Visual Target Measurement Method in Unknown Environment Based on Stereo SLAM

With the development of UAV technology and deep learning, the demand for 3D information acquisition of unknown environment is increasingly strong. The 3D information of unknown environment can be widely used in unmanned driving, augmented reality, disasters rescue and other scenes. Using computer vision to estimate the 3D information of the target in the scene has become the mainstream method. Because the single image lacks depth information, it is usually necessary to know the size of a reference object in the scene, or to shoot continuously by using SLAM algorithm to obtain the 3D information. The stereo vision can solve the shortcomings of monocular vision. It uses two cameras to shoot and uses the spatial geometric relationship to recover the 3D information. This paper uses stereo SLAM to track and map, and extracts feature points to calculate its 3D coordinates of the feature points. After a series of image processing operations, the feature points are used to estimate the size of target object. Finally, the accurate measurement of road width and tree height is realized.