Topological Mapping for Event-based camera using Fast-GNG and SNN

In recent years, research has been conducted on digital transformation toward the realization of Society 5.0. Among this research, there has been little research on monitoring that takes advantage of the features of 5G communications. To realize such ultra-real-time monitoring, it is necessary that predict and respond to the movement of the measurement target without large-scale computer simulations. Therefore, we focused on Event-based cameras. Unlike conventional cameras, event-based cameras are inspired by living organisms in which each pixel outputs luminance changes asynchronously and have excellent features such as high time resolution, high dynamic range, low latency, and low power consumption. These features have attracted attention in the fields of robot vision and computer vision, and the system is a compatible solution to realize such monitoring. We have also used topological mapping techniques to deal with various problems, and we are convinced that such an approach will also be effective for Event-based cameras. However, a lot of noise must be dealt with when trying to learn topological structures using Growing Neural Gas (GNG). However, Event-based cameras are, in principle, completely different from conventional cameras and require a novel methodology. Eventbased cameras work well with Spiking Neural Networks (SNN) because of their principle. Therefore, to realize ultra-real-time monitoring, this paper proposes a model to obtain the topological structure of an Event-based camera with noise reduction by using SNN and GNG.