Automatic disease diagnosis in plants using digital images

Abstract

The fast diagnosis of diseases in plants with economical value is essential to ensure food safety and avoid major losses caused by the spread of the malady. Two main problems may hamper such an objective: 1) permanently monitoring the whole crop by people capable of detecting the diseases is, in most cases, not viable; 2) in many cases the person that detects the symptoms does not have the necessary knowledge to identify its cause.Although there are some solutions that explore technology as a facilitator for a quick diagnostic, usually involving a reference database that can be consulted by the user, the arising of fully automated systems has been slow, especially in face of the importance of the problem. This situation is not due to lack of interest by the researchers, but to the lack of a database comprehensive enough to allow the development of really robust methods for disease diagnosis. Hence, as important as the development of the technique itself, is the need for databases capable of making it possible to completely validate the proposed methods.Therefore, this project has two main objectives: 1) to build a database containing images of symptoms and detailed descriptions of their causes and consequences; 2) to develop a method capable of providing a reliable diagnosis using the digital images provided by the users. The cultures to be considered initially will be cotton, soybean, maize, rice, bean, sorghum, sugarcane, citrus, grapevine, pineapple, coffee and cupuaçu, but others may be considered as the project evolves. (AU)