基于视觉的家居服务机器人物体感知研究

发布时间：2019-06-05 03:20

【摘要】：随着社会的发展和老龄化问题的日益严峻,人们对家居服务机器人的需求越来越迫切。良好的物体感知能力是家居服务机器人实现智能作业的前提,也是机器视觉领域的研究热点。家居环境的非结构化、物体种类的繁多、物体不同姿态的任意摆放、多个物体间的相互接触与遮挡,机器人运动带来的图像模糊性,以及实时性要求等都为机器人的物体感知带来了诸多困难。本文面向家居服务机器人的这些挑战性问题,主要研究内容如下:首先,基于SURF特征设计并实现了一套完整的家居服务机器人物体识别系统。主要包括特征点检测、特征点描述、特征点匹配三个过程,并且针对误匹配问题进行改进,采用RANSAC算法进行匹配提纯,有效消除了误匹配。搭建机器人硬件平台,在室内非结构化环境下进行识别实验,验证了该方法具有良好的鲁棒性,但是实时性有待提高。其次,为了进一步提高系统性能,提出了一种基于显著性区域指导的机器人物体识别算法。通过视觉选择性注意机制选择图像中显著区域,提取显著性区域SURF特征,通过特征点匹配,实现场景中目标物体的识别。实验证明,和传统SURF算法相比,本文改进算法的识别率和识别速度均得到有效提高。最后,为解决传统Camshift算法需要手动选取跟踪目标的问题,提出一种基于检测的无监督的家居服务机器人物体跟踪算法,利用SURF特征匹配自动寻找目标物体,实现了全自动跟踪。针对传统Camshift算法容易受相似颜色干扰的问题,将SURF特征融入到Camshift算法框架之中,提出一种结合SURF特征的Camshift物体跟踪算法。实验证明,该算法可以有效解决传统Camshift算法在与目标颜色相似的背景下跟踪失败的问题。
[Abstract]:With the development of society and the increasingly serious problem of aging, the demand for home service robots is becoming more and more urgent. Good object perception ability is the premise of intelligent operation for home service robots, and it is also a research focus in the field of machine vision. The unstructured home environment, the variety of objects, the arbitrary placement of objects in different positions, the mutual contact and occlusion between multiple objects, and the fuzziness of the images caused by robot motion. And real-time requirements have brought a lot of difficulties to robot object perception. The main research contents of this paper are as follows: firstly, a complete object recognition system of home service robot is designed and implemented based on SURF features. It mainly includes three processes: feature point detection, feature point description and feature point matching, and improves the mismatching problem. RANSAC algorithm is used for matching purification, which effectively eliminates the mismatching. The hardware platform of the robot is built and the recognition experiment is carried out in the indoor unstructured environment. The results show that the method has good robustness, but the real-time performance needs to be improved. Secondly, in order to further improve the performance of the system, a robot object recognition algorithm based on salient area guidance is proposed. Through the visual selective attention mechanism, the significant region in the image is selected, the SURF feature of the significant region is extracted, and the recognition of the target object in the scene is realized by matching the feature points. The experimental results show that compared with the traditional SURF algorithm, the recognition rate and recognition speed of the improved algorithm are effectively improved. Finally, in order to solve the problem that the traditional Camshift algorithm needs to select and track the target manually, an unsupervised object tracking algorithm for home service robot based on detection is proposed. SURF feature matching is used to automatically find the target object, and the automatic tracking is realized. In order to solve the problem that the traditional Camshift algorithm is vulnerable to similar color interference, SURF features are integrated into the framework of Camshift algorithm, and a Camshift object tracking algorithm based on SURF features is proposed. The experimental results show that the algorithm can effectively solve the problem of tracking failure of the traditional Camshift algorithm in the context of similar color to the target.
【学位授予单位】：河北工业大学
【学位级别】：硕士
【学位授予年份】：2015
【分类号】：TP391.41;TP242

【相似文献】