Utilizing hyperspectral and multispectral remote sensing and geographic information systems to identify and differentiate weed and crop species

Weed species are not evenly distributed across a field; thus, if remote sensing imagery could be utilized as a tool for locating and identifying these patches, herbicides could be applied according to species and spatial distribution. In order to utilize site-specific technology to apply herbicides, the spatial variability of weed populations within a field must be known. Research was conducted to determine if remote sensing could be utilized as a tool for identifying and separating weed species from cotton and corn. Additionally geographic information systems and herbicide decision aids were tested to determine if weed population mapping could result in accurate site-specific and multiple product herbicide applications.; Species separation based on hyperspectral data was successful in differentiating cotton and corn from weed species. By identifying the best spectral bands and utilizing vegetation indices, species classification ranged from 67 to 99% for cotton, 47 to 98% for corn and 95 to 98% for weed species. Species identification increased with data accumulation later in the growing season due to increased leaf area and biomass. Results indicate that weed and crop separation is possible when the best spectral bands are identified and utilized in conjunction with vegetation indices.; Multispectral imagery was also tested for species identification and was found to produce lower classification accuracies than hyperspectral data. Cotton and corn were classified 50 to 100%; however, weed species classification was poor (<50%) with multispectral imagery. Poor classification was observed because only 4 regions of the electromagnetic spectrum were utilized. Multispectral identification would likely improve if the best spectral bands identified in the hyperspectral research were utilized in sensors for aerial imagery.; Weed species and density information was mapped utilizing a geographic information system. Site-specific and broadcast herbicide applications were made based on results from the herbicide application decision support system (HADSS) program. Weed efficacy or cotton yield did not differ between site-specific or broadcast applications. However, herbicide product, along with associated costs, was reduced with site-specific applications.; Overall results from this research provide new technological advances in weed and crop identification as well as controlling weed patches with state of the art site-specific, point-injection technology. The identification of the best spectral bands for weed and crop separation and utilization of the point-injection site-specific sprayer may revolutionize future weed control practices.