| In order to improve the intelligent level of mobile robot further and make it serve the human better,people put forward higher requirements for the autonomy of the mobile robot.Localization is one of the key technology to autonomously complete all kinds of complex tasks for mobile robot,which has important theoretical significance and application value for improving its autonomy.With the development of computer vision and image processing technology,the technology of using vision to perceive indoor environment and realize localization of mobile robot has attracted more and more attention.In practical application,there are obvious differences between single feature and feature-rich work scenes,so the suitable visual localization methods are often different.In this paper,the basic theories of camera imaging model and camera calibration are studied firstly.Then,taking the indoor mobile robot equipped with small size and low cost monocular camera as research object,the artificial marker based monocular visual localization technology,and feature based monocular visual SLAM technology,are studied respectively in the actual scenarios.The main contents of this paper are as follows:1.Aiming at the problem that the previous artificial markers are easily affected by the interference of illumination change and motion blur,a new artificial marker and its recognition and localization algorithm are designed.The artificial marker has obvious characteristics of color,shape and size,and is easy to identify.According to the characteristics of the artificial marker,the mobile robot can recognize it by using image processing algorithms such as color space conversion,threshold segmentation and morphological filtering.The experiments show that the combination of the marker and the recognition algorithm can improve the accuracy and real-time performance,reduce the impact of illumination changes and motion blur,and thus has better anti-interference performance.2.In order to realize the autonomous localization of mobile robot,a localization model is established,and a monocular visual localization and navigation system based on artificial marker is constructed.According to the results of artificial marker recognition,combined with the localization model,the current pose and pose errors of the mobile robot can be determined.Using the correction strategy,the localization errors can be corrected to reduce the error accumulation.The software system of monocular visual localization and navigation based on artificial marker is designed and developed.The experiment shows that the system can guide the mobile robot to move stably along the planned path,and has strong practicability.3.Aiming at the problem that there are many mismatches in the front-end of monocular visual SLAM system that will affect the results of pose estimation,a series combination of mismatches elimination algorithm is designed to optimize features matching.The algorithms of distance threshold,cosine similarity and random sampling consistency are combined in series to eliminate the mismatches in turn.The feature based pose estimation and the ORB feature extraction algorithm with better real-time performance are studied and analyzed,and a monocular visual SLAM system is built based on ROS.Experiments show that the proposed algorithm can effectively integrate the advantages of the three original algorithms and optimize the results of matching.Compared with the experiment results only using one of the three original algorithms and the original algorithm in the system,the matching accuracy is significantly improved.At the same time,the system can draw the moving track and construct the environment map in real time,which shows good feasibility in reality. |
PDF Full Text Request