/DSMGenerator

Primary LanguagePython

Introduction

This repo aims to generate the 3D point cloud from two 2D images. All processes includes Tie Point Finding, Rectification, Dense Matching and Point Cloud Generating.

Result

CASE1: Aerial Image

  1. Source

    Left Right
    Aerial_Source_l Aerial_Source_r

  2. Result

    • SGBM_CloudPoint

Theory

Please see Pinhole Camera Model PPT

Scripts

  • TiePoints.py: Use SIFT, ORB and SURF to find the sparse matching point (same point in real world) from 2 images.

    python TiePoints.py
    • TP_00
    • TP_01
    • TP_02
    • TP_10
    • TP_11
    • TP_12
    • TP_20
    • TP_21
    • TP_22
    • TP_unFiltered
    • TP_Filtered

  • Rectification.py: Rectify the image for application of SGBM reduce the searching dimension from two dimensions to one dimension.)

    python Rectification.py
    • Rectification_EpipolarLine
    • Rectification_Rectification
    • Rectification_Add

  • SGBM_opencv.py: OpenCV version SGBM.

    python SGBM_opencv.py
    • SGBM_TP
    • SGBM_Rectification
    • SGBM_TP_Rectification
    • SGBM_Disparity
    • SGBM_CloudPoint

  • AeroTriangulation.py: Conversion between 2D image pixels into 3D object points.

    python AeroTriangulation.py

  • AeroTriangulation_tf.py: This is used for calibrate the relational extrinsic parameters of stereo camera. But its recommended to use the chessboard to calibrate the extrinsic parameters of stereo camera.

    python AeroTriangulation_tf.py
    • AereoTriangulation

  • io_aereo_params.py: This is used for get the saved aereo parameters (OPK, L_XYZ, DMC_ROWS_LABEL, DMC_COLS_LABEL, DMC_FOCAL_LENGTH, DMC_PIXEL_SIZE, XOFFSET, YOFFSET)

  • temp_refine_resolution.py: This is used to lower the resolution of original images in order to push the images on to the gitlab.

  • SGBM.py: This is jit implementation of SGBM which can be used to learn SGBM, but its efficiency is not terrible. currently not working.