摘要图像分割指的是将图像细分为构成它的子区域或物体的操作。这一操作极具 实用意义，因为图像处理、识别应用程序经常需要将准备进行进阶处理的图像部 分先行分割出来，这一操作对人工智能与机器人学的意义尤为重大。传统的摄像 设备只采集由红（Red）、绿（Green）、蓝（Blue）三路基色组成的光学色彩信号， 因而长期以来计算机图形界对图像分割算法的设计和研究都围绕着 RGB 这一图 像描述机制进行。但近些年，以微软 Kinect 为首的、同时采集光学色彩信息与深 度（Depth）信息的 RGB-D 摄像设备的推广让图像分割的效率与准确性迎来了一 个质的飞跃。本论文设计了一个基于 RGB-D 摄像设备的图像分割算法，藉此证 明：把色彩和深度信息综合起来，即能大大提高图像分割的准确度。68519

毕业论文关键词 数字图像处理 RGB-D 图像分割

Title   Research and Implementation of Image Segmentation Algorithm Based on RGB-D Camera

Abstract

Image segmentation pides an image into its constituent regions or objects. The operation of segmentation provides a large amount of practical significance, since image processing or recognition programs generally need to separate out image section that needs further processing, so it plays a large part in artificial intelligence and robotics. Classical camera devices, which most existing image segmentation algorithms are based on, only capture three optical primary colors including red(R), green(G) and blue(B). Recent years, the popularization of RGB-D camera device represented by Microsoft Kinect has brought a qualitative leap to image segmentation’s efficiency and accuracy. We designed a RGB-D based image segmentation algorithm to demonstrate that combining color and depth information substantially improves quality of segmentation results.

Keywords  Digital image processing, RGB-D, Image segmentation

目 次

1 引言 5

1.1 研究背景 5

1.2 研究现状 6

1.3 研究内容 7

1.4 论文结构 8

2 数据采集 8

2.1 Kinect 类 RGB-D 摄像机的深度信息采集原理 8

2.2 RGB-D 物体数据集 8

3 算法实现 10

3.1 算法框架 10

3.2 图像基本输入输出 11

3.3 基于色彩信息的分割 11

3.4 基于深度信息的分割 12

3.5 综合与除噪 14

4 实验结果 16

4.1