Advanced Image Sensing Methods for Traffic Surveillance

Problem

This study sought to assess advanced imaging technologies for application to roadway surveillance and detection. A major motivation for this study was the possibility of improved effectiveness under foggy or dusty conditions by using wavelengths longer than those of the visible spectrum.

Current electronic imaging methods used to detect vehicles on roadways utilize conventional video equipment which must have adequate lighting, whether natural or artificial, and fairly clear atmospheric conditions. During times of low light or during smoky, dusty, or foggy atmospheric conditions, these systems may be inadequate. Yet, it is in these conditions of low visibility that the greatest need for traffic detection may exist in order to advise motorists of impending dangerous traffic situations.

This project studied several alternatives to visible spectrum imaging technology which have superior ability to operate in conditions of low light or less than ideal atmospheric conditions. The most promising technologies are infrared (IR) sensitive cameras and passive millimeter-wave radiometric imaging.

In order to use these new technologies, computer vision applications were developed and used to compare videotaped or digitally-stored imagery from field work using these systems. The concept of using more than one of these technologies simultaneously was also evaluated.

Findings

The non-visible spectrum imaging technologies evaluated in this study both provide advantages for detection in low light and low visibility situations. However, the cost of these systems is currently very high, making their implementation unlikely for most situations. However, certain situations were identified for which non-visible spectrum imaging technologies may be good candidates:

• Roadways characterized by frequent, dense fog, smoke, or dust in combination with recurrent hazardous traffic patterns, where surveillance could reduce traffic incidents.
• Situations in which the temperature information provided by non-visible spectrum imaging is useful, such as for detection of overheated truck brakes.
• Machine vision applications in which the shadow and glare refection features of IR imaging are required for proper detection or measurement.