Research on Visual Inertia SLAM Technology with Additional Point and Line Features

With the rapid development of GNSS jamming technology, GNSS is facing increasingly severe challenges. Under the condition of GNSS rejection in complex electromagnetic environment, many devices cannot accurately perceive the surrounding environment and thus cannot work normally. Therefore, exploring autonomous navigation under the condition of GNSS rejection is one of the hot topics in the future research. With the rapid development of computer vision, visual inertial odometer (VIO), which is closely coupled with camera and inertial measurement unit (IMU), can obtain high precision local pose results in unknown environments, and is widely concerned for its low cost and miniaturization. In complex and changeable structured scenes, sparse and structured features are still the bottleneck of restricting the performance of visual navigation. In this paper, LSD line segment extraction algorithm is added on the basis of visual inertial odometer to extract more line features of environmental structure, and the sliding window strategy is used to achieve state optimization. In order to verify the effectiveness of this algorithm, this paper tests the proposed method using the open data set. The test results show that the proposed method can effectively provide position and attitude estimation when GNSS refuses. In the test, the error mean value of 0.119 m, the minimum error of 0.015 m, and the maximum error of 0.259 m can be achieved. Compared with the traditional point feature VIO, the precision is improved by 50.6%, and has strong real-time and robustness.