Optical Flow Motion Motion Perception Method Based On Multi-aperture Bionic Compound Eye

Biological vision as an important development prospect of navigation has attracted much attention.In recent years,multi-aperture imaging systems inspired by insect compound eyes have been greatly developed in scene motion perception,target detection and depth estimation,laying a foundation for visual navigation.In this paper,due to the low resolution of the multi-aperture bionic compound eye imaging system based on micro-lens array and its easy to be affected by illumination and outliers,the relationship between the sub-eye flow field and space of the bionic compound eye is constructed,and the ranging method based on the multi-aperture bionic compound eye is realized.Finally,a visual-based optical flow velocity measurement method for UAV is proposed by combining the two methods,and the experimental verification is carried out.The main work of this paper is as follows:Firstly,the multi-aperture bionic compound eye imaging system is classified and introduced,and the multi-aperture bionic compound eye imaging system based on micro-lens array is simulated by 300×400 array.The main optical flow estimation algorithms are introduced and their advantages and disadvantages are compared.Secondly,a 7×10 multi-aperture optical flow estimation method based on NSCT and guided filtering is proposed for bionic compound eye multi-aperture images.The idea of multi-scale analysis is used to preserve the motion details of the aperture edge and efficiently remove the outliers.The open data set and the real data captured by UAV are used for experimental verification,and the highly robust optical flow field and velocity field are established.Then,in order to solve the problem of multi-aperture parallax extraction and depth estimation and realize stereo ranging,the theory of binocular parallax extraction was used to establish a multi-plane ranging model to realize the scene depth estimation of the bionic compound eye with 3×3 aperture,and the parallax relationship was used to convert the real distance with the camera model,and the experimental verification was carried out for the real scene.The error of less than 8m can be achieved in the scenario of 100 m visual distance.Finally,in order to solve the limitations of ranging sensors in the traditional optical flow velocity measurement algorithm of traditional UAV in complex scenes,a vision-based optical flow velocity measurement method of UAV is proposed.The optical flow field is provided by the multi-aperture optical flow estimation method,and the distance to ground is measured by the bionic compound eye multi-plane model.Finally,the proposed method is verified by airborne experiments,and the velocity errors of different methods in the north and east direction are compared.The quantitative index proves that the proposed method has the highest accuracy in complex scenes.