Holo Pi: 便携式本地CV应用系统开题报告

1. 研究目的与意义

One of the points of performing this senior design is to familiarize me with both basic and advanced technologies in the modern software engineering field. The other point is that this design has a wonderful idea behind it. This project, after being finished, allows users wearing a glass-like device to capture a picture and get, read the information (e.g. what are the objects in the picture) of that picture in seconds.This senior design or project involves three different kinds of devices known as Microsoft Hololens (1st Generation), Raspberry Pi 3 model B and Intel Movidius Neural Compute Stick.To successfully accomplish my project, I need to develop two parts: a client UWP(Universal Windows Platform) application running on Microsoft Hololens (1st Generation) and a server Python application running on Raspberry Pi board computer connected with an Intel Movidius Neural Compute Stick. By doing this, I can learn how to code a functional UWP application using C# as well as a working script written in Python at the same time. It is challenging but on the other hand, has a positive effect on improving my technique skills.Besides, it is a cross-platform project (Hololens runs on Microsoft Windows 10 while Raspberry Pi runs on Raspbian, a Debian-based, operating system), therefore, I have to handle problems raised by this cause.The system allows users to snap a picture using Microsoft Hololens built-in camera and upload to Raspberry Pi, the board computer to which one or more Intel Movidius Neural Compute Sticks connect, for further image processing tasks. Users will retrieve processing results from Raspberry Pi and see them displayed on Hololens Screen. The whole system requires Hololens and Raspberry Pi to be associated with the same local network, which means if you turn on your smartphones hotspot, you can carry a portable CV system with you at anywhere you go. Besides, users are able to share the results among them only if they are under the same local network.

2. 国内外研究现状分析

Microsoft released the HoloLens for good reasons. One of the amazing things you can do with HoloLens is to detect objects in the real world and see the results in the virtual world right in front of your eyes using Microsoft cognitive service supported on Azure cloud. And that is what many individual developers do. They have developed and deployed some real working object detection applications on HoloLens based on Microsoft prototype. As to Raspberry Pi, developers find it powerful enough to perform deep learning procedures. However, peripherals are needed to present a complete detection process: collecting data with the camera and display the results with a monitor, which is not very flexible.

3. 研究的基本内容与计划

Tasks and Schedule1. Manage to transfer data between the HoloLens and the Raspberry Pi2. Manage to perform computer vision tasks utilizing the Raspberry Pi and the Intel Movidius Neural Compute Stick3. Manage to develop a UWP application on the HoloLens device4. Manage to implement service functions in a python application on the Raspberry PiSchedule is included in the attached files

4. 研究创新点

1. Utilize the mixed reality feature of HoloLens to present computer vision tasks without cloud services.2. Utilize the Raspberry Pi, a handy, single-board computer to simulate cloud service functions.3. Data sharing function4. Cross-device data transfer