The Electrophysics PV320 is a broadband thermal imaging system with several attractive features including low cost (about USD 25K including optics and software), small size, uncooled operation with a BST sensor array, spectral response from 0.6 to 14 μm, easily interchangeable warm optics, and on board USB 2.0 digital video output. In this paper we describe the technical challenges that were involved in integrating together two copies of the PV320L2Z camera variant to create an experimental dual-band IR data acquisition system for measuring targets, backgrounds, and clutter. The PV320 manufacturer-supplied software includes a user friendly, all-in-one application as well as software development kits providing camera control routines that are callable from C++, Visual Basic, and LabView. While this software works well for operating a single PV320 camera, it does not provide any direct support for simultaneously imaging with multiple cameras. The main technical issues are that the base software driver can connect to only one camera at a time and that multiple instances of the driver cannot be loaded simultaneously. Therefore, to achieve our goal of acquiring dual-band IR signatures, it was necessary to program a custom distributed algorithm capable of running two copies of the driver simultaneously on two separate computers with one PV320L2Z connected to each.
In this paper, we study unsupervised image segmentation using wavelet-domain hidden Markov models (HMMs), where three clustering methods are used to obtain the initial segmentation results. We first review recent supervised Bayesian image segmentation algorithms using wavelet-domain HMMs. Then, a new unsupervised segmentation approach is developed by capturing the likelihood disparity of different texture features with respect to wavelet-domain HMMs. Three clustering methods, i.e., K-mean, soft clustering and multiscale clustering, are studied to convert the unsupervised segmentation problem into the self-supervised process by identifying the reliable training samples. The simulation results on synthetic mosaics and real images show that the proposed unsupervised segmentation algorithms can achieve high classification accuracy.