Computer Vision

Human vision is the most flexible and powerful of the five senses when it comes to understanding the immediate environment. The human eye distinguishes between thousands of hues, provides acuity throughout a great range of ambient lighting, and perceives objects in three dimensions.

Computer vision can provide these functions, and more, using visible and infrared cameras, imaging scanners, ultrasonic probes, and other sensors to gather data for processing by machines. Data processing can use classic image analysis procedures and can computationally mimic human neural physiology.

Computer vision enables robots to navigate orchards and battlefields or to follow precisely the contours of a fighter aircraft. Image analysis and pattern recognition algorithms locate tumors in the body, identify customers at bank automatic teller machines, and detect hidden cracks in machined parts. The possible applications of computer vision are almost endless.

With its expertise in advanced computer technology, Southwest Research Institute (SwRI) conducts research and applied engineering projects in a wide range of computer vision technologies including:

• Image processing
  
• Scene analysis
  
• Pattern recognition
  
• Model-based vision
  
• Neural networks
  
• Fuzzy logic
  
• Electro-optics and lasers
  
• Custom sensor development
  
• Sensor fusion

Vision-Based Motion Control and Guidance

Computer vision enhances robot capabilities to accommodate variations in position and orientation of workspace objects. Computer vision permits object identification and workspace analysis in support of strategic motion planning, and it provides feedback to control repair, positioning, and assembly tasks.

SwRI's vision-assisted robotic systems include:

• Aircraft deriveter
  
• Canopy polisher
  
• Paint stripper
  
• Bottle sorter
  
• Fiber-optic cable assembly
  
• Forge automation

Using a stereo-based laser end effector and computer vision techniques, this robot locates and identifies flaws on the surface of canopy transparencies and removes them.

Industrial Automation and Quality Control

Computer vision systems provide quality control and real-time feedback for industrial processes, overcoming physical limitations and subjective judgment of humans. Automated systems capable of high-speed measurement of parts and flaws yield unprecedented quality.

Computer vision-based industrial quality systems at SwRI include:

• Flaw detection in glass containers
  
• Nuclear fuel pellet inspection
  
• Crack identification in cast and forged parts
  
• Web inspection
  
• Pit and void detection in cast and molded parts
  
• Microencapsulation process control

A high-speed imaging system uses a linescan camera to acquire images of reusable glass beverage containers. A pipelined image processor identifies flaws, cracks, chips, and wear at rates of 1,000 bottles per minute.

An inspection process allows flaw detection and measurement in cast and forged parts.

Automated Precision Measurement

Computer vision techniques automate measurement tasks that require high precision and accuracy. The resulting measurements are used to position and orient parts, provide numerical data for complex computations, and monitor the performance and efficiency of mechanical devices.

Institute-developed applications include:

• Biomechanical stress-strain measurement
  
• Die insertion into printed circuit boards
  
• Photogrammetric measurement of displacements to analyze cracks in materials
  
• Inspection of cutting tool inserts to measure edge sharpness and flaws
  
• Monitoring of blade tip positions in operating gas turbines
  
• Determining metrology of machined parts

Institute engineers obtain precision optical beam measurements to verify flatness and position.

Computer Vision Research

SwRI uses computer vision to investigate the discipline itself and to solve problems in other disciplines.

Institute research activities include:

• Thermal imaging
  
• Infrared image interpretation
  
• Sensor fusion
  
• Neural networks
  
• Fuzzy logic

Image-understanding algorithms identify, track, and complete missing sections of contour lines on topographic maps.

Additional Sensing and Computational Capabilities

SwRI's broad range of expertise and the interdisciplinary composition of project teams provide a synergistic advantage to engineering and development efforts. SwRI is continually expanding its engineering capability.

SwRI's computer vision engineers have contributed to a wide range of projects incorporating alternate sensors and computer technologies, including:

• Train speed monitor using radar and acoustic sensors
  
• Distributed computing system for modeling and visualization
  
• Enhanced robot control using measured tool-workpiece reaction forces
  
• Manufacturing process improvement and new process development

Institute staff members have supported technology assessments for:

• Railway crossing monitor
  
• High-voltage power distribution line inspection
  
• Sensor technology for industrial process control
  
• Materials for new manufacturing processes
  
• Vehicle-animal collision warning system

Radar and acoustic sensors monitor train speed to estimate arrival time at street crossings, permitting early warning of traffic delays to motorists and emergency vehicle drivers.

Facilities

SwRI supports computer vision research and development with modern laboratories that contain high-performance, dedicated image-processing systems, graphics workstations, and personal computer-based image processing systems. Support laboratory equipment includes optical benches, mounting fixtures, and precision-positioning stages, which enable convenient configuration prototyping. As an independent research and development organization, SwRI can use proven off-the-shelf hardware and software for production system application. If commercial equipment does not satisfy project requirements, the Institute offers comprehensive onsite development capabilities.

In an SwRI-funded effort, engineers develop a thermal-imaging system to detect material defects. When fully developed, this system may be used to monitor on-line components and assemblies to reveal hidden manufacturing defects.

Automation and Data Systems Division Brochures
SwRI Brochures SwRI Publications SwRI Home