Automatic Detection of Scattered Garbage Regions Using Small Unmanned Aerial Vehicle Low-Altitude Remote Sensing Images for High-Altitude Natural Reserve Environmental Protection
Weiyang Chen, Yiyang Zhao, Tengfei You, Haifeng Wang, Yang Yang, Kun Yang
Abstract
Recently, some famous high-altitude nature reserves have been shut down due to tourist garbage pollution. In order to clean up such garbage more conveniently and quickly, a novel detection framework is proposed to automatically detect scattered garbage regions using low-altitude remote sensing of small unmanned aerial vehicles (SUAVs), and it contains the following steps. First, high-resolution, low-altitude, multitemporal remote sensing images containing scattered garbage regions are collected by SUAVs, and two data augmentation methods are proposed to expand the training samples. Second, low-altitude remote sensing image registration and target-level image change detection are used to extract the candidate regions of garbage. Finally, a deep learning detection network is adopted to classify the scattered garbage regions. Experimental results show that the proposed detection framework achieves a mean accuracy of 96.94% and provides better performances on the real dataset compared with state-of-the-art methods.