抄録
This paper proposes the vision-based mowing boundary detection algorithm for an automatic lawn mower. The proposed algorithm utilizes the idea of a bank of filters, which consists of 36 Gabor filters, to capture the change of the features before and after mowing the lawn. The x^2 statistic is calculated as the feature value from the filter responses. The input image of lawn environment with two fields before and after mowing the lawn is classified into two classes by utilizing the x^2 statistic. The boundary between two classes is determined by RANSAC.
