Semantic map construction method based on target candidate box in any direction

The invention provides a semantic map construction method based on a target candidate box in any direction, and the method comprises the steps: S1, obtaining a depth image through a camera, and obtaining original point cloud data through the depth image; S2, using a camera to obtain an RGB color image; S3, designing an arbitrary direction target detector for the RGB color image, and detecting the obtained image to obtain an arbitrary direction candidate frame of the target; S4, projecting the candidate box in any direction obtained in the step S3 into the point cloud in the step S1, carrying out box extraction of the point cloud, and carrying out post-processing to obtain a relatively regular point cloud cluster; completing data association between the framing point cloud and the target category, and obtaining semantic information of the 3D target; and S5, converting the point cloud coordinates into coordinates in a world coordinate system by using visual odometer information, and storing the coordinates, so