Instant positioning and map construction method suitable for dynamic environment

Abstract

The invention discloses an instant positioning and map construction method suitable for a dynamic environment, and the method comprises the following steps: firstly carrying out the external parametercalibration of a depth camera and an encoder, and obtaining a transformation matrix from the coordinates of the encoder to the coordinates of the depth camera; layering each frame of image acquired by the depth camera by using an image pyramid, and extracting ORB feature points from each layer of the image pyramid; taking the data obtained by the encoder as a constant-speed model of the trackingmodule, then projecting the feature points of the key frame of the previous frame onto the current frame, matching the feature points with the feature points of the current frame, and performing triangulation on the successfully matched key points to obtain a pose between the two frames. According to the invention, the tracking part uses encoder data as an initial pose during camera re-projection,accurate positioning tracking in a dynamic environment is realized, and compared with a traditional semantic slam tracking module which needs to remove dynamic pixels from each frame, amount of timeis saved.

Description

technical field [0001] The invention relates to the technical field of mobile robot positioning, navigation and mapping, in particular to an instant positioning and map building method suitable for dynamic environments. Background technique [0002] Slam (Simultaneous Localization and Mapping) is widely used in fields such as mobile robots, drones and augmented reality. However, in dynamic environments, traditional slam systems are not robust. In a real scene full of dynamic objects, the tracking module in the classic slam system like ORB_SLAM will fail. What's worse, the classic slam system builds a static map, which cannot remove dynamic objects from the map, resulting in the map being completely Not available. In order to solve the problem of how to make the slam system applicable to dynamic scenes, many papers and documents have given different solutions. DS_SLAM is one of the effective solutions, which performs semantic segmentation on each frame of the input image. ...

AI Technical Summary

Problems solved by technology

However, due to the semantic segmentation of each frame of image, it is difficult for the system to achieve real-time performance on the mobile platform, and the dense point cloud map is established, which consumes a lot of memory.

Images