This electronic flyer highlights our capabilities and activities in the area of IMAPCAR® Development Capabilities. Please sign our guestbook. For additional information, e-mail Kevin J. Alley, Southwest Research Institute.

IMAPCAR® Development Capabilities

Southwest Research Institute® (SwRI®) engineers use methodologies based on thorough studies of cutting-edge research in industry and academia. The processing power of the NEC IMAPCAR is integrated into multiple approaches for image processing that produce greater robustness within a single application.

For example, rather than using template matching only for road sign recognition, SwRI engineers use highly capable processing platforms. This enables SwRI-developed algorithms to examine an image more thoroughly, searching for object properties such as symmetry, corner angles, size and color. By combining these metrics, SwRI engineers develop complete detection schemes rather than proposing a solitary method.

Capabilities

  • Highly parallelized image processing algorithms
  • Image processing with scale invariant features
  • Region of interest (ROI) detection and processing
  • User-friendly output
  • Color and spatial feature extraction
  • Sensor fusion

Algorithms

  • Lane tracking
  • Symmetry scanning
  • Template matching
  • Spatial-statistical background elimination
  • Vanishing point detection
  • Inverse perspective transformation
  • Image warping

Frame capture from SwRI's IMAPCAR lane detection algorithm

Lane Detection Technology

SwRI’s lane tracking algorithm for IMAPCAR constitutes a unique, comprehensive approach to lane detection. The algorithm is designed to fully recognize the relevant portion of a scene in order to find consistent lane boundaries.

Conventional algorithms are based on shape features.
Most algorithms avoid color information because it is unreliable due to variable lighting conditions and road surfaces. SwRI’s lane detection algorithm uses an adaptive technique combined with spatial data to extract relevant color information.

 

SwRI's statistical background elimination technique illuminates pedestrians at a busy intersection

Rather than focusing on detecting a lane marker, the SwRI-developed algorithm is designed to identify the drivable surface.
In both human and machine vision, context is a powerful tool. SwRI’s strong focus on context resulted in an algorithm that can be used to produce a more complete scene understanding.

Color information provides a source that is distinct from shape-based information.
Fusing multiple approaches based on spatial and color information is the key to performance of the lane detection algorithm.

 

Frame capture from SwRI's lane detection algorithm showing the estimated vanish point. Knowing the vanishing point enables the processor to quickly eliminate irrelevant information.

SwRI engineers design algorithms that use the final or intermediate results of other algorithms to minimize computational cost.

Design Guidelines

  • Avoid approaches that require absolute thresholds.
  • Incorporate optics predictions with end-product configurable parameters.
  • Focus on system integration.
  • Design transparent code, avoiding black-box solutions that may cause unexplainable results. Explainable results can be addressed in ways other than a complete rebuild.

Frame capture of SwRI's lane detection and vehicle detection algorithms working together. Detected lanes provide a context for vehicle detection.

Application Areas

IMAPCAR’s highly parallelized architecture allows SwRI engineers to develop robust sensor processing and fusion algorithms for visual, infrared, LIDAR and radar data. SwRI’s IMAPCAR development capabilities can be applied to any industry requiring robust, real-time, accurate sensor processing. Example applications include:

  • Automotive
  • Mining
  • Agriculture
  • Construction
  • Industrial
The Aerospace Engineering Department in the Aerospace Electronics and Information Technology Division designs and fabricates hardware for aircraft and ground support equipment, ranging from microcircuits to major subsystems. A variety of platforms support extensive software development. The Autonomous Systems and Controls Section provides engineering services to clients in the areas of autonomous controls for air and ground vehicles, systems integration, interface requirements analyses, prototype hardware and software development, and advanced model-based design and simulation.

 
This flyer was published in April 2009. For more information about IMAPCAR® Development Capabilities, please contact Kevin J. Alley, Phone (210) 522-4660, Aerospace Electronics and Information Technology Division, Southwest Research Institute, P.O. Drawer 28510, San Antonio, Texas 78228-0510.

Aerospace Electronics and Information Technology Division Brochures
SwRI Brochures separate.gif (834 bytes) SwRI Publications
SwRI Technical Divisions separate.gif (834 bytes) SwRI Home