A 470mV 2.7mW Feature Extraction-Accelerator for Micro-Autonomous Vehicle Navigation in 28nm CMOS

📋 论文概要

该论文提出了一种用于微型自主车辆导航的特征提取加速器，采用28nm CMOS工艺，在470mV低电压下仅消耗2.7mW功率。解决了高精度特征提取算法在功耗受限嵌入式系统中无法高效运行的问题。

💡 主要创新点

• 首次在亚0.5V电压下实现低功耗特征提取加速，适用于微型自主车辆导航。
• 通过优化的架构和电路设计，在极低功耗下保持足够的特征提取性能。

🏷 关键词

📄 原文摘要

University of Michigan, Ann Arbor, MI Recently, computer vision technologies are being applied to smaller systems such as cell phones, digital cameras, and unmanned surveillance platforms [1]. Feature extraction is a critical step in these applications. However, high-quality feature extraction algorithms require high performance and power-hungry processing, making them unsuitable for power-constrained embedded systems unless their scope is restricted based on a specific application [4]. On the other hand, algorithms with relatively low computational cost exhibit poor extraction performance and can be used only in limited application spaces, such as face detection. Hence, there is a need for high performance and energy-efficient feature extraction for use in emerging mobile applications. This paper proposes a power-efficient speeded-up robust features (SURF) extraction accelerator targeted primarily for micro air vehicles (MAVs) with

👥 作者与机构

Dongsuk Jeon, Yejoong Kim, Inhee Lee, Zhengya Zhang, David Blaauw, Dennis Sylvester