using a plurality of cameras with different viewing angles to image a symbol to obtain a symbol image, a reference object for camera calibration being called the symbol, the symbol comprising a plurality of markers, and each of the markers having a corresponding ID number;

performing gray-scale processing and binarization processing on the symbol image in sequence to obtain a binary image of the symbol; performing ROI area extraction and corner detection on the binary image of the symbol, and according to the extracted contour area and its concave and convex properties, and the number of corners to judge and obtain the marker; and

identifying the ID number of the marker according to values of black and white pixels in a coded area of the marker and searching for world coordinates corresponding to the marker according to the ID number;

computing an external parameter matrix of the camera according to marker coordinates of a camera coordinate system and marker coordinates of a world coordinate system, and unifying point clouds under the world coordinate system to obtain a plurality of point clouds under different viewing angles; and

stitching the plurality of point clouds together to obtain a 3D reconstructed image.