Rapid Road Edge Detection Based on Biomimetic Image Processing

Many soldiers’ lives and expensive equipment are lost annually due to rollover of Army vehicles. A recent Combat Readiness Center report revealed that a main cause of rollover is the driver’s inability to assess rollover threat (e.g., cliff, soft ground, water) on the passenger side of the vehicle. Physical Sciences Inc. and the University of Michigan Transportation Research Institute propose to develop an automated driver assistant system for road edge detection and rollover prevention. The system consists of a camera unit, a laser scanner, and a processing and display unit. The cameras provide visible and LWIR images of the roadside and the laser scanner measures terrain steepness. The processor parses the images and terrain steepness information to detect road edges, identify road boundary, detect rollover threat, and issue audio-video warnings to the driver. An LCD display also serves as interface with the driver. During this program, we will: (1) develop a sensor module by integrating the cameras and the laser scanner into a unit that meets MIL-STD-810, (2) develop a processing and display module by integrating a digital signal processor with the Driver Vision Enhancer, (3) test and characterize the system, and (4) analyze human factors and user variability issues.