In order to overcome the difficult acquisition of feature points and the low accuracy of ORB-SLAM algorithm in large-scene and weak-texture environment, a PAL-SLAM algorithm based on RGB-D camera is proposed. In this algorithm, first, on the basis of ORB-SLAM algorithm, a new framework of point-line feature fusion is designed. Next, by investigating the fusion principle of point feature and line feature, a re-projection error model of point-line fusion is deduced, and the Jacobian matrix analytical form of the model is obtained. Then, the PAL-SLAM algorithm framework of point-line feature fusion is proposed. Furthermore, a comparative experiment between PAL-SLAM algorithm and ORB-SLAM algorithm is carried out by using TUM data set. The results show that the positioning accuracy of PAL-SLAM algorithm is higher in large indoor scenes, and the standard error is far less than that of ORB-SLAM algorithm, which is only 18.9% of that of ORB-SLAM algorithm| and that PAL-SLAM algorithm reduces the positioning error of traditional visual SLAM algorithm in large-scene and weak-texture environment, and effectively improves the accuracy of the system. Moreover, the experimental results at a platform based on Kinect V2 show that the proposed PAL-SLAM algorithm can be well combined with the hardware platform.