Adaptive Least Research Articles

Concentration and velocity field measurements in turbulent flows using laser-induced fluorescence tomography (LIFT)

Laser-Induced Fluorescence is used to tomographically produce volume information of concentration distribution in a turbulent shear flow. Based on Adaptive Least Squares Correlation (ALSC) of grey level distributions in small patches cut out of consecutive tomographically constructed observation volumes, 3-D fields of velocity, vorticity and rate-of-strain are determined with high spatial resolution, satisfactory temporal resolution and high accuracy. This novel technique opens new perspectives for the study of mixing processes.

Fuzzy Adaptive Least Squares and Its Application to Structure-Activity Studies

Fuzzy adaptive least squares (FALS91), a pattern recognition method for analyzing structure-activity rating data to generate QSAR models, has been developed. A novel feature of FALS91 is that the degree to which each sample belongs to its activity class is given using a fuzzy membership function. This paper first describes the algorithm and calculation procedure of FALS91, and then shows its application to the correlation of structure with the activity rating of 31 calmodulin inhibitors and 29 α-methylene-γ-butyrolactones with allergenic activity. Considerably high reliability was shown in both recognition and leave-one-out prediction of the FALS91 analyses.

High-precision image matching for digital terrain model generation

The Adaptive Least Squares Correlation (ALSC), combining gray-level correlation with geometric conditions, was applied to a grid-sampling mode for generation of Digital Terrain Models (DTM) from aerial stereo models. The heights of targets of different geometric and contrast quality in large-scale digitized images were determined using this technique and compared to results from manual measurements. Both solutions showed an average difference of 0.016–0.038% and a maximum differences of 0.037–0.120% flying height (hg), depending on the height approximation. Most cases investigated required an initial height approximation at the 1–2% hg level. An average height error of 0.2% hg was cleared in each iteration step. The constraints imposed proved to define the movement of the gray-level image patches, thus limiting the search area and probability of false correlation, and increasing the precision and reliability of height determination.

Image Averaging Using the Adaptive Two-dimensional Least Mean Square Filter

Abstract This paper proposes that a two-dimensional least mean square (TDLMS) algorithm can be used in a multi-input single-output adaptive form for image averaging. The filter offers a number of advantages over earlier methods used for image averaging and does not require image alignment or any correlation operation before the averaging process. Results are included to illustrate the TDLMS averager's operation in comparison with the direct averager method. The results show that the adaptive averager performs as well as the direct averager when the images are perfectly aligned. When they are not aligned the adaptive averager has a better performance.

Adaptive Least Squares for Parametric Spectral Estimation and Its Application to Pulse Estimation and Deconvolution of Seismic Data

The method of recursive least squares, which has been used extensively in the field of system identification will be developed for adaptive parametric spectral estimation of digital signals. The method will be applied for adaptive pulse estimation and deconvolution of seismic data. Unlike the Levinson-type deconvolution method, the adaptive least-squares method does not need to estimate a priori the autocorrelation function and it avoids the windowing problem of a time-limited signal. Because of the recursive nature of the method, it is suitable for adaptive estimation and removal of time-varying pulses, and it is computationally simple and suitable for implementation on small computers with less memory requirements. The method has been implemented on different simulated examples, and the results are given and discussed.

A Robust and Efficient Least Squares Estimation Algorithm

A new modular structure of an adaptive Least Squares estimation procedure is formulated. The procedure consists of three main-blocks. The first block supervises the estimation scheme by checking whether the observations are persistently exciting. This condition is formulated in the geometrical framework of the singular value decomposition. And this leads to a very reliable and recursive way of checking this condition. This allows in the first block to detect parameter dependencies and to reduce the system model. The second block performs the least squares estimation. Different recursive schemes have been compared and evaluated for efficiency and reliability. An analysis of the propagation of the round-off errors through these different filter structure is performed. This lead to the conclusion that the persistently excitation condition is required to assure numerically stable computations. A last block detects parameter changes and prescribes typical forgetting patterns. The different building blocks and their interaction have been implemented in software. Simulations for aircraft model identification have already demonstrated its real-time usefulness.