Crowd-sourcing Applied to Photograph-Based Automatic Habitat Classification

摘要: Habitat classification is a crucial activity for monitoring environmental biodiversity. To date, manual methods, which are laborious, time-consuming and expensive, remain the most successful alternative. Most automatic methods use remote-sensed imagery but remotely sensed images lack necessary level of detail. Previous studies have treated habitat as an image-annotation problem developed framework that uses ground-taken photographs, feature extraction random-forest-based classifier to automatically annotate unseen photographs with their habitats. This paper builds on this previous two new contributions explore benefits applying crowd-sourcing methodologies collect, classify First, we Geograph, photograph website, collect larger geo-referenced database, over 3,000 11,000 We tested original much database show it maintains its success rate. Second, mechanism obtain higher-level semantic features, designed improve limitations visual features Fine-Grained Visual Categorization (FGVC) problems, such classification. Results inclusion these improves performance framework, particularly in terms precision.