The problem of calibrating two color cameras as a stereo pair has been heavily researched and many off-the-shelf software packages, such as Robot Operating System and OpenCV, include calibration routines that work in most cases. However, the problem of calibrating two infrared (IR) cameras for the purposes of sensor fusion and point could generation is relatively new and many challenges exist. We present a comparison of color camera and IR camera stereo calibration using data from an unmanned ground vehicle. There are two main challenges in IR stereo calibration; the calibration board (material, design, etc.) and the accuracy of calibration pattern detection. We present our analysis of these challenges along with our IR stereo calibration methodology. Finally, we present our results both visually and analytically with computed reprojection errors.
The Multiple Hypotheses Tracking (MHT) algorithm has been shown to have the best tracking performance among existing multi-target tracking algorithms using real world sensors with probability of detection less than unity and in the presence of false alarms. The improved performance of the Multiple Hypotheses Tracking comes at the cost of signicantly higher computational complexity. Most Multiple Hypotheses Tracking implementations only form the best global hypothesis. This paper compares the Linear Multitarget Integrated Track Splitting (LMITS) tracking algorithm with the Multiple Hypotheses Tracking algorithm. LMITS has a simpler structure than Multiple Hypotheses Tracking as it decouples local hypotheses and avoids the measurement to multi-track allocation entirely. The number of LMITS hypotheses equals the sum of the number of local hypotheses added to the number of initiation hypotheses. Thus LMITS can retain a deeper hypotheses subtree which can result in better performance. We compare tracking performances of LMITS and MHT algorithms using simulated data for multiple maneuvering targets in heavy and non-uniform clutter.
Despite the enhanced time-frequency analysis (TFA) detailing capability of quadratic TFAs like the Wigner and Cohen representations, their performance with signals of large dynamic range (DNR in excess of 40 dB) is not acceptable due to the inability to totally suppress the cross-term artifacts which typically are much stronger than the weakest signal components that they obscure. AMTI and GMTI radar targets exhibit such high dynamic range when microDoppler is present, with the aspects of interest being the weakest components. This paper presents one of two modifications of linear TFA to provide the enhanced detailing behavior of quadratic TFAs without introducing cross terms, making it possible to see the time-frequency detail of extremely weak signal components. The technique described here is based on subspace-enhanced linear predictive extrapolation of the data within each analysis window to create a longer data sequence for conventional STFT TFA. The other technique, based on formation of a special two-dimensional transformed data matrix analyzed by high-definition two-dimensional spectral analysis methods such as 2-D AR or 2-D minimum variance, is compared to the new technique using actual AMTI and GMTI radar data.