DSM) ) ) LiDAR ) )
LiDAR Digital Elevation Model( DEM) DSM ) Fig. Table. ( ) LiDAR( ) ( ) 0 ) ) LiDAR ) Fig. Multinomial Logit Model( MLM) 0 LiDAR MLM LiDAR DSM DEM DSM )
0 0 MLM ) Logit Model MLM MLM Seto and Kaufmann ) MLM (U ) 0 i n i ( U ) () in U in = β x + β x + L+ β x + ε in in k kin in () 0 k x inlx β kin Lβ k () ( V in ) ( ε in ) ( V in ) ( ε in ) n i P () in P in = e V 0n Vin e Vn + e + L+ e V in () P in
0 ( i = 0 ) ( i = ) ( i = ) k = ( ε in ) MLM Wald R Fig. 0 000m 00m 00 ) 00 IKONOS ( m IKONOS ) Space Imaging DN )
0 0 IKONOS LiDAR 00 LiDAR Table. m DSM DEM LiDAR DEM 0).m 00. IKONOS LiDAR IKONOS LiDAR LiDAR 00 IKONOS 00 00 IKONOS IKONOS Fig.
0 0 00 Fig. Table. DSM Wald 0.0% 0. Table. Fig. m 0 IKONOS IKONOS
0 0 Fig. Fig. ( ) IKONOS ( ) ( ) ( ) Fig. ( ) ( ) IKONOS ( ) Fig. IKONOS MLM ( ) ( ) IKONOS LiDAR
0 (User s accuracy) 0 0 (Producer s accuracy) Table. % 0%. 0 LiDAR LiDAR MLM ) ) LiDAR ) MLM 0
MLM GIS
SFC GIS ) :,,, 00. ),, :,, (), pp.-, 00. ) : IKONOS,, (), pp.-0, 00. ),, : LiDAR,, (), pp.-, 00. ),,, : LiDAR,, (), pp.-, 00. ) J. C. White, M. A. Wulder, D. Brooks, R. Reich and R. F. Wheate : Detection of red attack stage mountain pine beetle infestation with high spatial resolution satellite imagery. Remote Sensing of Environment,, pp.0-, 00. ) M. Schwarz, C. Steinmeier, F. Holecz, O. Stebler and H. Wagner : Detection of windthrow in mountainous regions with different remote sensing data and classification methods, Scandinavian Journal of Forest Research, (), pp.-, 00. ),, : LiDAR (LIDAR ),, (), pp.-, 00. ) K. C. Seto and R. K. Kaufmann : Using logit models to classify land cover and land-cover change from Landsat Thematic Mapper, International Journal of Remote Sensing, (), pp.-, 00.
0),,,,, : DEM,,, pp.-0, 00. ),, :,,, pp.-, 00. ) J. R. Dymond and J. D. Shepherd : Correction of the topographic effect in remote sensing, IEEE Transactions on Geosciences and Remote Sensing, (), pp.-,. Space Imaging, 00. Space Imaging Document IKONOS relative spectral response and radiometric calibration coefficients http://www.spaceimaging.com/products/ikonos/spectral.htm (accessed Jan. 00)
High Resolution Satellite imagery Result Withered Fallen Withered Withered No damage LiDAR Fallen Withered DSM DEM Fig. Forest damage categories which can be detected by high resolution satellite imagery and LiDAR data. Table. Forest damage categories which can be detected by high resolution satellite imagery, LiDAR and Integration of and. High resolution satellite imagery LiDAR Integration of and Withered Fallen No damage
. Processing for explaining variables High resolution LiDAR Satellite imagery Radiometric caribration Gap extraction Resampling Red band NIR band Gap extracted data.estimation of MLM and apply to whole image Selection of training Parameter Estimation Evaluation of estimated model Apply to whole image Forest Damage Map Fig. Flow chart for forest damage detection
Gifu prefecture Withered Kilometers Fallen Former Minami village (Gujo city) R,G,B =,, 0 0 00,00 m Fig. Test site and IKONOS imagery Table. Specifications of LiDAR data
Table. Estimated parameters Withered (Wald statistics) Fallen (Wald statistics) Intercept -.0 (. ) -. (. ) Band 0.00 (. ) 0. (. ) Band -.0 (. ) -. (. ) Gap -. (. ). (. ) 0.0% significance level Table. Accuracy assessment of training data Training data Withered Fallen No damage Withered Choice Result Fallen No damage Sum Accuracy.%.%.% Sum
0 0 0 0 0 m No Damage Withered Fallen Fig. Comparison of training data and results
Fig. Result of the test site
Result Aerial Photograph IKONOS(RGB=) Gap data 00m Fig. Comparison of a result and other data. Result of this study (top left), Aerial Photograph (top right), IKONOS (bottom left) and gap data (bottom right). Legend of the result (top left) is same as Fig.. Result Aerial Photograph IKONOS(RGB=) Gap data 0m Fig. Comparison of a result and other data. Result of this study (top left), Aerial Photograph (top right), IKONOS (bottom left) and Gap data (bottom right). Legend of the result (top left) is same as Fig.. Yellow circles show low tree height area. 0
0m IKONOS IKONOS + Gap data ( this study) Fig. Comparison of result used only IKONOS band and (left) and result of this study (right). Legend of the results is same as Fig.. Yellow circles show significant difference between left and right image. Table. Accuracy assessments User s Accuracy Producer s Accuracy Withered 0 % 0 % Fallen 0 % 0 %