Automate green coverage measure using a novel DIA method: UIP-MGMEP

Green coverage is an important indicator of health status of rangeland, and it´s meaningful to improve the efficiency, accuracy and objectivity of green coverage measurement. Digital image analysis (DIA) technique becomes more attractive; however, there have been few advances in automatic measurement methods. In this paper, we automate rangeland green coverage measure by DIA, which is a novel threshold segmentation algorithm named as UIP-MGMEP applied to the vegetation index (Excess Green (EXG) in this paper). UIP-MGMEP is not recommended to images whose fraction of vegetation is smaller than 0.5% or larger than 99.5%, which could be rounded to 0% or 100% visually. UIP-MGMEP optimizes EXG threshold by searching the upper inflexion point (UIP) of the M-Et curve (mean gradient magnitude of edge pixels (MGMEP) vs. EXG threshold), based on the assumption that EXG variance of either background or vegetation is suppressed. The results show that UIP-MGMEP works well, and fails when either background or vegetation is too complex. UIP-MGMEP achieves accurate and objective green coverage measure without any human intervention. However, UIP-MGMEP is only developed to extract green-leaved vegetation, not suitable for non-green (even grayish-green) leaved vegetation. Accuracy could be improved through using VIS-NIR camera instead of VIS camera.