Abstract:

A method on occupant detection under a single video camera was presented, which used a window classifier to explore occupant header in the image. The window classifier was constructed by histograms of oriented gradients and support vector machine classifier, and it was trained by occupant header and nonoccupant header. To get multiscale detection results, nonmaximum suppression was used to get the maxim point in the space of position and scale, which is occupant header region. In experiment, 200 occupant header samples and 200 nonoccupant header samples under occupant different pose condition were trained to get the classifier mode between occupant header and non occupant header, and occupant videos including different occupant poses were tested by the proposed method. The results show the method can satisfy the need of occupant detect in smart airbag application.

Key words: video detection, histograms of oriented gradients （HOG）, occupant detection, support vector machine （SVM）, road safety, smart airbag