毕业论文答辩准备工作

first ten minutes is important

• give a confident introduction
• give your credentials (who your are ,and what can you do for me)
• deliver your hook
• introduce your agenda(agenda,do not read the agenda,I will should you …the summary)
• give credible statement

一：毕业设计开场白

各位老师好！我叫XXX，来自XXX，我的论文题目是“xxxxxxx”。在这里，请允许我向XX老师的悉心指导表示深深的谢意，向各位老师不辞劳苦参加我的论文答辩表示衷心的感谢。下面我将从论文的思想内容、结构框架、遣词造句三个方面向各位老师作一大概介绍，恳请各位老师批评指导。

Good Morning,ladies and gentlemen,I am xxx,a student majored in master robotics from xxxxxx, before the presentation,I would like to express my appreciation to the people who help me during the internship,and I would like to thank you all for coming here today and listening me.（礼貌用语，致谢） The topic of my master thesis is research in SLAM using plane information, and this presentation is divided into three main parts, some background will be shared ,I will mention main methods and finally a small video could be shown .（ 演示的）

二：毕业设计答辩导言

Here gives an agenda for today, ( we do not read the slide,just for audience to read themselves), Here, Mainly introducing aspects: the research purpose, background for plane features and the method of study and some results we have achieved. first we will start from a introduction. Q:这里面有一个疑问，是先说 introduction 还是 research purpose?

Research purpose

In a human-made scence, planes commonly exist and contain additional information in the SLAM problem. They are predominant features that are less affected by measurement noise. Planar features can significantly help SLAM systems to operate in indoor environments. They can effectively reduce errors, help establish longer-term pose constraints, and effectively reduce cumulative errors. When we meet the situation of low-texture environments and unstable light, plane information could play a supporting role.

论文摘要

This master thesis deals with an open sub-problem simultaneous localization and mapping(SLAM) in three dimensional space.The goal is to learn typical SLAM structure with plane information and make a progress with point cloud generation,plane information detection and think about concentrate abouthow to utilize plane information in the SLAM system.Necessary Jacobians for the optimization are computed.The reason why we would like to explore plane information is that in daily life environment,no matter indoors or outdoors,plane could give high level information in the local or global map.On the other hand,the techniques lie in the proposed SLAM system can facilitate applications in virtual reality, augmented reality(VR/AR) devices，construction Robots and automatic drive.

（再次修改稿）
This master thesis deals with an open sub-problem, simultaneous localization and mapping(SLAM) in three dimensional space with plane information.The goal is to learn typical SLAM structure with plane information and make a progress with point cloud reconstruction,plane information detection and concentrate about how to utilize plane information in the SLAM system.Necessary Jacobians for the optimization of point feature and plane feature are computed.We then simulated with a artificial data for optimization part.Since(之前的太啰嗦) in common place scenes,no matter indoors or outdoors,plane could give high level information in the local or global map.The techniques lie in the proposed SLAM system can facilitate applications in virtual reality, augmented reality(VR/AR) devices，construction Robots and automatic drive.

(8月23日再次改稿)
This master thesis deals with an open sub-problem, simultaneous localization and mapping(SLAM) in three dimensional space with plane information. Since in common place scenes, no matter indoors or outdoors, plane could give high level information in the local or global map(研究问题的必要性一句话概括).The goal is to learn typical SLAM structure with plane information and make a progress with point cloud reconstruction and plane detection.Point cloud generation is the precondition of plane information detection. 为了解决这一问题，我们要解决什么，问题之间的逻辑关系要有介绍。Then we would concentrate about how to utilize these plane information in the SLAM back-end issue with mathematical model. Necessary Jacobians for the optimization of point feature and plane feature are computed. We then simulated with a synthetic（用词要准确） data for optimization part. The techniques lie in the proposed SLAM system can facilitate applications in virtual reality, augmented reality(VR/AR) devices, construction Robots and automatic drive.

导师建议，一个摘要要写的是一句话概括研究的必要性，研究的意义重要性一定要放在前面，为了研究解决这一问题，我们要几个步骤？步骤内容之间的关系讲清楚，后面有什么应用。

(8月25日再次修稿)
This master’s thesis deals with an open sub-problem of simultaneous localization and mapping(SLAM) in three dimensional space using plane information. Since in common-place scenes,whether indoors or outdoors, plane information could provides high level information for the
local or global map. The goal is to use the typical SLAM structure in conjunction with plane information and make progress by investigating point cloud reconstruction and plane detection.Point cloud generation is the precondition of plane information detection. Then we would concentrate about how to utilize better plane information in the SLAM back-end issue with mathematical model. Necessary Jacobians for the optimization of point feature and plane fea-
ture are computed. We then simulate synthetic data for the optimization part. The techniques of the proposed SLAM system could facilitate applications in virtual reality and augmented reality(VR/AR) devices and in robots for building and automatic driving.

论文介绍

Simultaneous Localization And Mapping(SLAM) comprises two parts:how to estimate the stateof a robot equipped with on-board sensors and how to construct a map of the environment thatthe sensors are perceiving.There are various directions researchers have chosen before.This master report is an introduction to the design of the SLAM system with plane information.We focus on previous researches using plane feature and a basic SLAM structure.For point, line and plane feature fusion,we should spend more time on the construction ofassociated pose constraint and joint optimization method of line and plane features,instead of focusing on which algorithm should be used for online feature extraction.
The rest of the paper is structured as follows.(我们在)

• Chapter one deals with state of the art.We provide an overview of the current accomplishments on the theoretical aspects of SLAM with plane information.
• Necessary properties of lie group and quaternion are presented in Charpter 2.The notation is first introduced and then Chapter two prepares geometric background and basic knowledge which helps reader to have deeper understanding. geometric background and basic knowledge which helps reader to have deeper understanding are described.
• In chapter 3 and 4,we discuss details of methodology which has three subsection:pointcloud generation,plane detection and bundle adjustment with plane features implemented with g2o library.We propose a new vertex and edge to describe point and plane feature in g2oframe.Implementation details and the experimental evaluations are presented.
• Finally,the paperis finished with the conclusion and the future work in the last chapter.(Conclusions and some future works will be mentioned in the last charpter.)

How to prepare for your master's thesis

• Know your thesis cover to cover(master thesis should be a story)why you do it and why you do not use it.
• compile a comprehensive list of possible questions
• (one question one slide)
• recording yourself when you ask it.
• download macro random the slide
• practice daily with brief responses
• practice in front of the people

The proposed system exploits geometrical structures (planes) of the environments and adopts the closest point (CP) for plane parameterization.

Introduction

Back-End Modeling :Simultaneous localization and mapping with infinite planes (2015)	Front-End Feature Detection : Pop-up SLAM: Semantic Monocular Plane SLAM for Low-texture Environments (2016)
Loop Closure Improvement : PlaneMatch: Patch Coplanarity Prediction for Robust RGB-D Reconstruction(2018)	Fusion : Monocular Object and Plane SLAM in Structured Environments(2019)

As the topic of our presentation shows,before we use plane information,we need to know about it,we begin with the introduction about the plane model firstly.

Different plane models found in the literature. Note that the most general model 1 has four parameters, although three parameters are sufficient to
describe a plane (see model 3). Model 2 is the popular Hesse notation formed by a unit normal vector n and the orthogonal distance to the origin d. Model 4 and 5 both have singularities and are therefore not to be used without restrictions.Model 6 [Kanatani, 1996] provides a way to include all plane parameters in single unit vector. Depending on the application, different plane models are appropriate. In this work, model 2 is used for evaluating the plane parameters and model 4 for error propagation.

general model	Hesse notation
${\pi }_{1}x+{\pi }_{2}y+{\pi }_{3}z+{\pi }_4=$