Monocular vision inertia odometer method based on point-line characteristics

Abstract

The invention discloses a monocular vision inertia odometer method based on point-line characteristics. A monocular camera in sensor equipment collects a current visual image frame, and an IMU collects the movement information of a robot, and extracts characteristic points and characteristic lines; a key frame is selected according to the obtained characteristic points and characteristic lines, and surrounding environment space point and space line drawing and platform positioning optimization are performed; and whether the motion trail of the platform forms a closed loop or not is determined,a correct closed loop key frame is acquired, overall optimization is performed on a global image attitude and a map, and a global map is output. According to the invention, a cylinder-like axis solving precision is improved by utilizing the advantages of high linear characteristic machining precision, good robustness, a high anti-noise capability, a large information amount, convenience in optimization and the like, and meanwhile, due to the fact that point and line characteristic individuals do not need to be accurately positioned, the combined characteristic pose solving precision cannot bereduced due to errors caused by assembly; and therefore, a split-type target also has high measurement precision, and the method has better stability.

Description

technical field [0001] The invention relates to the field of simultaneous localization and mapping (SLAM), in particular to a monocular visual inertial odometer method based on point-line features. Background technique [0002] Simultaneous positioning and mapping (SLAM) refers to the process in which a robot completes its own positioning and simultaneously constructs a surrounding map by detecting the environment in an unknown environment. In the past ten years, SLAM technology and related theories have made great progress. Since the purely visual SLAM method cannot work in places with little image texture, it will cause blurred images during fast movement, and the inertial sensor can obtain acceleration and angular velocity information when the camera is moving. Therefore, the two measurement methods are complementary. The inertial measurement unit can supplement the lack of visual information, and the visual information can correct the drift problem caused by the long-te...

AI Technical Summary

Problems solved by technology

[0004] Since feature extraction and matching consume a lot of calculations, image pixel-based visual SLAM has gradually attracted the attention of researchers. Image pixel-based visual SLAM algorithms directly use image grayscale information for image tracking without feature extraction. And description, directly use the gradient of the image for tracking and optimization, which can enhance the continuity of visual SLAM in areas with fewer features, but it completely relies on image pixels, and the effect is poor for areas with strong illumination changes, and only relies on images Pixel gradient, the calculated positioning and composition accuracy is relatively low

Images