Improvement on object detection accuracy by using two compound eye systems

Abstract

Compound eye is a multiple apertures imaging device, indicates that it can be applied for three-dimensional object detection. In our previous report, an artificial compound eye system was developed for 3D object detection. The system consists of a layer of plano-convex microlenses and a prism-like beam steering lens. An innovative multi-position calibration method is developed to relate the incident light rays and the relevant image points. Theoretically, one compound eye system alone is capable of 3D objects detection. However, the detection accuracy is limited due to the relatively small baseline between the adjacent microlenses. In this work, an equivalent large baseline is obtained by using a two compound eyes system. Preliminary experiments were performed to verify the improvement on the accuracy of 3D object detection. The experimental results with two compound eyes are compared with that obtained by only one compound eye. Experimental results show that the system with two compound eyes can detect an object much more accurately, indicating the feasibility and flexibility of the proposed method.