nthuy190991 / temp

Thesis: A Versatile and Efficient Data Fusion Methodology for Heterogeneous Airborne LiDAR and Optical Imagery Data Acquired Under Unconstrained Conditions

Matlab codes developed by TH Nguyen, Université Laval & IMT Atlantique

Main folders:

• phd_dev: common folder contains the coarse registration development, and other miscellaneous codes
• Snakes: dedicated to the early development of snake models
• SR: dedicated to the super-resolution and fine registration
• SRSM: dedicated to the super-resolution-based snake models used in Remote Sensing MDPI article
• R2Sonic: dedicated to the R2Sonic suspended sediment challenges

Registration dev:

Building Segment Extraction from LiDAR

• building_region_seg: extraction of building regions from LiDAR data, resulting a building mask
• buildingBoundaryExtraction: extracting boundaries given a mask (as seeding regions)

Building Segment Extraction from Image

• meanshift_seg_final: run meanshift segmentation and filtering, call MeanShiftCluster
• MeanShiftCluster: meanshift algorithm
• segmentFiltering: filtering segments based on their area
• building_img_seg_meanshift: refinement of meanshift resulting segments with the MBR filling percentage

Segment Matching

• compare_segs: compute segment MBR filling percentage (in preparation for the segment matching)
• compare_segs_b4regis: running GTM/RANSAC for the segment matching
• minBoundingBox: MBR detection algorithm
• GTM_algo_original: GTM algorithm. Sub-functions for GTM in the GTM folder

Transformation Model Estimation

• TME_GS_algo: transformation model estimation, calling gold_standard_algo
• gold_standard_algo: Gold Standard algorithm for the estimation of the general camera matrix
• gold_standard_algo_affine: Gold Standard algorithm for the estimation of the affine camera matrix
• normalize_pts: centering of a set of points, by their center, i.e. x_i = x_i-mean(x_i)
• runGoldStandard: Gold Standard algorithm resulting in different camera components K, C, R

Fine Registration

• regis_6params_9patches: fine registration on patches, involving the maximization of MI/NCMI
• calc_mi_with_P: calculate MI given a camera matrix P
• calc_mi_with_theta_6params: calculate MI given 6 exterior parameters
• calc_ncmi_with_P: calculate NCMI given a camera matrix P
• SR_for_MI: super-resolution to generate i-image (later used for the maximization of MI), calling SR_by_grad.m
• SR_for_NCMI: super-resolution to generate i-image and z-image (later used for the maximization of NCMI), calling SR_by_grad.m
• SR_by_grad: run the FISTA algorithm for the super-resolution
• mi: measure Mutual Information between two images
• ncmi: measure Normalized Combined Mutual Information between three images
• patch_based_interpolation: IDW-based interpolation for the smoothing between patches

Super-resolution (SR)

• depth_map_SR and run_SR_by_grad: scripts for the super-resolution of a depth-map
• cost_func_Dxy: cost function defined by the sum squared directional gradients (SSGD)
• gradient_Dxy: gradient of the cost function SSGD
• gradient_Dxy_fast: gradient of the cost function SSGD (faster version)
• soft_thres: shrinkage operator

Snake Model

(The implementation of original snake models was developed by D. Kroon University of Twente. To download the original codes, please visit link)

• building_seg_isprs: preliminary extraction of building points on ISPRS Vaihingen dataset
• BE_by_snakes_ISPRS: call snake model for building extraction
• Snake2D: run snake model algorithm
• ExternalForceImage2D: compute external image-based energy function
• GetContourNormals2D: compute the normals of the contour points using the neighbouring points (used in traditional balloon force)
• GVFOptimizeImageForces2D: gradient vector flow (GVF)
• ImageDerivatives2D: gaussian based image derivatives
• InterpolateContourPoints2D: resamples a given contour to a smooth contour of uniform sampled points
• MakeContourClockwise2D: make contour clockwise
• SnakeInternalForceMatrix2D: compute internal energy function
• SnakeMoveIteration2D: compute the snake movement at iteration
• snake_polygonization: apply a polygonization (Douglas-Peucker or Dutter) on snake model results
• polygonization_Dutter: polygonization algorithm proposed by (Dutter 2007)

SRSM

(Many of the codes are the same with the ones in folder Snakes and SR)

• SRSM_prototype: run SRSM to extract buildings
• SRSM_prototype_ISPRS: run SRSM to extract buildings on ISPRS Vaihingen dataset
• SR_z: call super-resolution algorithm to generate z-image
• SR_by_grad: super-resolution algorithm
• SnakeMoveIteration2D_pixel_wise: compute the snake movement at iteration with option for modified balloon force
• las2txt: convert LAS files to TXT

Misc.

• calc_IoU: calculate Intersection-over-Union metric
• calcPolisMetric: calculate PoLIS metric
• convertGeoCoorMatCoor: function to convert georeferenced coordinates into matrix coordinates
• convertMatCoorGeoCoor: function to convert matrix coordinates into georeferenced coordinates
• countObject: function to count an object (for object-based accuracy) if it overlaps 50% of its reference boundary.
• display_TP_FP_FN: display true positives, false positives, and false negatives, for building extraction results evaluated with a ground truth
• dist_euclid: compute Euclidean distance from a point to a line
• dpsimplify: Douglas-Peucker line simplification algorithm
• findNearestPoints: find the nearest points within a cloud to a given location (x,y)
• findPoints: find the points within a cloud that have horizontal coordinates within a rectangle [xmin xmax ymin ymax]
• HausdorffDist: measure Hausdorff distance between two sets of points
• load_datasets: load Quebec datasets (image and LiDAR)
• load_datasets_isprs: load Vaihingen datasets (image and LiDAR)
• maxk_myfunc: find k largest elements of an array (if maxk is not avaible)
• mink_myfunc: find k smallest elements of an array (if mink is not avaible)
• project_pts_given_TM: project points given a transformation model (i.e. camera matrix P)

R2Sonic Challenge

• readXTFFiles_R2Sonic_Truepix: process XTF files to read and extract TruePix data, calling R2SONIC_TRUEPIX

• R2SONIC_TRUEPIX: function to process XTF files and read TruePix data header

• R2SONIC_TRUEPIX_D: function to process XTF files and read TruePix data packets

• XTFPINGHEADER: function to process XTF files and read the sonar header (metadata, config, etc.)

• XTFATTITUDEDATA: function to process XTF files and read attitude data

• XTFPOSRAWNAVIGATION: function to process XTF files and read raw navigation data

• createTruePix_sediments: create bathymetric point cloud from TruePix data

• createWCI_from_Truepix: create water column fan-shaped images from TruePix data

• truepix_sediments_Sv: attempt to compute backscattering strength per unit volume of ensonified water, inspired by [Simmons et al 2010]

• localisation_gage: try to locate the gage and extract the TruePix data in proximity

• physical_samples_analysis: read and plot physical samples, in attempt to analyze them and construct historical data