tools to parse bundle adjustment results
The code is based on the following prerequisites:
- Eigen (tested on 3.3.4)
- vcglib (optional) get vcglib
prerequisites: cmake version >= 3.0
1. git clone https://github.com/StOriJimmy/libBATools.git
2. cd libBAToolsin the cmake-gui
3. set CMAKE_INSTALL_PREFIX to the directory you what to install the compilation
4. set EIGEN_INCLUDE_DIR if no environment variable is given
5. cmake generate
6. open the vs solution, and install
A demo project is attached for test
- .xml: ContextCapture BlocksExchange XMLformat
<BlocksExchange version="2.1">
<!-- Definition of one or several useful spatial reference systems (SRS). -->
<!-- Here, we need the French Lambert 93. -->
<SpatialReferenceSystems>
<SRS>
<!-- In the sequel, the SRS is referred to by this ID. -->
<Id>0</Id>
<!-- User-friendly SRS name. -->
<Name>Lambert 93</Name>
<!-- The spatial reference system definition can be either a Well Known Text definition (e.g. WGS84), "EPSG: ", a PROJ.4 definition, or the path to a file containing such a definition. -->
<Definition>EPSG:2154</Definition>
</SRS>
</SpatialReferenceSystems>
<!-- Optional: a path to prefix image paths when they are not absolute (default: use the directory of the exchange file). -->
<BaseImagePath>D:\data\Paris2012sample\Images</BaseImagePath>
<!-- Block definition. A single file can define several blocks. -->
<Block>
<!-- Block name and description are optional. -->
<Name>Paris2012</Name>
<Description>Small sample of the Paris 2012 dataset</Description>
<!-- The type of block is optional. Currently, there is a specific behavior for the "Aerial" type only. -->
<Type>Aerial</Type>
<!-- Photo downsampling rate can be optionally provided with <PhotoDownsamplingRate> tag -->
<!-- Minmum and maximum viewing distance can be optionally provided with <MinimumViewingDistance> and <MaximumViewingDistance> tags -->
<!-- ID of the base SRS used in the block. By default, the spatial reference system is assumed to be a local Cartesian coordinate system. -->
<SRSId>0</SRSId>
<Photogroups>
<!-- All photos taken using the same physical camera, with identical focal length and dimensions must be gathered in a photogroup. -->
<Photogroup>
<!-- Optionally, a different spatial reference system can be specified for each photogroup with the <SRSId> tag. By default, the SRS of the block is assumed. -->
<!-- Optionally, a name with the <Name> tag, and a description with a <Description> tag. -->
<Name>UCX</Name>
<!-- Image dimensions, in pixels. -->
<ImageDimensions>
<Width>9420</Width>
<Height>14430</Height>
</ImageDimensions>
<!-- Optional camera model type Perspective or Fisheye (Perspective type is used if this tag is unspecified). -->
<CameraModelType>Perspective</CameraModelType>
<!-- Sensor's largest dimension, in millimeters. -->
<!-- Sensor's pixel size, in millimeters, can be substituted for this field. -->
<!-- For this dataset, it would be: -->
<!-- <PixelSize>0.0072</PixelSize> -->
<SensorSize>103.896</SensorSize>
<!-- Focal length, in millimeters. This can be either the nominal specification or a closer estimate. -->
<FocalLength>100.735601903992</FocalLength>
<!-- Optionally, focal length, in pixels. Can replace FocalLength and SensorSize data. -->
<!-- For this dataset, it would be: -->
<!-- <FocalLengthPixels>13991.05582</FocalLengthPixels> -->
<!-- Optionally, specification of the xy-axis of the camera sensor for pose rotation and for any position given in millimeter (i.e. <PrincipalPoint> and <Measurement> when using <xmm> and <ymm>) -->
<!-- Possible values: XRightYDown (default), XRightYUp (more frequent in photogrammetry), XLeftYDown, XLeftYUp, XDownYRight, XDownYLeft, XUpYRight, XUpYLeft -->
<CameraOrientation>XRightYDown</CameraOrientation>
<!-- Principal point, with respect to the sensor center. This field is optional: by default, the sensor center is used as an initial estimate of the principal point. -->
<PrincipalPoint>
<!-- 2D image position of the principal point, in pixels. The origin is the center of the upper-left image pixel, the x-axis is oriented to the right side of the image, and the y-axis is oriented to the bottom of the image. -->
<x>4683.755692</x>
<y>7223.0141002</y>
<!-- Alternatively, a 2D position with respect to the sensor center can be specified, in millimeters. In this case, the xy-axis are oriented according to CameraOrientation. -->
<!-- For this dataset, it would be: -->
<!-- <xmm>-0.1853590176</xmm> -->
<!-- <ymm>0.06130152144</ymm> -->
</PrincipalPoint>
<!-- Lens distortion coefficients. This field is optional: by default, no distortion is assumed as an initial estimate. -->
<!-- Distortion coefficients correspond to the classical Brown's distortion model, as described in http://en.wikipedia.org/wiki/Distortion_(optics). -->
<Distortion>
<K1>-0.0010041516639</K1>
<K2>0.0056007056563</K2>
<K3>-0.009874171117100001</K3>
<P1>0</P1>
<P2>0</P2>
</Distortion>
<!-- Fisheye focal matrix used only for the camera model type Fisheye. -->
<FisheyeFocalMatrix>
<M_00>1960</M_00>
<M_01>0</M_01>
<M_10>0</M_10>
<M_11>1960</M_11>
</FisheyeFocalMatrix>
<!-- Fisheye distortion coefficients used only for the camera model type Fisheye. -->
<FisheyeDistortion>
<P0>0</K1>
<P1>1</K2>
<P2>0</K3>
<P3>0</P1>
<P4>0</P2>
</FisheyeDistortion>
<Photo>
<!-- The ID of a photograph must be unique over the whole block (in other words, even across other photogroups). -->
<Id>146</Id>
<!-- Path of the image. If not absolute, it is considered to be relative to BaseImagePath if defined, or otherwise to the directory of the exchange file. -->
<ImagePath>071_2810.jpg</ImagePath>
<!-- Optional path of the image mask. If not absolute, it is considered to be relative to BaseImagePath if defined, or otherwise to the directory of the exchange file. -->
<!-- <MaskPath>071_2810_msk.jpg</MaskPath> -->
<!-- Exterior orientation of the camera, as estimated by navigation devices or aerotriangulation. This field is optional. -->
<!-- Optional component Id: 1 is the main component (default), 0 is no component. Photos with incomplete pose cannot belong to the main component -->
<!-- <Component>1</Component> -->
<Pose>
<!-- 3D rotation of the camera in the spatial reference system defined above. -->
<!-- It defines the rotation matrix transforming world coordinates (in the specified SRS) into camera coordinates (with axes defined by the CameraOrientation of the photogroup). -->
<!-- The rotation is optional: the pose can be limited to the 3D position of the camera center (e.g. when GPS position only is available). -->
<Rotation>
<!-- It can be specified in one of the three representations below: -->
<!-- 1. A 3x3 matrix, using the <M_ij> tags, where M_ij denotes the coefficient of the (i+1)th row and the (j+1)th column of the rotation matrix. -->
<M_00>-0.9999982912233401</M_00>
<M_01>-0.001636319085375301</M_01>
<M_02>-0.0008602425863163225</M_02>
<M_10>-0.001631068695467463</M_10>
<M_11>0.9999802528616577</M_11>
<M_12>-0.00606906089589293</M_12>
<M_20>0.0008701565192966738</M_20>
<M_21>-0.006067647409696231</M_21>
<M_22>-0.9999812130648239</M_22>
<!-- 2. Omega/Phi/Kappa angles, in degrees, using the <Omega>, <Phi>, <Kappa> tags. Please refer to the formula in the camera model document. -->
<!-- For this dataset, it would be: -->
<!-- <Omega>179.6523471469934</Omega> -->
<!-- <Phi>0.04985630236313049</Phi> -->
<!-- <Kappa>179.9065465708369</Kappa> -->
<!-- 3. Heading/Pitch/Roll angles, in degrees using the <Heading>, <Pitch>, <Roll> tags. Please refer to the formula in the camera model document. -->
</Rotation>
<!-- 3D position of the camera center in the spatial reference system defined above. -->
<!-- Depending on the SRS, x/y may stand either for longitude/latitude, for map projection coordinates, or for something else. -->
<!-- Depending on the SRS, z may stand either for a geoid height, for an orthometric height, or for something else. -->
<Center>
<x>651999.7159189156</x>
<y>6863073.633923346</y>
<z>1318.897690166719</z>
</Center>
</Pose>
<!-- Optional depth information can be provided with <NearDepth>, <MedianDepth>, and <FarDepth> tags -->
<!-- Optional Exif data (each exif data property is optional) --><!--
<ExifData>
<ImageDimensions>
<Width>4000</Width>
<Height>3000</Height>
</ImageDimensions>
<PixelDimensions>
<Width>0.02</Width>
<Height>0.02</Height>
</PixelDimensions>
<GPS>
<Latitude>45.5686684</Latitude>
<Longitude>2.6551104</Longitude>
<Altitude>1589.5469</Altitude>
</GPS>
<FocalLength>24</FocalLength>
<FocalLength35mmEq>36</FocalLength35mmEq>
<Make>FUJIFILM</Make>
<Model>FinePix S3Pro</Model>
<LensModel>FinePix S3Pro</LensModel>
<DateTimeOriginal>2015-04-29T06:40:26</DateTimeOriginal>
</ExifData> -->
</Photo>
...
</Photogroup>
</Photogroups>
<!-- Control points are highly recommended for accurate positioning. -->
<ControlPoints>
<!-- Optionally, a different spatial reference system can be specified with the <SRSId> tag. By default, the SRS of the block is assumed. -->
<ControlPoint>
<!-- Control point name and description are optional. -->
<Name>Control point #1</Name>
<!-- Optionally, a different spatial reference system can be specified for each control point with the <SRSId> tag. By default, the SRS of the control points container is assumed. -->
<!-- 3D position of the control point in the active spatial reference system. -->
<!-- Depending on the SRS, x/y may stand either for longitude/latitude, for map projection coordinates, or for something else. -->
<!-- Depending on the SRS, z may stand either for a geoid height, for an orthometric height, or for something else. -->
<Position>
<x>652788.0525588237</x>
<y>6863015.362218254</y>
<z>78.07000000122935</z>
</Position>
<!-- Image measurements. -->
<Measurement>
<!-- ID of the photo where the measurement is taken. The measurements of a control point must have different photo IDs. -->
<PhotoId>151</PhotoId>
<!-- 2D image position of the measurement, in pixels. The origin is the center of the upper-left image pixel, the x-axis is oriented to the right side of the image, and the y-axis is oriented to the bottom of the image. -->
<!-- Alternatively, a 2D position with respect to the principal point, in millimeters. In this case, the xy-axis are oriented according to the CameraOrientation of the corresponding photogroup. -->
<!-- For this dataset, it would be: -->
<!-- <xmm>18.6231910176</xmm> -->
<!-- <ymm>-4.48973352144</ymm> -->
<x>7270.31</x>
<y>6599.44</y>
</Measurement>
...
</ControlPoint>
<ControlPoint>
<Name>Control point #2</Name>
<Position>
<x>652123.2211166573</x>
<y>6863245.976366176</y>
<z>80.07999999914318</z>
</Position>
<!-- If specified, a control point may be a check point (default: false)-->
<CheckPoint>true</CheckPoint>
<!-- If specified, a control point may have accuracies -->
<HorizontalAccuracy>0.01</HorizontalAccuracy>
<VerticalAccuracy>0.10</VerticalAccuracy>
<Measurement>
<PhotoId>146</PhotoId>
<x>3296.56</x>
<y>9253.75</y>
</Measurement>
...
</ControlPoint>
<ControlPoint>
<Name>Control point #3</Name>
<!-- If specified, a control point might be Full (default), Horizontal or Vertical -->
<Category>Horizontal</Category>
<Position> <!-- no z position -->
<x>652365.1205012415</x>
<y>6863549.148163618</y>
</Position>
<HorizontalAccuracy>0.01</HorizontalAccuracy> <!-- no vertical accuracy -->
<Measurement>
<PhotoId>95</PhotoId>
<x>3178.26</x>
<y>4020.21</y>
</Measurement>
...
</ControlPoint>
...
</ControlPoints>
<!-- Known tie points may be used to refine an aerotriangulation. -->
<!-- Uncomment this section before import to get a single tie point in the imported block. Please note that a reconstruction created from a such block will fail. -->
<!--<TiePoints>
--><!-- Optionally, a different spatial reference system can be specified with the <SRSId> tag. By default, the SRS of the block is assumed. --><!--
<TiePoint>
--><!-- Tie point name and description are optional. --><!--
--><!-- Optionally, a different spatial reference system can be specified for each tie point with the <SRSId> tag. By default, the SRS of the tie points container is assumed. --><!--
--><!-- Optional 3D position of the tie point in the active spatial reference system. --><!--
--><!-- Depending on the SRS, x/y may stand either for longitude/latitude, for map projection coordinates, or for something else. --><!--
--><!-- Depending on the SRS, z may stand either for a geoid height, for an orthometric height, or for something else. --><!--
<Position>
<x>652119.8871409688</x>
<y>6863304.305716386</y>
<z>87.79328384995461</z>
</Position>
--><!-- Optional tie point color. --><!--
<Color>
<Red>0.59</Red>
<Green>1.0</Green>
<Blue>0.0</Blue>
</Color>
--><!-- Image measurements. --><!--
<Measurement>
--><!-- Optional measurement type User or Automatic (Automatic type is used if this tag is unspecified). --><!--
<Type>Automatic</Type>
--><!-- ID of the photo where the measurement is taken. The measurements of a tie point must have different photo IDs. --><!--
<PhotoId>146</PhotoId>
--><!-- 2D image position of the measurement, in pixels. The origin is the center of the upper-left image pixel, the x-axis is oriented to the right side of the image, and the y-axis is oriented to the bottom of the image. --><!--
--><!-- Alternatively, a 2D position with respect to the principal point, in millimeters. In this case, the xy-axis are oriented according to the CameraOrientation of the corresponding photogroup. --><!--
<x>3324.26001</x>
<y>9930.269531</y>
</Measurement>
<Measurement>
<PhotoId>158</PhotoId>
<x>9079.006836000001</x>
<y>9902.772461000001</y>
</Measurement>
<Measurement>
<PhotoId>162</PhotoId>
<x>6240.366211</x>
<y>9896.118164</y>
</Measurement>
</TiePoint>
</TiePoints>-->
<!-- Optional block's positioning constraints based on user tie point (0 based indices). -->
<!-- Each constraint is optional. -->
<!--<PositioningConstraints>
<OriginConstraint>
<O>2</O>
</OriginConstraint>
<ScaleConstraint>
<A>0</A>
<B>3</B>
<DistanceAB>56.350</DistanceAB>
</ScaleConstraint>
<AxisConstraint>
<A>0</A>
<B>3</B>
<AxisAB>z</AxisAB> --><!-- x, y, or z --><!--
</AxisConstraint>
<OrientationConstraint>
<A>0</A>
<B>3</B>
<C>2</C>
<AxisAB>z</AxisAB> --><!-- x, y or z --><!--
<AxisSideC>upperX</AxisSideC> --><!-- lowerX, upperX, lowerY, upperY, lowerZ, or upperZ --><!--
</OrientationConstraint>
</PositioningConstraints>-->
</Block>
</BlocksExchange>
- .info: las and image registration result, cv system
--------------------------------Camera Information--------------------------------
CameraID x0 y0 f format_X format_Y pixelSize k0 k1 k2 k3 p1 p2 b1 b2 Attrib
<num_cameras>
<camera1>
<camera2>
<camera3>
...
注：采用CV畸变系数
--------------------------------Image Information--------------------------------
ImageID ImageName Xs Ys Zs Phi Omega Kappa StripID Attrib CameraID bFlag BlockID
<num_images>
<image1>
<image2>
...
- bundle.out: Bundler file v0.3
# Bundle file v0.3
<num_cameras> <num_points> [two integers]
<camera1>
<camera2>
...
<cameraN>
<point1>
<point2>
...
<pointM>
Each camera entry <cameraI> contains the estimated camera intrinsics and extrinsics, and has the form:
<f> <k1> <k2> [the focal length, followed by two radial distortion coeffs]
<R> [a 3x3 matrix representing the camera rotation]
<t> [a 3-vector describing the camera translation]
more information can be found in the official website
http://www.cs.cornell.edu/~snavely/bundler/bundler-v0.3-manual.html#S6
- image list
Ascii file, recording image path (relative or absolute) and size. If no image size information in the list, set the image info by yourself
For each line:
image_path width height
- convert world to image coordinates and backwards
- convert world to camera coordinates and backwards
- check point in image
- get viewing direction and view point (camera center)
- ray tracing
- point-wise visibility check
If you found bugs or have new ideas, do not hesitate to pull request 😬
If you have trouble compiling or using this software, email to [email protected]