One of the most important ecosystems in the Amazon rainforest is the Mauritia flexuosa\nswamp or â??aguajalâ?. However, deforestation of its dominant species, the Mauritia flexuosa palm,\nalso known as â??aguajeâ?, is a common issue, and conservation is poorly monitored because of the\ndifficult access to these swamps. The contribution of this paper is twofold: the presentation of a\ndataset called MauFlex, and the proposal of a segmentation and measurement method for areas\ncovered in Mauritia flexuosa palms using high-resolution aerial images acquired by UAVs. The method\nperforms a semantic segmentation of Mauritia flexuosa using an end-to-end trainable Convolutional\nNeural Network (CNN) based on the Deeplab v3+ architecture. Images were acquired under different\nenvironment and light conditions using three different RGB cameras. The MauFlex dataset was\ncreated from these images and it consists of 25,248 image patches of 512 Ã? 512 pixels and their\nrespective ground truth masks. The results over the test set achieved an accuracy of 98.143%,\nspecificity of 96.599%, and sensitivity of 95.556%. It is shown that our method is able not only to\ndetect full-grown isolated Mauritia flexuosa palms, but also young palms or palms partially covered\nby other types of vegetation
Loading....