A Visual Slam Loop Closure Detection Method Fused with Geometric Information

Abstract

The invention discloses a visual SLAM loopback detection method that combines geometric information, including S1: obtaining real key frames; S2: determining the change of the viewing angle of the virtual camera, and calculating the pose of the virtual camera in the SLAM system; S3: according to the real key frame Frame reconstruction of the 3D point cloud in the SLAM system; S4: Calculate the pixel coordinates of each 3D point of the 3D point cloud in the virtual key frame according to the pinhole camera model and projection geometry; S5: Render the virtual key of the virtual camera Frame; S6: extract the feature descriptors of real key frames and virtual key frames; S7: calculate the BoV of real key frames and virtual key frames and add them to the database, and search for all real key frames and virtual key frames in the database when performing loopback detection frame to obtain loopback detection information. The visual SLAM loop detection method fused with geometric information of the present invention can not only directly use monocular camera information to generate loop detection information from different perspectives, but also has higher efficiency than traditional loop detection methods.

Description

technical field [0001] The invention belongs to the field of simultaneous localization and map building (SLAM), in particular to a visual SLAM loop detection method fused with geometric information. Background technique [0002] Simultaneous localization and mapping (SLAM) is widely used on small mobile platforms. In recent years, with the development of SLAM technology and the enhancement of computing power of small platforms, mobile phones, drones, robots and other small motion platforms have more and more demands for SLAM applications. Small platform application scenarios pose a series of challenges to related algorithms in SLAM. The limitations of computing power, sensor types, complex scenes and other aspects put forward higher requirements for the implementation of SLAM algorithms. At this stage, SLAM is developing rapidly towards high-speed computing, data fusion and stronger scene adaptability. Visual SLAM is favored by many small mobile platform applications due ...

AI Technical Summary

Problems solved by technology

Although some of these methods can provide relevant solutions to the problem of viewing angle transformation, it is difficult to load high-performance devices such as large lidars and omnidirectional cameras on existing small mobile platforms due to cost constraints.

Images