Dengue larvae detection and tracking using CNN and kalman filtering

Article type: Research Article

Authors: Sandhu, Muhammad Abdullah^{; *} | Amin, Asjad

Affiliations: Department of Information and Communication Engineering, The Islamia University, Bahawalpur, Punjab, Pakistan

Correspondence: [*] Corresponding author. Muhammad Abdullah Sandhu. E-mail: [email protected].

Abstract:  During the last decade, dengue fever has emerged as a life-threatening disease. Dengue fever is caused by the bite of the dengue mosquito, and it spreads rapidly especially in the rainy season due to the availability of water carriers inside and outside the living vicinity. In this work, we propose an automated model for dengue larvae detection and tracking using Convolutional Neural Network (CNN) and Kalman filters. Despite substantial literature available on object tracking, no model has been proposed for dengue larvae. We started our work by collecting water areas and dengue larvae datasets as no public datasets were available. Our water areas dataset has 30 videos of different containers and environments. The dengue larvae dataset has 50 short videos of dengue larvae having different locations, backgrounds, and textures. In the first step, we used CNN to detect water areas, and the detected water area is then processed for the detection and tracking of larvae. Next, we propose a Kalman filter-based workflow for dengue larvae detection and tracking. A Gaussian Mixer Model with background subtraction is applied for foreground and object detection. Then we used Kalman filters to track the moving larvae in the experimental videos. The proposed model shows excellent results considering the small size of larvae and the challenging dataset. Subjective and objective experimental results clearly show the superior performance of the proposed model. The feedback received from the health authorities has been encouraging and the work is expected to facilitate the health department in eliminating the dengue.

Keywords: Dengue larvae, Detection, Tracking, CNN, Kalman Filtering

DOI: 10.3233/JIFS-223660

Journal: Journal of Intelligent & Fuzzy Systems, vol. 44, no. 4, pp. 6387-6401, 2023