A Time-Delay Feedback Neural Network for Discriminating Small, Fast-Moving Targets in Complex Dynamic Environments

Abstract

Discriminating small moving objects within complex visual environments is a significant challenge for autonomous micro robots that are generally limited in computational power. By exploiting their highly evolved visual systems, flying insects can effectively detect mates and track prey during rapid pursuits, even though the small targets equate to only a few pixels in their visual field. The high degree of sensitivity to small target movement is supported by a class of specialized neurons called small target motion detectors (STMDs). Existing STMDbased computational models normally comprise four sequentially arranged neural layers interconnected via feedforward loops to extract information on small target motion from raw visual inputs. However, feedback, another important regulatory circuit for motion perception, has not been investigated in the STMD pathway and its functional roles for small target motion detection are not clear. In this paper, we propose an STMD-based neural network with feedback connection (Feedback STMD), where the network output is temporally delayed, then fed back to the lower layers to mediate neural responses. We compare the properties of the model with and without the time-delay feedback loop, and find it shows preference for high-velocity objects. Extensive experiments suggest that the Feedback STMD achieves superior detection performance for fast-moving small targets, while significantly suppressing background false positive movements which display lower velocities. The proposed feedback model provides an effective solution in robotic visual systems for detecting fast-moving small targets that are always salient and potentially threatening. Index Terms-Neural system modeling, visual neural systems, small target motion detection, time-delay feedback, complex background. I. INTRODUCTION

Bio-Inspired Small Target Motion Detection With Spatio-Temporal Feedback in Natural Scenes

“…As shown in Fig. 6 and 7 of the reference [16], the time-delay feedback shows a preference for fast-moving objects. Specifically, after applying the time-delay feedback, the neural response to a slow moving object would be largely suppressed while that to a high-velocity object would be properly maintained.…”

“…• We demonstrate the proposed model overcomes the limitation of the time-delay feedback model [16] and performs well without requirement of that small targets are faster than surrounding backgrounds. We organize the rest of this paper as follows.…”

“…Although these feedforward models exhibit parts of selectivities the same as the STMD neurons, they generally yield a number of false positives in dealing with small target detection against complex dynamic environment. To suppress false positive background movement, a time-delay feedback mechanism was incorporated into the STMD-based neural network [16]. However, it can only filter out slow-moving background motion and would be powerless against those false positives with high speed.…”

“…It is worthy mentioning that the time-delay feedback model [16] only uses the time-delay version of the neural output as feedback signal, ignoring the spatio-temporal background dynamics φ(t, s) and ψ(t, s). As shown in Fig.…”

“…The robustness of these models needs to be improved in dealing with small target detection against complex natural background. To suppress false positive background movement, time-delay feedback connection was introduced into the STMD-based methods [16]. However, the time-delay feedback model makes certain assumption that small targets should be faster than background, which means it can only filter out those slow-moving background false positives.…”