Evidential Reasoning for Control of Smart Automotive Air-Bag Suppression

M.E. Farmer (USA)


Evidential reasoning, Dempster-Shafer, image classifica tion


Evidential reasoning provides a framework for managing beliefs in an environment where information may be un certain, imprecise, and occasionally inaccurate. Image processing and pattern recognition applications provide excellent examples of systems where the outputs are often uncertain and inaccurate. In real-time image processing applications a sequence of images is collected and classi fied over time. During operation of the system, it may experience a variety of environmental conditions that may cause great variability in the robustness of the classifica tion decision. This variability may be due to changes in object orientation, illumination, or occlusion. A process ing framework is required to manage the classification decisions made in this environment in order to reduce the classification error. We propose a framework based on the Dempster Shafer theory of evidential reasoning to process a real-time sequence of image classification results. We demonstrate the performance of our proposed sequential classification processing in the application of a real-time vision system for smart automotive airbags. Our proposed approach leads to nearly a 20% increase over the original classification accuracy of the system, and provides a 10 15% improvement over traditional classifier combination methods.

Important Links:

Go Back