Enhancing performance of traffic safety guardian system on Android by task skipping mechanism

Abstract

The Traffic Safety Guardian system is a real-time object recognition system based on Android mobile devices to assist car drivers for safety driving. It can analyze driving conditions to remind the drivers with related warning notices. The computing overhead of the system is large because the captured frames have to be processed to recognize interested objects and to track them for analysis of the driving status. In order to enhance the performance of the TSG, we skip some tasks to reduce the computing overhead as well as keep the tracking results correct. The experimental results show that the proposed approach can enhance the performance approximated to 32%.