×

Dynamic Bayesian Networks for vehicle classification in video

  • US 9,466,000 B2
  • Filed: 05/21/2012
  • Issued: 10/11/2016
  • Est. Priority Date: 05/19/2011
  • Status: Active Grant
First Claim
Patent Images

1. A method for vehicle classification comprising:

  • detecting at least three subcomponents, the at least three subcomponents comprising vehicle detection, license plate extraction, and tail light extraction;

    using a Gaussian mixture model approach for detection of a moving object, wherein the Gaussian mixture model comprises Gaussian distributions to determine if a pixel is more likely to belong to a background model or not, and an AND approach, which determines a pixel as background only if the pixel falls within three standard deviations for all the components in all three R, G, and B color channels;

    validating detected moving objects by using a simple frame differencing approach;

    removing shadows and erroneous pixels by finding a vertical axis of symmetry using an accelerated version of Loy'"'"'s symmetry and readjusting a bounding box containing a mask with respect to an axis of symmetry, wherein if the shadow is behind the vehicle, removing the shadow using geometrical assumptions such as camera location, object geometry, and ground surface geometry, and wherein given the vehicle rear mask, measuring a height and width of the bounding box, and area of the mask;

    inputting the license plate corner coordinates into an algorithm for license plate extraction, and adding a Gaussian noise with constant mean 0 and variance 0.2 times width to the license plate width measurement; and

    performing a Bayesian network analysis on the at least three detected subcomponents for a plurality of vehicle classifications, wherein the Bayesian network analysis is defined as a directed acyclic graph G=(V, E), where nodes represent random variables from a domain of interest and arcs symbolize direct dependencies between the random variables.

View all claims
  • 0 Assignments
Timeline View
Assignment View
    ×
    ×