Consumo de recursos de algoritmos de subtração de fundo

Abstract

There are several applications related to smart cities where you have cameras with low or medium resolutions, but there is no clarity about the impact of the resolution with the quality of the offer, and there should be more studies that balance resource consumption and quality of offer an image resolution. The following work presents a study on the consumption of computational resources and the impact on segmentation to perform a subtraction in the background of videos with 10 measures. The objective is to identify a classification that increasing the definition affects the consumption of resources and the quality of quantity. For this study, 10 measures, 38 algorithms (available in the literature) and 2 videos were adopted, and the average behavior of each resolution, execution time, memory consumption, processor consumption will be studied, and various metrics are assigned to evaluate the behavior of each resolution. As a partial result of this research, an exponential behavior of resource consumption was observed considering the 3 lowest resolutions, and that not always higher resolutions provide better quality of supply.