Aedes larvae surveillance system on smartphones and algorithms
P. Kukieattikool, L. Klinkusoom, Watcharakon Noothong, Piruin Panichphol, Porntipa Choksungnoen, J. Wisanmongkol, T. Demeechai · International Journal of Infectious Diseases · 2019
Purpose: The dangerous diseases: dengue fever, yellow fever, Zika virus and Chikungunya, which affect hundreds of millions of people each year, are transmitted by female Aedes mosquitoes. To prevent the spreading of the Aedes mosquitoes, the Aedes larvae should be under close surveillance, so that the responsible agencies can make a warning or prepare for the epidemic. In order for anyone without the expert knowledge to detect and report the existence of the Aedes larvae, which are very responsive to the change of light intensity and move abruptly, we propose the Aedes Larvae Surveillance System using Smartphones. Methods & Materials: The system comprises an application on smartphones that record a video at the time of flash-on and then the video is processed for counting the number of the Aedes larvae at the server. The video processing algorithms are based on the detection and tracking of the moving objects, which can be simply differentiated from the unmoved background. The proposed algorithm contain the following steps: frame extraction, frame stabilization to get rid of the jerky image motion caused by platform movement, background subtraction to capture the moving Aedes larvae when they expose to the flashlight, images segmentation using local threshold, modified Hungarian algorithm for the track assignment, larvae path classification based on pattern of the movement, and finally counting the valid paths. In order to investigate the performance of the algorithm, the 120 recorded videos of Aedes larvae in the light-color container taken by the developed smartphone application are processed in MATLAB. The detected Aedes Larvae obtained from the algorithm are compared with the ground truth obtained from visual inspection. Results: The result of the processing algorithm by means of Recall (Hit Rate) = TP/(TP+FN) is 88.02%, where TP is the true positives and FN is the false negatives. The False Alarm Rate (FAR) = FP/(FP+TP) is 43.29%, where FP is the false positives. Conclusion: The proposed system has a promising performance in detecting the Aedes larvae in the test environment. However the FAR is still quite high, thus we will try reduced it in the future.