AUTOMATIC DETECTION OF MOVING WILD ANIMALS IN AIRBORNE REMOTE SENSING IMAGES Yu Oishi 1 and Tsuneo Matsunaga 2 1 University of Tsukuba 2 National Institute for Environmental Studies 1. INTRODUCTION The coexistence of wild animals with people needs proper conservation and management measures of wild animals including population control. However, there is not enough necessary information for management of large-sized mammals such as population sizes, habitat, etc [1]. This is partly because many of the mammals are nocturnal animals, and it is difficult to observe the large-sized mammals because of their large habitat area [2]. Therefore in Japan population densities of large mammals have been estimated using direct and indirect methods, although these methods are labor-intensive and require long-term research. Thus, it is expected to estimate population densities of large mammals using remote sensing. In this study, we developed an algorithm for automatic detection of moving wild animals in the snow in airborne remote sensing images with 60 % overlap taken in Sarufutsu Village, Hokkaido in Japan [3][4]. This is the first report on automatic detection of wild animals from remote sensing images except infrared thermal imagery in Japan. 2. DATA The study area is Sarufutsu Village, northern Hokkaido in Japan. We used airborne remote sensing images taken in the basin in the Sarufutsu River that flows in the south of Sarufutsu Village (45°12’N, 142°2’E, Fig. 1). Fig. 2 shows birds-eye view of the taken area from the Okhotsk Sea generated from ASTER 3A01 product. A geographical feature of this area could be as follow: 100 ± 60 m above the sea, low difference of elevation, many meandering flows of river, and an artificial forest. The 68 airborne remote sensing images used in this study were collected May 3, 2006 at an altitude of 750 m with 60 % image overlap at an interval of 3 seconds. The pixel resolution of the imagery is about 8 cm. The image size is about 600 m by 1 km. And these images have four bands in the visible – near infrared region. The target species in this study are brown bear, sika deer, fox, and human which can be recognized in 8 cm resolution imagery.