Laser lidar and vision sensors are the two mainstream three-dimensional sensing techniques in the applications of robot location and navigation. However, existing low-cost laser lidar usually has low location accuracy and cannot achieve loop closure detection in large areas. In this paper, an indoor robot equipped with low-cost laser lidar and camera was used for experiment. And a novel localization and mapping method was introduced by combing both lidar and image information. An optimization method based on sparse pose adjustment was used to optimize the robot pose by fusing laser points cloud and image feature points as constraints. At the same time, the bag of words model based on visual features was used for loop closure detection. The grid map was optimized by loop closure constraints. Real experimental results show that, the proposed method has better localization accuracy than either laser lidar or vision sensors, and loop closure detection also can be realized.