The purpose of this study was to assess the potential for estimating important parameters of forest stands with small footprint LiDAR (Light Detection and Ranging) data. The study area was a plantation of Sugi (Cryptomeria japonicaD. Don) and Hinoki (Chamaecyparis obtusa Endl.) with plots of differing densities. We explored the relationships between field survey data and LiDAR data using scatter plots and regression analysis. The targeted parameters were stem number, mean tree height, mean DBH (diameter at breast height) and mean stem volume. DBH was indirectly obtained from the tree crown surface area, and the timber volume was indirectly obtained from DBH. We confirmed that the coefficients of determination were 0.80-0.94 and there was a high correlation in stand parameters, except for Hinoki mean tree height. However, the relative error of Hinoki mean tree height was only 0.22m. Hinoki relative error of DBH and timber volume was greater than that of Sugi. We assume that this was due to a decrease in accuracy of DTM (Digital Terrain Model), as the Last-pulse did not reach the ground because of interference by Hinoki tree crown structure. In addition, it is possible that there are inaccuracies in the tree crown surface area data due to other reasons.
The forest snow damage is the disaster which is caused by load of the snow that adheres to a tree canopy, and results in fallen trees. When snow damages occur, an administration needs to identify damaged areas. However the current investigation method relies on a ground survey, which is difficult to grasp the conditions of wide areas. In this study, we developed a forest snow damaged area detection method using high resolution satellite optical sensor imagery and LiDAR data. The method consists of following procedures, (1) detection of damaged areas using satellite optical sensor imagery by a discreet choice model, (2) detection of gap areas using DSM and DEM generated by the LiDAR data, and (3) assimilation of (1) and (2) . The assimilation of (1) and (2) enables the mutual complementation of each other's defects. The method was examined on the IKONOS multispectral imagery and the LiDAR data in the test area. Accuracy assessment was conducted from the aspect of omission and commission. From the aspect of omission, accuracy was evaluated by comparing the 50 randomly selected pixels of the result with aerial photograph interpretation. 47 pixels of 50 (94%) were correctly detected. From the aspect of commission, accuracy was evaluated by examining the result of (3) detection in 56 randomly selected pixels which damage was observed in aerial photography. 46 pixels of 56 (82.1%) were correctly detected. From these results, the method achieved high accuracy, and the effectiveness of the combination was demonstrated.
Convenient 3D measurements using consumer grade digital cameras are enormously expected in various fields with appearance of low cost and high resolution consumer grade digital cameras. In these circumstances, software for digital photogrammetry“3DiVision”was designed to perform convenient 3D measurement using consumer grade digital cameras. However, there are still problems for efficient digital photogrammetry. These problems include distance measurement for absolute orientation and previous interior orientation procedure, and these restrictions should be removed for ideal convenient photogrammetry using digital cameras. With this objective, integrated sensor with CCD and laser called as Image Based Integrated Measurement (IBIM) system was developed by the authors. The most remarkable point of the system is its ability to calculate both of exterior and interior orientation parameters without scale distance or ground control points which have exact 3D coordinates in object field. This paper focuses the IBIM system and its performance evaluations. Furthermore adaptability of this system to 3D modeling of topography and historical structures in architecture and archaeology are discussed from the view point of 3D noncontact measurement.
A program to produce DEM and orthoimage from the PRISM sensor is developed. A calculation of the pixel and line number on the raw images from the ground coordinates needs long time for the PRISM by the used method. The program reduces the computation volume by interpolating the pixel and line number in ground grid space and along ground height axis. The program worked properly using 1.6m grid test data obtained by the ADS40, which projection geometry is like the PRISM. Observing control points settled mainly in open sky area, root mean square error was 1.13m in vertical for the DEM, and 1.30m in horizontal for the orthoimage. These values represent errors of the DEM and orthoimage production procedure, and pointing error on the orthoimage. Comparing the DEM with DSM obtained by LIDAR measurement, root mean square error of the DEM was 3.59m in a mixed area of high buildings, low housing, woods, and crop fields. The error corresponds 5.7m for the PRISM.
The effectiveness of object-based classification using high resolution satellite data was examined to establish in applying to vegetation mapping. We compared object-based and pixel-based classifiers for secondary forests in a rural area in the east part of Chiba prefecture. The minimum distance classifier as the object-based classification, and the maximum likelihood classifier and the ISODATA classifier as the pixel-based classification were applied. The results showed that the overall classification accuracy and Kappa statistics of object-based classification were higher than those of pixel-based, ISODATA and maximum likelihood classifications (overall classification accuracy of object-based : 64.17%, maximum likelihood : 60.17%, ISODATA : 53.64% and Kappa statistics of object-based : 0.551, maximum likelihood : 0.497, ISODATA : 0.388, respectively) . Boundaries of each plant community were well extracted by object-based classification. This research clarified that the object-based classification method is useful and has high potentiality in vegetation mapping.