TODO(oalexan1): This must be integrated with ISIS

Delete vertices in a given box

Compute which camera pixel observes a DEM pixel.

Given an orbit given by the first and last camera center positions in projected coordinates, a real number t describing the position along this line,

If the input folder contains only a single folder, return that contained folder. Otherwise return the input folder.

Detects features, matches them between image pairs, and builds tracks. This function handles the entire feature detection and matching pipeline: 1. D

Detect interest points. Return also the rough homography that aligns the right image to the left.

Determine the output nodata value. Read from first DEM if not user-specified, and validate that it can be represented exactly as a float.

Determine the common shift value to use for the output files

Filter D_sub by reducing its spread around the median. Read from disk, filter, and write back to disk.

This is some logic unrelated to triangulation, but there seems to be no good place to put it. Unalign the disparity, and/or create match points from d

Disparity norm

TODO(oalexan1): Use std::abs() instead of fabs(). May change the result.

When distortion params are shared, their number must agree

The main fetch function. Returns the number of failures.

Do all of the processing.

Use Ceres to do bundle adjustment.

Draw the images on the screen

Convert keypoints from Eigen format

This is needed to ensure distortion coefficients are not so small that they don't get optimized. This modifies the camera models in place. This does n

Given a pixel (x, y), and a length l, enumerate all the pixels on the boundary of the square of side length 2*l centered at (x, y).

Erode the DEM based on grassfire weights and recompute weights using centerline algorithm. This is used for priority blending where weights follow the

TODO: This function should live somewhere else! Erode this many pixels around invalid pixels

Given the quaternions in a camera model, calculate the curvature for each coordinate. This cannot be done at the starting and ending coordinate, as we

Find the best-fit exposure and haze given the input sampled image and reflectance. Also find the sampled albedo along the way. The albedo will be opti

Initialize the exposure as mean(intensity)/mean(reflectance)/albedo. Initialize the haze to 0. Skip images with zero exposure. If the user provided in

Sample the image and get generous estimates (but without outliers) of the maximum triangulation error and of the 3D box containing the projected point

Given a set of points in 3D, heuristically estimate what it means for two points to be "not far" from each other. The logic is to find a bounding box

Given the normal (height) to the SfS DEM, find how different a height can be from this before the computed intensity due to that height is bigger than

Estimate the valid height range in a region based on input height info. Also generates the lidar DEM.

Get a generous estimate of the bounding box of the current set while excluding outliers

Estimate the centroid of the reference points

Estimate the region in the right image corresponding to left_trans_crop_win based on ip in the current box and also by creating ip from D_sub. If cann

Estimate surface resolution by looking at four immediate neighbors. Return 0 if at least 2 neighbors are missing or the current point.

TODO(oalexan1): This needs more work

Save the depth clouds and optimized transforms needed to create a mesh with voxblox (if depth clouds exist).

Export the latest xyz values from param_storage to a vector of Vector3

Extract data from HDF file to temporary directory.

Extract raw data from the parsed XML for direct CSM population. No resampling is done here. The raw positions, velocities, and quaternions are returne

Extract the 3x3 rotation matrix R and 3D translation vector T from a 4x4 ECEF transformation matrix.

Collect the cam2world matrices and camera centers in vectors.

Extract data from an Icebridge hdf5 file. Use this ASP format code to read: --csv-format 1:lat,2:lon,3:height_above_datum

Trace rays from pixel corners to DEM to see where they intersect the DEM. Save a pointer to the camera, georef, and interp_dem, as we may need these l

Given an input string having algorithm options, like "-v 4", and environmental variables, like "VAL=5", possibly with repetitions, so VAL=5 and VAL=6

Extract data from a n Icebridge qi file. Use this ASP format code to read: --csv-format 1:lat,2:lon,3:height_above_datum

TODO(oalexan1): Move this to a better place Given a set of orbital positions acquired at with uniform time spacing, corresponding velocities, the time

Download all data needed to process a run

Fetch a camera folder from the archive if it exists. Returns True if we got the file. If more than one, return the latest by modification tim

Fetch all the nav data for a flight.

Fetch all the files for a HiRISE image and return the list of files.

Compute alignment using FGR

Make sure file exists and is non-empty.

Make a copy of the batch list file which only contains incomplete batches.

Filter GCP outliers based on mean reprojection error per triangulated point. Adds to the existing outlier set (does not clear it).

Keep only image files, shape files, and csv files

Filter matches by projection window. TODO(oalexan1): Use this in jitter_solve. TODO(oalexan1): This needs to be done before subsampling the matches

Filters out all points from point_cloud with an error entry higher than cutoff

Filter D_sub. All alignment methods are supported.

Filter ip by triangulation error and height range. This assumes the interest points are for the original images, without alignment or mapprojection. T

Unalign the ip, filter them using the cameras, align them back, and return the inlier ip.

Points in source_point_cloud farther than opt.max_disp from the reference cloud are deleted.

Find how to map cids which may have repetition, to new unique cid. Also sort the new cid by image name.

TODO(oalexan1): Decrease the complexity of this function. Use binary search. Also process each rig group separately.

Given the direction going from first_proj to last_proj, and knowing that we are at curr_proj, find if we are before or after ref_proj. Return 1 if aft

Estimate the transform from B_xyz_vec to A_xyz_vec using RANSAC. A lot of outliers are possible.

Find the best matching lidar file from a list.

Find the interpolated disparity value at a given pixel. If not valid, look in a small neighborhood. That should not be overused.

Get a list of matched IP based on an input disparity

Hillshade the reference and source DEMs, and use them to find interest point matches among the hillshaded images. These will be used later to find a

Function to find the mean camera position

Find the point cloud center TODO(oalexan1): Move this to some low-level point cloud utils file

Compute the 3D locations at the shape corners based on interpolating into the DEM and converting to ECEF and to local projected stereographic coordina

Read a set of lon-lat measurements in CSV format, then interpolate into the DEM with bilinear interpolation to find the height.

Read a set of measurements in CSV format, to use later to fit the water surface

Estimate the projection and convert llh_vec to projected coordinates

Find shortest distance from camera center to the origin

Computes a system solution from a seed and returns the final error number.

Find the tile at given location for a parallel_stereo run with local epipolar alignment.