Performance Assessment of a Side Object Detection System 1999-01-1235
A great deal of research and development has been invested in collision avoidance systems. A few such systems are now available commercially, including backup warning systems, backup vision systems for heavy vehicles, forward impact warning systems for heavy vehicles, and side collision avoidance systems for heavy vehicles.
Relatively little data are published regarding basic sensor performance. A limited amount of performance data have been presented in the form of fleet accident avoidance, predictive reliability models, and individual driver accuracy (“Right clear?” questioning samples). However, few instances of actual sensor accuracy measurements are presented in the literature.
This paper discusses parameters which can be used to characterize sensor performance. Methods of data collection are also considered, and two measures are recommended: Detection Accuracy and Mean Time Between False Alarms. Real-world data collection on public roadways is recommended.
Finally, performance data are presented for a sample sensor. This performance measurement is based on approximately 75 hours of data collected on public highways using prototype and early production samples of a side collision avoidance system.
Overall detector performance showed Accuracy of Detection of 99.32%, with a Mean Time Between False Alarms of approximately 3 minutes. Performance showed a correlation with time of day which is reflective of decision factors built into the algorithm used by the system.