Focused on the intelligent vehicle safety issues under complex traffic environment, this paper presents an algorithm based on an improved SUSAN edge detection to extract the vehicle boundary characteristics. First, we introduce the principle of SUSAN edge detection operator, then present an improved SUSAN edge detection what treats crude extract of pixels and uses an adaptive threshold selection method for the candidate of edge detection to extract the edge. The experimental results indicate that the algorithm can identify in front of the vehicle in the complex image, have higher accuracy.