说明：  针对目前嵌入式视觉系统的图像识别算法复杂、鲁棒性差等问题，提出了一种嵌入式智能视觉系统。该系统以进阶精简指令集机器 Cortex-A53 为核心搭建硬件平台，并搭载 Linux 嵌入式操作系统。基于 V4L2 接口设计图像采集程序，完成图像的采集及存储。通过移植学习框架，采用 Inception-V3 神经网络模型实现图像的智能化识别。实验结果表明: 该系统具有良好的采集效果，对不同物体和同一物体在不同姿态下的识别准确度均可达 100%，满足智能视觉系统的设计要求。
(Aiming at the problems of complex image recognition algorithms and poor robustness of embedded vision systems, an embedded intelligent vision system is proposed. The system is built on the hardware platform of the advanced reduced instruction set machine Cortex-A53 and is equipped with a Linux embedded operating system. The image acquisition program is designed based on the V4L2 interface to complete image acquisition and storage. Through the transplantation learning framework, the Inception-V3 neural network model is used to realize the intelligent recognition of images. The experimental results show that the system has good acquisition effect, and the recognition accuracy of different objects and the same object in different attitudes can reach 100%, which meets the design requirements of intelligent vision system.)