基于Kinect的红外单目图像测距

摘 要

随着Kinect的兴起，深度图像处理迅速变成图像处理和计算机视觉领域的热点。基于Kinect的应用也日渐增多。但是，由于Kinect深度测距原理所限，Kinect的深度图像存在以下四个方面的问题：与距离的平方成正比的噪声、由无效像素组成的空洞、深度图像与彩色图像边缘不匹配和深度图像在时域上的闪烁。这些因素阻碍了Kinect在诸如人脸识别、3D重构等需要高质量的深度数据的领域上的应用。红外单目图像测距应用广泛，比如人工智能系统、机器人以及电子设备等行业。更高效率和更高精度的红外单目图像测距设备一直是研究的热点。在我们日常生活中，红外单目图像测距随处可见，因为其相比其他测距而言结构相对简单，运行稳定且便于维修等优势，最重要的是红外单目图像测距在人工智能方面具有很好的优势。随着视觉算法术和微电子技术的不断革新，目前的技术水平为实现红外单目图像测距调节控制打下坚实的基础，提供扎实的理论依据。

本次设计主要以红外单目图像测距控制系统设计应用作为研究背景，深度测距是系统核心功能。基于Kinect 深度相机工作原理，研究Kinect 获取深度距离算法，开发深度测距程序，在不同光照条件下对检测到的人体中心位置进行测距实验，结果表明Kinect 测距具有抗光照不足的优势。为解决测量误差随物体到传感器的距离增大而增大的问题，分析了误差的成因，利用像素三维欧氏距离与对应深度补偿值的权重占比关系，建立了的深度误差补偿模型，将测量误差降低至±10 mm，验证了误差补偿的有效性。本次设计主要以红外单目图像测距控制系统设计应用作为研究背景，运用MATLAB仿真工具编写相应的仿真程序。红外单目图像测距控制系统拥有很好的准确性，运行稳定性高、测距的范围较大，性能可靠等，在实际生产制造中被广泛的应用。该仿真设计对红外单目图像测距控制方法，通过 Matlab仿真软件运行仿真，并对仿真结果进行了分析。经过对控制系统进行仿真，可以明显的看出，红外单目图像测距控制系统响应快，测量精准，运行稳定，抗光照不足能力强等特点。

本文介绍了红外单目图像测距的基本工作原理，对各种工作模式进行了研究和理论分析，并建立了其数学模型。深入学习红外单目图像测距的整体结构和原理，针对红外单目图像测距控制系统作为研究对象。在对红外单目图像测距控制系统工作原理和各个部分的数学模型进行深入研究的基础上，在MATLAB对不同部件的传递原理和数学模型进行了深入的研究，建立红外单目图像测距控制系统，通过仿真结果验证，可以比较各种策略与方案，优化并确定相关参数。最后，对红外单目图像测距控制系统分析方法进行了详细比较分析，为科学决策提供可靠的依据。

关键词：红外单目，Kinect，图像测距，误差

Abstract

With the rise of Kinect, deep image processing has quickly become a hot spot in the field of image processing and computer vision. Kinect-based applications are also increasing. However, due to the limitation of Kinect depth ranging principle, Kinect depth image has the following four problems: noise proportional to the square of distance, void composed of invalid pixels, edge mismatch between depth image and color image, and depth image flicker in time domain. These factors have hampered Kinect's use in areas such as face recognition and 3D reconstruction, which require high-quality deep data. Infrared monocular image ranging is widely used in artificial intelligence systems, robotics and electronic equipment. Infrared monocular image ranging equipment with higher efficiency and accuracy is always a research hotspot. In our daily life, infrared monocular image ranging can be seen everywhere, because compared with other ranging, its structure is relatively simple, stable operation and easy maintenance and other advantages, the most important is that infrared monocular image ranging has a good advantage in artificial intelligence. With the continuous innovation of visual computing and microelectronics technology, the current technical level for the realization of infrared monocular image ranging control to lay a solid foundation, provide a solid theoretical basis.

The second design mainly takes the infrared monocular image ranging control system design and application as the research background, depth ranging is the core function of the system. Based on the working principle of Kinect depth camera, this paper studied the algorithm of obtaining depth distance of Kinect, developed depth ranging program, and carried out ranging experiments on the detected human center position under different lighting conditions. The results show that Kinect ranging has the advantage of anti-insufficient lighting. In order to solve the problem that the measurement error increases with the increase of the distance from the object to the sensor, the cause of the error was analyzed, and the depth error compensation model was established based on the weight ratio between the three-dimensional Euclidean distance of the pixel and the corresponding depth compensation value. The measurement error was reduced to ±10 mm, and the validity of the error compensation was verified. This design mainly to infrared monocular image ranging control system design and application as the research background, using MATLAB simulation tools to write the corresponding simulation program. Infrared monocular image ranging control system has good accuracy, high operation stability, ranging range is large, reliable performance and so on, in the actual production is widely used in manufacturing. The simulation design of infrared monocular image ranging control method, through Matlab simulation software simulation, and the simulation results are analyzed. After the simulation of the control system, it can be clearly seen that the infrared monocular image ranging control system has the characteristics of fast response, accurate measurement, stable operation and strong anti-illumination ability.

In this paper, the basic working principle of infrared monocular image ranging is introduced, various working modes are studied and analyzed theoretically, and its mathematical model is established. The overall structure and principle of infrared monocular image ranging are deeply studied, and the infrared monocular image ranging control system is taken as the research object. On infrared monocular image distance control system working principle and the mathematical model of each part, on the basis of in-depth study in MATLAB to the transmission principle and mathematical model of different parts of in-depth research, to establish the infrared monocular image distance control system, through the simulation results show, you can compare all kinds of strategy and plan, optimize and related parameters were determined. Finally, the analysis methods of infrared monocular image ranging control system are compared and analyzed in detail to provide reliable basis for scientific decision making.

Key words: Infrared monocular, Kinect, image ranging, error

目录

摘  要................................................................................................. 1

第一章 绪论.................................................................................... 7

1.1  概述...................................................................................... 7

1.2  课题研究的现状............................................................... 8

1.3  课题研究的意义............................................................. 10

1.4  课题的研究内容及章节安排...................................... 10

第二章 Kinect标定方法研究................................................. 12

2.1  Kinect 硬件结构分析.................................................. 12

2.2.1 Kinect 性能参数........................................................ 12

2.2.2 Kinect 工作机制........................................................ 13

2.3 Kinect 相机标定方法...................................................... 13

2.3.1 构建相机透视投影模型......................................... 14

2.3.3 Harris 算子角点检测............................................... 14

第三章 Kinect人体检测算法研究....................................... 16

3.1  Kinect 图像获取方法.................................................. 16

3.2  相机标定.......................................................................... 16

3.3  图像校正.......................................................................... 17

3.4  Kinect 滤波方法分析.................................................. 17

3.4.1  线性滤波................................................................... 18

3.4.2  形态学滤波.............................................................. 19

3.5  图像分割算法................................................................. 20

第四章 Kinect的单目红外测距算法研究......................... 23

4.1   Kinect 测距算法.......................................................... 23

4.1.1 Kinect V1 测距算法................................................. 23

4.1.2 Kinect V2 测距算法................................................. 24

4.2   基于Kinect 测距算法实现与仿真.......................... 25

4.3  本章小结........................................................................... 33

第五章 .....................................

完整论文点击如下链接下载：

https://download.csdn.net/download/weixin_45905610/88536904

基于Kinect的红外单目图像测距资源-CSDN文库

程序及matlab仿真分析点击如下链接下载：

https://download.csdn.net/download/weixin_45905610/88536905

基于Kinect的红外单目图像测距-matlab仿真资源-CSDN文库