|
|
SVS SoftwareSRI International's Small Vision System is an accessible Developer Kit for stereo applications. The SVS™ libraries calibrate a stereo head, rectify the images to account for distortion, perform stereo correlation to compute a range image, and display the range image in OpenGL 3D form. These algorithms are 3 to 4 times faster than similar algorithms, and have high-quality filtering to reject false stereo matches. The SVS™ can be used with any of the Videre Design stereo heads or on images available in the computer memory, under both MS Windows and Linux. SVS™ is the most efficient, high-quality stereo range software available for standard or embedded PCs. Request Quote & Order |
* Vista compatible for all cameras sold after 12/01/08.
NOTE: For MDCS cameras, the interface uses SSE instructions. These instructions are available in PIII-type processors, including AMD K7 processors and VIA Eden processors
SRI International's patent-pending algorithms are the most efficient ones available -- up to 6 times faster than competing solutions. Realtime capable: 320 x 240, 32 disparities at frame rates (30 Hz) with a PIII 700 MHz.
Excellent range resolution and accuracy, using imagery from Videre Design cameras. Range is interpolated to 1/16 pixel, and texture and consistency filters remove unreliable results.
Simple, fast, accurate stereo head calibration using a printable planar target. You can change lenses, change baseline, even verge your cameras, and still calibrate them accurately. Lens distortion, all internal camera parameters, and external camera parameters are computed.
Easy-to-use C++ library API for MS Windows or Linux makes programming applications simple. Cross-platform display with the FLTK GUI and OpenGL 3D windows let you display results immediately.
SVS has an interface to MatLab, so stereo images and disparity results can be sent directly to MatLab for further analysis.
The XVision2 system, developed by Greg Hager at Johns Hopkins University, has an interface to SVS™. XVision2 concentrates on realtime segmentation and tracking in video images.
SVS™ comes bundled with Videre Design digital stereo heads. It is also available as a software system for a modest licensing fee.
This 640x480 color image on the left, taken with a STH-MDCS stereo camera, shows the fine detail of the imagery. On the right, the SVS™ system has converted the stereo pair into an accurate 3D point cloud. The right image is an animation of the point cloud. The red ray is the camera line-of-sight.
SVS™ was run in standard mode, with no post-processing of the disparity or 3D image, other than the standard texture and L/R filters.
|
|
To fully appreciate the impact of full-motion stereo, you can download these Quicktime movies, which show several people moving in a corridor -- click on the images to show the movies. The left-hand movie is the video image, the right is the stereo disparity image. The frame size is 320 x 240, with 24 disparities, at 15 fps.
 |
 |
XVision2 is the newest version of the XVision system developed by Greg Hager, now at Johns Hopkins University. It is an open-source vision system, concentrating on realtime segmentation and tracking algorithms for video data. The new version incorporates an interface to SVS™, enlarging its scope with realtime stereo input. XVision2 is a convenient way to experiment with new algorithms and applications using stereo.
XVision2 also has a MatLab interface to SVS data, so that the power of MatLab analysis can be brought to bear on your stereo data.
Relevant subdirectories for SVS users are:
| Feature | Specifications |
|---|---|
| Standard Routines |
Calibration of stereo heads using a simple planar target Capturing video streams using standard frame grabbers Computing dense stereo range images at video rates Displaying video images and range information |
| Characteristics |
Two-image stereo computation Arbitrary frame sizes Area correlation algorithm Video-rate implementation at up to 320 x 240 frames on standard PCs |
| Calibration |
Calibration routines by presentation of a simple planar target Internal parameters - radial and tangential distortion, lens decentering, focal length, pixel aspect ratio External registration - baseline, orientation of each camera Based on Tsai's algorithms [Tsai 1991] |
| Rectification | Bilinear interpolation using the calibration parameters |
| Disparity computation |
Laplacian of Gaussian image filter Correlation: sum of absolute differences over a square window Correlation window sides from 5 to 21 pixels Disparity search from 8 to 128 pixels Subpixel interpolation to 1/16 pixel |
| Post-filtering |
Low-texture confidence check Uniqueness check |
| 3D Reconstruction | Transform routines to generate 3D points from image point and disparity |
The Stereo Engine code is written in optimized MMX assembly code for Pentium-based PCs running Linux or MS Windows. The recommended hardware configuration for best performance is a Pentium III/IV processor and a PCI bus, and a display card with at least 8 MB of video memory.
If you have your own cameras, then you must use frame grabbers to digitize the stereo video stream and place it in main memory, where the Stereo Engine can process it. You must write your own code to do this; SVS™ provides function calls to take images from memory and process them.
Videre Design has developed several stereo head assemblies that have direct interfaces to SVS™.
The SVS™ algorithms are optimized for Pentium processors with MMX instructions. Frame rates are a function of frame size (number of pixels) times the number of disparities (search range). Here are some timings on Pentium M and IV processors:
| Processor | Speed | OS | Memory | Resolution | Disparities | FPS | FOM (Mp*d/s) |
|---|---|---|---|---|---|---|---|
| Pentium M | 1.4GHz | MSW | 500MB | 512 x 384 | 48 | 28 | 264 |
| Pentium M | 1.4GHz | MSW | 500MB | 640 x 480 | 64 | 15 | 295 |
| Pentium M | 2GHz | Linux | 1GB | 512 x 384 | 48 | 43 | 405 |
| Pentium M | 2GHz | Linux | 1GB | 640 x 480 | 64 | 22 | 432 |
| Pentium 4 | 2.5GHz | Linux | 500MB | 640 x 480 | 64 | 15 | 295 |
For demanding stereo applications, the recommended PC configuration is a Pentium M. These processors, besides being power-efficient, are better than Pentium IV's at executing integer and MMX/SSE instructions, which are used heavily by the algorithm.
The Figure of Merit (FOM) is the best direct comparison of the efficiency of the algorithm on different systems. It gives the number of pixel-disparities processed per second.
Because the stereo algorithms are storage-efficient, performance scales linearly with increasing frame sizes. The algorithms execute almost entirely from L1 cache, so that future increases in processor speed will translate directly to higher frame rates. The diagram below, which normalizes different frame sizes and disparity ranges to a common scale based on the pixel-disparity law, shows how the amount of processing needed per pixel-disparity stays relatively constant across frame sizes and disparity ranges.
| Part Number | Description |
|---|---|
| SVS-MSW | SRI Small Vision System software – MSW version |
| SVS-LINUX | SRI Small Vision System software – Linux version |