This paper presents two approaches for despeckling of SAR and InSAR data using an optimal dual based l1 analysis. The energy term was modeled using optimal dual based l1 analysis for both approaches. The data fidelity term for amplitude of SAR data was modeled using a Nagakami distribution. The InSAR data were modeled using the likelihood function of an InSAR model, which considered reflectivity, coherence and phase. The minimization of cost function was solved using a quasi Newton approach and forward backward splitting algorithm for amplitude and InSAR data, respectively. The experimental results showed promising results in despeckling of amplitude SLC SAR data compared to the PPB and SARBM3D methods. The proposed method removed noise well in the interferometric phase of SAR data and it preserved structures within the interferometric phase and achieved better results compared with the Non-local means method for despeckling of InSAR data.
COBISS.SI-ID: 21562646
The motivation and goal of this paper was to assess damage caused by fires in Ireland using Sentinel 1 data. To achieve this goal a Change Detection (CD) algorithm is proposed. The novelty in this paper is a feature extraction within Tunable $Q$ Discrete Wavelet Transform using higher order log-cumulants of Fractional Fourier Transform (FrFT), which were fed into a Stacked Auto-Encoder (SAE) to distinguish changed and unchanged areas. The extracted features were used train the SAE layer-wise using an unsupervised learning algorithm. After training the decoding layer was replaced by a logistic regression layer in order to perform supervised fine tuning and classification. The proposed algorithm was compared with the algorithm for classification of temporal changes which used log cumulants of FrFT within the oriented dual tree wavelet transform using Support Vector Machine and SAE for classifier. Experimental results showed that both methods are very efficient in CD for mid-resolution SAR data. The results were verified using the in-situ data, and experimental results showed that post fire damage assessment is possible using the proposed algorithm.
COBISS.SI-ID: 21153558
This paper presents SAR patch categorization. The novelty of this paper is an optimal dual based sparse coding, applied to SAR patch categorization with complex-valued data. The motivation of this paper was to achieve better performances by using dual based $l_1$ analysis than standard $l_1$ analysis. The sparse categorization is implemented within the sparse framework and compared with the bag of visual words (BoW) categorization. A quasi Newton approach was used to solve minimization problems within framework of dual based $l_1$ analysis first on the amplitude part and then on phase part of the SAR data. Experimental results showed that sparse classification achieves very similar categorization results as BoW categorization using manually generated database and real Spotlight TerraSAR-X product.
COBISS.SI-ID: 21153302
This paper proposes an all-optical-fiber sensor for continuous measurements of liquid levels. The proposed sensor utilizes an optically absorbing vanadium doped optical fiber, which is configured as a long-gauge, optically-heated, fiber-optic, Fabry-Perot Interferometer that is immersed into the measured liquid. The sensor is excited cyclically by a medium-power 980 nm optical source, which induces periodic temperature variation and, consequently, optical pathlength modulation within the vanadium doped fiber. The amplitude of this pathlength variation depends on the liquid level and is measured by an interferometric approach. The relation between the liquid level and the amplitude of optical path length modulation caused by the fiber’s temperature variation were investigated analytically, and the theoretical model proved to be in good agreement with the experimental results. Two versions of level sensors are demonstrated experimentally, the first with single-side optical heating power delivery and 0.45 m measurement range, and the second with dual-side power delivery and 1 m of operational measurement span. Experimental measurement level resolutions achieved for these sensors were xxx and xxx mm respectively. The simple and efficient design of sensor and signal integrator system, the latter is based solely on a few widely available telecom components, provides straightforward opportunities for use of the proposed system in a variety of industrial applications
COBISS.SI-ID: 21659926
This paper analyses disturbances and errors in wireless sensor networks using standard IEEE 802.15.4, when co-channel interferences occurs. The main contribution of this paper is a novel analytical model of data level error probability parameters.
COBISS.SI-ID: 20870422