METHOD OF REAL-TIME VEHICLE RECOGNITION AND TRACKING USING KALMAN FILTER AND CLUSTERING ALGORITHM BASED ON HAAR-LIKE FEATURE AND ADABOOST

摘要

Vehicle recognition is an essential technique for making a vehicle as an important part of intelligent vehicle research for stability and convenience. A vehicle recognition technique is in the limelight as a technique for notifying a driver of risk factors such as vehicle collision in advance or automatically controlling the vehicle. The present invention detects a vehicle candidate area by using Haar-like and Adaboost algorithms through a pre-learned detector after setting up an interest area which is possible to show the vehicle in an input image, applies a clustering technique to remove an overlapped area, traces the vehicle in a frame image through a Kalman filter, and applies the clustering technique again in the overlapped area.