Performance of haar and LBP features in cascade classifiers to whiteflies detection and counting

This paper presents a comparison between the performances of LBP and Haar features in a Cascade Classifier, which aims automatically to count immature whiteflies (nymphs). The relevance of counting these nymphs is related to pest management, which leads to an increase in food production. Evaluate differences in weak classifiers can lead to better understanding and implementation of object detection using these classifiers. Photos of infested leaves of beanstalks were taken in order to set samples. Then, models were trained according to documentation and available literature. Using both Haar and LBP features results in high performance for these samples. The conception of LBP features as binary gives to LBP a rapid training and prediction performance, therefore LBP features is a better option for rural applications. Those results, which are presented and discussed.