An intelligent multimedia system for identification of weed seedlings

An intelligent multimedia tool for identifying weeds was developed in order to facilitate weed identification in the seedling stage. The identification is based on a visual image approximation method and it is implemented using multimedia, hypertext, object oriented design and knowledge based technologies. The method of image approximation is based on a progressive approach to the actual seedling through a sequence of images with increasing definition. The implementation of this method has required the design of a visual identification key that includes the set of images with different abstraction levels and their logic relationship. Trials with small groups comprising of people with botanical knowledge and others without that knowledge showed that the method of image approximation makes it possible to reduce the botanical knowledge necessary for identification of weeds, as well as that, multimedia technology has a positive effect in the intuitiveness and the facility of use of the tool.