Vision detection of vehicle occupant classification with Legendre moments and Support Vector Machine

Intelligent airbag system can control its deploy time and force according to different types of occupant in different sitting position. The accurate detection of vehicle occupant is the precondition and plays an important role in such system. This paper presents a vision detection method using low-cost CMOS camera and pattern recognition algorithm for the classification of different occupant classes. First, in view of the relative fixed passenger room, pre-processing algorithm and occupant edge detection method for the measurement space of occupant image are established; then, principal feature of occupant image edge in the measurement space is described using Legendre moments; at last, classifier of vehicle occupant is obtained using Support Vector Machine so that the classification of vehicle occupant is made possible.