We proposed a method to improve accuracy and spatial resolution of urban extent map by integration of ASTER/VNIR images and existing urban extent maps. The proposed method mainly consists of 1) unsupervised clustering on ASTER/VNIR images, 2) extracting urban clusters from clustered ASTER/VNIR images by overlaying existing urban extent maps, 3) calculating conditional probability under condition of urban or non-urban on the extracted urban clusters and existing urban extent maps against ground truth. We developed urban extent map with ASTER/VNIR images, MOD12Q1 and GRUMP by proposed method at 101 sites of the world. As the result, spatial resolution is improved to 15m, and user's accuracy and producer's accuracy of developed map is 56% and 62%, better than those of MOD12Q1 and GRUMP. The proposed method will be useful to improve accuracy and spatial resolution of urban extent map. In addition, the conditional probability indicates effect of the integration. Investigation based on the conditional probability will clarify way of improving accuracy and spatial resolution.
In order to address the issues arising due to the various environmental problems that are currently attracting attention, it is necessary to devise a method that enables wide-area monitoring of fluctuations in vegetation conditions such as variations in moisture and temperature and land cover changes. By taking advantage of both the high temporal resolution and the wide swath mode of multi-temporal satellite data, such as NOAA/AVHRR, MODIS, and SPOT/vegetation, it is possible to perform high-frequency monitoring of wide-area, land cover changes. However, since the multi-temporal satellite data are influenced by clouds and system noise, in many cases, they must be processed in order to accurately represent the actual surface conditions. This study describes the development of a multi-temporal, spectrum anomaly-detection method, which takes into consideration wide-area seasonal changes, based on SPOT/vegetation S10 products. To reduce the effect of clouds on the reference-year data, the spectral information of the pixels was first converted to characters, and the influence of clouds was eliminated through a time-series modeling using hidden Markov models. Since the anomaly-detection method requires a clustering of character strings, a dedicated software based on the self-organizing map algorithm was developed. The data for anomaly detection is not dependent on the information of neighboring pixels, and it is possible to detect an anomaly even if there is only one pixel. By applying this method, we were able to detect a burned scar in Far East Russia. Once the parameters necessary for the calculation of the anomaly-detection score are obtained, anomaly-detection processing at 10-day intervals can be performed using a personal computer.
Specular reflection on the surface of target material causes huge error in the measured spectral reflectance. It is important to eliminate the specular reflection components for estimation of the diffuse reflection components that shows the true spectral characteristics. Today, a study about separating specular and diffuse reflection components has been vigorously conducted in the field of computer vision and the computer graphics. Various methods have been proposed to separate these two components using the polarizing filters and so on. But this method should install the polarizing filters in the instruments to polarize the incident light. Therefore, this method is not suitable to use for a spectral measurements because the optical characteristic is changed. In this paper, we propose a novel method based on dichromatic reflection model with multiple viewing angle reflection data to estimate the diffuse reflection components from the measured spectral reflectance data of the material surface with luster such as leaves of plant and plastics without any additional filters and instruments. Finally, an experiment using a color sample sheet is executed for demonstrating the feasibility of the proposed method.