Deep Learning Using Symmetry, FAST Scores, Shape-Based Filtering and Spatial Mapping Integrated with CNN for Large Scale Image Retrieval

This article presents symmetry of sampling, scoring, scaling, filtering and suppression over deep convolutional neural networks in combination with a novel content-based image retrieval scheme to retrieve highly accurate results. For this, fusion of ResNet generated signatures is performed with the innovative image features. In the first step, symmetric sampling is performed on the images from the neighborhood key points. Thereafter, the rotated sampling patterns and pairwise comparisons are performed, which return image smoothing by applying standard deviation. These values of smoothed intensity are calculated as per local gradients. Box filtering adjusts the results of approximation of Gaussian with standard deviation to the lowest scale and suppressed by non-maximal technique. The resulting feature sets are scaled at various levels with parameterized smoothened images. The principal component analysis (PCA) reduced feature vectors are combined with the ResNet generated feature. Spatial color coordinates are integrated with convolutional neural network (CNN) extracted features to comprehensively represent the color channels. The proposed method is experimentally applied on challenging datasets including Cifar-100 (10), Cifar-10 (10), ALOT (250), Corel-10000 (10), Corel-1000 (10) and Fashion (15). The presented method shows remarkable results on texture datasets ALOT with 250 categories and fashion (15). The proposed method reports significant results on Cifar-10 and Cifar-100 benchmarks. Moreover, outstanding results are obtained for the Corel-1000 dataset in comparison with state-of-the-art methods.

Unscented Kalman Filtering for Nonlinear Systems with Colored Measurement Noises and One-Step Randomly Delayed Measurements

Traditional unscented Kalman filtering (UKF) cannot solve the filtering problem for nonlinear systems with colored measurement noises and one-step randomly delayed measurements. To fix this problem, a new UKF algorithm is proposed in this paper. First, a system model with one-step randomly delayed measurements and colored measurement noises is established, wherein a first order Markov sequence model for whitening colored noises and an independently identical distributed Bernoulli variable for modeling one-step randomly delayed measurements is introduced. Second, an UKF is proposed for the above established models through unscented transformation by calculating the nonlinear states posterior mean and covariance based on the Bayesian filter framework. Specially, the proportional symmetric sampling method is used in the new UKF algorithm. Finally, the effectiveness and superiority of the proposed method is verified via simulation.

The Square Root UKF-SLAM Algorithm Based on the Smallest Proportion of Skewness in Single Line Sampling

In view of the distortion in the filter gain matrix calculation as well as the high computational complexity and the nonlocal effect of symmetric sampling that exists in the UKF-SLAM algorithm, the square root UKF-SLAM algorithm based on the smallest proportion of skewness in single line sampling was proposed. According to the mended algorithm, the square root of covariance matrix is brought into iteration operation instead of covariance matrix, moreover, the smallest proportion of skewness in single line sampling is utilized for the optimization of sampling strategy. The results of simulation show that the algorithm can effectively improve the position accuracy in robot as well as the estimation accuracy of feature map. Furthermore, the computational complexity is reduced and the algorithm stability is improved.

Soft Sensing of the Lysozyme Mycelium Bacteria Concentration Based on SUKF Algorithm

In order to measure lysozyme biomass activity concentration accurately and in real-time in the fermentation process of marine biological enzyme preparation, soft sensing with the nonlinear state-estimation based on SUKF has been used, the method uses KF framework, embedded in SUT. In fact the fermentation bacteria is lysozyme, which is fermented in a fermenter of KRH-100L according to process requirements. The statistical properties of variables through the nonlinear transformation has been calculated and the degradation effects of aggregation of high-dimensional and nonlinear fermentation model would be effectively settled in sample. By using σ-point set with symmetric sampling strategy, the mean points increased, according to the fermentation of priori information of each dimension mean. By using cross-validation method to select model parameters, compared with the support vector machine SVM with RBFNN algorithm, the experimental results show that the smallest root mean square statistical error of training and testing in soft Sensing with SUKF reduced by 2% or so.

3D symmetric sampling of sparse acquisition geometries

3D symmetric sampling introduced in the 1990s is characterized by dense sampling of two of the four spatial coordinates. The two sparsely sampled coordinates determine the periodicity of the geometry and the dimension of the offset-vector tiles that can be used to generate pseudocommon-offset-vector gathers. These gathers turn out to be useful for prestack processing applications, such as regularization, migration velocity analysis, and azimuthal anisotropy analysis. Although single-point acquisition is the ideal acquisition method, it is not necessarily better than array-based acquisition. Field arrays are still useful in suppressing noise and need not harm signal in most practical cases. In hybrid geometries three spatial coordinates are sampled densely. In all published cases at least two of the three are sampled quite coarsely and may not provide the best quality for the given trace density. Coil geometry (sailing in circles) is a special case of wide-azimuth towed streamer acquisition. It is essentially a random geometry that should be modifiable into a geometry with regularly sampled midpoints, absolute offsets and azimuths. Despite recent technological developments, the basic idea of 3D symmetric sampling still is a highly useful principle for the design of land and marine 3D seismic surveys.