Methodology for urban vegetation segmentation based on cartesian hue and saturation characteristics using UAV images

Abstract

A renewed importance in the urban vegetation segmentation task has been growing over the last years, mainly due to new urban planning and management projects, which require proper data on green spaces. Although multi-spectral sensors and complex systems for image acquisition that support this task are standard in some works, they show disadvantages in cost and spatial resolution. Unmanned aerial vehicles (UAV) offer an affordable alternative so that they can get high-quality images. Such images deliver some challenges for the urban vegetation segmentation job, such as the color dispersion that urban green spaces present, vegetation greenness, and lightning conditions, like those seen in related works that use the most advanced devices. In this research, a cartesian chromatic histogram-based algorithm is proposed for urban vegetation segmentation in UAV images. The developed method uses morphological operators to enable the reduction in histogram color discontinuities. The tests that were carried about over sample images resulted in accuracy up to 98 %, surpassing the state-of-the-art tested techniques. The results validated the robustness and the accuracy of the proposal against different conditions presented in study cases.