Submitted to: International Conference on Precision Agriculture Abstracts & Proceedings
Publication Type: Proceedings
Publication Acceptance Date: 7/2/2006
Publication Date: 7/23/2006
Citation: Wiles, L., Waltermire, R., Giles, T., Bobbitt, R. 2006. Mapping Weed Presence in Dryland Crops. International Conference on Precision Agriculture Abstracts & Proceedings. Minneapolis MN, July 23-26, 2006. CD Will be released in December Interpretive Summary: Growers need maps of the distribution of weeds in their fields to reduce herbicide use with site-specific weed management (SSWM). Remote sensing is key to successful weed mapping and sophisticated hyper- and multi-spectral image-based systems appear promising for detecting weed patches and identifying species. However, methods of varying cost and complexity are needed for the widest adoption of SSWM. Our objective is to develop a system for mapping weed cover in dryland cropping that is low cost, easy to use, and does not require expensive GIS, GPS, and image analysis software. The premise of our design is based on the following: (1) growers typically make 3 to 5 trips across a fallow field to control multiple flushes of weeds with tillage or herbicides; (2) these flushes include most weed species of the crop rotation; and (3) weed patches are spatially stable. A field is mapped during fallow to avoid the difficulty of distinguishing crop and weed cover. The system includes a consumer digital camera and GPS unit, commercial software to map locations of images and view the image at each location, and image analysis and map viewing software that we developed. Weed cover is calculated as percent of pixels classified as green in an image. Locations are displayed using a color legend for percent weed cover and are linked for viewing original and analyzed images. Locations can also be mapped on aerial photographs, topographical maps, or shape files of information such as crop yield.
Technical Abstract: Weed maps are useful to monitor the effectiveness of weed control, detect new invasions, plan preventative weed management for following crops and use site-specific weed management to reduce herbicide use. Sophisticated methods to map weed populations are being developed with slow progress. We developed a system to map weed cover in fallow fields that is automated, inexpensive, easy to use and does not require expensive and complicated GIS, GPS or image analysis software. Our low cost system uses only readily available technology including a consumer digital camera and GPS unit plus commercial software to map the location of images. We developed software to estimate and map weed cover from digital images. In tests in 15 fallow fields, we were able to detect all weed species, including small seedlings. Also, weeds in shadows were recognized. A map of the variability of weed cover in a field can be displayed in our software with an aerial photograph or other maps for a field such as a yield map as the background. We have demonstrated that expensive and sophisticated technology is not always needed to map weeds in growers' fields. We will be able to distribute our software for free.