无人机光电吊舱系统设计

DESIGN OF UNMANNED AERIAL VEHICLE ELECTRO-OPTICAL PAYLOAD SYSTEM

张彬

11649159

硕士

电子与通信工程

郝祁

2018-06-05

2018-07-06

哈尔滨工业大学

深圳

光电吊舱是现代智能无人机的关键的组成部件。军用级无人机已有不少高精度、高稳定性、多功能的光电吊舱解决方案。但是现有工业级无人机或消费级无人机各类光电吊舱的功能相对比较单一、精度较低、价格昂贵。其核心问题是缺乏软硬件一体化的设计框架以及对各类传感与地理信息缺乏有效的利用。随着近年来深度学习与计算机视觉技术的高速发展，传统机电一体化的光电吊舱能够融合更多的智能算法，实现进一步的信息化与智能化，应用于更广泛的领域。本篇毕业论文提出了一个增强型I2C 总线的光电吊舱技术方案，在软硬件一体化设计的框架下，充分利用各类传感器信息，应用了最新的计算机图像算法与深度学习算法，大大提高了光电吊舱的稳定性、精确性、及快速响应性，最终实现了基于运动平台的高准确率图像多目标检测与主动视觉追踪等功能。本课题的创新点主要包括以下三点：（1）提出并实现了一种增强型I2C 总线的光电吊舱技术方案（2）设计并实现了基于高精度编码器的三相无刷直流电机驱动电路（3）设计并实现了光电吊舱控制与视觉及深度学习算法的一体化通过本项目，对于提出的光电吊舱系统完成了软件和硬件一体化的开发，并利用3D 打印机制造了产品原型，在无人机与机械臂等运动平台上，对所提出的方案和想法进行了多方面可行性与系统性能验证。本项目的完成，不仅能够实现低成本、高性能、多功能的工业与消费级光电吊舱，还将大大推进智能无人机以及智能机器人在安保监控、巡线测绘、抗灾救援、精准农业等领域的应用，对今后在运动平台上感知融合、主动视觉以及定位导航等各方面的深入研究与产品化都将发挥着积极的作用。

The electro-optical payload (EOP) is the key component for UAVs. Various highly accurate, robust and multifunctional EOPs have been developed for military-grade UAVs. However, industry-grade or consumer-grade EOPs are known for their simple functionalities, low precisions and costly prices. The main problem lies in the short of a hardware-software co-design framework, as well as the underuse of multi-sensor fusion techniques and GIS databases. With the development of deep learning and computer vision technologies in recent years, the traditional electrical-mechanical-componentcentric EOPs will utilize more intelligent algorithms, and hence become more information-centric and intelligence-oriented for a much wider range of applications.This thesis presents an enhanced I2C bus based solution for EOP systems with the paradigm of hardware-software co-design. In addition to the multi-sensor fusion techniques, state-of-the-art image processing and deep learning algorithms are utilized to achieve the functionalities of high-accuracy multi-object detection and active object tracking.The novelties of this paper include:(1) developing an enhanced I2C bus based EOP solution,(2) developing a high-precision encoder based BLDC motor driver circuit,(3) developing a set of integrated EOP control, image processing and deep learning algorithms.This project develops a co-design framework for EOP systems, and the functionalities and reliabilities of a fast prototype have been tested and validated with UAVs and robotic arms. The outcomes of this project will help develop low-cost, highperformance, multifunctional industry-grade and consumer-grade EOP products, promote applications of intelligent UAVs within the areas of security and surveillance, monitoring and mapping, search and rescue, precise agriculture etc., and significantly benefit the research in many areas including multi-sensor fusion, active computer vision, localization and navigation.

光电吊舱 云台 无人机 计算机视觉

electro-optical payload Unmanned Aerial Vehicle computer vision

中文

联合培养

南方科技大学

张彬. 无人机光电吊舱系统设计[D]. 深圳. 哈尔滨工业大学,2018.