Cogongrass is recognized as one of the most significant invasive species in the U.S. Cogongrass is a growing threat as it continues its rapid spread across the Southeast, reducing forest and pasture productivity, destroying wildlife habitat, impacting rights of ways and presenting an extreme fire hazard. Throughout the World it has gained notoriety as the "Worst Invasive Plant of Non-crop Lands". Treatment methods by chemical application can be effective. Detailed highly accurate mapping of cogongrass boundaries can be used to form a spatial prescription of chemical applications. The objective of this research is to develop cost-effective mapping methods using airborne imagery. The imagery can be used to classify the exact cogongrass extent and boundaries and to establish chemical application rates and geographic placement to achieve maximum effectivness at minimum cost. Followup imagery can be used to determine the effectiveness of the treatment. Cogongrass grows in distinctive patterns and has a characteristic color. Very high resolution imagery will allow distinguishing the well-defined cogongrass regions by a visual inspection. However, imagery at very high resolution will be too expensive for wide area use. Thus linear stips of high-resolution imagery will be collected simultaneously with more afforadble coarse contiguous data as a way to enable interpretation of the coarse data. OBJECTIVES: Use airborne remote sensing methods to classify amd spatially map ares of cogongrass growth. Use high resolution imagery transects as a means to support the training of a classifier that uses coarse, more affordable, imagery flown at higher altitudes. APPROACH: Use an existing airborne multispectral imagery collection system to collect coarse resolution (e.g., 1m) imagery over a large area suspected to have cogongrass infestation. Simultaneously fly very high resolution (e.g., 3in) imagery in strips over the same area using the same airborne camera system. The high resolution imagery strips will be used to visually classify the coarse resolution imagery that is collected at approximately the same time and environmental conditions. The imagery will include standard red, green, blue, and NIR as well as four additional bands to instrument the red-edge spectral region. A classifier will be developed that uses the high-resolution strips as truth training data to allow classification from the low resolution data. The objective is to enable an affordable classification approach from the combined coarse/fine imagery
