Sensor-Based Nonlinear and Nonstationary Dynamic Analysis of Online Structural Health Monitoring
The proposed framework improves the diagnostic and prognostic schemes for damage-state awareness and structural life prediction in civil engineering structures. The underlying research achieves three main objectives, namely, (1) sensor placement optimization using partial differential equation modeling and Fisher information matrix, (2) structural damage detection using quasi-recursive correlation dimension (QRCD), and (3) structural damage prediction using online empirical mode decomposition (EMD).
The research methodology includes three research tasks: Firstly, to formulate the optimal criteria for the sensor placement optimization damage detection problem based upon a partial differential equation (PDE) analytical model. The PDE model is derived and then validated through experimental results using correlation analysis. Secondly, to develop a novel Quasi-recursive correlations dimension method for structural damage detection. The QRCD algorithm is integrated with an attractor analysis and overlapping windowing technique. Thirdly, to design an online structural damage prediction method based on empirical mode decomposition. The proposed SHM prediction scheme consists of two steps: prediction based change point detection using Hilbert instantaneous phase, and damage severity prediction using the energy index of the most representative Intrinsic Mode Function.
Study results show that; (1) the proposed optimal sensor placement method leads to an optimal spatial location for a collection of sensors, which are sensitive to structural damage, (2) the proposed damage detection algorithm can significantly alleviate the complexity of computation for correlation dimension to approximate O(N), making the online monitoring of nonlinear/nonstationary processes more applicable and efficient; and (3) the proposed empirical mode decomposition method for online damage prediction overcomes the boundary effects of the sifting process, and it has significant prediction accuracy improvement (greater than 30%) over other commonly used prediction techniques.
In this research work, an implementation of neural networks and image processing methods to detect and classify the value of arecanuts. A neural network classifier for back-propagation was used to classify the value of the arecanuts. Arecanut is often infected with different pathogens, including plants, bacteria, viruses, and damaging insects. The HSI system is used to portray the object color and the contrast curve increases the variations in brightness evenly across the images dynamic spectrum. In the sector of arecanut advertising, computer vision technology provides an alternative to replace manual sorting.
Selective Image Encryption of Medical Images Based on Threshold Entropy and Latin Square Image Cipher (LSIC)
In medical image applications, selective image encryption plays an important role as it reduces computational cost and time. Lot of existing full image encryption algorithms may be more complex and uses traditional techniques. In this paper, a new approach developed for protection of medical images. The algorithm is based on a combination entropy calculation and LSIC. In order to reduce the blocking artifacts after the partition of medical image, neglect the insignificant pixels present in the image sub blocks based on the thresholding of Lower four binary planes. The proposed technique is projected to reduce the execution time of the encryption process and increase the robustness of the encrypted image.