Position and direction estimation of wolf spiders, Pardosa astrigera, from video images

Abstract

The wolf spider is an important predator of some agricultural pests. The final objective in this research is to characterize hunting locomotion of the spider and to design a hunting robot that mimics the spider's hunting locomotion. To this end, this paper proposes an image processing procedure to estimate the two-dimensional position and direction of a wolf spider in an observation box from video images. A spider is first extracted from each captured image by background subtraction and extraction of the largest connected component. The extracted spider image is divided into the body part and the leg part in order to estimate the position and the direction in the following steps. The position of the spider is determined as the center of the body part. The direction of the spider is estimated by a combination of an initialization operation and a sequential computation. The initialization is executed by using image moments and the difference between the centers of the body part and the leg part. The sequential computation is performed by using directions at the previous or subsequent time step. The proposed estimation procedure is applied to three wolf spider individuals. This paper demonstrates that the proposed estimation procedure is valid for the three individuals and it is unaffected by sexual differences or differences in leg arrangements. The proposed estimation procedure is available for locomotion analysis of spiders.