This paper describes a novel non-contact strain measurement method defined as deep learning based on digital image correlation (DDIC). In particular, it is very difficult to measure directly displacement of gauge length during tensile testing of thin films. Therefore, we obtained the image data continuously to observe the behavior of the material during tensile testing. The sequential image data obtained at a specific position is assigned to a multi-channel input to train the deep neural network. As a result, the multi-channel image is composed of sequential images obtained along the time domain. Since these images have a correlation with each other along the time domain in each pixel, the neural network learns displacement, including temporal information. The DDIC method originates from a 3D convolutional neural network, which can extract both the spatial and the temporal domain features at the same time. A 3D convolutional filter is used as the feature extraction part of the network in order to effectively learn the input data. The deep learning is an end-to-end learning method and the deformation between two images can be measured regardless of the parameters for the nonlinear deformation of the images. An elastic modulus of 118 GPa, 0.2% yield strength of 941 MPa, ultimate tensile strength of 1108 MPa and fracture strain of 0.02414 are estimated by applying the DDIC method during a tensile test of BeCu thin film. The results of the DDIC method are compared with the displacement sensor data and digital image correlation data.

No abstract

Discrete Wavelet Transform has proved to be powerful for image compression because it is able to compact frequency and spatial localization of image energy into a small fraction of coefficients. For a long time it was assumed that there is no compression gain when coding the sign of wavelet coefficients. However, several attempts were carried out and several image encoders like JPEG 2000 include sign coding capabilities. In this paper, we analyze the convenience of including sign coding techniques in tree-based wavelet image encoders, showing their benefits (bit-rate saving). In order to exploit the scarce redundancy of wavelet coefficients sign, we propose the use of machine learning approaches, like evolutionary algorithms, to find the best sign prediction scheme that maximizes the resulting compression rate. We have developed a sign prediction module based on the results provided by the evolutionary algorithms, which it is able to work with whatever the tree-based wavelet encoder like SPIHT, LTW, and others. After performing several experiments, we have observed that, by including the proposed sign coding capabilities, the sign compression gain is up to 17%. These results show that sign coding techniques are of interest for improving compression rate, especially when working with large images (2 Mpixel and beyond) at low compression rates (high quality).

Potential maps recorded from the body surface during PTCA-induced inflation provide a controlled study of the electrocardiogmphic changes caused by severe ischemia. We apply the tempoml wavelet transform to this data to separate smooth and rapidly varying parts of the waveforms. We hope to reduce alignment sensitivity, retain temporal resolution and dynamic information, and extmct features which reflect the ischemia-induced changes.

To obtain a more complete picture of cardiac electrical activity, data is aquired from an array of electrodes on the torso surface during a clinical procedure in which a balloon is inflated in an artery to clear obstructions. The goal is to analyze the data and extract features which reveal the occurrence and location of the balloon inflation, and to relate these features to underlying physiological changes. A multidimensional time and space Wavelet Transform is applied to the data to enable extraction of relevant features. Both scaling and wavelet coeffiecients contain useful information.