Density Weighted Connectivity of Grass Pixels in Image Frames for Biomass Estimation.

Accurate estimation of the biomass of roadside grasses plays a significantrole in applications such as fire-prone region identification. Currentsolutions heavily depend on field surveys, remote sensing measurements andimage processing using reference markers, which often demand big investments oftime, effort and cost. This paper proposes Density Weighted Connectivity ofGrass Pixels (DWCGP) to automatically estimate grass biomass from roadsideimage data. The DWCGP calculates the length of continuously connected grasspixels along a vertical orientation in each image column, and then weights thelength by the grass density in a surrounding region of the column. Grass pixelsare classified using feedforward artificial neural networks and the dominanttexture orientation at every pixel is computed using multi-orientation Gaborwavelet filter vote. Evaluations on a field survey dataset show that the DWCGPreduces Root-Mean-Square Error from 5.84 to 5.52 by additionally consideringgrass density on top of grass height. The DWCGP shows robustness tonon-vertical grass stems and to changes of both Gabor filter parameters andsurrounding region widths. It also has performance close to human observationand higher than eight baseline approaches, as well as promising results forclassifying low vs. high fire risk and identifying fire-prone road regions.