Abstract: Autonomous exploration is a research direction to improve the intelligence of robots, and it is dedicated to realizing the efficient exploration of unknown areas by robots. Most of the current research on autonomous exploration algorithms only considers geometric information such as images, boundaries, and distances in the environment. At the same time, the greedy strategy of the algorithm tends to ignore small areas, making the explored volume smaller than the actual volume. To this end, this paper proposes an indoor autonomous exploration algorithm for mobile robots that integrates semantic information. Based on the idea of independent exploration of the next optimal viewpoint, the algorithm extends some viewpoints to exploration points with semantic information. On this basis, the proposed new algorithm introduces the reward value based on semantic information as the basis for selecting the robot's next target point, guiding the robot to retrieve the neglected area. Multiple experiments conducted in the open source data set Matterport3D environment show that the exploration area of the autonomous exploration algorithm in this paper increases by at least 14.6% when compared with the Dual-Stage Viewpoint(DSV) Planner algorithm, and increases by at least 15.7% when compared with the TARE Planner algorithm. The results show that the new algorithm for fusing semantic information proposed in this paper can effectively improve the robot's exploration volume of the unknown environment.