Identification of plants for wild flower garden

摘要

The use of wildflowers has become popular for creating aesthetic public landscaping. As weeding costs are rathr expensive, automation of weeding would help reduce maintenance costs. In this study, as a step towards automation of weeding, identification of plants for wildflower gardens was performed by images processing. Shape feature analysis was applied to the binary images to identify the plant as one of four species (aster, pansy, treasure flower, and California Poppy). As features, roundness, first invariant central moment, area, and perimeter were studied for each plant type. Mahalanobis' generalized distance was used in identification of the plants. The results showed that the best shape features for identifing these plants were roundness and first invariant central moment.