Bio-inspired artificial vision and neuromorphic imaging devices

Abstract

In the past decade, significant advancements have been accomplished in image sensor technology for achieving efficient object detection with high signal to noise ratio that consists of image acquisition and image data processing. This advancement in image capturing and image processing has promoted major development in the field of mobile electronics and applications in machine vision. However, since the conventional imaging technology uses a planar image sensor, a multi-lens structure is required for wide visual field, high-quality and aberration-free imaging. This multi-lens structure makes the camera module large and bulky, which hinders the miniaturization of overall system. In this context, by mimicking a single lens and a curved retina structure of natural eyes that evolved to be optimized for their habitat, such technological limitations can be efficiently resolved. Also, by mimicking their strategies to obtain high quality visual information in their habitat, specialized camera with high object detectability and high signal to noise ratio can be developed. Here, we describe three different types of the next-generation bio-inspired artificial vision systems. Firstly, we develop a novel wide-field-of-view camera by integrating a tailored monocentric lens and a hemispherical silicon nanorod photodiode array inspired by the eye of an aquatic animal. Such an aquatic-vision-inspired camera offers the wide FoV, miniaturized module size, minimal optical aberration, deep depth-of-field, facile accommodation, and enhanced light sensitivity in one simple integrated device. Theoretical analyses and imaging demonstrations have corroborated the validity of the developed artificial vision system. Secondly, inspired by the cuttlefish-eye, we develop a high-contrast polarizing camera that specializes in the low-contrast shallow-water environment. By mimicking the W-shaped pupil of cuttlefish-eye, an artificial vision system compensates uneven vertical light profile. In addition, the cylindrical silicon photodiode array which contains high-density belt-like pixel profile, allowing to have wide horizontal-field-of-view inspired by curved retina of cuttlefish-eye. The flexible, linear polarizing film on top of the cylindrical silicon photodiode array enables high-contrast polarization imaging. Thirdly, inspired by the structural and functional features of the human visual recognition system, we report a curved neuromorphic image sensor array with reduced noise by employing a heterostructure of MoS2 and poly(1,3,5-trimethyl-1,3,5-trivinyl cyclotrisiloxane) (pV3D3). Such an artificial vision system features the photon-triggered electrical responses similar to the synaptic plasticity of neural network that consists of short-term plasticity and long-term potentiation. Such photon-triggered synaptic plasticity of the MoS2-pV3D3 phototransistor was experimentally and theoretically analyzed. The developed system can efficiently recognize the original image by reducing the noises of input images.