Support system for surveying moving wild animals in the snow using aerial remote-sensing images

Japan is one of the most diverse zoogeographic regions in the world. It includes subtropical to cool temperate zones, high land and low land, and a high ratio of forest and mountain or hill. Owing to their low densities, large geographical ranges, and the large size of forested and mountainous areas, population size information on large mammals in Japan is generally insufficient. To address this problem, we developed an automated support system for detecting wild animals moving over snow using two overlapping aerial images. The system reduces the number of man-hours required to survey moving wild animals from a large amount of aerial image data. The system consists of three newly developed algorithms which perform the following tasks: (1) feature point extraction for registering two images, (2) corresponding point identification by determining the correspondence between feature points, and (3) automatic detection of moving wild animals by comparing two images using a computer-aided detection of moving wild animals (DWA) algorithm. We applied the proposed algorithms to several types of aerial images to automatically extract cattle, deer, and a walking human. Furthermore, we conducted a survey of wild animals using the system and used it to detect a walking human. The number of man-hours required to conduct the survey was reduced by 90%. A further advantage of the system was that, since relief displacement effects do not cause false detection, the system can be employed in forested areas during the leaf-off period.