Vehicle recognition apparatus

A vehicle recognition apparatus which recognizes a preceding vehicle from road images input by use of vehicle-mounted video cameras and measures a distance to the recognized vehicle, comprising:

• stereo image input means mounted on a vehicle for picking up stereo road scenes in front of the vehicle;
• image storage means for storing the road scenes information input from said stereo image input means;
• edge extraction means for applying a differential processing to the road scenes information stored in said image storage means, to thereby extract edges;
• traffic lane region extraction means for extracting a traffic lane region from the road scenes information stored in said image storage means;
• vehicle edge searching means for producing a histogram by projecting edges which are extracted by said edge extraction means and are included within the traffic lane region extracted by said traffic lane region extraction means, to thereby produce axes of the image, said vehicle edge searching means further provided for extracting a position at which the histogram converges as an edge of a preceding vehicle existing region candidate;
• symmetric region extraction means for extracting a symmetric region within the preceding vehicle existing region candidate defined by said vehicle edge searching means and for further limiting the preceding vehicle existing region candidate;
• initial model setting means for setting an initial contour model for the preceding vehicle from a ratio of a width and height of the preceding vehicle existing region candidate defined by said symmetric region extraction means, said initial model setting means further provided for selecting an initial model having a most adequate configuration among prepared models having various configurations, said initial model setting means further expanding or compressing the selected initial model based on a size of the preceding vehicle existing region candidate, such that a center of the expanded or compressed initial model coincides with a center of the preceding vehicle existing region candidate;
• contour extraction means for defining various features of the preceding vehicle as an energy function, the features being present in the road scenes information stored in said image storage means, said contour extraction means further provided for dynamically transforming the initial model obtained from said initial model setting means by evaluating the energy function, to obtain a closer contour of the preceding vehicle;
• and
• intervehicle distance measuring means for measuring a distance to the preceding vehicle on the basis of the contour of the preceding vehicle extracted by said contour extraction means.