A Driving Support System Base on Traffic Environment Analysis

Objectives: Vehicle driving are complex action, when drivers couldn’t focus on driving the accident will happen at any moment, we build a system to help driver prevent from traffic accident. Methods/Statistical analysis: In this paper we build a system based on computer vision techniques using support vector machine training detection pattern include distance computing and alert module. At this research we use visual data instead traditional ultrasonic wave data as source data because visual data will provide more environment detail and range than sound wave can, so using the visual data make long rang driving environment alert possible. Findings: In this research we using Haar-like features as system detection feature, because Haar-like feature is a fast and stable detection feature, in the followed test this algorithm total fit our system requires. After small sample Haar-like feature test we use Support Vector Machine to training large sample for system detection pattern and it include 3 parts: pedestrian, vehicle and motorcyclist. Base on driving video sample analysis we found there’s a “hot zone” in camera that most of valuable target will appear in this area, it improved our system computing effectiveness. Application/Improvements: In research we achieve our goal, system can help driver handing complex driving environment to avoid accident happen, future research we will focus on detection accuracy and distance accuracy.