DIRSIG

As many of you may have read in past posts by Niek , we have been using the Blender software heavily in the construction of scenes so they can be ingested into DIRSIG . While we have many Blender tools that work in the editing and exporting of existing scenes, they only utilize a small subset of features that exist in Blender. Blender can be viewed as the vi (or emacs) of the 3D animation world which means it can be very powerful but at times opaque to the newly initiated. After paying our dues in the apprenticeship of Blender knowledgedom, we wanted to make accessible to DIRSIG users the capabilities that have made Blender the de facto standard of the 3D animation domain. In an ongoing project that addresses the need of data sets for training machine learning algorithms and assessing the performance of target detection/exploitation algorithms, DIRS has focused on implementing tools for producing numerous instantiation of simulated scenes. Blender has been an integral part ...

A new "refinery" scene will be included in the next DIRSIG release. The scene was inspired by the Ras Lanuf oil refinery in Libya.  It features an industrial site sitting on a desert-water interface.  Two DIRSIG renderings are shown below. The highlight is the refinery itself, featuring pipes and general industrial clutter.  The plant connects to a railroad track and an access road.  A small oil carrying ship is anchored nearby.  The desert itself is filled with shrubs. As with the Warehouse scene, the geometry was placed using our Blender scripts .  This refinery scene took roughly a week to construct.

The next DIRSIG Basic Training course will be held at the RIT Campus on May 25-27, 2011. The course is open to everyone. More information is available on the DIRSIG training webpage

The RIT Digital Imaging and Remote Sensing (DIRS) Lab has a booth (#3014) in the Exhibition Hall at this year's SPIE Defense, Security, and Sensing symposium in Orlando, FL (April 25-27). Adam Goodenough, Niek Sanders and myself will be hanging out at our booth talking to conference goers about DIRSIG. Other lab personel will be on hand to discuss RIT's other research interests ranging from MSI and HSI algorithm development to rapid data collection and delivery. In addition to the exhibition, we will be milling about the Technical Conference. Specifically, I am chairing Session #9 entitled "Landsat Data Continuity Mission" in the "Algorithms and Technologies for Multispectral, Hyperspectral, and Ultraspectral Imagery XVII" conference on Wednesday morning. In that session, I will be showing some of the work we are doing with DIRSIG to support pre-flight modeling of the next-generation Landsat system. These new features to streamline the modeling of space-ba...

The Final Release of DIRSIG 4.4.1 is out. This release is primarily a maintenance release to address a few bugs in the 4.4.0 release. The following is a summary of fixes and features added in this release: First release to include a Windows 64-bit version Improvements to correlated deviate models used for jitter The frequency spectrum can now (optionally) include the phase New demonstrations Indoors1 ExtinctionProps1 Blender plug-ins for DIRSIG file formats Various LIDAR related improvements Speedups for atmospheric backscatter returns Bug fix for Linear-mode APD LIDAR detector model Various GUI bug fixes and improvements Improvements to streamline integration with SUMO and CityEngine SUMO vehicle track import utility included on all platforms Simplified shared library requirements for Solaris 9 and Solaris 10. Initial MuSES support built-in for the following platforms Windows 32-bit Windows 64-bit We wish to thank those early adopter users for the...

In a previous post, we discussed the advantages of using OBJ files in DIRSIG. This tutorial video shows how Blender can assign DIRSIG materials to an OBJ geometry mesh.  Basic familiarity with Blender is a prerequisite. Attributing an OBJ with DIRSIG materials (video) The idea is simple: in the OBJ file, the "use material" ( usemtl ) string will specify a DIRSIG material id rather than just an arbitrary material name.  The OBJ's "material library" ( mtllib ) string is ignored by DIRSIG.

Many users want just raw radiance data from DIRSIG.  Rather than apply a built-in instrument model, they do their own post-processing of the radiance results.  The "Raw" capture method can be used to achieve this.  It writes raw DIRSIG results without applying any spatial or spectral integration. Here's a snapshot of the Raw Capture Method configuration dialog.  We'll now work through each section.   The Spectral Range controls the wavelengths for which DIRSIG will calculate and output results.   In this example, DIRSIG will compute from 0.40 microns to 2.50 microns in 0.01 micron increments.  The output image will have 211 bands of result data.  The delta must divide the range in to a whole number of bins.  For example, the range 0.4 to 0.7 microns with a 0.2 micron delta will result in an error message when DIRSIG runs. The Spatial Response controls how many results are computed and written out per detector array element....

For quite some time, the DIRS lab has been using Blender to build small DIRSIG scenes. Blender is an advanced, open source, cross-platform, 3-D modeling package. It is available free of charge, but it has a fairly steep learning curve. We've created Blender plug-ins allowing us import and export some of DIRSIG's native formats, including GDB and ODB. Blender also allows us to use geometry with the OBJ format . We can now create new scenes from scratch by assembling geometry, placing instances using Blender, and then saving out a DIRSIG ODB file. Traditionally, we have distributed the Bulldozer application for both attributing materials to geometry and constructing scenes. This set of plug-ins allows Blender to be used as an alternative to Bulldozer for constructing scenes. Starting with the official DIRSIG 4.4.1 release, we will be distributing these Blender scripts to the DIRSIG user community. All future DIRSIG distributions will include these scripts in the dir...

Last week, Brian Flusche (one of our students) defended his Ph.D. dissertation entitled "Analysis of Multi-Modal Fusion for Target Detection in an Urban Environment". Brian's research reflects how many people are using DIRSIG to simulate data in an effort to explore the value of new and novel exploitation schemes. In this case, Brian was trading the value of hyper-spectral imaging vs. polarimetric imaging vs. a combination of the two for doing target detection in an urban environment. In addition to his great use of DIRSIG (you can't expect us to be unbiased!), Brian was able to draw some very interesting and valuable conclusions. It you would like too watch his defense, it was recorded and is available here . A copy of his dissertation is available here . The DIRSIG Team would like to congratulate Brian on his work and the successful defense of his Ph.D research. Good job, Dr. Brian.

We've just put together another DIRSIG "demo" simulation. It shows a camera placed indoors. The scene is illuminated by both a secondary source at the ceiling and through a window cut-out at the far end of the room. This demonstration will be included in the next DIRSIG release.

The next DIRSIG Basic Training course will be held at the NASA Ames Research Center on Feb 23-25, 2011. The course is open to everyone (not just NASA employees). More information is available on the DIRSIG training webpage

We have placed an update to the MegaScene1 distribution on myDIRSIG. The major aspect of this release is that it comes with a set of DIRSIG4 .scene files. This release also includes updates to material properties including the conversion of the tree leaf optical properties from extinction to transmission. We have found that changing this optical property provided speed ups because a large number of calls to the exponent function (to compute the transmission from the extinction) are eliminated. You can also download an update to MicroScene1. Like MegaScene1, we had neglected to push out an update of MicroScene1 with DIRSIG4 .scene files that were as simple to use as "unzip and use". Both updated scene files can be downloaded by registered users from the myDIRSIG website.

This release is primarily a maintenance release to address a few bugs in the 4.4.0 release. A "release candidate" is a preview of the next release of DIRSIG. This version of the software is produced when we feel the software is ready for release, but we would like some last-minute user feedback. Our early adopters always help spot a few lingering issues that our release quality control process misses. We expect the official version of 4.4.1 to be released around Jan 14th, 2011 The following is a summary of fixes and features added in this release: First release to include a Windows 64-bit version Improvements to correlated deviate models used for jitter The frequency spectrum can now (optionally) include the phase Various LIDAR related improvements Speedups for atmospheric backscatter returns Bug fix for Linear-mode APD LIDAR detector model Various GUI bug fixes and improvements Improvements to streamline integration with SUMO and CityEngine SUMO v...

A common question we get asked is how to visualize the point cloud data produced by either the Linear-mode or Geiger-mode LIDAR simulations. First, you should remember that the point cloud files produced by the "APD Processor" are simple ASCII/text files. Each line is the entry for a single return or "point" in the point cloud, including the point location and some attributes that vary depending on whether you modeled a Linear-mode or Geiger-mode system. For a Linear-mode system, a point cloud file will generally look like the example below: 12.7388 -45.3612 -0.0256 5.0290 0 0 0 0 12.8169 -45.3612 -0.0264 4.8362 0 1 0 0 12.8950 -45.3612 -0.0271 4.8362 0 2 0 0 ... 32.4008 -25.5446 10.5945 4.6783 0 65533 0 0 32.4781 -25.5446 10.5953 5.8959 0 65534 0 0 32.5360 -25.5640 12.5408 5.9185 0 65535 0 0 The first three columns are the X/Y/Z location of the point return. The 4th column is the intensity (in photons). Since Linear mode can support multiple returns per pulse, t...

We often want the ability to add clouds to a large scale scene in order to see the reflective, transmissive and shadowing effects on the scene. In these cases we are not so much concerned with the absolute radiometry of the clouds (true multiple scattering, exact modeling of limb regions, etc..), as we are with capturing bulk effects (single-scattering contributions, transmission effects, realistic spatial distributions, etc..). With these cases in mind, and to support modeling of the next generation Landsat sensor, we are developing a simplistic and very fast large-scale cloud model based on using the USAF Cloud Scene Simulation Model ( CSSM ) and OPAC data to generate voxelized representations of inherent optical properties. DIRSIG uses this data to drive a simple radiative transfer model that captures the bulk of the radiometric effects on the scene with very little additional computation overhead. The process is currently going through a period of validation and refinement and sh...

Three weeks ago we announced a "release candidate" for DIRSIG 4.4.0. In the time since, we had quite a few early adopters try out 4.4.0 and provide us with feedback. Those improvements and bug fixes have now been incorporated into the official release of DIRSIG 4.4.0, which is now ready for download on myDIRSIG for registered users. Like we said at the time of the release candidate, we are very excited about this release as it packs in a bunch of new features and performance improvements: New User Interface Tools Graphical simulation preview Basic image viewer Component browser SGP 4 Orbiting Platform Motion Wizard Geometry Improvements Added affine transform INFO option to ODB Ray-trace optimization for dynamic (moving) scene geometry Support for vertex normals (normal "shading") on OBJ input geometry Atmosphere Added the GUI for the built-in "uniform" atmosphere model User-defined MODTRAN profile support (make_ adb config ...

Over the past few years we have been doing more simulations of video systems. Perhaps you have seen the video simulation of MegaScene1 with the cars driving down the streets as would be acquired from a circling UAV: So how do we make these videos? It is really a simple tool chain involving a tool that is distributed with DIRSIG and another tool that can be downloaded from the web. For this example, we are going to deal with a 2D framing array type system and that has a red, green and blue (RGB) channel set being captured (read out) on a regular interval: DIRSIG itself doesn't generate video files directly, but you can employ DIRSIG to generate the individual frames for a video and them combine them. This workflow breaks down into a three (3) step process: Instruct DIRSIG to output each capture (or "frame") of the focal plane to an individual file. Convert each floating-point radiance image file to a 24-bit (8-bits per channel) RGB image file. Encode the RGB image files...

When we first started on the DIRSIG model in the late 1980s, 3D computer graphics was still in its infancy. At the time, standard file formats for 3D geometry were still being established. In addition, the DIRSIG model had the need for a specific set of per-facet attributes (material ID, thickness, override temperature, etc.) that are still somewhat unique in the 3D modeling community. As a result, the DIRSIG specific Geometric Database (GDB) format was created and is still used to this day. In the mean time, a variety of 3D geometry file formats have become widely used. Although the GDB format serves our needs well, it requires the user to translate the 3D geometry assets from other formats in order to use them with DIRSIG. Lately we have been experimenting with using the Alias/Wavefront OBJ file format as a possible replacement for GDB. The important features of the OBJ format are: It is widely supported by 3D geometry authoring tools (Rhino, SketchUp, Blender, etc.). It supports per...

In the DIRSIG 4.4.0 release users will find that we made some modest improvements to the classic Urban and Forest scenes. Primarily, we improved the scene geometry by replacing the decade old, low facet count trees with the high fidelity trees we have been using in other scenes. We have also cleaned out a bunch of unused files and tried to improve the organization of the respective scene directories. The first thing you will notice in the updated Urban scene (seen above) is the new trees and the fact that there is more than one tree species. The scene also has some new textures for the grass and asphalt areas. This scene now has both day/night and summer/winter configurations. The "night" .scene files have the streetlights automatically enabled. The "winter" .scene files use defoliated trees. You might ask why the river running down the middle of the scene is so green in color. The river is modeled as a flat surface with an effective reflectance measured from s...

One of the ways we try to teach users about specific features is by making a "bare bones" simulation as a demonstration of that feature. These "demos" don't usually generate a very exciting output data product, but they are fully self-contained (no external data required) simulations that focus on how to configure a specific feature. We have been distributing our growing set of demos with DIRSIG releases for some time. In DIRSIG-4.4.0, we tried to make demos easier to use. The "demos" folder now includes a top-level HTML index file. On a Windows installation, this HTML index is can be opened by clicking the "Demonstrations" item in the "DIRSIG 4" sub-menu of the Start menu. On Mac and Unix/Linux, just open $DIRSIG_HOME/demos/index.html in your browser. This shows a thumbnail image for each demo and gives a brief description of what the example demonstrates. Note the "Download this demo" link under each description. Whe...