柔索机器人环境视觉关键技术研究

发布时间：2018-03-02 19:01

本文选题：混合噪声　切入点：噪声检测　出处：《西南科技大学》2017年硕士论文　论文类型：学位论文

【摘要】：柔索驱动并联机器人作为一种新型柔索结构起机器人,具有重量小,结构简易,工作空间大等特点使其能够应用于环境多变的复杂的工作场景。如何在该复杂环境下,获取到有效、准确的机器人工作环境图是柔索机器人环境视觉技术必须首要解决的视觉问题。在获取柔索机器人工作环境图像过程中,经常会遇到相机传感器发热现象产生高斯白噪声和图像传输过程中受信道内噪声影响产生脉冲噪声,所以获取到的柔索机器人环境图像通常都是夹杂了混合噪声。针对柔索机器人工作环境中获得混合噪声图像(高斯白噪声和脉冲噪声所组成的混合噪声图像)本文提出了一种去噪算法。对于脉冲噪声的检测首先提出了用两个阈值更有效地对噪声进行检测,然后在现有基于加权编码的算法上将图像稀疏表示以及非局部相似先验融入到去噪模型,最终得到去噪图像。具体工作如下:1、分析了高斯噪声、脉冲噪声、混合噪声的的数学模型,进而构建出本文的混合噪声模型。2、详细分析了非局部相似和稀疏表示进行图像去噪的原理、模型、算法,并进行实验获得了算法去噪性能的相关指数。3、依据去噪模型,提出一种适用性更高的脉冲噪声检测算法。4、针对本文的混合噪声模型,有效地将非局部相似和稀疏表示先验融合在一个去噪模型中,并在模型中添加了正则限制项,完成了去噪模型算法的优化,通过求解去噪模型,最终获得去噪效果图。5、对本文算法和几种典型去噪算法进行了图像去噪实验,获得的客观数据和去噪效果图都显示了本文算法的优越性。
[Abstract]:As a new type of robot with flexible cable structure, the flexible cable driven parallel robot has the characteristics of small weight, simple structure and large workspace. Obtaining an effective and accurate robot working environment map is the most important visual problem that must be solved in the rope-cable robot environment vision technology. In the process of obtaining the flexible robot working environment image, It is often encountered that the heating phenomenon of camera sensor produces Gao Si white noise and the impulse noise caused by the noise in the channel during the image transmission. So the environment image of cable robot is usually mixed with mixed noise. In order to obtain mixed noise image (the mixed noise image composed of Gao Si white noise and pulse noise) in the working environment of cable robot. In this paper, a denoising algorithm is proposed. For impulse noise detection, two thresholds are proposed to detect noise more effectively. Then the image sparse representation and nonlocal similarity priori are incorporated into the denoising model based on the existing weighted coding algorithm, and finally the denoising image is obtained. The detailed work is as follows: 1. Gao Si noise and pulse noise are analyzed. The mathematical model of mixed noise is constructed, and then the mixed noise model. 2. The principle, model, algorithm of image denoising based on nonlocal similarity and sparse representation are analyzed in detail. The correlation index of denoising performance is obtained by experiments. According to the denoising model, a more applicable impulse noise detection algorithm .4 is proposed, which is aimed at the mixed noise model in this paper. The nonlocal similarity and sparse representation are effectively fused into a denoising model, and a regular restriction item is added to the model to optimize the de-noising model algorithm, and the de-noising model is solved by solving the de-noising model. Finally, the denoising effect figure .5 is obtained. The image denoising experiments are carried out on this algorithm and several typical denoising algorithms. The obtained objective data and denoising effect graph show the superiority of this algorithm.
【学位授予单位】：西南科技大学
【学位级别】：硕士
【学位授予年份】：2017
【分类号】：TP391.41;TP242

【相似文献】