Robust monocular vision SLAM method suitable for shadow environment

Abstract

The invention discloses a robust monocular vision SLAM method suitable for a shadow environment, and the method comprises the following steps: 1, carrying out the shadow detection of an image frame obtained at a visual front end through employing an image processing algorithm; step 2, dividing image frame subareas; 3, matching image frame subareas; using a subarea matching algorithm to find othersubareas corresponding to the subareas in the non-negative area for each subarea in the shadow area; 4, removing shadows in the image frames; for each matched subarea pair, removing shadows in the subareas in the Lab color space by using an adaptive illumination conversion method; after shadow removal processing is completed, converting a result to the RGB image; 5, extracting image frame features; and 6, carrying out camera tracking and constructing a map.

Description

technical field [0001] The invention belongs to the field of computer vision, relates to visual positioning, image processing and three-dimensional reconstruction, and can realize real-time positioning and mapping in shadow environments. Background technique [0002] The full English name of SLAM is Simultaneous Localization and Mapping, that is, real-time synchronous positioning and mapping. It usually refers to a robot or a mobile rigid body, equipped with a specific sensor, to estimate its own motion and build a model of the surrounding environment without prior information. Due to its simple structure, the monocular vision SLAM system can be directly used by various small devices such as mobile phones and drones, and can be applied to various scenarios. Since such small devices work in complex lighting environments at any time [1], it is very important to ensure that the visual front end of the SLAM system can work stably in real time in various environments. [0003] I...

AI Technical Summary

Problems solved by technology

However, since the system completely relies on pre-built maps, this method is only suitable for visual localization problems in fixed environments.

[0006] To sum up, the existing SLAM methods are less robust to shadows in natural environments, and are susceptible to interference from shadow areas during runtime, resulting in problems such as decreased accuracy, lost tracking, and interrupted operations.

However, there are often a lot of environmental shadows in the actual application scenarios of monocular vision SLAM, so it is of considerable significance to propose a monocular vision SLAM method suitable for shadow environments.

Images