Feature point line-based monocular camera pose estimation and optimization method and system

Abstract

The invention discloses a feature point line-based monocular camera pose estimation and optimization method and system. The method comprises the following steps of: S1, constructing an overdeterminedlinear system on the basis of a point constraint and a line constraint, and establishing a relationship between rotation matrixes and translation vectors; constructing a total cost function by utilizing a least square method according to cost functions of sub-constraints of the point constraint and the line constraint, obtaining a final cost function through importing Kaili parameter expression, and solving a rotation matrix and a translation vector of camera pose by utilizing the final cost function; and S2, importing a Sampson error in the point constraint and the line constraint to construct a target function, and optimizing the rotation matrix and the translation vector by adoption of an FNS iterator by taking the rotation matrix and the translation vector obtained in the step S2 as initial values. The method and system are unlimited to monocular photo obtaining equipment and photo scenes, are suitable for photos of vehicle-mounted monocular cameras, onboard monocular cameras and handheld monocular cameras, and are also suitable for photos in road scenes, aviation scenes and indoor scenes.

Description

technical field [0001] The invention belongs to the technical field of real-time positioning and map construction, and in particular relates to a monocular camera pose estimation and optimization method and system based on feature point lines. Background technique [0002] Visual measurement is one of the important research topics in real-time localization and map construction technology. In recent years, visual measurement has been well-studied and widely used in robotics, autonomous driving, wearable computing, and augmented reality, and is especially applicable in some special cases such as indoor and underwater environments where GPS is ineffective. Among them, the visual photometer using a monocular camera for monocular vision measurement has the advantages of broader application prospects, cheaper, more compact, and easier calibration. The most standard procedure for monocular vision measurement is to first extract a set of image features and match them in consecutive...

AI Technical Summary

Problems solved by technology

The problem with this method is that these methods cannot be applied or become very inefficient when the system contains a large number of observations, and they do not work well in low-texture scenes, especially artificial environments

The problem with this method is that the accuracy is not high, and other constraints and optimizations need to be added to obtain more accurate results

The disadvantage of these methods is that they will be ineffective when features are scarce

Even if collinearity and coplanar constraints are used to solve extreme conditions, this method uses different models for different situations, and the viewpoint and line cannot be equipotential conditions, the real point-line combination cannot be achieved, and it is difficult to expand for redundancy Observed

Images