Uma proposta integrada de software e hardware para monitoramento de insetos-praga em plantações na forma de uma armadilha inteligente

Abstract

Integrated pest management (IPM) is a technique that seeks to keep pests always below the level at which they cause damage to crops. IPM also seeks to carry out pest control that is practical, efficient, economical and does not harm the health of the population, aiming at reducing the use of pesticides in plantations. A technique widely employed in monitoring the population of pests is based on the use of traps. However, the usual method consists of regular field surveys with the visual observation of traps by a human operator. This method has drawbacks because it is intensive, unhealthy and costly for the operator, and it is not possible to obtain real-time and synchronized information of all traps. An alternative option is to use intelligent traps that are capable of collecting and processing images, identifying and counting insects and sending this information to a central server. However, the task of identification is not trivial, since many factors can impair this process, such as insect overlap, high variety of captured insects, low illumination, among others. Based on this context, this work aims to develop an intelligent trap prototype to automate the whole process of pest insect counting in plantations. To achieve this goal, a neural network was trained to identify the insect pests Ceratitis capitata and Grapholita molesta in images obtained inside the developed traps. In addition to the identification, the counting of the captured pest insects is also performed, in which the counting result is sent directly to a remote system, where the plant pest insect population will be monitored. The identification step presented satisfactory results, with an accuracy higher than 93%. The counting step also presented promising results, however, as a result of the factors mentioned above, some errors were introduced in this step, showing that there is still room for improvement.