Automatic weed detection system and smart herbicide sprayer robot for corn fields

The goal of this paper is to develop a new weed detection and classification method that can be applied for autonomous weed control robots. In order to achieve this goal plants must be classified into crops and weeds according to their properties which is done by a machine vision algorithm. Plants growing between rows are considered as weed, while inside a row, where crops are mixed with weeds, a classification method is required. Accordingly in the initial step, plants pixels were segmented from background with an adaptive method which is robust against variable light conditions as well as plant species. After that, crops and weeds were classified according to features extracted from wavelet analysis of the image. Finally, based on positions of weeds, herbicide sprayers are told to spray right on desired spots. The whole algorithm is implemented in LabVIEW software which is appropriate for real-time in-field purposes. In order to evaluate the performance of the algorithm 73 corn field images have been taken and selected, overall classification accuracy of 95.89% was achieved.