The Research Of Monocular Simultaneous Localization And Mapping

Simultaneous localization and mapping are key technologies for realizing the autonomous movement of intelligent robots.The SLAM can provide robots with environmental information and their own position and posture information in order to achieve real-time positioning and navigation.Due to the low cost,ease of use,and rich environmental information that the camera has with respect to other sensorsu.In recent years,more and more attention has been paid to related research based on simultaneous localization and mapping.Especially in applications such as driverless vehicles(or advanced driver assistance systems),service robots,etc.,vision-based simultaneous localization and mapping technology can meet the high-precision positioning requirements of the robot in a complex driving or working environment.This paper takes the driverless vehicle as the background,and studies several problems of monocular vision simultaneous positioning and map construction technology.The main work of this article is as follows:1.The basic framework of simultaneous localization and mapping algorithm is implemented,and experimental verification is performed.The tracking of camera trajectory and reconstruction of road information are completed.2.At the feature extraction stage,various features such as RGB,HSV,gray level co-occurrence matrix,and lane lines are used.The AdaBoost algorithm is used to segment the road surface area,and then the feature points on the road surface and nearby areas are collected and passed.Allocating different response intensity weights makes the extracted feature points more evenly distributed on the image,which reduces the calculation error caused by unreliable feature points and improves the effect of subsequent matching.3.At the stage of camera pose estimation,the holographic matrix is mainly used to estimate the camera pose,the triangulation is used to estimate the spatial position of the points,the BA algorithm is used to optimize the tracking results over a period of time,and the distribution of feature points is used to optimize the calculation flow.4.An algorithm is designed to detect the motion of the camera.The correlation between the distribution of the feature point pairs and the camera motion in the top-down view of the road surface.The interference of the bumpy state to the pose estimation is eliminated,and the features can be used under specific conditions.The distribution of point pairs directly estimates the camera pose transformation,which improves the computational efficiency and tracking effect.