视觉导航在扑翼飞行器避障中的应用

Visual Navigation for Obstacle Avoidance of Flapping Wing Micro-Air Vehicles

摘要: 视觉传感器本身具有独立性、准确性和可靠性等优点，使得它在无人机视觉导航的应用领域中具有独特的作用，但是传统的视觉导航设备质量过大，无法应用在扑翼飞行器上。针对此问题，本项目利用FPGA芯片设计基于Census变换的局部立体匹配算法的立体视觉模块，该模块质量轻，并且能够进行图像的实时处理，满足避障场景的实时性要求。通过扑翼飞行器实际的飞行实验结果表明，本项目设计的视觉导航模块可以为无人机的自主避障飞行提供技术支持和决策依据。

Abstract: The visual sensor itself has the advantages of independence, accuracy and reliability, which makes it play a unique role in the application field of UAV visual navigation. However, the quality of traditional visual navigation equipment is too heavy to be applied in the bionic bird micro aircraft. In view of this problem, this project uses FPGA chip to design a stereo vision module based on Census transform local stereo matching algorithm. This module is light in weight and can carry out real-time image processing to meet the real-time requirements of obstacle avoidance scene. The actual flight experiment results of the Flapping Wing Micro-air Vehicles show that the visual navigation module designed in this project can provide technical support and decision-making basis for the autonomous obstacle avoidance flight of UAV.