The scene matching based navigation is an important precision navigation technology for unmanned aerial vehicles (UAV). Selection of interest area where reference image is made has an important influence on the precision of matching result besides the performance of match algorithm. In this paper, a method to select interest area based on structured edge detection is proposed. We use a data driven approach that classifies each pixel with a typical structured edge label. We propose a method that combines these labels into a feature measuring suitable to match of a region. Then a SVM classifier is trained to classify the features and get the final result of the selection of interest area. The experimental result shows that the proposed method is valid and effective.
For the problem of detecting and tracking a varying number of dim small target in IR image sequences, multitarget
track-before-detect approach based on mixture models of probability densities is proposed and mixtures
of t distribution particle filters (MTPF) are developed for the implementation of the proposed approach in this
paper. The existence of each tracked target is detected by using the sequential likelihood ratio test estimated by
the output of component particle filter. New targets are detected by the appearance probabilities in the discrete
occupancy grid in the image frame. The algorithm explicitly handles the instantiation and removal of filters in
case new objects enter the scene or previously tracked objects are removed. The proposed approach overcomes
the curse of dimensionality by estimating each target state independently by using separate particle filter and
avoids the exponential increase in the estimation complexity. Simulation experiments illustrated that the MTPF
algorithm can detect and track the variable number of dim small targets in the IR images, and simultaneously
detect the disappearance and appearance of targets.