Over the past few decades, passive millimeter-wave (PMMW) sensors have emerged as useful implements in transportation and military applications such as autonomous flight-landing system, smart weapons, night- and all weather vision system. As an efficient way to predict the performance of a PMMW sensor and apply it to system, it is required to test in SoftWare-In-the-Loop (SWIL). The PMMW scene simulation is a key component for implementation of this simulator. However, there is no commercial on-the-shelf available to construct the PMMW scene simulation; only there have been a few studies on this technology. We have studied the PMMW scene simulation method to develop the PMMW sensor SWIL simulator. This paper describes the framework of the PMMW scene simulation and the tentative results. The purpose of the PMMW scene simulation is to generate sensor outputs (or image) from a visible image and environmental conditions. We organize it into four parts; material classification mapping, PMMW environmental setting, PMMW scene forming, and millimeter-wave (MMW) sensorworks. The background and the objects in the scene are classified based on properties related with MMW radiation and reflectivity. The environmental setting part calculates the following PMMW phenomenology; atmospheric propagation and emission including sky temperature, weather conditions, and physical temperature. Then, PMMW raw images are formed with surface geometry. Finally, PMMW sensor outputs are generated from PMMW raw images by applying the sensor characteristics such as an aperture size and noise level. Through the simulation process, PMMW phenomenology and sensor characteristics are simulated on the output scene. We have finished the design of framework of the simulator, and are working on implementation in detail. As a tentative result, the flight observation was simulated in specific conditions. After implementation details, we plan to increase the reliability of the simulation by data collecting using actual PMMW sensors. With the reliable PMMW scene simulator, it will be more efficient to apply the PMMW sensor to various applications.
Proc. SPIE. 5789, Passive Millimeter-Wave Imaging Technology VIII
KEYWORDS: Wavelets, Microelectromechanical systems, Deconvolution, Radiometry, Passive millimeter wave sensors, Point spread functions, Electronic filtering, Super resolution, Image filtering, Signal to noise ratio
A poor inherent resolution capability of the passive millimeter-wave (PMMW) imaging becomes a problem in many applications. The need for efficient post-processing to achieve resolution improvement is being increasingly recognized. To obtain high- and super-resolution PMMW imaging, many restoration methods have been developed and evaluated. In this paper, two recent advanced wavelet based methods are discussed; Fourier-wavelet regularized deconvolution (ForWaRD) and multiscale entropy method. The ForWaRD is a linear deconvolution algorithm that performs noise regularization via scalar shrinkage in both the Fourier and wavelet domains. The ForWaRD has been reported to be efficient and applicable to all ill-conditioned deconvolution problems. The multiscale entropy method, which generalized the wavelet-regularized iterative methods, is advance of the maximum entropy method (MEM), which is more effective and leads to efficient restoration. These two methods have not been applied and analyzed in the PMMW images which were highly blurred and low signal to noise circumstance. We have studied the restoration performance of wavelet-based methods in the PMMW imaging comparing with particular reference to the Lorentzian method. The evaluation has been performed with actual radiometer imaging with the 94 GHz mechanically scanned radiometer as well as simulation. In the actual radiometer imaging, a simple blind restoration method was exploited with blur identification. To compare the restored image fidelity, objective and subjective criteria were used, and the super-resolution capability was also checked. Comparison of the linear and non-linear methods revealed the preferable bandwidth extension of the non-linear methods. In the non-linear methods, the multiscale entropy and Lorentzian, they showed their strength and weakness.