Binocular vision inertia odometer method utilizing parallel computing acceleration

Abstract

The invention provides a binocular vision inertia odometer method utilizing parallel computing acceleration, is used for real-time position and attitude estimation on a low-performance processor, andbelongs to the technical field of positioning. Based on a VINS-Fusion algorithm, a visual front-end algorithm of the method is improved, and the method is mainly characterized in that calculation redundancy removal of multi-time optical flow tracking, parallel execution design of corner detection and optical flow tracking and implementation of a corner detection algorithm based on an FPGA are carried out. The method is a set of visual inertia odometer method based on optimization, the operation time consumption of the algorithm is effectively shortened, the method can be used for real-time pose estimation on a low-performance processor, the positioning precision is high, the robustness is good, and the method has good engineering application and popularization value.

Description

technical field [0001] The invention relates to the technical field of positioning, in particular to a binocular visual inertial odometer method utilizing parallel computing acceleration. Background technique [0002] Visual-inertial odometry is an autonomous localization technique based on monocular / binocular cameras and an inertial measurement unit (IMU). This technology can accurately and quickly estimate the pose of the device by fusing the visual information provided by the monocular / binocular camera and the device motion information provided by the IMU. This technology is dedicated to solving the problem of estimating the self-pose of an unmanned mobile platform in an unknown environment when external positioning technologies (such as GPS and UWB positioning technologies) cannot be used. In the fields of autonomous driving, drones and augmented reality, this technology has spawned many commercial applications. For example, Apple's augmented reality solution ARkit and ...

AI Technical Summary

Problems solved by technology

However, this solution does not accelerate the time-consuming image processing part, nor does it use an optimization method with higher positioning accuracy for pose estimation

In 2013, Korbinian Schmid, a researcher of Roboception GmbH in Germany, and others proposed a visual-inertial mileage calculation method based on binocular stereo matching, in which the SGM algorithm (semi-global matching algorithm) was realized by using FPGA (programmable logic block). The depth of the scene can be quickly recovered from the binocular image, but the Harris corner detection algorithm in the image processing part is still executed on the CPU, and the speed is not improved

Images