Interactive system for integrated weed management. I. Weed identification

Abstract

The success of sustainable agroecosystems is closely linked to the availability of high quality information. This sort of information will optimize management and decision making, improving profits and lowering the agrochemicals impact in the agroecosystems. This is the aim of SITRAMA (Interactive System for Weed Management), a system that deals with the weed management within soybean crop context, using knowledge engineering techniques. SITRAMA comprises four modules. The first one deals with weed recognition at seedling stage; the second is concerned with the different control and management strategies which take into account the crop system variables; the third deals with crop losses taken into account weed density & distribution; and the fourth is concerned with economic analysis under conventional techniques (revenues/costs) or novel ones (margin maximisation, risks & uncertainty). The goal of this paper is to present the developement of the first module, - weed recognition -, stressing the strategies followed to achieve the identification goal. The strategies used simplify the conventional dicotomic classification, focusing into relevant characters of the seedling. In such a way, information required to the user is minimized, emulating an expert reasoning.