Biomimetic target detection: modeling 2nd order correlation of OFF and ON channels

Abstract

Flying insects have proven to be valuable animal models in elucidating the computational processes underlying visual motion detection. For example, wide-field motion processing (optic flow) in the insect visual system has been investigated from both behavioral and physiological perspectives [1] resulting in computational models with diverse applications [2,3]. However, insects must also extract the movement of prey or conspecifics from their environment, therefore have the capability of detecting and interacting with small moving targets, even amidst swarms [4,5]. Using electrophysiological techniques, we record from small target motion detector (STMD) neurons in the insect brain, likely to underlie these behaviors. We previously proposed an 'elementary' small target motion detector (ESTMD) model of STMDs that accounts for their size and velocity tuning, as well as their ability to discriminate targets against cluttered surrounds [6-8]. However, other properties such as directionality [9], facilitated responses [10] and high contrast sensitivity [11] are not accounted for in this model. Here we propose two model variants that cascade the ESTMD with a traditional motion detection model, the Hassenstein Reichardt 'elementary motion detector' (EMD) [12]. We show that these elaborations maintain core STMD properties as well as model directionality and highly expansive contrast sensitivity. By encapsulating the properties of STMD neurons we aim to develop computational models that can simulate these insects remarkable capability of target discrimination and pursuit for application in robotics and artificial vision systems. © 2013 IEEE.