Development and testing of image processing algorithm to estimate weed infestation level in corn fields

Abstract

The challenges of modern agriculture have led to the development of localized management tools which allow the rationalization of the use and application of pesticides, a reduction in production costs and the optimization of agricultural processes. This study was carried out to develop an algorithm capable of orienting weed control in the management of a corn crop, using digital image analysis to identify the level of weed infestation in the field. The seeds of six species of weed were sown in an experimental plot of corn, and daily images were captured for 40 days for the evaluation of the level of weed infestation (low, intermediate or high). The algorithm developed was able to target information about the plants and soil accurately and discriminate the residual information as referring to either the culture or weeds. The proposed algorithm has achieved 90% accuracy in identifying the level of infestation from images already evaluated by experts. The results can thus be used as part of weed control strategy, with the incorporation of the geographic coordinates of the image making possible the construction of a map of the level of weed infestation in the different areas where the crop is growing.