Digital Signal Processing Reference
In-Depth Information
An Effective TBD Algorithm for the Detection
of Infrared Dim-Small Moving Target in the Sky Scene
Lisha He, Lijun Xie, Tian Xie, Haibin Pan, and Yao Zheng
Center for Engineering & Scientific Computation, and School of Aeronautics and Astronautics,
Zhejiang University, Hangzhou, Zhejiang, 310027, PR China
helszju@163.com
Abstract.
An effective algorithm for the detection of dim-small moving target in
the infrared (IR) image sequence is described in this paper, which is based on the
idea of Track-Before-Detect (TBD). To deal with the low signal to noise ratio
(SNR) and high false alarm rate of the IR target detection in the sky scene, two of
the Track-Before-Detect (TBD) methods are introduced: dynamic programming
(DP) for the SNR enhancement by energy accumulation, and multistage hypothe-
sis testing (MSHT) to lower the false alarm rate by threshold judgment. Further-
more, constraints as the stabilization of the energy and the continuity of the
movement of IR dim-small target are applied to avoid the energy scatter. And
based on MSHT, most of the false trajectories are eliminated to reduce the calcu-
lated amount and save the storage space. Simulation shows good results for the
detection of IR dim-small moving target based on the algorithm we proposed.
Keywords:
Track-Before-Detect, dynamic programming, multistage hypothesis
testing, dim-small moving target.
1
Introduction
Detect-Before-Track (DBT) and Track-Before-Detect (TBD) are two categories of
detection algorithm for IR dim-small moving target [1]. Research [1] [2] [3]
shows that DBT algorithm adopts a "single-frame detection and multi-frame con-
firmation" strategy, and works well with the image sequence of high SNR by
placing emphasis on the small target's spatial character rather than the temporal
character during the detection procedure. Compared with TBD, the detection
based on DBT is more fast, simpler and easier to implement in real-time, but may
fail in the case of low SNR and target/background contrast. Figure 1 shows the
flow chart of DBT algorithm.
Meanwhile, TBD algorithm adopts a "multi-frame detection" strategy [4-6] [10] to
achieve the target, and both spatial and temporal information are needed when it
works. The algorithm keeps tracking more than one candidate trajectories in the de-
tecting process, and estimates a posterior probability for each one, which will be
compared with a certain threshold at the end of the process. If one's posterior proba-
bility exceeds the threshold, it will be predicted as a target trajectory. It is suggested
that TBD algorithm has a more complex structure and needs more computation and
storage than DBT algorithm, but extremely effective in the low SNR environments.
Figure 2 shows the flow chart of TBD algorithm.
