Feature Learning Based Approach for Weed Classiﬁcation Using High Resolution Aerial Images from a Digital Camera Mounted on a UAV
The development of low-cost unmanned aerial vehicles (UAVs) and light weight imaging sensors has resulted in signiﬁcant interest in their use for remote sensing applications. While signiﬁcant attention has been paid to the collection, calibration, registration and mosaicking of data collected from small UAVs, the interpretation of these data into semantically meaningful information can still be a laborious task. A standard data collection and classiﬁcation work-ﬂow requires signiﬁcant manual eﬀort for segment size tuning, feature selection and rule-based classiﬁer design. In this paper, we propose an alternative learning-based approach using feature learning to minimise the manual eﬀort required. We apply this system to the classiﬁcation of invasive weed species. Small UAVs are suited to this application, as they can collect data at high spatial resolutions, which is essential for the classiﬁcation of small or localised weed outbreaks. In this paper, we apply feature learning to generate a bank of image ﬁlters that allows for the extraction of features that discriminate between the weeds of interest and background objects. These features are pooled to summarise the image statistics and form the input to a texton-based linear classiﬁer that classiﬁes an image patch as weed or background. We evaluated our approach to weed classiﬁcation on three weeds of signiﬁcance in Australia: water hyacinth, tropical soda apple and serrated tussock. Our results showed that collecting images at 5–10 m resulted in the highest classiﬁer accuracy, indicated by F1 scores of up to 94%.
UAV remote sensing,serrated tussock,tropical soda apple,water hyacinth,weed classiﬁcation