Rancang Bangun Model Haar Cascade Classifier untuk Deteksi Jentik Nyamuk Otomatis pada Citra Digital

Abstract

Mosquito larvae are an important indicator in disease vector surveillance activities, such as dengue fever (DHF). However, the relatively small size of larvae and their random movement in water media make manual observation difficult, time-consuming, and potentially lead to errors in identification. These conditions encourage the need to develop technology-based detection methods that can provide faster, more accurate, and consistent results. The purpose of this study is to build a Haar Cascade model to detect the presence of mosquito larvae in digital images. The research stage begins with collecting a dataset in the form of positive images containing mosquito larvae objects and negative images containing water backgrounds without larvae. Next, the training process is carried out using the Haar Cascade algorithm that utilizes Haar feature extraction through integral images and a classification process with AdaBoost. The resulting model has successfully detected mosquitoes and met the minHitRate target (HR ≥ 0.995), however, the false alarm rate is still quite high, averaging 0.41, so further optimization is still needed to reduce the False Alarm Rate. This study shows that the Haar Cascade method can be used to detect mosquito larvae with a fairly good success rate, but false detections still occur so the model still needs to be improved.