RandNet Based Rapid Closed Loop Detection For VSLAM

The issue of Simultaneous Localization and Mapping(SLAM)is a key technology for autonomous positioning and navigation of mobile robots in an unknown environment.It is widely used in the fields of unmanned driving,smart home,and 3D reconstruction,and has theoretical significance and research value.Loop closure detection means that the robot recognizes whether the current position has been accessed,whose result can be used to correct the global correction of the SLAM,effectively reduces the cumulative error of building the map,and plays a key role in updating the map in real time and avoids the introduction of wrong map nodes.It is a key module in SLAM.Visual data is rich in information,which is convenient for mathematical methods and image processing technologies.It is inexpensive and easy to use.Therefore,it is used in loop closure detection.So this paper studies the loop closure detection of visual SALM system.The main research work of this article is briefly described as follows:The loop detection problem in this thesis is regarded as an im age retrieval issue,which is addressed by the detection technique of the closed loop vision.And based on the idea of deep learning,we apply the Rand Net neural network to the loop detection problem,and a threshold-based local-sensitive hashing algorithm is also used to accelerate the image feature matching process,and finally we virify these algorithms with the datasets City Center and New College.The test results show that the fast loop detection method based on Rand Net proposed in this paper can improve the feature extraction speed by more than 3 times on the premise of high accuracy,and the speed of feature matching is 10 times higher,and it can better meet the real-time requirement of loop detection.