In the present research, a model of three-dimensional structure of rice was developed. Three-dimensional position of rice leaves were measured along the edges of the leaves, and the position of stems were measured along the center line. The measured data were contaminated by systematic error due to the device and random error due to the wind and device. The proposed methodology reduces such errors through the following steps. First, the positional data of the center line of leaf was estimated from a pair of data measured at both leaf edges. Next, in xy-plane, principal component analysis (PCA) produced the first principal component, named as u-axis. Then, in uz-plane, a quadratic curve of the leaf was estimated by minimizing the sum of squared minimum distances between the points and the curve. The leaf width was modeled as a function of the leaf relative length to the full length. Finally, three-dimensional structure of rice was modeled. The estimated length and area of rice leaves were compared with the measured data. It was revealed that root mean square of errors (RMSE) of the estimated center length of all leaves was 7.4%, the relative error of the estimated areas of all leaves per rice individual was -6.8%, which is acceptable for the analysis of microwave scattering.
Phenology is closely related to carbon, water and other material cyclings on the earth's surface, and one of the most important indicators to understand vegetation responses to climate change. Satellite data provides us with a tool to monitor and assesses spatio-temporal variations in phenology. However, phenology based on satellite observation does not validated extensively with green-up or leaves dropping date acquired by ground based observation because of the lack in ground truth data for validation. By the way, Japanese Meteorological Agency (JMA) have observed the timing of flowering, budding, leaves coloring and dropping at more than 100 meteorological stations in Japan since 1953. In this study, green-up date detected from NOAA AVHRR observation was validated with JMA's phenological data, and four algorithms to detect green-up date were evaluated. As a result, spatio-temporal variation in green-up dates detected by newly developed algorithm was highly consistent with the ground data compared to another algorithms, and RMSE between the retrieved green-up dates and in situ data was 6.6 days. And, it was found that the algorithm was able to detect responses of green-up date to changes in annual temperature.
Recently orthophoto has been spreading in public, as the rectifying processes become easier under the development of digital photogrammetric workstation. Though it might be seemed familiar for public as the images are just like real, but some professional knowledge of photogrammetry are required to make interpretation relevant. This paper discusses geometric characteristics such as distortion and accuracy of large scale orthophotos. And these characteristics are analyzed as (1) distortion by height, (2) remaining distortion caused by central projection, (3) miss recognition without stereo view and (4) limitation of recognition of aerial photos. In this experiment, 30% of road is occluded and houses are misinterpreted by (1) low contrast, (2) low image quality, (3) surrounding trees and (4) structure misunderstood easily. And six instructions are presented to improve interpretation for houses. As the result, misinterpretation of houses on orthophotos is reduced from 25% to 9%. and become close to stereo interpretation.
Recently, pixel numbers of consumer grade digital cameras are amazingly increasing by modern semiconductor and digital technology, and there are many low-priced consumer grade digital cameras which have more than 10 mega pixels on the market. In these circumstances, convenient digital photogrammetry using consumer grade digital cameras are enormously expected in various application fields. With this motive, performance evaluations of consumer grade digital cameras for digital are photogrammetry a investigated using 24 kinds of cameras in this paper.
In recent years, plane type and ground type laser scanners have become to be widely used for constructing three-dimensional terrain data. However, when occlusion is occurred and lasers are not fully returned, lost of data happens and measurement noises contaminate terrain data. Terrain data are commonly represented with DEM (Digital Elevation Model) or TIN (Triangulated Irregular Network) . We propose, in this letter, an interpolation method for filling in lost DEM data. Through the experiments using sample data of lost DEM, we demonstrate the effectiveness of our interpolation method.