SLAM method based on semantic bundle adjustment method

Abstract

The invention provides an SLAM method based on a semantic bundle adjustment method, and belongs to the field of mobile robot simultaneous localization and mapping (SLAM). The method is characterized in that the method combines a 6DOF object and the camera attitude through new semantic global optimization, and can work under a 2D or 3D sensor. As semantic information is added, a target detection channel can be seamlessly integrated into a BA type optimization system of an SLAM system based on BA without peripherals. The method of the invention is simple and easy to implement, and has strong practicability. SLAM constraints can be used in robust target detection, and can adapt to a more complicated environment.

Description

technical field [0001] The invention relates to a SLAM (simultaneous localization and mapping) method based on a semantic beam adjustment method, belonging to the field of simultaneous localization and map creation (SLAM) of mobile robots. Background technique [0002] The visual SLAM (Simultaneous Localization and Mapping) problem involves the ability to gradually reconstruct maps and simultaneously localize sensing devices based only on visual cues. In the past decade, there has been remarkable progress in this field with the application of more efficient tools such as AR (Augmented Reality) and machine navigation and composition. [0003] Traditional solutions to SLAM problems are based on filtering techniques such as Kalman filters, in which visual features are tracked in terms of frames and their estimated 3D positions and uncertain camera positions. Since only a small fraction of the image pixels are tracked, this method can only create sparse maps. As an alternative...

AI Technical Summary

Problems solved by technology

In some schemes, although the detection scheme is closely combined with the SLAM framework, it does not attempt to estimate the actual position of the target object; the rest uses geometric information for continuous target detection, but does not effectively use the original detected objects to construct known maps; there are also algorithms that use standard feature-based pipelines to detect objects and estimate relative poses for positional representations that use laser and odometry data to detect objects on maps, but these algorithms still do not The position of the camera is combined for evaluation; although some algorithms use joint detection results and reconstruction methods, they are either limited to cars and pedestrians and assume the position of the environment and the camera, or propose a common pixel label that requires a small baseline for calibrating the camera and dense stereoscopic reconstruction methods; a semantic structure based on motion tricks that evaluates the problem of camera pose and recognizes objects from a series of image sequences, but when the algorithm extends the SFM detector and creates all frames at once , must resort to an external object detector to create hypotheses and measure

Images