Application of neuromorphic visual processing in pedestrian detection technology

There have been many researches on computer vision for the purpose of vehicle safety applications, making use of diverse methodologies like complex vision algorithms. However, despite its effectiveness, computer vision algorithms sometimes lack the robustness of mammalian visual system for the application in dynamic environments in vehicle driving conditions. We propose that the neuromorphic visual information processing offers an alternative which mimics the robustness and flexibility of the primary visual cortex, based on Hubel and Wiesel´s experimentation and Hodgkin-Huxley formalism. The elements of implementation of neuromorphic visual system is introduced with the orientation tuned function of synaptic connections and the spiking neurons. The effectiveness of proposed neuromorphic visual processing system is evaluated for the safety enhancement via the 95% of pedestrian detection rate for electric vehicle´s virtual engine sound system.