Identification of Fruit Fly in Intelligent Traps Using Techniques of Digital Image Processing and Machine Learning

The integrated pest management (IPM) is an approach that seeks to keep pests always below the level at which they cause damage to crops. A technique widely employed in monitoring the population of pests is based on the use of traps and consists in regular field visits to make visual observations of this traps by a human operator. This method is intensive, unhealthy and costly for the operator. Based on this context, this work aims to take the first steps into the development of an intelligent trap, with its scope focused on developing a classification system of the insects Ceratitis capitata and Anastrepha fraterculus, using digital image processing and machine learning techniques. In this work, was built a dataset of collected images consisting of 99 images in which 4338 regions were extracted. Each one of these regions has been transformed into a feature vector based on the bag-of-words model. For the classification, the algorithms support vector machine (SVM), k-nearest neighbors (KNN), decision tree (DT) and Gaussian Naive Bayes (GNB) were used. SVM performed better than other approaches, achieving an accuracy of 86.38%. Due to its satisfactory accuracy, it is possible to state that it is feasible to classify different insects in intelligent traps.