Moving Object Detection Package

Jean-Marc Petit, Matt Holman
Hans Scholl, JJ Kavelaars, Brett Gladman

  • Download the detection package now

    This package is describe in the following paper:
  • A highly automated moving object detection package
  • README file from the package:
    README file for the Moving Object Detection package.

    This packacge is available under the Gnu Public License, without guarantee or support.

    The package can be retrieved from ""

    This package has been tested on several Linux distributions including RedHat 7.0, 7.2, 8.0 and 9.0. It ought to work on most recent distributions of Linux. It has also been tested some times ago on SunOS.

    Content of the package:

    +-- clean-dist: a small script to clean up the distribution
    +-- install_all: the installation scrip
    +-- jmp/: the Wavelet path directory
    +-- pipematt/: the S-Extractor path directory
    +-- README: this very file

    How to install:

    1: retrieve the package "detect-pack-yyyymmdd.tgz" where yyyy stand for the year, mm for the month and dd for the day of the release. Put the file in the installation directory .

    2: untar the file in the installation directory:

    ...$ tar xzf detect-pack-yyyymmdd.tgz


    ...$ gzcat detect-pack-yyyymmdd.tgz | tar xf -

    This will create a directory called "detect-pack" under .

    3: Run the install script "install_all"

    ...$ cd detect-pack
    ...$ ./install_all [-p |--prefix ]

    where is the directory where the package will be installed. Usually, this is "." or /detect-pack

    4: Add some directories to the PATH and LD_LIBRARY_PATH environment variables:

    ...$ PATH=${PATH}:/bin/LINUX:/bin
    ...$ export PATH LD_LIBRARY_PATH

    Be sure to respect the order of directory in PATH.

    At this point, you should have a working detection package.
    How to run the detection package:

    This package assumes that there exists a directory tree structure that contains the 3 images of a same field in each directory. This structure is listed in a file using the script "". This script assumes a directory tree structure like

    < field-name >/< chip-name=c??|chip??|im?? >
    But this can easily adapted to your personnal needs.

    This creates a file named "directory" (not a good choice obviously) in the current directory.

    Using this directory list, one runs "" to find the correct fits images and determine the FWHM of each image, producing a file called "proc-these-file" in each directory and a list of directories to process, "find.input" in the current directory.

    ...$ [-d < dir >] [-p < prefix >] [-s < suffix >] [-f < seeing >] [-h|-?]

    < dir > is the directory where the file "directory" is located.
    < prefix > and < suffix > are the prefix and suffix of the image names: "${prefix}*${suffix}.fits"
    is the default FHWM of the images.
    -h or -? give a very limited help.

    Now you can run the "" script that will run the detection pipeline on each field.

    ...$ [-d < dir >] [-rn < min_rate >] [-rx < max_rate >] [-ang < angle >] [-aw < openning angle >] [-h|-?]

    < dir > is the directory where the file "find.input" is located.
    < min_rate > is the minimum rate of motion for searching.
    < max_rate > is the maximum rate of motion for searching.
    < angle > is the mean direction of motion, expressed in degrees, anticlockwise, from the right horizontal direction.
    < openning angle > is the openning angle about the mean direction for searching.
    -h or -? give a very limited help.

    For each directory where things go right, the script write its name in the file "processed" in the directory < dir >, and touches a file "processed" in the field/chip directory.

    The result of the S-Extractor path is given in "< image1 >.moving.matt", the result of the Wavelet path is in "< image1 >", and the combination of both is in "< image1 >.cands.comb".


    This page is maintained by