A Bag of Visual Words-Based Approach to Identify Scenarios Suspected of Being Breeding Sites of the Aedes aegypti Mosquito from Aerial Images Acquired by UAVs

Resumo

In addition to the programs of the Brazilian Ministry of Health to prevent and combat the Aedes aegypt
mosquito, the use of unmanned aerial vehicles − UAVs (known as drones) has proved to be an important alternative
to assist the work of health surveillance teams. However, the analysis of aerial images acquired by such equipment
are usually manual, which can be time-consuming for health workers. Thus, it becomes important the proposition
of computational approaches able to recognize and interpret patterns in such images. This work proposes a
computer vision approach to identify scenarios that represent potential breeding sites of the Aedes aegypti
mosquito from aerial images acquired by UAVs. The proposed approach employs the Bag of Visual Words −
BoVW technique combined with the Support Vector Machine − SVM classifier (BoVW + SVM), taking into
account two descriptors based on the SIFT − Scale Invariant Feature Transform algorithm (Transformed color
SIFT and RGB-SIFT), and the descriptors Color Level Co-occurrence Matrices − CLCM and Color Histograms.
For conducting the experiments we compose a database of images, acquired in urban regions of the city of São
Paulo, which contemplate real and simulated scenarios suspected of being breeding sites of the mosquito (gutters
and roofs with accumulation of objects, open-air garbage containing old tires, old tires, pet bottles, plastic and
paper packaging and other open containers that can accumulate water). The results obtained in the experiments
with BoVW+SVM, in terms of hit rate, were: RGB-SIFT (91,37%), Transformed color SIFT (85,34%), CLCM
(64,65%) and Color Histograms (86,20%).